sentinel event case study

Lippincott ® Nursing Center ®

Ewen, Brenda M. MSN, RN, CPHRM
Bucher, Gale MSN, RN, COS-C

Adverse events, including sentinel events, require comprehensive review to improve patient safety and reduce healthcare errors. Root cause analysis (RCA) provides an evidence-based structure for methodical investigation and comprehensive review of an event enabling appropriate identification of opportunities for improvement. Use of RCA is described in the home care setting.

Article Content

Every day, serious adverse events occur in healthcare systems across the country resulting in injury to tens of thousands of people annually ( Institute of Medicine, 1999 ). Home care is not immune. Lack of staff supervision, communication, coordination of care, reduced ability to engage in double checks, lack of care environment control, and a heightened reliance on patient and family cooperation are situations unique to home care that contribute to serious adverse events. Some of these events will rise to the level of a sentinel event as defined by The Joint Commission.

Figure. No caption available.

Sentinel Event

The Joint Commission defines a sentinel event as "an unexpected occurrence involving death or serious physical or psychological injury, or the risk thereof" (The Joint Commission, 2012, p. 1). "Risk thereof" refers to incidents for which a recurrence would involve a significant risk of serious adverse outcome. The Joint Commission (2012) further defines reviewable sentinel events as occurrences that result in "an unanticipated death or major permanent loss of function not related to the natural course of the patient's illness or underlying condition" (p. 1). Permanent loss of function may refer to sensory, motor, physiologic, or intellectual impairment requiring continued treatment or change in lifestyle not present at the start of care.

The Joint Commission's policy on sentinel events includes a subset of events that are considered reviewable regardless of death or serious injury ( The Joint Commission, 2013b ). In the past, these events have included occurrences involving patients or those receiving services. In July 2013, this list expanded to include certain "harm events" to staff, visitors, or vendors that occur on the healthcare organization's premises (The Joint Commission, 2012).

Root Cause Analysis

The Joint Commission designates events as sentinel because they require an immediate investigation and response. Accredited organizations are expected to respond to sentinel events with a "thorough and credible root cause analysis [RCA] and action plan" ( The Joint Commission, 2013a , p. 12). RCA can be defined as "a process for identifying the basic or causal factors that underlie variation in performance ( Anderson et al., 2010 , p. 8). RCA is a powerful tool used to improve systems, mitigate harm, and prevent recurrence of adverse events without directing individual blame. These goals are accomplished through in-depth examination of an organization's processes and systems with the purpose of answering three questions:

1. What happened?

2. Why did it happen?

3. What can be done to prevent it from happening again?

Identifying the RCA Team

Preparation for RCA begins immediately after the event is declared sentinel. The Joint Commission allows 45 days for completion of the analysis and development of an action plan. Delays in beginning the process could result in unnecessary stress to meet the deadline. The first step in the RCA process is the identification of team members.

A multidisciplinary team, which includes staff members with knowledge of the processes and systems, allows for an effective analysis of the event. Leadership needs be involved to bring decision-making authority to the table. Individuals able to implement change are needed. The decision to involve staff directly related to the sentinel event should be made on a case-by-case basis. Individuals emotionally traumatized by an event may be further distressed through inclusion on the team.

Teams are most effective when members are chosen for their willingness to participate and cooperate. Honed listening and communication skills are key ( Anderson et al., 2010 ). Members must be motivated with time to attend meetings and accomplish assignments. Members may attend all meetings or do so on an as needed basis.

The team needs to have a designated team leader and facilitator. Leaders with authority in the organization, knowledge of the event, and the ability to build consensus are most capable. The facilitator must be experienced with conducting RCA as well as managing groups. Small teams allow for the greatest efficiency ( Croteau, 2010 ).

Gathering Information

Gathering appropriate information is vital to the team's ability to define the problem and determine what happened. Witness information needs to be gathered quickly before memories begin to fade. Staff must be reassured that RCA is confidential and not used for discipline. Individual interviews can provide information that has not been influenced by others. Clinicians may feel more comfortable discussing the event in private. Group interviews can be used to increase the exchange of ideas and the development of problem-solving strategies. Open-ended questions are an effective means of encouraging staff to share, clarify, or elaborate information.

Pertinent medical records, photographs, notes, and phone logs should be gathered. Relevant policies, procedures, training or education records, time sheets, and schedules should be collected. A literature review, pertaining to the process in question, conducted early in the RCA helps to identify the root cause, strategies, and actions.

If a device or piece of equipment is involved, secure it for examination. Gather manufacturer guidelines, directions for use, and maintenance logs. It should be determined if the Safe Medical Devices Act requires reporting ( http://www.fda.gov/downloads/MedicalDevices/DeviceRegulationandGuidance/Guidance ).

Organizing Information

RCA often involves large amounts of information. It is critical to the success of the analysis that all information is well organized and easy to access. Team charters, agendas, and project plans can be used to outline objectives, set target dates, assign responsibility, and keep the team on track. A brief, factual summary of the event, written early in the process, will keep the team focused. Timelines and flow sheets improve understanding and identify disciplines.

Flow charts, affinity charts, or fishbone diagrams can be used to organize information in a visual format. Flow charts outline a process as it is designed as well as how it is commonly carried out. A comparison between a written process and the way it is implemented provides insight into process failures. Fishbone diagrams highlight contributing factors and causes. Affinity charts organize potential causes. The Joint Commission developed tools, including a RCA framework and action plan template, ensure comprehensive review of the event, and organize findings. Tools can be found at http://www.jointcommission.org/sentinel_event.aspx .

Contributing Factors

After information is gathered and organized, the team starts to identify factors that contributed to the event. Contributing factors are system failures that produce consequences ( Croteau, 2010 ). They are the causes of the event, although not necessarily the main cause. The key to the discovery of contributing factors is the question, "Why?"

When determining contributing factors, discussion needs to focus on outcomes and processes not on individual behavior(s). Examine processes to determine if they are inherently flawed or if a variation in the process occurred leading to the event. All possible contributing factors must be considered. Examples of possible factors include:

1. Human factors (human limitations and capabilities): Human limitations and capabilities such as fatigue, distraction, or inattentional blindness. (See Box 1 .)

Box 1. Additional Resources

2. Patient assessment: Timeliness, accuracy, link to plan of care, documentation, communication.

3. Equipment: Availability, function, condition, appropriate maintenance and calibration.

4. Environmental: Lighting, accessibility, privacy, safety.

5. Information: Accessibility, accuracy, completeness.

6. Communication: Technology, documentation, timing, handoff.

7. Training/competency: Education, scope of practice, competency assessment, qualifications, effectiveness.

8. Procedural compliance: Compliance, availability of procedures and policies, barriers.

9. Care planning: Individualized, effectiveness.

10. Organizational culture: Response to risk and safety issues, communication of priorities related to safety, and prevention of adverse outcomes.

The Joint Commission offers a "Minimum Scope of Root Cause Analysis for Specific Types of sentinel events," which can aid the team in conducting a thorough review of contributing factors (The Joint Commission, n.d.). Members need to participate in conversation analyzing contributing factors. The importance of exchanging thoughts without criticizing must be emphasized. Whiteboards and flips charts are an excellent way to group ideas and ensure that all team members can visualize information. Once the team has identified all possible contributing factors, the root cause can be identified.

Identifying the Root Cause

To identify the root cause, the team will drill down the contributing factors until the root cause, or most fundamental causal factor of the event, is determined. Success depends on the team's ability to remain focused on system issues instead of human error. When a human error is involved, the cause of the error must be identified. It is the cause of the error, not the error, which must be corrected to prevent recurrence.

There are many tools available to assist teams. "Five Whys" is easily used to isolate a root cause ( Anderson et al., 2010 ). The team starts with listing a contributing factor on a white board. They then ask, "Why?" The answer is listed on the white board and becomes the next factor requiring an answer to "Why?" This process continues until no new answer occurs.

For example, in the case of a wound infection, the team may start with the contributing factor of an unintended retention of a dressing.

There was a retained dressing. Why?

The count was not reconciled. Why?

Clinician A was unable to reconcile the dressing count. Why?

Clinician B had not documented the count. Why?

Clinician B forgot to document. Why?

Clinician B didn't have her laptop during that visit and was unable to document until later.

In this example, it takes many "Whys" before the root cause (a delay in documentation) is determined.

Identifying the root cause may be accomplished by asking three questions ( Croteau, 2010 ):

1. Is it likely that the problem would have occurred if the cause had not been present?

2. Is the problem likely to recur due to the same causal factor if the cause is corrected?

3. Is it likely that a similar condition will recur if the cause is corrected or eliminated?

If the answer to each question is "No," then the team has identified the root cause. In the above example, it is not likely that the clinician would have forgotten to document the count if she had been able to document immediately in the home. Nor is it likely a similar problem would occur if the root cause were corrected.

It is essential that the RCA team does not prematurely stop asking "why," so that the true root cause can be identified. The team may consider whether the identified cause is actionable to prevent recurrence ( Croteau, 2010 ). If it is, it may be acceptable to stop questioning. Teams must also recognize that more than one root cause is possible. Interactions between root causes cannot be overlooked and may be the actual precipitators of the event ( The Joint Commission, 2013b ). The correction of one cause does not necessarily mean the recurrence of the event will be prevented. All root causes must be corrected.

The root cause statement needs to be succinct. The Veteran's Health Administration (n.d.) suggests considering the following guidelines while developing the statement:

1. Clearly demonstrate cause and effect.

2. Avoid negative words such as "poor" or "negligent."

3. Every human error has a preceding cause.

4. Procedure violations have a preceding cause; they are not root causes.

5. Failure to act is only a root cause if there is a preexisting duty to act.

Action Plans

After determining the root cause, the team focuses on identifying strategies to reduce the risk of recurrence. Although the goal is to implement interventions to prevent a repeat of the event, the team must understand that failures and errors do occur. Design strategies to minimize the risk a process failure will reach the patient and to mitigate the effects of the failure if it does (The Joint Commission, 2010). Strategies directed at system and process issues, not individual performance or behavior, are most effective in preventing reoccurrence.

Actions that are concrete, easily understood, and clearly linked to the root cause or a contributing factor are most valuable. To avoid work-arounds, make the safest thing to do the easiest thing to do. The plan needs to clearly define who is responsible for implementing each action and a time line for completion. Action plans may include pilot testing. Determine strategies for measuring the effectiveness of each action.

Actions can vary in effectiveness. The National Center for Patient Safety (n.d.) provides a recommended Hierarchy of Actions on their Web site. Stronger actions are thought to be the most successful. Actions are divided into three categories:

* Physical changes to the work environment,

* Forcing functions,

* Simplification of the process, and

* Standardization.

Intermediate:

* Increase staffing,

* Software modifications,

* educe distractions,

* Checklists/cognitive aids,

* Read back,

* Eliminate look and sound alikes,

* Enhanced documentation or communication, and

* Redundancy.

* Double checks,

* New procedures,

* Training, and

* Warnings.

Once proposed actions are decided, cost, resources, long-term sustainability, and barriers to implementation must be considered. Buy-in from leadership and those on the front lines who will be impacted is critical. Those assigned individual actions must take ownership.

Sharing results of the RCA with leadership is necessary. Reports include a brief description of the event, analysis, the root cause, contributing factors, and the action plan. Share lessons learned with all staff. Transparency demonstrates that RCAs are not punitive, but a method to change processes and improve patient safety.

RCA is an excellent tool for identifying causes of sentinel events. The focus on systems and processes instead of performance brings with it a welcome change from past practices of placing blame on individuals. RCA can be used any time a home care agency has a serious adverse event. (See Figure 1 .) RCA can also be used proactively to examine near misses. Instead of asking "what happened," the team asks "what might have happened?" Either way, RCA can improve systems and processes and keeps patients safer.

Figure 1. Process for responding to patient safety events.

RCA Case Study: Retained Foreign Object

A 75-year-old female patient was readmitted to the hospital with a wound infection post abdominal excision of a large seroma and delayed primary wound closure. Negative pressure wound therapy (NPWT) was initiated on January 5 and replaced with a wet to dry dressing prior to hospital discharge on January 8. The patient was admitted to home care and NPWT was reinitiated by Nurse 1. Information on packing count was not made available to the agency and there was no follow-up contact with the hospital staff.

Later that day, the patient complained that the NPWT system was not functioning. Nurse 1 determined the NPWT was defective, and packed wet to dry pending delivery of a new NPWT device. According to the electronic medical record, the wound was packed with six, 4 4 gauze pads, topped with three, 4 4 gauze pads (nine total) and four large abdominal gauzes pads secured with tape during the interim. The packing count removed, packing placed, and description for this dressing was documented in the clinical note.

On January 9, Nurse 2 removed and counted seven pieces of gauze and packed the wound with white foam, covered with black foam, and initiated the new NPWT system with no documentation of packing reconciliation. Seven pieces of gauze removed did not reconcile with the previous note, but went unnoticed. Once the NPWT was in place, the patient received home visits 3 days a week (Monday, Wednesday, and Friday) for wound assessment and dressing changes.

On January 11, Nurse 1 removed the NPWT dressing, including black and white foam as noted and one 4 4 gauze pad found in the wound bed. The nurse made a thorough exam of the wound bed using a sterile Q-tip and flashlight to visualize the deep wound bed. The patient was experiencing an increase in pain and had a temperature of 99.1[degrees]F. The nurse reported the findings immediately to the supervisor and the surgeon. The patient was accompanied by the home care nurse to the surgeon's office for further wound exploration. The patient was started on antibiotics in response to a positive wound culture.

The Joint Commission's policy on sentinel events includes retained foreign body as a reviewable event. This event warranted an immediate RCA. A timeline was created using the medical record. Inpatient records were reviewed to pinpoint when packing could have been retained. Review of inpatient and home care records indicated that it was a possibility that the gauze was retained during the inpatient stay. Because of the lack of documentation reconciliation and/or error in removing all dressings from the wound, the time of packing retention could not be pinpointed.

As one can see from the documentation, the investigation and "what-ifs" can be complex. If the reader is counting, one gauze pad is still unaccounted. The first opportunity missed was communication of packing from the hospital. The second missed opportunity occurred on January 9 when the nurse did not document that the count of packing removed was reconciled with the documentation from January 8. The gauze pads could have been retained at any point where there was no communication and/or reconciliation. A gauze pad could have been saturated in a large wound and gone unnoticed. Do staff count and reconcile cover dressings? How thoroughly are staff checking the wound bed to ensure there are no retained dressings?

The team consisted of the agency's chief nursing officer as leader, medical advisor as champion, risk manager as facilitator, wound ostomy continence nurse, supervisor, and staff nurse representatives. Members were selected to provide expert opinions and offer solutions. The chief nursing officer was essential for decision making and implementation of change. The team began the investigation by finding out what happened from interviews and documentation review. An immediate action was to send an alert to staff regarding the importance of adhering to procedures on packing reconciliation and documentation. It is imperative that staff are notified to reduce likelihood of recurrence even during investigation. The team developed an affinity chart to identify possible cause(s) and contributing factors. (See Figure 2 .)

Figure 2. Causal events chart.

Contributing factors were as follows:

* Process for documenting wound packing and cover dressings was not standardized.

* Lack of available Kerlix for single length packing of wounds.

* Risk of retained packing increases with use of multiple dressings.

* Variation in wound assessment; wounds are inconsistently probed and examined with high-quality lighting.

* Large wound with copious drainage made it more likely that dressings would become saturated and invisible in the wound bed.

* Reconciling counts was inconsistent among staff. This was a new process and nurses were still integrating it into practice.

The team learned that secondary cover and packed dressing materials can saturate and stick together, making it difficult to differentiate from cover and packed materials. The root cause determined by the team: Gauze used to cover wounds are not included in the count and reconciliation process; this practice increases the potential for the cover dressing to be counted as wound packing in large wounds with copious drainage resulting in a retained foreign body . This shows that the cause-and-effect relationship, if controlled or eliminated, will prevent or minimize future events. The root cause statement includes a specific description for the preceding cause, not human error or procedure violation.

Risk reduction strategies/actions were identified to eliminate or reduce the chance that the event would recur. There should be an action for each cause and contributing factor. The following actions were implemented:

* Policy : Referrals involving packed wounds must include packing count for reconciliation.

* Procedure : Revision of wound packing process included a process for counting packing and cover dressings, limiting use of multipieces used for packing and documenting dressings materials on the outside of the dressing. The nurse will immediately notify the supervisor when packing is not reconciled.

* Availability of equipment : Supply a dressing kit including single length Kerlix for use on all NPWT cases in the event that NPWT is interrupted. Upgrade quality of flashlights for wound exploration.

* Communication : Develop a log for patients and family members who change or reinforce dressings. Standardize clinical documentation and evaluate potential for customizing documentation software to include alerts. Adherence is evaluated during record review and shared with supervisors and staff.

* Training/competency : Instruct staff on the rationale for accounting for all dressing materials. Simulation training was utilized for demonstration of NPWT dressings and new documentation requirements.

The actions listed include stronger actions such as simplification (use of single length of packing material) and forcing function (software alerts). Although routine staff training is considered a weaker action, use of simulation is considered highly effective. Each action was assigned to an individual who was accountable.

Equally important was sharing lessons learned with the organization. Home healthcare agencies that are part of a healthcare system may have a structure that requires broader sharing results of the RCA. The committee may include members from other care settings and community experts. In our example, new handoff procedures from one level of care to another can result in increased patient safety.

The use and understanding of RCA is essential to healthcare risk management. Healthcare professionals who master RCAs offer valuable expertise to the organization. Experts drive direct care staff to identify best strategies for patient safety.

Anderson B., Fagerhaug T., Beltz M. (2010). Root Cause Analysis and Improvement in the Healthcare Sector . Milwaukee, WI: ASQ Quality Press. [Context Link]

Croteau R. J . (Ed.). (2010). Root Cause Analysis in Health Care: Tools and Technique . Oakbrook Terrace, IL: Joint Commission Resources. [Context Link]

Department of Veterans Affairs National Center for Patient Safety (NCPS). (n.d). Root cause analysis tools . Retrieved April 5, 2013, from http://www.patientsafety.gov/CogAIds/RCA/index.html

Gosbee J. (2010). Handoffs and communication: The underappreciated roles of situational Awareness and inattentional blindness. Clinical Obstetrics and Gynecology , 53(3), 545-558.

Green M. (2004). "Inattentional blindness" and conspicuity. Retrieved from http://www.visualexpert.com/Resources/inattentionalblindness.html

Institute of Medicine. (1999). To Err is Human: Building a Safer Health System . Retrieved April 3, 2013, from http://www.nap.edu/openbook.php?isbn=0309068371 [Context Link]

Sentinel event policy expanded beyond patients. Joint Commission Perspectives, 32 (12), 1-3.

The Joint Commission. (2013a). Comprehensive Accreditation Manual for Home Care . Oakbrook Terrace, IL: Joint Commission Resources. [Context Link]

The Joint Commission. (2013b). Responding to sentinel events conducting an effective root cause analysis. The Source, 32 (12), 12-14. [Context Link]

The Joint Commission. (n.d.). Sentinel events . Retrieved April 3, 2013, from http://www.jointcommission.org/assets/1/6/2011_CAMLTC_SE_(2).pdf

Connect with NursingCenter

Connect with us on Facebook, Twitter, Linkedin and Google+

All Content

Sentinel events.

Joen Pritchard Kinnan

In July, a teenage mother-to-be entered a Madison, Wis., hospital to give birth. Within hours she was dead, though her baby survived.

An investigation by the Wisconsin State Department of Health revealed that the young woman had died after receiving an intravenous dose of an epidural anesthetic instead of the penicillin she was supposed to be given. Shortly after receiving the injection, the teenager had a seizure. She died less than two hours later.

In explaining what had happened, a nurse told investigators that the patient had been nervous about how she was to be anesthetized during the birth. To ease her concerns, the nurse brought out the epidural bag and told her how it worked. Unfortunately, it was one bag too many; the nurse later confused the epidural bag with the penicillin bag. The consequences were fatal.

The Human Toll

Such sentinel events are all too common. According to a just-released report, Preventing Medication Errors , prepared by the Institute of Medicine (IOM) at the behest of the Centers for Medicare and Medicaid Services, medication errors harm 1.5 million people yearly in the U.S. and kill thousands. The annual cost: at least $3.5 billion. But medication mistakes are just part of the picture.

Sentinel events—unexpected occurrences that result in death or serious physical or psychological injury, or the risk of their later occurrence—can happen anywhere along the healthcare continuum, in any setting. Statistics from the Joint Commission on Accreditation of Healthcare Organizations (JCAHO), however, show that 68% occur in general hospitals and another 11% in psychiatric hospitals. JCAHO tracked the sentinel events they reviewed from 1995 to March of 2006 and found that the most commonly reported sentinel events were patient suicide, wrong-site surgery, operative/postoperative complications, medication errors, and delay in treatment—in that order. Of the total number of cases reviewed, 73% resulted in the death of the patient and 10% in loss of function.

Hard-and-fast statistics on sentinel events are difficult to come by, however. Information from the JCAHO covers only the incidents reviewed by that organization, and experts agree that almost all types of sentinel events are under-reported. Researchers cite a number of reasons that many incidents go unreported; among them are lack of time, fear of punishment, and confusion about the severity of events that require notification. For example, do near misses count? (See “Near Misses,” The Hospitalist , May, p. 34.) Others see no benefit to themselves or their institutions from reporting.

Studies have attempted to define the true incidence of sentinel events, but a lack of consistent language and definitions makes it difficult to put the whole puzzle together, even when sentinel events do come to the surface.

Focus on Medication Errors

Share: , comment on this article cancel reply.

Your email address will not be published. Required fields are marked *

Subscribe to journal Subscribe
Get new issue alerts Get alerts

Secondary Logo

Journal logo.

Colleague's E-mail is Invalid

Your message has been successfully sent to your colleague.

Save my selection

Root Cause Analysis

Responding to a sentinel event.

Ewen, Brenda M. MSN, RN, CPHRM; Bucher, Gale MSN, RN, COS-C

Brenda M. Ewen, MSN, RN, CPHRM , is a Risk Manager at the Christiana Care Visiting Nurse Association, New Castle, Delaware.

Gale Bucher, MSN, RN, COS-C , is the Director of Quality/Risk Management at the Christiana Care Visiting Nurse Association, New Castle, Delaware

The authors wish to acknowledge Gale Faraone and Michelle Campbell for their support and guidance

The authors declare no conflicts of interest

Address for correspondence: Brenda M. Ewen, MSN, RN, CPHRM, Christiana Care Visiting Nurse Association, 1 Read's Way, Suite 100, New Castle, DE 19720 ( [email protected] )

For 24 additional continuing nursing education articles on Quality Improvement topics, go to nursingcenter.com/ce .

Sentinel Event

The Joint Commission defines a sentinel event as “an unexpected occurrence involving death or serious physical or psychological injury, or the risk thereof” (The Joint Commission, 2012, p. 1). “Risk thereof” refers to incidents for which a recurrence would involve a significant risk of serious adverse outcome. The Joint Commission (2012) further defines reviewable sentinel events as occurrences that result in “an unanticipated death or major permanent loss of function not related to the natural course of the patient's illness or underlying condition” (p. 1). Permanent loss of function may refer to sensory, motor, physiologic, or intellectual impairment requiring continued treatment or change in lifestyle not present at the start of care.

The Joint Commission's policy on sentinel events includes a subset of events that are considered reviewable regardless of death or serious injury ( The Joint Commission, 2013b ). In the past, these events have included occurrences involving patients or those receiving services. In July 2013, this list expanded to include certain “harm events” to staff, visitors, or vendors that occur on the healthcare organization's premises (The Joint Commission, 2012).

The Joint Commission designates events as sentinel because they require an immediate investigation and response. Accredited organizations are expected to respond to sentinel events with a “thorough and credible root cause analysis [RCA] and action plan” ( The Joint Commission, 2013a , p. 12). RCA can be defined as “a process for identifying the basic or causal factors that underlie variation in performance ( Anderson et al., 2010 , p. 8). RCA is a powerful tool used to improve systems, mitigate harm, and prevent recurrence of adverse events without directing individual blame. These goals are accomplished through in-depth examination of an organization's processes and systems with the purpose of answering three questions:

What happened?
Why did it happen?
What can be done to prevent it from happening again?

Identifying the RCA Team

Gathering Information

Organizing Information

Contributing Factors

Human factors (human limitations and capabilities): Human limitations and capabilities such as fatigue, distraction, or inattentional blindness. (See Box 1 .)
Patient assessment: Timeliness, accuracy, link to plan of care, documentation, communication.
Equipment: Availability, function, condition, appropriate maintenance and calibration.
Environmental: Lighting, accessibility, privacy, safety.
Information: Accessibility, accuracy, completeness.
Communication: Technology, documentation, timing, handoff.
Training/competency: Education, scope of practice, competency assessment, qualifications, effectiveness.
Procedural compliance: Compliance, availability of procedures and policies, barriers.
Care planning: Individualized, effectiveness.
Organizational culture: Response to risk and safety issues, communication of priorities related to safety, and prevention of adverse outcomes.

The Joint Commission offers a “Minimum Scope of Root Cause Analysis for Specific Types of sentinel events,” which can aid the team in conducting a thorough review of contributing factors (The Joint Commission, n.d.). Members need to participate in conversation analyzing contributing factors. The importance of exchanging thoughts without criticizing must be emphasized. Whiteboards and flips charts are an excellent way to group ideas and ensure that all team members can visualize information. Once the team has identified all possible contributing factors, the root cause can be identified.

Identifying the Root Cause

There are many tools available to assist teams. “Five Whys” is easily used to isolate a root cause ( Anderson et al., 2010 ). The team starts with listing a contributing factor on a white board. They then ask, “Why?” The answer is listed on the white board and becomes the next factor requiring an answer to “Why?” This process continues until no new answer occurs.

For example, in the case of a wound infection, the team may start with the contributing factor of an unintended retention of a dressing.

There was a retained dressing. Why?

The count was not reconciled. Why?

Clinician A was unable to reconcile the dressing count. Why?

Clinician B had not documented the count. Why?

Clinician B forgot to document. Why?

Clinician B didn't have her laptop during that visit and was unable to document until later.

In this example, it takes many “Whys” before the root cause (a delay in documentation) is determined.

Identifying the root cause may be accomplished by asking three questions ( Croteau, 2010 ):

Is it likely that the problem would have occurred if the cause had not been present?
Is the problem likely to recur due to the same causal factor if the cause is corrected?
Is it likely that a similar condition will recur if the cause is corrected or eliminated?

If the answer to each question is “No,” then the team has identified the root cause. In the above example, it is not likely that the clinician would have forgotten to document the count if she had been able to document immediately in the home. Nor is it likely a similar problem would occur if the root cause were corrected.

It is essential that the RCA team does not prematurely stop asking “why,” so that the true root cause can be identified. The team may consider whether the identified cause is actionable to prevent recurrence ( Croteau, 2010 ). If it is, it may be acceptable to stop questioning. Teams must also recognize that more than one root cause is possible. Interactions between root causes cannot be overlooked and may be the actual precipitators of the event ( The Joint Commission, 2013b ). The correction of one cause does not necessarily mean the recurrence of the event will be prevented. All root causes must be corrected.

The root cause statement needs to be succinct. The Veteran's Health Administration (n.d.) suggests considering the following guidelines while developing the statement:

Clearly demonstrate cause and effect.
Avoid negative words such as “poor” or “negligent.”
Every human error has a preceding cause.
Procedure violations have a preceding cause; they are not root causes.
Failure to act is only a root cause if there is a preexisting duty to act.

Action Plans

Physical changes to the work environment,
Forcing functions,
Simplification of the process, and
Standardization.

Intermediate:

Increase staffing,
Software modifications,
educe distractions,
Checklists/cognitive aids,
Eliminate look and sound alikes,
Enhanced documentation or communication, and
Redundancy.
Double checks,
New procedures,
Training, and

RCA is an excellent tool for identifying causes of sentinel events. The focus on systems and processes instead of performance brings with it a welcome change from past practices of placing blame on individuals. RCA can be used any time a home care agency has a serious adverse event. (See Figure 1 .) RCA can also be used proactively to examine near misses. Instead of asking “what happened,” the team asks “what might have happened?” Either way, RCA can improve systems and processes and keeps patients safer.

RCA Case Study: Retained Foreign Object

On January 11, Nurse 1 removed the NPWT dressing, including black and white foam as noted and one 4 4 gauze pad found in the wound bed. The nurse made a thorough exam of the wound bed using a sterile Q-tip and flashlight to visualize the deep wound bed. The patient was experiencing an increase in pain and had a temperature of 99.1°F. The nurse reported the findings immediately to the supervisor and the surgeon. The patient was accompanied by the home care nurse to the surgeon's office for further wound exploration. The patient was started on antibiotics in response to a positive wound culture.

As one can see from the documentation, the investigation and “what-ifs” can be complex. If the reader is counting, one gauze pad is still unaccounted. The first opportunity missed was communication of packing from the hospital. The second missed opportunity occurred on January 9 when the nurse did not document that the count of packing removed was reconciled with the documentation from January 8. The gauze pads could have been retained at any point where there was no communication and/or reconciliation. A gauze pad could have been saturated in a large wound and gone unnoticed. Do staff count and reconcile cover dressings? How thoroughly are staff checking the wound bed to ensure there are no retained dressings?

Contributing factors were as follows:

Process for documenting wound packing and cover dressings was not standardized.
Lack of available Kerlix for single length packing of wounds.
Risk of retained packing increases with use of multiple dressings.
Variation in wound assessment; wounds are inconsistently probed and examined with high-quality lighting.
Large wound with copious drainage made it more likely that dressings would become saturated and invisible in the wound bed.
Reconciling counts was inconsistent among staff. This was a new process and nurses were still integrating it into practice.

Policy : Referrals involving packed wounds must include packing count for reconciliation.
Procedure : Revision of wound packing process included a process for counting packing and cover dressings, limiting use of multipieces used for packing and documenting dressings materials on the outside of the dressing. The nurse will immediately notify the supervisor when packing is not reconciled.
Availability of equipment : Supply a dressing kit including single length Kerlix for use on all NPWT cases in the event that NPWT is interrupted. Upgrade quality of flashlights for wound exploration.
Communication : Develop a log for patients and family members who change or reinforce dressings. Standardize clinical documentation and evaluate potential for customizing documentation software to include alerts. Adherence is evaluated during record review and shared with supervisors and staff.
Training/competency : Instruct staff on the rationale for accounting for all dressing materials. Simulation training was utilized for demonstration of NPWT dressings and new documentation requirements.

View Full Text | PubMed | CrossRef
+ Favorites
View in Gallery

Find Accredited Organizations

Revised definition of suicide in Sentinel Event Policy

The Joint Commission revised its definition of suicide in the Sentinel Event Policy, effective Jan. 1, 2024. The original definition, developed more than 10 years ago, focused on inpatient and “staffed around-the-clock” care settings or suicides within 72 hours of discharge. Data and evidence-based literature support extending the time frame and services in which a patient receives care when considering suicide as a sentinel event.

Additional revisions to the Sentinel Event Policy clarify expectations regarding a health care organization’s partnership and collaboration with The Joint Commission’s Office of Quality and Patient Safety (OQPS). Organizations are strongly encouraged to report sentinel events to The Joint Commission. This gives organizations the opportunity to use the expertise and experience of Joint Commission patient safety specialists to analyze root causes, redesign processes, and monitor performance improvement practices and other aspects of the sentinel event process.

The revisions apply to all Joint Commission accreditation and certification programs, except the Health Care Staffing Services Certification, Integrated Care Certification, Sustainable Healthcare Certification, and Maternal Levels of Care Verification programs.

The following is the current definition of suicide in the Sentinel Event Policy: Suicide of any patient receiving care, treatment, and services in a staffed around-the clock care setting or within 72 hours of discharge, including from the health care organization’s emergency department (ED).

The following is the revised definition of suicide in the Sentinel Event Policy: Death caused by self-inflicted injurious behavior if any of the following apply:

While in a health care setting
Within 7 days of discharge from inpatient services
Within 7 days of discharge from emergency department (ED)
Day Treatment/Partial Hospitalization Program (PHP)/Intensive Outpatient Program (IOP)
Residential
Transitional Supportive Living

The Joint Commission extensively reviewed current evidence-based literature to evaluate the time of highest risk for self-injurious behavior while receiving health care services or post discharge from a health care facility. As a result, the following reflects how the definition of suicide was revised:

Aligns criteria with times of highest risk for suicide.
Fosters a shared mental model among stakeholders through phases of treatment.
Highlights the health care organization’s continued responsibility of ongoing assessment as the patient, resident, or individual served progresses through their treatment plan.

Both the current and revised Sentinel Event Policies are available on The Joint Commission’s Sentinel Event Policy and Procedures webpage . For further clarification, contact The Joint Commission via the ASKOQPS tab on your organization’s Joint Commission Connect ® extranet site.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Publications
Account settings
My Bibliography
Collections
Citation manager

Save citation to file

Email citation, add to collections.

Create a new collection
Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

Search in PubMed
Search in NLM Catalog
Add to Search

Sentinel events. In memory of Ben--a case study

Affiliation.

1 Martin Memorial Health System, USA.
PMID: 10177138

PubMed Disclaimer

Publication types

Search in MeSH

LinkOut - more resources

MedlinePlus Health Information
Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Publications
Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

Advanced Search
Journal List
BMC Health Serv Res
PMC10024850

The relationships between patient safety culture and sentinel events among hospitals in Saudi Arabia: a national descriptive study

Samar binkheder.

1 Medical Informatics and E-Learning Unit, Medical Education Department, College of Medicine, King Saud University, Riyadh, 12372 Saudi Arabia

2 Technical Affairs, Saudi Patient Safety Center (SPSC), Riyadh, 12264 Saudi Arabia

Yasser A. Alaska

3 Emergency Medicine, College of Medicine, King Saud University, Riyadh, 12372 Saudi Arabia

Alia Albaharnah

Rawan khalid alsultan, nawaf mubarak alqahtani, anas ahmad amr.

4 Saudi Critical Care Society, Riyadh, 12243 Saudi Arabia

Nawfal Aljerian

5 Department of Emergency Medical Services, King Saud Bin Abdulaziz University for Health Sciences (KSAU-HS), Riyadh, 14611 Saudi Arabia

6 Medical Referrals Center, Ministry of Health, Riyadh, Saudi Arabia

Rabab Alkutbe

Associated data.

Data that support the findings of this study are available from the Saudi Patient Safety Center (SPSC) and the authors upon request.

Sentinel events (SEs) can result in severe and unwanted outcomes. To minimize the fear of sentinel events reporting and the occurrence of sentinel events, patient safety culture improvements within healthcare organizations is needed. To our knowledge, limited studies explored the relationships between patient safety culture and sentinel events on a local level and no research has been conducted at the national level in Saudi Arabia.

This study aimed to explore the relationships between the patient safety culture and the reported-SEs on a national level during the year 2020 in Saudi hospitals.

This was a descriptive study. We utilized two data sources (the reported-SEs and the patient safety culture survey) that were linked using hospitals information. To explore the relationships between patient safety culture and reported-SEs rates, we performed descriptive statistics, a test of independence, post-hoc analysis, correlation analysis, and multivariate regression and stepwise analyses.

The highest positive domain scores in patient safety culture domains in the Saudi hospitals ( n = 366) were “Teamwork Within Units” (80.65%) and “Organizational learning-continuous improvement” (80.33%), and the lowest were “Staffing” (32.10%) and “Nonpunitive Response to Error” (26.19%). The highest numbers of reported-SEs in 103 hospitals were related to the contributory factors of “Communication and Information” (63.20%) and “Staff Competency and Performance” (61.04%). The correlation analysis performed on 89 Saudi hospitals showed that higher positive patient safety culture scores were significantly associated with lower rates of reported-SEs in 3 out of the 12 domains, which are “Teamwork Within Units”, “Communication Openness”, and “Handoffs and Transitions”. Multivariate analyses showed that “Handoffs and Transitions”, “Nonpunitive Response to Error”, and “Teamwork Within Units” domains were significant predictors of the number of SEs. The "Staff Competency and Performance" and "Environmental Factors" were the most contributory factors of SEs in the number of significant correlations with the patient safety culture domains.

This study identified patient safety culture areas of improvement where hospitals in Saudi Arabia need actions. Our study confirms that a more positive patient safety culture is associated with lower occurrence of sentinel events. To minimize the fear of sentinel events reporting and to improve overall patient safety a culture change is needed by promoting a blame-free culture and improving teamwork, handoffs, and communication openness.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-023-09205-0.

Medical errors were reported as the third major cause of death in the United States (US) in 2016 and it has been estimated that avoidable errors contribute to 1708 death per year in the United Kingdom [ 1 – 3 ]. Medical errors can result in financial costs estimated at billions of dollars yearly as well as psychological and physical harm to patients leading to a lack of trust in the healthcare systems [ 4 ]. To Err Is Human report [ 5 ] was published by the Institute of Medicine (IOM) in 2000 and highlights the responsibility of healthcare professionals in improving the patient safety culture aiming for better patient outcomes [ 6 ]. This problem is not blamed on people but rather on healthcare systems that need to be safer [ 5 ]. The most severe errors are called sentinel event (SE), which is defined by the Joint Commission as “the patient safety event that results in death, permanent harm, or severe temporary harm” [ 7 ]. Furthermore, the Saudi Patient Safety Center explains SE as “an adverse event in health care delivery or other services, which either leads to or has the potential to lead to catastrophic outcomes” [ 8 ]. In Saudi Arabia, the number of complaints and claims relevant to adverse events against healthcare providers has increased [ 9 ]. During the years from 2016 to 2019, there were 727 SEs reported in Saudi Arabia where the top SEs were 38.4% unexpected patient death, 19.4% maternal death, and 11.7% unexpected loss of limb of function [ 10 ]. Thus, the reduction of SEs and harm to patients is one of the groundwork of patient safety that needs to be continuously improved. Therefore, the Saudi Patient Safety Center [ 11 ] was established in 2017 as one of Saudi’s National Transformation Vision of 2030 initiatives to improve patient safety in a national level in Saudi Arabia.

There are several actions to be taken in the case of suspected or confirmed SEs at hospital-level, including a reporting system that helps to identify a root cause analysis, followed by a corrective action plan [ 12 ]. Although reporting medical errors is mandated in many healthcare systems, in the US, for example, only 10% of the events are reported [ 13 ]. According to Anderson and Abrahamson (2017), 15% of the hospitals that have reported errors had a corrective action plan that involved system changes [ 14 ]. An observational study conducted by Hamilton et al. (2017) to evaluate the variance reporting systems, found that during six weeks five observed adverse events were not reported by any reporting systems whereas only two adverse events were handwritten reported [ 15 ]. A scoping review of patient safety in Saudi hospitals found that there is evidence of some cultural aspects that might negatively influence patient safety such as the tendency to report errors with no harm more than those with harm, punitive response to error, and a need for confidential incident reporting [ 16 ]. These findings in agreement with other studies suggested that adverse events require a culture change to minimize the fear of reporting [ 15 , 17 ]. Therefore, one of the critical processes is assessing areas of weaknesses and strengths in patient safety culture among healthcare organizations, including the reporting culture, to maintain better healthcare outcomes and safer organizations.

Patient safety culture (PSC) is one of the vital elements that is evident to affect the patient safety level within healthcare organizations. A systematic review found that the key gaps in patient safety culture in Saudi Arabia are ineffective leadership, a blame culture, inadequate staffing, and poor communication. On the other hand, they identified the s patient safety culture strength factors which are supportive organizational attitudes to continuous improvement, good teamwork within units, and support from management [ 9 ]. Some studies that investigated the impact of PSC on reporting events agreed that positive culture improves patient safety standards in collaborating with the capacity and fearlessness to report errors [ 18 , 19 ]. For instance, Mardon et al. [ 20 ] hypothesized that higher scores on the PSC survey were associated with lower rates of adverse events. Among the top factors that contributed to underreporting of SEs and lower PSC in Saudi Arabia and worldwide are blame culture, poor leadership, lack of staffing, poor communication, fear of reporting sequel, human and work environment factors, the reporting system used, compliance with policy and procedure, fear of punitive action, and lack of understanding of patient safety events [ 16 , 21 – 27 ]. A survey conducted by the Netherlands Federation of University Medical Centers (2020) to assess the handling and learning from SEs found that 5 out of 8 hospitals reported that the reoccurrence of similar events was caused by culture and communication [ 28 ]. Similarly, a study conducted in Saudi Arabia to investigate the most common SEs reported by the hospitals found that policies and procedures, ineffective communication, shortage of staff, and lack of competencies are the common causes [ 10 ]. Therefore, overcoming the barriers and developing a non-blaming, non-punitive learning culture may have a significant role in the reporting system and enhance patient safety initiatives [ 29 ].

Despite the challenges of reporting the SEs that face the health systems, monitoring and analyzing these reports significantly contributes to the effectiveness of improving patient safety [ 12 ]. A study assessed PSC and included 13 hospitals located in Riyadh city in Saudi Arabia found that underreporting of errors due to the fear of blame culture even when harm occurs can result in threatening patient safety and neglecting valuable information on errors as well as inability to analyze these errors [ 30 ]. PSC is a crucial part of organizational culture in which healthcare providers recognize patient safety performance as the highest priority measure to prevent patient harm [ 31 ]. Data generated from PSC is multi-dimensional and can be used to explore relationships and correlations among their attributes and linked to patients’ outcomes; however, there is a gap in the literature due to limited studies in this context [ 32 ]. More specifically, linking the aggregated PSC data with the occurrence of SEs and their contributory factors might have an impact on identifying cultural factors that can positively or negatively influence patient safety within healthcare systems and processes. A systematic review aimed to identify connections between PSC and patients’ outcomes found that there is evidence of relationships between them, such as the negative correlations between PSC and mortality rates [ 6 ]. We hypothesized that hospitals with a more positive PSC have lower numbers of SEs. Furthermore, this study can contribute to providing nationwide hospital-level aggregated PSC and reported-SEs results that can help healthcare leaders and decision-makers to gain insights about areas that need improvements. To our knowledge, limited studies explored the relationships between patient safety culture and sentinel events on a local level and no research has been conducted at the national level in Saudi Arabia.

This study aimed to describe and explore the relationships between the PSC and the sentinel events (number of reported SEs, the categories of contributory factors, and the level of harm) by linking the two data sources during the year 2020 in Saudi hospitals.

Study design

This was a descriptive and exploratory study. Ethical approval was obtained by the institutional review board at King Saud Medical City (IRB Project No. E-22–6551).

Study setting and data sources

The study utilized secondary data from two data sources collected at a national level in Saudi Arabia at the Saudi Patient Safety Center, which are the patient saftey culture survey and the sentinel events reports. The hospitals included all healthcare sectors, which are the ministry of health (MOH), government non-MOH, and private.

Dataset 1: The patient safety culture survey

The Hospital Survey on Patient Safety Culture (HSPSC) questionnaire is a tool developed by the Agency for Healthcare Research and Quality (AHRQ) [ 33 ] to evaluate the PSC [ 34 ]. The survey is distributed yearly through the Saudi Patient Safety Center targeting all healthcare workers anonymously in Saudi hospitals in both Arabic and English languages [ 35 ]. Hospitals were required to register their interest in participating with the Saudi Patient Safety Center by creating an account on the designated electronic survey platform through the authorized point of contact who is an authorized staff at each hospital. The point of contact updated the facility and staff information such as the number of beds, and the number of staff per department. After that, the surveys were distributed electronically to the hospitals by unique links. Surveys were filled in real-time with the ability to visualize the progress by the Saudi Patient Safety Center team, governance sectors, and hospitals. The maximum number of filled surveys was determined based on the total number of staff for each hospital as self-reported by the hospital, where the system did not allow extra surveys beyond the total number of staff. After closing the 8-week cycle of data collection, the hospitals that did not meet the Saudi Patient Safety Center criteria as the following: non-verified registration, non-completed registration (at the end of each cycle), and hospitals with less than 10 completed surveys were excluded.

We utilized the HSPSC survey responses that were collected by the Saudi Patient Safety Center during the period from January 2021 and March 2021. The final dataset included 124,891 responses from 366 hospitals. The average hospital response rate during this cycle was 63.81% (ranging from 2 to 100%). We removed duplicated records and calculated the PSC domains according to the AHRQ guidelines that facilitate comparison between hospitals and the entire database per item, per domain, and per hospital’s domains. The used version in this cycle was version 1 of the HSPSC, which consists of 42 items measuring 12 PSC domains, which are (D1) teamwork within units, (D2) supervisor/manager expectations and actions promoting patient safety, (D3) organizational learning—continuous improvement, (D4) management support for patient safety, (D5) overall perceptions of patient safety, (D6) feedback and communication about error, (D7) communication openness, (D8) frequency of events reported, (D9) teamwork across units, (D10) staffing, (D11) handoffs and transitions, and (D12) nonpunitive response to error. The scoring of the items was a five-point Likert scale of agreement (1 = ‘strongly disagree’ to 5 = ‘strongly agree’) or a scale of frequency (1 = ‘never’ to 5 = ‘always’). Furthermore, two additional PSC measures were included in our study, the patient overall safety grade (POSG) and the positive events reported (PER). For the calculation of the percent positive per hospital for these PSC measures, we counted “Excellent” and “Very Good” as positive responses for “Please give your work area/unit in this hospital an overall grade on patient safety” and counted “1 to 2 event reports”, “3 to 5 event reports”, and “6 to 10 event reports” as positive responses for “In the past 12 months, how many event reports have you filled out and submitted”. In addition, demographic and hospital data were also collected including region, bed capacity, staff position, work area, weekly working hours, and interaction with patients.

Dataset 2: The sentinel events reports

The Saudi Patient Safety Center provides a framework to standardize the root cause analysis process and direct the healthcare facility to improvement. The reporting healthcare facility should complete and submit the root cause analysis and a corrective action plan within 45 days of the SE. The nine categories of contributory factors (CF) (Process Issues, Human Factors, Equipment / Technology, Environmental Factors, Staff Competency and Performance, Manpower Planning Issues, Leadership and Safety Culture, Communication and Information, and Others) were based on the root cause analysis template [ 8 ]. Each category repeatedly asks a series of triggering questions and a checklist of sub-CFs until the root systemic causal factors that culminated in the SE are identified as described in the Saudi healthcare sentinel event manual [ 8 ]. The Saudi Patient Safety Center received the SE reports from healthcare providers and reviewed them internally according to the Saudi healthcare sentinel events manual criteria [ 8 ], including the root cause analyses that were reported by the hospitals for each event and the corrective action plan. Afterward, all data were manually entered after cleaning and reviewed on an excel sheet. All reports with no personal identifiers from January 2020 to December 2020 were included. The number of reported-SEs was 231 from 103 hospitals. The variables included in this study were event location, who was affected (patient, staff, or organization), the level of harm (death, permanent harm, severe temporary harm, no harm) or event outcome, and the categories of CFs.

Data analysis

Descriptive statistics were performed by using measures of frequency to calculate frequencies and percentages [ 36 ] for both data sources (the PSC and the reported-SEs). All the PSC measures were calculated based on the hospital-level positive percentages and then were aggregated for all variables by calculating their averages. To assemble the final dataset, we merged the two data sources by matching hospitals to assess the relationships between the reported-SEs and PSC measures. The test of independence, F-test for numeric variables, was used to test if variables were dependent on the bed size of the hospitals and we reported the following: mean, standard deviation, F-test, and p-value. In addition to the F-test, the Tukey post-hoc test was used for pairwise comparisons between bed size groups ("50–100 beds", "101–200 beds", "201–300 beds", "301–500 beds", and "501 + beds") among all variables with significant differences. Bivariate analysis using spearman correlation was used to measure the relationships between the following variables: the number of reported-SEs, CFs, level of harm, and the PSC domains and measures. We visualized these correlations using a heatmap for all the variables. Multivariate regression was used to identify statistically significant PSC domain predictors for each of the following dependent variables: the number of sentinel events, percent positive of overall patient safety grade, and percent positive of events reported. Multivariate regression was also performed using a stepwise algorithm bidirectional selection [ 37 ]. R [ 38 ], Tableau [ 39 ], SPSS [ 40 ], and Microsoft Excel [ 41 ] were used for data analysis and visualization. The following R functions and packages were used: “vtable” [ 42 ], “TukeyHSD” [ 43 ], “cor” [ 44 ], “corrplot” [ 45 ], “ggcorrplot” [ 46 ], and “stepAIC” [ 47 ]. P -value < 0.05 was considered to be significant.

Descriptive summary of the two data sources

Table Table1 1 shows the percentages of reported-SEs based on the CFs and the positive percentages results of the PSC survey for the whole dataset. The highest CFs were “Communication and Information” (63.20%) and “Staff Competency and Performance” (61.04%). The reported-SEs that were categorized in either of these CFs were received from 75.73% of the reporting hospitals (Table (Table1). 1 ). “Others” (11.26%) was the lowest CF which includes the unavailability of some policies, procedures, or resources. For the PSC survey (Table (Table1), 1 ), the three highest-scored domains within the database were: “Teamwork Within Units” (80.65%), “Organizational learning-continuous improvement” (80.33%), and “Feedback and Communication About Errors” (65.31%). On the other hand, the three lowest-scored domains in the database were: “Communication Openness” (53.43%), “Staffing” (32.10%), and “Nonpunitive Response to Error” (26.19%). The database's average percentage of positive responses across 12 domains was 58.38%.

A summary of the average percent positive responses of the patient safety culture (PSC) across 366 hospitals and the reported sentinel events (SEs) based on the category of contributory factors


		[ ]
Teamwork Within Units	80.65%	82%
Organizational Learning—Continuous Improvement	80.33%	72%
Feedback & Communication About Error	65.31%	69%
Supervisor/Manager Expectations & Actions Promoting Patient Safety	64.39%	80%
Management Support for Patient Safety	64.10%	72%
Teamwork Across Units	59.92%	62%
Frequency of Events Reported	59.45%	67%
Overall Perceptions of Patient Safety	59.17%	66%
Handoffs & Transitions	55.50%	48%
Communication Openness	53.43%	66%
Staffing	32.10%	53%
Nonpunitive Response to Error	26.19%	47%
Across 12 Domains	58.38%	65%
Patient overall safety grade (POSG)	75.71%	78%
Positive events reported (PER)	50.00%	45%


Communication and Information	146 (63.2%)	78 (75.73%)
Staff Competency and Performance	141 (61.04%)	78 (75.73%)
Process Issues	137 (59.31%)	77 (74.76%)
Human Factors	131 (56.71%)	74 (71.84%)
Manpower Planning Issues	108 (46.75%)	73 (70.87%)
Leadership and Safety Culture	89 (38.53%)	59 (57.28%)
Equipment / Technology	58 (25.11%)	44 (42.72%)
Environmental Factors	36 (15.58%)	28 (27.18%)
Others	26 (11.26%)	25 (24.27%)
Not Categorized	35 (15.15%)	27 (26.21%)

The total number of SEs reported during 2020 was 231 from 103 hospitals across Saudi Arabia. The highest percentage of SEs (31.60%, n = 73) was reported in Riyadh region followed by Makkah region (22.94%, n = 53). The number of reported-SEs in our dataset was the highest (29%, n = 67) reported from 25 hospitals with a bed size of “301–500 beds”. For the PSC survey, the total number of respondents was 124,891 from 366 hospitals across different regions in Saudi Arabia. The PSC results showed that 22.92% ( n = 28,622) of respondents were from hospitals with a bed size of “301–500 beds” (9.56%, n = 35). In addition, hospitals with a bed size of “50–100 beds” were the highest (54.37%, n = 199) with 25,403 (20.34%) respondents. The highest number of respondents were working in the emergency department (8.79%, n = 10,979). The highest number of respondents were registered nurses (34.02%, n = 42,485). Appendix 1 and 2 show a descriptive summary of the reported-SEs and the PSC survey before merging the two datasets.

Linking the patient safety culture and the reported sentinel events

Table Table2 2 shows the PSC measures (average positive events reported, average patient overall safety grade, and average percent positive across 12 PSC domains) linked to the hospitals with reported-SEs. After merging the PSC dataset with the sentinel events dataset, the number of reported-SEs was 195 which occurred in 89 hospitals. The highest number of SEs was found in “301–500 beds” bed size with 60 SEs (30.77%), which were reported from 21 different hospitals (23.6%). For the CFs, “Communication and Information” was the highest reported category ( n = 128, 65.64%), followed by “Staff Competency and Performance” ( n = 122, 62.56%), then “Process Issues” ( n = 121, 62.05%). Among the 195 reported-SEs, 132 (67.69%) of the cases had resulted in death, and 13 (6.67%) of the cases resulted in permanent harm.

A descriptive summary after linking the patient safety culture (PSC) measures to the reported sentinel events variables

Variable	Sentinel Events		Percent Positive Response of Patient Safety Culture (PSC) Measures
Variable

	23 (11.79%)	20 (22.47%)	53.14%	74.12%	56.84%
	42 (21.54%)	23 (25.84%)	50.78%	75.17%	56.57%
	25 (12.82%)	14 (15.73%)	48.83%	73.83%	54.70%
	60 (30.77%)	21 (23.6%)	48.24%	74.08%	53.09%
	45 (23.08%)	11 (12.36%)	53.70%	73.90%	56.32%

	128 (65.64%)	70 (78.65%)	50.66%	74.31%	55.28%
	122 (62.56%)	70 (78.65%)	50.22%	73.84%	54.96%
	121 (62.05%)	70 (78.65%)	50.44%	73.68%	55.14%
	115 (58.97%)	66 (74.16%)	50.74%	73.75%	55.04%
	98 (50.26%)	69 (77.53%)	50.56%	74.02%	55.28%
	83 (42.56%)	55 (61.8%)	50.13%	74.06%	55.05%
	47 (24.1%)	38 (42.7%)	49.83%	73.72%	54.58%
	31 (15.9%)	24 (26.97%)	50.11%	70.98%	53.56%
	23 (11.79%)	22 (24.72%)	49.01%	72.24%	53.70%
	28 (14.36%)	21 (23.6%)	54.08%	75.91%	56.98%

	132 (67.69%)	72 (80.90%)	49.89%	73.65	54.83%
	13 (6.67%)	10 (11.24%)	47.16%	73.49	52.59%
	26 (13.33%)	21 (23.60%)	53.75%	76.34	58.17%
	24 (12.31%)	17 (19.10%)	51.55%	74.78	55.47%

	195 (100%)	89 (100%)	50.77%	74.31%	55.49%

Table Table3 3 shows the results of test of independence based on bed size for each of the following: number of reported-SEs, category of CFs, level of harm, and PSC domains. There was a statistical significance difference among hospitals according to bed size for all reported-SEs and CFs, except for “Equipment/ Technology”, “Environmental Factors”, and “Others”. Using post-hoc analysis, we found that the difference in the reported-SEs rates was between “501 + beds” with each of “50–100 beds” ( p = 0.0011),“101–200 beds” ( p = 0.017), and “201–300 beds” ( p = 0.034) as well as between “50–100 beds” with “301–500 beds” ( p = 0.048). Furthermore, “Communication and Information” showed a difference between “301–500 beds” with “50–100 beds” ( p = 0.017) and with “101–200 beds” ( p = 0.025). The “Process Issues” showed that the difference was between “501 + beds” with “50–100 beds” ( p = 0.004), “101–200 beds” ( p = 0.004), and “201–300 beds” ( p = 0.029). The “Human Factors” showed that the difference was between “501 + beds” with “50–100 beds” ( p = 0.041) and with “101–200 beds” ( p = 0.049). The level of harm showed a statistical significance difference only in “Death” ( p = 0.0003). Among hospitals according to bed size and using post-hoc analysis the difference in “Death” was between “501 + beds” with each of “50–100 beds” ( p = 0.0007), “101–200 beds” ( p = 0.0009), and “201–300 beds” ( p = 0.004). As for PSC domains, the results showed a statistical significance difference among hospitals according to bed size in 4 domains. The post-hoc analysis showed that the difference in “Teamwork Within Units” domain was between “301–500 beds” with “50–100 beds” ( p = 0.0001) and with “101–200 beds” ( p = 0.032). The difference for “Organizational Learning—Continuous Improvement” domain was between “50–100 beds” with “301–500 beds” ( p = 0.02). The difference for “Communication Openness” domain was between “50–100 beds” and “301–500 beds” ( p = 0.05). Lastly, “Handoffs and Transitions” domain showed a difference between “50–100 beds” and “301–500 beds” ( p = 0.03).

The relationship between bed size and the number of sentinel events, category of contributory factors, level of harm, and patient safety culture measures

Bed Size	50–100 beds	101–200 beds	201–300 beds	301–500 beds	501 + beds	F-Test	-value
Variable						F-Test	-value

	20 (1.15 ± 0.37) c,d	23 (1.83 ± 1.15) a	14 (1.79 ± 1.19) b	21 (2.86 ± 2.29) d	11 (4.09 ± 4.06) a, b, c		0.001

	15 (1.2 ± 0.41) a	21 (1.38 ± 0.74) b	10 (1.6 ± 0.84)	14 (2.79 ± 1.76) a, b	10 (2.6 ± 2.46)		0.005
	12 (1.17 ± 0.39)	20 (1.35 ± 0.67)	11 (1.27 ± 0.65)	17 (2.41 ± 1.54)	10 (2.6 ± 2.46)		0.009
	12 (1.08 ± 0.29) a	21 (1.29 ± 0.64) b	10 (1.4 ± 0.7) c	18 (2.11 ± 1.18)	9 (3.22 ± 3.07) a,b,c		0.002
	13 (1 ± 0) a	18 (1.17 ± 0.38) b	11 (1.45 ± 0.82)	15 (2.53 ± 1.81)	9 (3 ± 3.54) a, b		0.009
	15 (1 ± 0) a	19 (1.37 ± 0.68)	11 (1.27 ± 0.65)	16 (1.75 ± 0.93) a	8 (1.88 ± 0.99)		0.021
	9 (1 ± 0) a	16 (1.06 ± 0.25) b	6 (1.83 ± 0.98)	14 (2 ± 0.96) a,b	10 (1.8 ± 0.92)		0.002
	7 (1 ± 0)	10 (1.1 ± 0.32)	5 (1.2 ± 0.45)	10 (1.5 ± 0.71)	6 (1.33 ± 0.52)	= 1.462	0.236
	3 (1 ± 0)	7 (1 ± 0)	4 (1 ± 0)	5 (2 ± 1.22)	5 (1.4 ± 0.89)	= 1.94	0.145
	3 (1 ± 0)	6 (1 ± 0)	3 (1 ± 0)	7 (1.14 ± 0.38)	3 (1 ± 0)	= 0.483	0.748
	2 (1 ± 0)	7 (1.29 ± 0.49)	4 (1 ± 0) a	3 (1 ± 0)	5 (2 ± 0.71) a		0.024

	14 (1.21 ± 0.43) a	19 (1.37 ± 0.96) b	12 (1.42 ± 0.79) c	18 (2.28 ± 1.27)	9 (3.44 ± 2.51) a, b,c		< 0.001
	0	6 (1 ± 0)	3 (1.33 ± 0.58)	4 (2.5 ± 3)	4 (1 ± 0)	= 1.01	0.42
	1	2 (1 ± 0)	1	4 (1 ± 0)	2 (2.5 ± 2.12)	= 1	0.486
	5 (1 ± 0)	6 (1.33 ± 0.52)	3 (1 ± 0)	3 (1.67 ± 0.58)	4 (1.25 ± 0.5)	= 1.541	0.238

	20 (81.28 ± 5.55) a	23 (78.53 ± 5.88) b	14 (76.59 ± 5.7)	21 (73.59 ± 5.67) a,b	11 (76.99 ± 4.08)		< 0.001
	20 (63.21 ± 5)	23 (63.68 ± 7.02)	14 (62.31 ± 4.82)	21 (60.54 ± 5.75)	11 (64.10 ± 5.02)	= 1.129	0.349
	20 (80.14 ± 5.37) a	23 (79.13 ± 6.41)	14 (77.82 ± 3.46)	21 (75.34 ± 3.63) a	11 (77.46 ± 4.09)		0.03
	20 (60.57 ± 8.73)	23 (60.12 ± 9.43)	14 (60.52 ± 10.09)	21 (57.68 ± 8.62)	11 (61.80 ± 5.61)	= 0.511	0.728
	20 (57.71 ± 4.48)	23 (58.93 ± 5.71)	14 (57.73 ± 5.08)	21 (56.11 ± 6.09)	11 (56.90 ± 3.65)	= 0.844	0.501
	20 (64.41 ± 6.78)	23 (64.89 ± 7.57)	14 (62.54 ± 6.02)	21 (60.82 ± 5.25)	11 (64.83 ± 4.53)	= 1.523	0.203
	20 (53.30 ± 5.64) a	23 (52.88 ± 6.25)	14 (49.59 ± 3.6)	21 (48.28 ± 6.49) a	11 (49.56 ± 5.13)	=	0.022
	20 (57.61 ± 6.56)	23 (59.22 ± 8.5)	14 (56.71 ± 6.99)	21 (57.39 ± 5.24)	11 (62.95 ± 4.71)	= 1.742	0.148
	20 (57.62 ± 9.61)	23 (54.65 ± 9.45)	14 (52.05 ± 8.65)	21 (50.16 ± 9.35)	11 (55.03 ± 4.19)	= 2.045	0.095
	20 (30.06 ± 2.98)	23 (31.63 ± 3.75)	14 (30.88 ± 4)	21 (30.22 ± 7.09)	11 (32.27 ± 5.67)	= 0.593	0.668
	20 (53.08 ± 9.7) a	23 (50.81 ± 9.74)	14 (47.78 ± 7.26)	21 (44.92 ± 8.8) a	11 (49.45 ± 5.81)		0.047
	20 (23.08 ± 5.15)	23 (24.4 ± 9.09)	14 (21.95 ± 3.66)	21 (22.04 ± 8.3)	11 (24.48 ± 6)	= 0.504	0.733
	20 (56.84 ± 5.12)	23 (56.57 ± 5.53)	14 (54.70 ± 4.56)	21 (53.09 ± 5.59)	11 (56.32 ± 3.42)	= 1.929	0.113
	20 (53.14 ± 8.28)	23 (50.78 ± 10.29)	14 (48.83 ± 8.88)	21 (48.24 ± 7.72)	11 (53.69 ± 4.71)	= 1.361	0.254
	20 (74.12 ± 9.13)	23 (75.17 ± 8.44)	14 (73.83 ± 6.42)	21 (74.08 ± 6)	11 (73.90 ± 6.59)	= 0.104	0.981

Statistical significance markers: * p < 0.05; ** p < 0.01; *** p < 0.001

The significant pairwise comparison differences in means ( p < 0.05) using Tuckey post-hoc analysis are denoted by the same letters (a,b,c,d)

The relationship between the patient safety culture and the reported sentinel events

Patient safety culture domains and reported sentinel events.

Figure 1 shows the correlation matrix as a heatmap to explore the relationships between the PSC domains and the reported-SEs, CFs, and level of harm. We found negative correlations between PSC domains and reported-SEs with r coefficients ranging from –0.30 to –0.001. Higher positive scores in the PSC domains were associated significantly with lower reported-SEs rates, which are “Teamwork Within Units” ( r = -0.30, p = 0.004), “Handoffs and Transitions” ( r = -0.29, p = 0.006), “Communication Openness” ( r = -0.23, p = 0.03). Overall higher PSC aggregated average percent positive across domains were associated with lower reported-SEs rates ( r = − 0.905, p < 0.01). By examining the relationships between reported-SEs and PSC self-reported outcome measures, POSG “overall perception of safety” and PER “positive events reported (at least one event during the last 12 months)” were not significantly related to reported-SEs. The bed size was negatively related to all of the PSC domains except “Frequency of Events Reported” domain with significant correlations between bed size and each of “Teamwork Within Units” ( r = -0.40, p < 0.001), “Communication Openness” ( r = -0.32, p = 0.002), "Organizational Learning—Continuous Improvement" ( r = -0.31, p = 0.003), and “Handoffs and Transitions” ( r = -0.24, p = 0.02). Supplementary 1 shows the spearman correlation coefficient and p-values for all of the variables.

An external file that holds a picture, illustration, etc.
Object name is 12913_2023_9205_Fig1_HTML.jpg

Heatmap of spearman correlations matrix. Legend: The magnitude of the r value is denoted by blue color for positive correlations and red color for negative correlations. CF1: Communication and Information; CF2: Staff Competency and Performance; CF3: Process Issues; CF4: Human Factors; CF5: Manpower Planning Issues; CF6: Leadership and Safety Culture; CF7: Equipment / Technology; CF8: Environmental Factors; CF9: Others; CF10: Not Categorized; LH1: Death; LH2: Permanent Harm; LH3: Severe temporary harm; LH4: No harm; D1: Teamwork Within Units; D2: Supervisor/Manager Expectations and Actions Promoting Patient Safety; D3: Organizational Learning—Continuous Improvement; D4: Management Support for Patient Safety; D5: Overall Perceptions of Patient Safety; D6: Feedback and Communication About Error; D7: Communication Openness; D8: Frequency of Events Reported; D9: Teamwork Across Units; D10: Staffing; D11: Handoffs and Transitions; D12: Nonpunitive Response to Error; D_All: Average percent positive across domains; PER: % Positive Events Reported (at least one event during the last 12 months); POSG: % Patient Overall Safety Grade. Statistical significance markers: * p < 0.05; ** p < 0.01; *** p < 0.001

Patient safety culture domains and contributory factors

We found 33 significant negative correlations (Fig. 1 ) between CFs of the reported-SEs and PSC measure. “Staff Competency and Performance”, “Human Factors”, and “Communication and Information”, showed significant negative associations with average percent positive across PSC domains ( r = -0.26, -0.24, and -0.23, respectively). “Staff Competency and Performance” was correlated negatively with the following 6 PSC domains from the highest to the lowest: “Handoffs and Transitions” ( r = -0.33, p = 0.002), “Teamwork Across Units” ( r = -0.29, p = 0.006), “Teamwork Within Units” ( r = -0.27, p = 0.010), “Overall Perceptions of Patient Safety” ( r = -0.25 p = 0.018), “Organizational Learning—Continuous Improvement” ( r = -0.22, p = 0.037), and "Feedback and Communication About Error" ( r = -0.22, p = 0.041). Furthermore, "Human Factors" was also correlated negatively with 4 PSC domains, which were "Teamwork Within Units" ( r = -0.33, p = 0.001), "Overall Perceptions of Patient Safety" ( r = -0.29, p = 0.007), "Handoffs and Transitions" ( r = -0.28, p = 0.007), and "Communication Openness" ( r = -0.24, p = 0.023). “Environmental Factors” were negatively associated with 8 measures of the PSC, notably the highest and most significant correlations among the 8 measures were "Teamwork Across Units" ( r = -0.30, p = 0.004), "Management Support for Patient Safety"( r = -0.30, p = 0.005), "Patient Overall Safety Grade" ( r = -0.29, p = 0.007), and "Organizational Learning—Continuous Improvement" (r = -0.28, p = 0.008). Remarkedly, the “Handoffs and Transitions” correlates negatively on a significant level with all the CFs apart from the “Manpower Planning Issues”. The bed size was positively correlated with five CFs, which were “Process Issues”, “Leadership and Safety Culture”, “Staff Competency and Performance”, “Human Factors”, and “Communication and Information”.

Patient safety culture domains and level of harm

The results (Fig. 1 ) showed that “Death” was negatively correlated with the following PSC domains: “Teamwork Within Units” ( r = -0.28, p = 0.008), “Organizational Learning—Continuous Improvement” ( r = -0.24, p = 0.02), “Feedback and Communication About Error” ( r = -0.23, p = 0.03), “Communication Openness” ( r = -0.25, p = 0.02), “Teamwork Across Units” ( r = -0.26, p = 0.02), “Handoffs and Transitions” ( r = -0.31, p = 0.003), and average percent positive across domains ( r = -0.27, p = 0.01). For “Permanent Harm”, the results showed that it was negatively correlated with the following PSC domains: “Teamwork Within Units” ( r = -0.27, p = 0.009), “Overall Perceptions of Patient Safety” ( r = -0.23, p = 0.03), “Handoffs and Transitions” ( r = -0.29, p = 0.006), “Nonpunitive Response to Error” ( r = -0.23, p = 0.03), and ( r = -0.27, p = 0.0.009). The bed size was positively related to the reported-SEs ( r = 0.44, p < 0.001) and to “Death”( r = 0.37, p < 0.001).

The influence of patient safety domains on sentinel events, patient overall safety grade (POSG), and positive events reported (PER)

Multivariate regression (Table (Table4) 4 ) was calculated to identify the influence of the 12 PSC domains on the positive percent of the number of reported-SEs (SE model), the patient overall safety grade (POSG model), and the positive events reported i.e. at least one event reported during the last 12 months (PER model). For SE Model, the 12 PSC domains explain a significant amount of variance in the number of reported-SEs which was 28.21% (F (12, 76) = 2.49, p = 0.008). “Handoffs and Transitions” and “Nonpunitive Response to Error” domains were statistically significant predictors of the number of SEs. “Handoffs and Transitions” domain had a greater beta with a significant negative influence on the SE (β = − 0.672, p = 0.012), while “Nonpunitive Response to Error” domain positively predicted the SE (β = 0.458, p = 0.021). Five PSC domains accounted for 25.38% of the variance (F (5, 83) = 5.65, p < 0.001) in the final SE model using stepwise approach. The significant PSC predictors of the SEs were similar to the multivariate model with one more significant domain which is “Teamwork Within Units”. For POSG model, we found that the PSC domains explain a significant amount of the variance in the positive percent of the overall safety grade as a whole was 58.45% (F (12, 76) = 8.911, p < 0.001). The analysis showed that “Overall Perceptions of Patient Safety” and “Nonpunitive Response to Error” made a statistically significant contribution to the prediction of patient overall safety grades. “Overall Perceptions of Patient Safety” recorded a higher beta value than “Nonpunitive Response to Error”; however, “Overall Perceptions of Patient Safety” positively contributed with (β = 0.648, p = 0.001), and “Nonpunitive Response to Error” had a significant negative influence on the overall safety grade (β = − 0.375, p = 0.013). The final POSG model using stepwise approach showed a variance of 57.65% (F (7, 81) = 15.75, p < 0.001) accounting for 7 PSC domains. In addition to the two significant domains that appeared in the multivariate regression, “Staffing” domain was also significant in the stepwise regression model. For the PER model, we found that the 12 PSC domains explain a significant amount of the variance in the percent positive events reported as a whole was 36.91% (F (12, 76) = 3.705, p < 0.001). There were no statistically significant domains in this model. The final PER model showed a variance of 33.01% (F (5, 83) = 8.18, p < 0.001). Unlike multivariate regression, the stepwise regression was influenced by 5 PSC domains where two of which were statistically significant predictors which are “Organizational Learning—Continuous Improvement” and “Frequency of Events Reported” (see Table Table4 4 ).

The multivariate linear regression and multivariate stepwise regression analyses (standardized β coefficients)

Model	Sentinel events (SE) Model		Patient overall safety grade (POSG) Model		Positive events reported (PER) Model

D1: Teamwork Within Units	-0.437		-0.255	-0.184	0.297	0.290
D2: Supervisor/Manager Expectations and Actions Promoting Patient Safety	0.115		-0.183	-0.148	0.007
D3: Organizational Learning—Continuous Improvement	0.328	0.245	0.176	0.200	-0.267
D4: Management Support for Patient Safety	0.235	0.276	0.219	0.255	0.058
D5: Overall Perceptions of Patient Safety	-0.244				-0.284	-0.201
D6: Feedback and Communication About Error	0.101		0.054		0.430	0.368
D7: Communication Openness	-0.178		0.121		-0.187
D8: Frequency of Events Reported	-0.062		-0.039		0.351
D9: Teamwork Across Units	0.168		0.098		-0.405
D10: Staffing	0.061		0.211		0.011
D11: Handoffs and Transitions			0.023		0.297
D12: Nonpunitive Response to Error					0.224

SE Sentinel events, PER % Positive Events Reported (at least one event during the last 12 months), POSG % Patient Overall Safety Grade

This study investigated the relationships between PSC and SEs utilizing two data sources collected on a large scale in Saudi Arabia, which are PSC survey and SEs reports. The overall percentage of positive response of PSC (58.38%) is acceptable when compared to the overall percentage (65%) of the United States in 2018 [ 48 ]. Our study identified PSC domains as areas of strength (positive areas) and areas of weakness (negative areas) in Saudi hospitals that are consistent with the other studies [ 9 , 49 – 51 ]. The top areas of strengths in PSC domains among 366 hospitals in Saudi Arabia were “teamwork within units” and “organizational learning". The top areas of weaknesses in PSC domains were “nonpunitive response to error”, “staffing”, and “communication openness”, which may be also indicative of risky factors affecting the occurrence of SEs. On the other hand, there was a total of 231 SEs reported from 103 hospitals in Saudi Arabia. Two factors mostly contributed to these reported-SEs which are “communication and information” followed by “staff competency and performance”, consistent with previous studies in Saudi and internationally [ 10 , 16 , 52 , 53 ]. As our objective was to investigate the relationships between the PSC and reported-SEs, merging the two datasets resulted in 89 hospitals with 195 reported-SEs that were matched by hospitals to their corresponding PSC measures. The correlation results showed a significant increase in the rates of reported-SEs among the hospitals with larger bed size. Hospitals with larger bed size might be at increased risk of more adverse levels of harm.

When we explored the relationships between PSC and the numbers of reported-SEs, our results indicated that more positive PSC is associated with lower rates of reported-SEs which was consistent with previous studies [ 20 , 54 , 55 ]; with 3 out of the 12 PSC domains were significantly related to lower rates of reported-SEs. These results suggested that improving positive culture in “teamwork within units”, “handoffs and transitions”, and “communication openness” were significantly associated with lower rates of reported-SEs that are consistent with Najjar et al. findings on adverse events [ 54 ]. These three PSC domains showed also lower positive culture among hospitals with bed size “301–500 beds” than in hospitals with lower bed size. Similarly, this pattern was found also with death as an outcome; expectingly, the rate gradually increased in the hospitals with larger bed size. Therefore, it is important to create a safe environment for healthcare workers to communicate openly and report mistakes without the fear of blame or punishment. Our regression analyses showed similar findings for the association between SEs and “handoffs and transitions” and “teamwork within units” domains. However, it also showed that more positive culture around “nonpunitive response to error” was influenced by higher numbers of reported-SEs. The “nonpunitive response to error” domain was one of the domains that gained the most attention as a crucial topic in research locally among Saudi Arabian hospitals and internationally [ 9 , 16 , 51 , 56 ]. This is also evident in our national-level survey results as it was one of the lowest-scored domains among the PSC domains among 366 hospitals in Saudi Arabia. On the other hand, the association of “nonpunitive response to error” with higher numbers of reported SEs might indicate a positive culture of reporting without a fear of blame indicating a safer organization, which is a topic of debate [ 57 – 59 ]. This is evident in the influence of “frequency of events reported” domain on the positive events reported (PER) model as per our stepwise regression analysis. Therefore, it is suggested that caution should be taken when interpreting the results of reported incidents, such as sentinel events, and consideration of supporting data and contextual analysis is critical [ 59 ]. The overall perception of safety and the events reported (at least one event during the last 12 months) were not significantly related to a lower rate of reported-SEs.

The contributory factors explain the root causes of the reported-SEs. We found that more positive culture of "handoffs and transitions" was associated with lower rates of reported-SEs explained by 8 contributory factors, including "staff competency and performance", "process issues", "human factors", and "communication and information". This might indicate that improving handoffs and transitions among staff and healthcare professionals help in decreasing the incidence of SEs. Communication is one of the most common issues that can be triggered during and after the handoff of care. For instance, verbal communication can carry more risks during handoffs when compared to written communication [ 60 ]. The “communication and information” factors justify subfactors such as ineffective handover communication, lack of information, failure to seek support, and misunderstanding of the information. Our findings are also consistent with another study regarding communication during patient handoffs [ 61 ] which found that the language either the usage of the English language, which is not the native language, or the usage of medical terms negatively impacts communication during the handoffs. Moreover, the channel that is used during communication such as facial expression, body language, and eye contact may affect the interpretation of exchanging information [ 61 , 62 ]. Furthermore, senior staff might be busy with leadership tasks, which can affect the performance and the communication of information. Additionally, the environment where this information is communicated among staff can have an impact on the quality and accuracy of the information, such as a noisy environment [ 60 ].

Among the top contributory factors in the number of significant correlations with PSC measures were “staff competency and performance” (e.g., lack of knowledge/skills/competence and inadequate supervision) and "environmental factors" (e.g., poor or inappropriate area design and noise), which indicate that these factors might also be potential areas to be improved among hospitals for a better PSC and lower reported-SEs rates. In addition to its correlation with "handoffs and transitions", “staff competency and performance” have a negative association with the domains that are linked with the teamwork aspects (“teamwork within units” and “teamwork across units”) and error prevention and positive change aspects (“organizational learning”, “overall perceptions of patient safety”, and “feedback and communication about error”. "Environmental factors" was the only contributory factor that had a negative association with the "patient overall safety grade" measure of PSC. Hayashi et al. [ 63 ] examined the relationship between PSC and the working environment and suggested that managing the work environment among healthcare workers can improve PSC. Furthermore, they found that high number of night shifts might increase the number of adverse events. Our results are in agreement with previous findings that the following factors are critical towards a more positive PSC: leadership, teamwork, evidence-based, communication, learning, just, and patient-centered [ 64 , 65 ].

Such identified PSC domains and sentinel events relationships in this study might be considered risky or negative domains require further actions including mitigation actions and interventions to minimize or eliminate their negative impact on patient outcomes. Examples of interventions that hospitals can implement to improve culture and minimize unpredictable events are implementing effective reporting systems to improve “nonpunitive response to error”, appointing a safety champion for every unit to improve and conduct satisfaction surveys “staffing”, using safety briefings to improve “communication openness”, implementing Situation-Background-Assessment-Recommendation (SBAR) technique to improve “teamwork within units”, and relaying safety reports at shift change to improve “handoffs and transitions” [ 66 ]. Other examples of strategies with leadership responsibility to improve response to error, such as “Just culture”, staff supporting programs, and “good catches” [ 67 ]. “Just culture”, for example, can address two of the weakest domains which are “nonpunitive response to error” and “communication openness”. To improve communication during handoffs, the following strategies are suggested: standardization and simplification, avoiding interruptions during handoffs, limiting the use of intermediaries, using a common communication style, implementing a readback or hearback communication process, and keeping communication patient-focused. Staff and healthcare workers must feel safe during communication to speak up and actively participate during the handoffs [ 60 ]. Furthermore, the implementation of the TeamSTEPPS teamwork concept on patient safety culture among hospitals might have an impact in improving some dimensions, including “teamwork within units” and “communication openness” [ 54 , 68 ].

Several areas of patient safety culture and underreporting that have been identified as weaknesses still exist among hospitals in Saudi Arabia. First, we would like to emphasize that hospital management support is critical and showed evidence of improved patient safety culture and reporting culture [ 65 ]. Second, even with the most advanced analytical methods and tools, addressing the underreporting problems requires a culture that engages the acknowledgment of errors [ 56 ]. Learning from the reported-SEs is an essential phase to improve the patient safety culture in the hospital and improve prioritizing of the recommendations [ 28 ]. 7% of the hospitals reported the process issues including stages of the task not being designed or lack of prioritization of guidelines were sub-contributors to the events; another study concluded that prioritizing recommendations with subjective criteria from the reported-SEs is an essential phase of the learning process [ 28 , 69 ]. Therefore, assessing the patient safety culture and its connection to adverse events and sentinel events is a continuous learning process. Third, we believe that adopting technologies as a part of the clinical workflows and reporting process might help in providing structured information resources that can assist in more standardized reporting and better learning processes. For instance, electronic health records (EHR) nowadays can enhance the standardization of handoff communication when used effectively with greater efficiency, accountability, timeliness of communication, and data accuracy and completeness [ 70 ]. In fact, it is highly recommended for hospitals to move to data-driven approaches with transparent reporting of SEs [ 64 ] that can be facilitated by technologies. Fourth, we also would like to highlight the importance of government initiatives to improve patient safety through data-driven approaches, such as those established by the Saudi Patient Safety Center [ 11 ], which has implemented electronic-based tools to facilitate reporting adverse and sentinel events as well as participating in the patient safety culture initiatives. Fifth, our results indicated that “human factors” was among the top contributory factors, it is important to give attention to human factors during communication, such as stress and rushing to complete a task. Developing accurate and efficient healthcare systems and processes is necessary to ensure that these processes are carried out safely during situations, such as handoffs [ 60 ].

Given how patient safety culture is new in Saudi and very few studies have been published on a large scale in this area, we believe our research currently represents the first study with this context of linking patient safety culture to sentinel events and its contributory factors. One of the strengths of this study is that the study was conducted on a national level among participating hospitals in Saudi Arabia when compared to previous studies [ 9 , 51 ]. Additionally, we investigated the contributory factors of the reported-SEs among different healthcare sectors (MOH, government non-MOH, and private hospitals) in Saudi Arabia. A limitation of this study is that the number of reported-SEs was relatively small in comparison to the patient safety culture survey data and merging the two datasets led to a decrease in the number of analyzed hospitals from 366 to 89, which might affect the generalizability of the results on other hospitals and other countries. This might be because the occurrence of reported-SEs is much lower than the occurrence of other adverse events in general. Moreover, data was collected nationally at the hospital level where there might be additional data that might not have been shared. Underreporting is a known issue that we hope in this study we encourage hospitals to report. This study was performed on a one-year duration, future studies can focus on change over the years in both safety culture and the rates of reported-SEs. Future studies can focus on implementing interventions and measuring their impact on PSC and the occurrence of sentinel events. Since this study focused on quantitative approaches, we suggest that future studies should focus also on qualitative approaches to analyze the relationships between patient safety culture and sentinel events. Moreover, future studies can focus on more focused contexts, such as other adverse events or specific contributory factors, and can utilize predictive approaches and machine learning to predict patient safety outcomes.

Conclusions

At a national level in Saudi Arabia, there were limited efforts carried out to measure, unify, analyze, and generate aggregate reports and analytics that addressed patient safety culture domains and sentinel events. Communication is the most highlighted negative domain and has a negative association with the reported sentinel events. To minimize the fear of sentinel events reporting and to improve overall patient safety in Saudi Arabia a culture change is needed by promoting a blame-free culture and improving teamwork, handoffs, and communication openness. Furthermore, there was evidence that a more positive patient safety culture was associated with lower numbers of sentinel events. Lastly, we identified evidence-based areas of strengths and weakness in patient safety culture and their relationships to the contributory factors of the sentinel events that can facilitate implementing future interventions, encourage data-driven approaches to patient safety, and require attention by health organizations in Saudi Arabia and worldwide. We hope that the results of this study can help leaders and decision-makers to prioritize the efforts of improving patient safety culture among health professionals and within healthcare organizations.

Acknowledgements

This work was supported by Saudi Patient Safety Center (SPSC). However, this study does not represent any official document from the SPSC, and the contents are solely the responsibility of the authors.

Abbreviations

AHRQ	Agency for Healthcare Research and Quality
CF	Contributory factor
HSPSC	Hospital Survey on Patient Safety Culture
MOH	Ministry of Health
PER	Positive events reported
POSG	Patient overall safety grade
PSC	Patient safety culture
SE	Sentinel event

Authors’ contributions

All authors had full access to all the study data and took responsibility for the integrity of the data and the accuracy of the analysis. All authors read and approved the final manuscript. Study conception and design: SB, YA, AA, and NA. Development of study materials: AA, RKA, and AAA. Acquisition of data: SB, RKA, and AAA. Analysis or interpretation of the data: SB, RKA, AAA, NMA, and RA. Drafting of the manuscript: SB, RKA, NMA, AAA, and RA. Critical revision of the manuscript for important intellectual content: SB, YA, AA, NA, and RA. Statistical and data analysis: SB and NMA. Administrative, technical, or material support: RKA, AAA, and NMA. Supervision: SB, YA, AA, and RA.

Not applicable.

Availability of data and materials

Declarations.

Ethical approval was obtained by the institutional review board at King Saud Medical City (IRB Project No. E-22–6551). We confirm that all methods were carried out in accordance with relevant guidelines and regulations. We confirm that informed consent was obtained from all subjects during the main data collection of the patient safety culture survey by the Saudi Patient Safety Center (SPSC). All data used in this study are aggregated secondary data with no personal identifying information. The confidentiality of data was maintained anonymously.

None to declare.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

DIAMANTE: A data-centric semantic segmentation approach to map tree dieback induced by bark beetle infestations via satellite images

Open access
Published: 04 September 2024

Cite this article

You have full access to this open access article

Giuseppina Andresini 1 , 2 ,
Annalisa Appice 1 , 2 ,
Dino Ienco 3 , 4 &
Vito Recchia 1

Forest tree dieback inventory has a crucial role in improving forest management strategies. This inventory is traditionally performed by forests through laborious and time-consuming human assessment of individual trees. On the other hand, the large amount of Earth satellite data that are publicly available with the Copernicus program and can be processed through advanced deep learning techniques has recently been established as an alternative to field surveys for forest tree dieback tasks. However, to realize its full potential, deep learning requires a deep understanding of satellite data since the data collection and preparation steps are essential as the model development step. In this study, we explore the performance of a data-centric semantic segmentation approach to detect forest tree dieback events due to bark beetle infestation in satellite images. The proposed approach prepares a multisensor data set collected using both the SAR Sentinel-1 sensor and the optical Sentinel-2 sensor and uses this dataset to train a multisensor semantic segmentation model. The evaluation shows the effectiveness of the proposed approach in a real inventory case study that regards non-overlapping forest scenes from the Northeast of France acquired in October 2018. The selected scenes host bark beetle infestation hotspots of different sizes, which originate from the mass reproduction of the bark beetle in the 2018 infestation.

Explore related subjects

Artificial Intelligence

Avoid common mistakes on your manuscript.

1 Introduction

Forests and woodlands cover roughly one-third of Earth’s surface and play a critical role in providing many ecosystem services, including carbon sequestration, water flow regulation, timber production, soil protection, and biodiversity conservation. However, the accelerating pace of climate change and its impact on species distribution and biome composition are leading to an increase in various types of disturbances, whether biotic, abiotic, or a combination of both, which are now affecting this vital natural resource and resulting in forest loss. Consequently, the decline in key forest ecosystem services is becoming more and more apparent. Among all the disturbances, insect infestations and disease outbreaks (e.g., bark beetle infestations) can induce massive tree dieback and, subsequently, significantly disrupt ecosystem dynamics (Gomez et al., 2023 ). This is why forest surveillance is crucial to monitor, quantify and possibly prevent outbreak diseases and enable foresters to perform informed decision-making for effective environmental management. Nevertheless, common strategies used to evaluate the health of forested regions primarily rely on laborious and time-consuming field surveys (Bárta et al., 2021 ). Consequently, they are restricted in their ability to cover extensive geographical areas, thereby preventing large-scale analysis across vast territories. To this end, the substantial amount of remote sensing information collected today via modern Earth observation missions constitutes an unprecedented opportunity to scale up forest dieback assessment and surveillance over large areas. As an exemplar, the European Space Agency’s Sentinel missions (Berger et al., 2012 ) provide a set of quasi-synchronous synthetic aperture radar (SAR) and optical data, systematically acquired worldwide, at high spatial (order of 10m) and temporal (an acquisition up to every five/six days) resolution. This information can be of paramount interest to support large-scale forest dieback assessment and surveillance systems.

While the research community is investigating the benefit related to exploiting multisensor remote sensing information via recent deep learning approaches (Hong et al., 2020 ; Li et al., 2022 ), there is still the necessity to design effective and well-tailored approaches to get the most out of multisensor remote sensing information (Hollaus & Vreugdenhil, 2019 ). This is the case for the large-scale assessment of tree dieback events induced by insect infestations and disease outbreaks where, to the best of our literature survey, existing works (e.g., Andresini et al., 2023 ;Bárta et al., 2021 ;Candotti et al., 2022 ;Dalponte et al., 2022 ;Fernandez-Carrillo et al., 2020 ;Zhang et al., 2022 ) mainly focus on optical data analysis, while no works exist that achieve improvements by leveraging multisensor remote sensing data (e.g., SAR and optical data). In particular, the literature studies to monitor bark beetle infestation in optical data pay high attention to both the data engineering step, through the synthesis of spectral vegetation indices, and the model development step, through the test of various machine learning and deep learning algorithms. On the other hand, similar to research communities where data play a major central role (e.g., computer vision, machine learning, information retrieval), also researchers coming from the remote sensing field are investing efforts towards more systematic and effective exploitation of available data sources. To this end, research actions in this direction have been proposed under the umbrella of data-centric Artificial Intelligence (AI) (Zha et al., 2023 ). Under this movement, the attention of researchers and practitioners is gradually shifting from advancing model design (model-centric AI) to enhancing the quality, quantity and diversity of the data (data-centric AI). Moreover, when remote sensing data are considered, the data-centric AI perspective is even more important since it can steer the community towards developing a methodology to provide further improvements related to the use of highly heterogeneous information to ameliorate the generalization ability with impact on real-world relevant problems and applications (Roscher et al., 2023 ). Nevertheless, the two perspectives (model-centric and data-centric AI) play a complementary role in the larger remote sensing deployment cycle, since standard approaches still struggle to manage and exploit valuable data coming from different and heterogeneous sources as, for instance, in the case of leveraging multisensor complementary information.

With the objective to find a trade-off between data-centric and model-centric achievements in remote sensing and map bark beetle-induced tree dieback events in remote sensing data adopting a semantic segmentation approach (e.g., categorization of pixels into a class), in this paper, we propose DIAMANTE ( D ata-centr I c sem A ntic seg M entation to m A p i N festations in sa T ellite imag E s): a data-centric semantic segmentation approach to train a U-Net like model from a labelled remote sensing dataset prepared using both SAR Sentinel-1 (S1) and multi-spectral optical Sentinel-2 (S2) remote sensing data sources. In particular, for the model development, we compare the achievements of several multisensor data fusion schema that are performed via early, middle or late stages fusion in an underlining U-Net architecture (Ronneberger et al., 2015 ). The U-Net is considered thanks to its wide versatility and increasing popularity, as well as due to the fact that it has been recently used to map bark beetle-induced tree dieback in Sentinel-2 images (Andresini et al., 2023 , 2024 ; Zhang et al., 2022 ). In addition, in this study, we consider that model recycling is one of the achievements to be evaluated in developing a data-centric AI approach. Hence, we start a preliminary investigation of how the multisensor fusion approaches considered in this study may allow us to train a semantic segmentation model for bark beetle detection, which still achieves good performance in a future data setting. The following are the main contributions of this work:

The definition of a remote sensing data collection and curation pipeline to prepare multisensor, Sentinel-1 and Sentinel-2 images of forest areas for which the ground truth map of the bark beetle infestation is available at a specific time. The defined pipeline pays particular attention to the quality of the Sentinel-1 and Sentinel-2 data prepared for the model development.

The adoption and comparison of several multisensor data fusion schemes to combine Sentinel-1 and Sentinel-2 data via early, middle or late stages fusion considering the underlying U-Net architecture.

The extensive assessment of our proposal using a ground truth map of tree dieback induced by bark beetle infestations in the Northeast of France in October 2018. The evaluation examines the performance of models trained and tested using images acquired over non-overlapping scenes in the same period, as well as the temporal forecasting and transferability of the model to an upcoming data setting.

The rest of the manuscript is organized as follows. Related literature is reviewed in Section 2 . The study site and the associated multisensor remote sensing dataset are introduced in Section 3 , while the proposed methodology is described in Section 4 . Section 5 reports the experimental evaluation and it discusses the related findings. Section 6 concludes.

2 Related work

This related work overview is organized into two main fronts. Firstly, we delve into recent remote sensing studies that incorporate machine learning and deep learning to map bark beetle infestation in Sentinel-1 (S1) and Sentinel-2 (S2) images. On the other front, we address the recent achievements of the data-centric artificial intelligence paradigm in remote sensing applications.

2.1 Bark beetle detection in remote sensing

Remote sensing studies to map forest stress related to bark beetle attacks have mainly focused on the analysis of Sentinel-2 data (Estrada et al., 2023 ). These studies are mainly inspired by the analysis conducted in Abdullah et al. ( 2019 ) to explore the effect of several forest disturbances sources (comprising bark beetle infestation) on S2 data. This study shows that the bark beetle infestation, which may affect the biophysical and biochemical properties of trees, is commonly visible via Sentinel-2 multi-spectral imagery. In particular, the chlorophyll degradation and nitrogen deficiency lead to an increase in reflectance spectrum in the visible region (particularly, red and green bands). Changes caused by the reduction of chlorophyll and leaf water have also an effect on Near Infrared (NIR) and Water vapor bands, while diseased and insect attacks affect red-edge bands. This analysis has boosted a plethora of studies (Andresini et al., 2023 , 2024 ; Bárta et al., 2021 ; Candotti et al., 2022 ; Dalponte et al., 2022 ; Fernandez-Carrillo et al., 2020 ; Huo et al., 2021 ; Jamali et al., 2023 ; Zhang et al., 2022 ) that explore the ability of various spectral vegetation indices to enhance the accuracy of decision models trained on Sentinel-2 data. Notice that explored spectral vegetation indices mainly combine red, green, NIR and SWIR (short wave infrared) bands.

Regarding the classification algorithms used to map bark beetle infestations in Sentinel-2 images, the most recent studies have mainly used machine learning algorithms such as Random Forest (Andresini et al., 2023 , 2024 ; Bárta et al., 2021 ; Candotti et al., 2022 ; Huo et al., 2021 ), Support Vector Machine (Andresini et al., 2023 ; Candotti et al., 2022 ; Dalponte et al., 2022 ) and XGBoost (Andresini et al., 2023 , 2024 ). Instead, (Andresini et al., 2023 , 2024 ; Zhang et al., 2022 ) explore the performance of deep learning algorithms under semantic segmentation settings such as U-Net (Andresini et al., 2023 , 2024 ; Zhang et al., 2022 ) and FCN-8 (Andresini et al., 2023 ). To handle the data imbalance situation, (Andresini et al., 2023 , 2024 ; Dalponte et al., 2022 ) use a cost-based learning strategy in combination with Random Forest and Support Vector Machine, while (Andresini et al., 2023 , 2024 ) use the Tversky loss in combination with U-Net and FCN-8. Finally, some studies consider Sentinel-2 time series data to train either Random Forest (Andresini et al., 2024 ; Bárta et al., 2021 ; Fernandez-Carrillo et al., 2020 ) or U-Net models (Andresini et al., 2024 ).

On the other hand, only recently, few remote sensing studies have started exploring the potential of Sentinel-1 data to detect bark beetle infestations. Sentinel-1 data are traditionally used in deforestation detection on Hoekman et al. ( 2020 ). However, (Hollaus & Vreugdenhil, 2019 ) has recently hypothesized that the joint exploitation of Sentinel-1 and Sentinel-2 satellite information can disclose useful information to detect bark beetle infestation hotspots. In particular, this study finds significant differences between Sentinel-1 values measured in infested and healthy sites, respectively. Similar conclusions are drawn in Alshayef and Musthafa ( 2021 ). However, (Alshayef & Musthafa, 2021 ; Hollaus & Vreugdenhil, 2019 ) perform a statistical analysis of Sentinel-1 data distribution without exploring how the use of the Sentinel-1 information can contribute to learning accurate decision models to characterise bark beetle infestations. In general, based on the literature survey, (Hollaus & Vreugdenhil, 2019 ) highlights that significant research effort is still needed to explore the full potential of multisensor data in insect-induced forest disturbance mapping. In this direction, (Huo et al., 2021 ) shows that the joint analysis of Sentinel-1 and Sentinel-2 data marginally contributes to improving the performance of Random Forest models. This conclusion has been recently confirmed also by Konig et al. ( 2023 ) where poor performances have been achieved for bark beetle infestation mapping exploiting only Sentinel-1 radar data and negligible amelioration by the joint exploitation of multisensor (Sentinel-1 and Sentinel-2) data considering both Bayesian and Random Forest classification models. Notably, in Konig et al. ( 2023 ), the multi-sensor data are stacked in a single feature vector that is used as input space for training a classification model. This corresponds to an early fusion schema that concatenates pixel-wise the feature vectors which are acquired with the Sentinel-1 and Sentinel-2 sensors before starting the training stage.

On the other hand, some recent studies have started to investigate how to combine multisensor remote sensing data (e.g., Sentinel-1 and Sentinel-2 data) for the underlying task of land use land cover mapping under a semantic segmentation setting (Sainte Fare Garnot et al., 2022 ). The authors of Li et al. ( 2022 ) have surveyed recent deep learning architectures developed to handle multisensor data comprising Sentinel-1 and Sentinel-2 data. However, this survey mainly considers problems of change detection and biomass estimation without any attention to bark beetle detection problems. In addition, this study points out that the majority of deep neural architectures trained with multisensor satellite data adopt an early fusion mechanism to concatenate pixel-wise data acquired with the Sentinel-1 and Sentinel-2 satellites. The output of the concatenation step is subsequently used as input space for the deep neural model development. In particular, the authors of both Muszynski et al. ( 2022 ) and Solórzano et al. ( 2021 ) learn a U-Net model for land cover classification and flood detection via an early fusion of the Sentinel-1 and Sentinel-2 data. The authors of Altarez et al. ( 2023 ) introduce the Principal Component Analysis (PCA) to combine stacked Sentinel-1 and Sentinel-2 imagery before training a U-Net model for the downstream task of tropical mountain deforestation delineation. On the other hand, a few studies have recently started the investigation of late fusion mechanisms to combine Sentinel-1 and Sentinel-2 data through a deep learning architecture. For example, the authors of Hu et al. ( 2017 ) describe a two-branch architecture that separately extracts features from data acquired with the two distinct satellites and perform the late convolutional fusion before the final decision. A similar late fusion schema is also investigated in Hafner et al. ( 2022 ) for a problem of urban change detection. This study describes an architecture composed of two separate, identical U-Net architectures that process Sentinel-1 and Sentinel-2 image pairs in parallel, and lately fuses extracted features from both sensors at the final decision stage. A middle fusion mechanism is introduced in Audebert et al. ( 2018 ) to perform the fusion of Infrared-Red-Green (IRRG) images and Digital Surface Model (DSM) data extracted from the Lidar point cloud through a SegNet model. Middle fusion is performed at the encoder layers with a simple summation. Imagery data fusion schemes are also discussed in the survey paper (Zhang et al., 2021 ).

In any case, to the best of our knowledge, no previous studies have been proposed yet to explore the opportunity of combining Sentinel-1 and Sentinel-2 data via modern deep learning architecture (i.e., U-Net) for the downstream bark beetle detection task. In addition, this is the first study that frames the investigation of different multisensor fusion schemes (i.e., early fusion, middle fusion and late fusion) in a U-Net development step performed under the umbrella of data-centric AI. On the other hand, neither previous studies have experimented with a fusion mechanism that operates at the encoder level of semantic segmentation models trained on Sentinel-1 and Sentinel-2 data, nor these studies have started the investigation of achievements of data fusion schemes for model development done under the possible lens of model recycling.

2.2 Data-centric artificial intelligence in remote sensing

Data plays a fundamental role in several remote sensing problems, comprising satellite imagery-based forest health monitoring. As a consequence, the emerging data-centric artificial intelligence paradigm (Zha et al., 2023 ) has recently started receiving attention in remote sensing where the big satellite image collections (e.g., the Earth Sentinel-1 and Sentinel-2 image collections acquired via the Copernicus programme) are freely available. Roscher et al. ( 2023 ) describe the main principles of the data-centric artificial intelligence paradigm in geospatial data applications by highlighting that data acquisition and curation should receive as much attention as data engineering and model development and evaluation. This study describes one of the first data-centric remote sensing pipelines experimented for land cover classification in satellite imagery. Phillips et al. ( 2022 ) describe a data-centric approach that uses deep feature extraction to prepare a Sentinel-2 dataset to improve the performance of insect species distribution models. de Carvalho et al. ( 2023 ) describe a data-centric approach that combines semantic segmentation and Geographical Information Systems (GIS) to obtain instance-level predictions of wind plants by using free orbital satellite images. Specifically, this study achieves an improvement of the model performance by including the wind plant shadows to increase the mapped area and facilitate target detection. Ferreira de Carvalho et al. ( 2023 ) investigate the application of iterative sparse annotations for semantic segmentation in remote-sensing imagery, focusing on minimizing the labor-intensive and costly data labeling process. Finally, Schmarje et al. ( 2022 ) describe a data-centric approach for RGB imagery dataset creation that reduces annotation ambiguity for RGB images by combining semi-supervised classification and clustering. To the best of our knowledge, no previous studies have explicitly defined a data-centric semantic segmentation approach that pays specific attention to the data curation step, in addition to the model development step, to support bark beetle infestation mapping considering multisensor remote sensing data provided by Sentinel-1 and Sentinel-2 satellites.

3 Study area and data preparation

This section describes the pipeline realised to prepare the datasets used to train and test the semantic segmentation models. We used Microsoft Planetary Computer Footnote 1 that provides the API to access petabytes of environmental monitoring data comprising Sentinel-1 and Sentinel-2 images from 2016 to the present. Datasets are accessed via Azure Blob Storage. The study site denoted as Northeast France , situated in the northeastern region of France, is predominantly covered by coniferous forests. In 2018 and 2019, a significant proliferation of bark beetles occurred, leading to an estimation by the French National Forestry Office in late April 2019 that approximately 50% of spruce trees in France were infested, contrasting with the typical rate of 15% for dead or diseased trees under normal circumstances. Notably, preceding 2018, there were no instances of substantial windthrows in this area, suggesting that the observed regional-scale attacks were likely spurred by the hot summer droughts experienced in 2018. Satellite data covering the Northeast France study site consists of a Synthetic Aperture Radar (SAR) image acquired via the Sentinel-1 sensor and an optical multi-spectral image acquired via the Sentinel-2 sensor.

3.1 Sentinel-1 and Sentinel-2 data collection

The Sentinel-1 satellite constellation collects polarization data via a C-band synthetic-aperture radar instrument. The C-band denotes a nominal frequency range from 8 to 4 GHz (3.75 to 7.5 cm wavelength) within the microwave (radar) portion of the electromagnetic spectrum. Imaging radars equipped with C-band are generally not hindered by atmospheric effects. They are capable of imaging in all-weather (even through tropical clouds and rain showers), day or night. The constellation is composed of two satellites (Sentinel-1A and Sentinel-1B), and it offers a 6-day exact repeat cycle. This means that, over the same geographical area, one SAR can be accessed every 6 days. Due to the nature of the radar signal, the raw information needs calibration correction related to the terrain topography. For this reason, we adopt the level-1 Radiometrically Terrain Corrected (RTC) product available via the Microsoft Planetary platform Footnote 2 . This product provides SAR images at 10m of spatial resolution. Here we consider the two polarizations VV (Vertical-Vertical) and VH (Vertical-Horizontal). In particular, VV is a mode of radar polarisation where the microwaves of the electric field are oriented in the vertical plane for both signal transmission and reception by means of a radar antenna. VH is a mode of radar polarisation where the microwaves of the electric field are oriented in the vertical plane for signal transmission and where the horizontally polarised electric field of the back-scattered energy is received by the radar antenna. The list of Sentinel-1 bands considered in this study is reported in Table 1 .

The Sentinel-2 satellite constellation retrieves multi-spectral radiometric data (13 bands) in the visible, near infrared, and short wave infrared parts of the spectrum through two satellites (Sentinel-2A and Sentinel-2B). The Sentinel-2 constellation permits covering the majority of the Earth’s surface with a repeat cycle of 5 days. The optical imagery is acquired at high spatial resolution (between 10m and 60 m) over land and coastal water areas. The mission supports a broad range of services and applications such as agricultural monitoring, emergency management or land cover classification. Similarly to the SAR signal, also the optical signal collected by the Sentinel-2 sensors requires corrections. To this end, we adopt the level 2A product available via the Microsoft Planetary platform Footnote 3 that provides atmospherically corrected surface reflectances. Here we consider all the multi-spectral bands at a spatial resolution of 10m. While bands B2, B3, B4 and B8 are originally at a spatial resolution of 10m, for all the other bands we downscale them at 10m of spatial resolution via the nearest-neighbor resampling based interpolation (Patil, 2018 ). This technique selects the value of the pixel that is nearby the surrounding coordinates of the intended interpolation point. Finally, we ignore the B10 (SWIR - Cirrus) band that is reserved for atmospheric corrections. The final list of Sentinel-2 bands considered in this study is reported in Table 2 . In particular, for each Sentinel-2 band, we report the spatial resolution, the central wavelength, and the band name. The central wavelength refers to the midpoint wavelength at the centre of the spectral band range (barycenter) that the satellite sensor captures. For example, for the B1 band that captures wavelengths from 433 to 453 nanometers (nm), the central wavelength is 443 nm.

3.2 Multisensor data alignment

Let us consider a collection of scenes in Northeast France for which we know the coordinates of each scene geometry and the timestamp in which scenes were observed using both Sentinel-1 and Sentinel-2 sensors. For each scene, we perform two geospatial queries to select a Sentinel-1 and a Sentinel-2 image acquired in a given time interval. The two queries are performed over the Sentinel-1 and Sentinel-2 collections, respectively, using the coordinates of the selected scenes and the selected time interval as query filters. The queried Sentinel-1 and Sentinel-2 images are recorded in the World Geodetic System 1984 ensemble using metric units. As each query may return a resultset of images, we adopt a pipeline to select a representative image from each resultset.

In particular, images are downloaded from Planetary using the STAC API. Footnote 4 For each scene in the study area, we first retrieve the Sentinel-2 image of the scene in a given month by formulating a STAC API query with the parameters “catalogue”, “bbox” and “datetime” set as follows: the value “sentinel-2-l2a” is used as “catalogue”, the “list of the coordinates of the four vertices of the rectangular box of the scene” is used as value for “bbox”, and the “date interval from the first day to the last day of a given month” is used as value for “datetime”. As the Sentinel-2 satellite may record images of the Earth every five days, the resultset of such query may contain several Sentinel-2 images recorded in the sentinel-2-l2a catalogue, covered by the given bbox, and acquired by the satellite within the selected datetime interval. The motivation for querying the sentinel-2-l2a catalogue with a time interval (one month in this study) is that cloud cover, shadows and defective pixels are among the main issues that may affect the Sentinel-2 imagery. The assumption for the success of a model development step performed with Sentinel-2 images is that images have to be as much as possible cloud and defective pixels-free. For this reason, we query Sentinel-2 imagery on a time interval (of one month in this study), to improve the possibility of choosing low-affected Sentinel-2 images in terms of clouds and defective pixels. Hence, we select the Sentinel-2 image of the resultset that achieves the lowest value of “cloud index”. If several images achieve the minimum value of the cloud index in the resultset, then we select the most recent Sentinel-2 image of this selection. The cloud index is computed based on the output of the Scene Classification Level (SCL) algorithm (Louis et al., 2016 ). This information is also recorded as a band in the sentinel-2-l2a catalogue. Specifically, the SCL algorithm uses the reflectance properties of imagery bands to establish the presence or absence of clouds or defective pixels in an image. In this way, it identifies clouds, snow and cloud shadows thus, generating a classification map, which consists of three different cloud classes (including cirrus), together with six additional classes covering shadows, cloud shadows, vegetation, not vegetated, water and snow land covers. For a candidate Sentinel-2 image, the index of cloud is computed as the percentage of imagery pixels that the SCL algorithm recognises as noise, defective, dark, cloud, cloud shadow or thin cirrus.

Given the Sentinel-2 image retrieved for a given scene in the given month, then we formulate the STAC API query to retrieve the Sentinel-1 image that is co-located in space and time with this Sentinel-2 image. The new query is performed by setting the “bbox” parameter as in the query performed to obtain the Sentinel-2 image while setting “catalogue” equal to “sentinel-1-rtc” and “datetime” equal to the “interval from three days before the date of the Sentinel-2 image and three days after the date of the Sentinel-2 image”. The time interval of this query depends on the fact that we would extract a Sentinel-1 image that should be roughly co-located in time with the Sentinel-2 image. On the other hand, Sentinel-1 images are collected every three days with any weather by using a technology not affected by clouds or weather. In addition, we note that noise has been already removed from the Sentinel-1 images that are recorded in the “sentinel-1-rtc” catalogue of Planetary thanks to the application of the Radiometrically Terrain Corrected (RTC) process. This process has been performed before recording the images in the “sentinel-1-rtc” catalogue by using the Ground Range Detected (GRD) Level-1 products produced by the European Space Agency with the RTC processing performed by Catalyst Footnote 5 . Hence, we limit to search the Sentinel-1 images potentially collected before and after the Sentinel-2 image and select the Sentinel-1 image that is the closest in time to the respective Sentinel-2 image.

3.3 Ground truth data, datasets and statistics

We use the ground truth map of the bark beetle infestation hotspots that caused tree dieback in the Northeast of France in October 2018. Footnote 6 This map was commissioned by the French Ministry of Agriculture and Food to Sertit (University of Strasbourg), to assess the damage in spruce forests of the Northeast of France following the 2018 bark beetle outbreak. The remote sensing company WildSense assessed and fixed the infestation hotspot polygons of this map. In particular, to avoid mixed reflectance from various causes in discoloration and defoliation of conifer, WildSense manually selected 87 squared, imagery tiles, covering spruce forestry areas fully under bark beetle attacks in October 2018. The scenes of the final collection cover 1004020 pixels at 10 square meters resolution. The size of the scenes varies from 27 $\times $ 16 to 296 $\times $ 319 pixels at 10 square meters resolution, while the percentage of infested territory per scene varies from 0.35% to 34.4% of the scene surface. The total percentage of damaged territory of the entire scene collection is 2.92%. For the experimental evaluation of this research work, we consider 71 scenes (covering 772844 pixels at 10 squared meters resolution) as training scene set and 16 scenes (covering 231176 pixels) as testing scene set. A map of the study scene location and their partitioning in the training set and testing set is depicted in Fig. 1 .

In addition, WildSense identified an extra scene covering spruce forestry areas fully under bark beetle attacks, according to a ground truth map acquired in March 2020. The geographic location of this scene is shown in Fig. 2 . This scene is a tile with size 205 $\times $ 135 covering 27675 pixels with 10 squared meters as spatial resolution with a percentage of infested territory equal to 3.55%.

Location of the centroids of the study 87 scenes in the Northeast of France area. The red circles correspond to scenes considered for training semantic segmentation models, while the blue circles correspond to scenes considered for evaluating semantic segmentation models

Location of the scene for which the ground truth mask of the bark beetle infestation was acquired in March 2020. The yellow patches map the forest areas with tree dieback caused by the bark beetle

In this study, we prepare four multisensor, satellite datasets populated with both the Sentinel-1 and Sentinel-2 images acquired for each scene in the study area in the Northeast of France. Hence, each dataset is populated with 87 Sentinel-1 images and 87 Sentinel-2 images roughly co-located in time within the same month. Specifically, the four multisensor satellite datasets were obtained by considering Sentinel-1 and Sentinel-2 images acquired monthly for the 87 study scenes in July 2018, August 2018, September 2018 and October 2018, respectively. We partition each imagery dataset into a training set and a testing set by using the same split ratio for each month. In particular, as mentioned above, we select 71 multisensor images as the training set and 16 multisensor images as the testing set for each of these four datasets. Notably, the multisensor images assigned to the four training sets were acquired for the same 71 training scenes although in different months. Similarly, the multisensor images assigned to the four testing sets were acquired for the same 16 testing scenes although in different months.

Box plot distribution of the polarization values measured for the Sentinel-1 bands and the radiometric values measured for the Sentinel-2 bands recorded in the datasets of Sentinel-1 and Sentinel-2 images acquired in the study site in July, August, September and October 2018. Bands are plotted independently with respect to the two opposite classes in the logarithmic scale

The dataset collected in October 2018 – the time at which the ground truth map of the bark beetle-induced tree dieback of the study scenes was produced – is elaborated to analyse the ability to map bark beetle-induced tree dieback in October, while datasets collected for the same scenes from July to September 2018 are elaborated to analyse the ability to predict as earlier as possible signs related to the bark beetle infestation (before trees start dying). Notice that the analysis of satellite imagery data collected in October 2018 follows some communications with foresters reported by Bárta et al. ( 2021 ), according to the beginning of the autumn, i.e., October in this study, may be considered the most suitable period for proactive measures, i.e., for looking for areas of infested trees and removing them from the forest before next spring. On the other hand, the analysis of satellite imagery data collected in July, August and September 2018 is done to explore the performance of the proposed approach in predicting where bark beetle infestation disturbance events are likely to cause future tree dieback. This evaluation is done according to the considerations reported in Kautz et al. ( 2022 ) that the early detection symptoms of bark beetle infestation, which comprise the presence of entrance holes, resin flow from entrance holes and boring dust that occur when the beetles attack the tree, penetrate the bark, and excavate mating chambers and breeding galleries that can be observed through terrestrial fieldwork inventory. So, counting on manually produced labels in the summer months may help the training of semantic segmentation models for automated early detection in scenes uncovered by the forestry fieldwork.

Spearman’s rank correlation coefficient computed between Sentinel-1 and Sentinel-2 bands in the images acquired in the study site in July, August, September and October 2018

Figure 3 shows the box plots of Sentinel-1 and Sentinel-2 data collected in the datasets prepared for this study. All bands are plotted independently of each other for the two opposite ground truth classes (“damaged” and “healthy”). The box plots show that the range of both Sentinel-1 and Sentinel-2 data changes over time. Sentinel-2 data, particularly B5, B6, B7, B8, B8A and B9, show a greater divergence between the opposite classes than Sentinel-1 data, over all the datasets. So, this visual data exploration confirms the general idea that Sentinel-2 contains the most important information to recognize bark beetle infestation hotspots, while Sentinel-1 data can be considered ancillary data that may be used to support analysis of Sentinel-2 data, to gain accuracy in the bark beetle infestation inventory.

In addition, Figure 4 shows the results of the bivariate correlation analysis performed by computing the Spearman’s rank correlation coefficient between Sentinel-1 and Sentinel 2 bands in images acquired between July and October 2018. Spearman’s rank correlation coefficient is a non-parametric measure of rank correlation that assesses how well the relationship between two compared variables can be described using a monotonic function. It varies between -1 and +1 with 0 implying no correlation, -1 implying an exact monotonic relationship with negative correlation and +1 implying an exact monotonic relationship with positive correlation. This correlation analysis shows that the Sentinel-1 bands VV and VH are negatively correlated to the Sentinel-2 bands B1, B2, B3, B4, B5, B11 and B12, while they are positively correlated to Sentinel-2 bands B7, B8, B8A and B9. The Sentinel-2 band B6 passes from showing a low negative correlation with the Sentinel-1 bands VV and VH in July to showing a low positive correlation with the same Sentinel-1 bands in August, September and October. In general, the intensity of the correlation between the Sentinel-2 bands B6, B7, B8, B8A and B9 and the Sentinel-1 bands VV and VH increases from July to August, September and October. In any case, the correlation is close to zero independently of the sign, especially on the bands B6, B7, B8, B8A and B9, which are the Sentinel-2 bands that better separate the opposite classes in the box plot analysis of the same data. Hence, this visual inspection of the collected data confirms a limited correlation between Sentinel-1 and Sentinel-2 data, which is one of the prerequisites for taking advantage of a multisensor approach in model development.

Figure 5 shows the box plot of the cloud index of the Sentinel-2 images selected for this study. This plot shows the high quality of Sentinel-2 images selected in each month. In fact, we are unable to select images with a cloud index lower than 5% only in one image in August 2018 and two images in October 2018. We also note that differences between the box-plot quartiles are slightly higher in October 2018 than in the period July-September 2018. This depends on the expected increase in the frequency of cloudiness as autumn advances.

Box plot of cloud index of Sentinel-2 images acquired in the study site in July, August, September and October 2018

Finally, we collect and prepare the pair of Sentinel-1 and Sentinel-2 images of the scene for which the ground truth map was acquired in March 2020. This pair of images is used in the evaluation stage only, to explore the transferability of the semantic segmentation model learned in October 2018 to subsequent periods. The Sentinel-2 image acquired for this scene in March 2020 and selected in this study has a low noise and cloud index equal to 0.16%. Finally, Figure 6 shows the box plots of both Sentinel-1 and Sentinel-2 data collected in March 2020 for this scene. We note that the outliers of Sentinel-1 data are spread across a lower heat range than that observed in the images collected in the summer and autumn months of 2018. On the other hand, B5, B6, B7, B8, B8A and B9 of Sentinel-2 data still show a remarkable divergence between the opposite classes as in the images collected in the summer and autumn months in 2018.

Box plot distribution of the polarization values measured for the Sentinel-1 bands and the radiometric values measured for the Sentinel-2 bands recorded in the Sentinel-1 image and the Sentinel-2 image acquired in March 2020 for the scene seen in Fig. 2 . Bands are plotted independently to the two opposite classes in the logarithmic scale

4 Semantic segmentation model development

The model development step is performed by leveraging the aligned Sentinel-1 and Sentinel-2 images of scenes for which the ground truth mask of bark beetle infestation is available. Let us consider $\mathcal {D} = \{ \left( \mathbf {X_{S1}}, \mathbf {X_{S2}}, \textbf{Y}\right) | \mathbf {X_{S1}} \in \mathbb {R}^{H\times W\times 2}, \mathbf {X_{S2}} \in \mathbb {R}^{H\times W\times 12}, \textbf{Y} \in \mathbb {R}^{H\times W\times 1} \}$ a collection of labelled Sentinel-1 and Sentinel-2 images of forest scenes, where every ground truth mask $\textbf{Y}$ is associated with the images $\mathbf {X_{S1}}$ and $\mathbf {X_{S2}}$ , acquired from Sentinel-1 and Sentinel-2 satellites, respectively. For each scene, H and W denote the spatial extent of the monitored scene in terms of scene height and scene width, respectively. The model development step trains a semantic segmentation network from $\mathcal {D}$ through a U-Net-like architecture that is also in charge of learning the data fusion.

The U-Net architecture is composed of an encoder part and a decoder part. The encoder extracts features. It consists of multiple blocks, where each block is composed of a Batch Normalization layer and a 2D Convolutional layer followed by Max-Pooling for downsampling. At each downsampling step, the height and width of the tensor halves, while the number of channels remains unchanged. The decoder part upsamples the encoded feature maps to the original input shape. It consists of one transposed Convolutional layer for upsampling, followed by multiple blocks, each of which each block consists of a Batch Normalization layer and a 2D Convolutional layer. Skipping connections between the decoder part and the encoder part are used to propagate the spatial information from the earlier layers to the deeper layers to alleviate the vanishing gradients problem (Wu et al., 2019 ). The final classification of each imagery pixel is obtained by using the Sigmoid activation function. The U-Net used in this study is trained via the Tversky loss, which is commonly used to handle imbalanced data (Hinton et al., 2015 ).

Early fusion . Abbreviations: 2D Conv = 2D Convolutional layer; BN=Batch Normalization; S1=Sentinel-1; S2=Sentinel-2

Middle fusion . Abbreviations: 2D Conv = 2D Convolutional layer; BN=Batch Normalization; S1=Sentinel-1; S2=Sentinel-2

Late fusion . Abbreviations: 2D Conv = 2D Convolutional layer; BN=Batch Normalization; S1=Sentinel-1; S2=Sentinel-2

The data fusion mechanism is implemented through three different strategies, namely, Early fusion , Middle fusion and Late fusion , which are defined according to the general classification of multimodal data fusion methods reported in the survey of Zhang et al. ( 2021 ). The Early fusion strategy is the first mechanism adopted in literature for the multimodal data fusion in the deep neural scenario (Couprie et al., 2013 ). It is implemented via a simple concatenation, performed at an early stage, of features from different modalities (i.e., sensors in this study). The concatenation produces a single input space for the model development. In our study, the Early fusion strategy, shown in Fig. 7 , concatenates each pair of images $\mathbf {X_{S1}}$ and $\mathbf {X_{S2}}$ obtaining a single hypercube with dimension ${H\times W\times 14}$ . A traditional U-Net architecture is trained on the newly stacked hypercubes.

The Middle fusion strategy combines features learned with the separate branches of a multi-input deep neural network that takes data acquired from different modalities as separate inputs. The fusion is performed at an intermediate layer of the deep neural network. The output of this combination performed at the fusion layer is processed across the subsequent layers of the network until the decision layer. In our study, the Middle fusion strategy, depicted in Fig. 8 uses an architecture with two encoder branches, each taking $\mathbf {X_{S1}}$ and $\mathbf {X_{S2}}$ as input, respectively. The output of these branches is fed into a single decoder. The two encoder branches are mapped into a common feature space via a fusion operation and the fusion output is used for the skipping connections. Two fusion operators, named SUM and CONC , are considered in this work for the middle fusion. The SUM operator performs an element-wise summation between the outputs of two parallel blocks in the encoder parts. The CONC operator produces a single hypercube by stacking the outputs of two parallel blocks in the encoder parts. Subsequently, it employs a 2D Convolutional layer to halve the channel size of the output hypercube. This is done to align with the number of channels of the corresponding decoder block for skipping connections. Both the concatenation (Couprie et al., 2013 ; Zhou et al., 2023 ) and the element-wise summation (Park et al., 2017 ; Qian et al., 2023 ) are two common fusion operators used in the literature to fuse multimodal features enclosed in RGB images and Depth images by using CNN-based algorithms. We select these two operators for the Middle fusion strategy performed in this study since they implement two different mechanisms in terms of information retention. In particular, the concatenation operator ( CONC ) allows us to keep all the information from both Sentinel-1 and Sentinel-2 data, where each feature is entirely preserved. On the other hand, the summation operator ( SUM ) provides a more compact representation than the concatenation. In fact, it fuses the features originated from the two sensors into a single vector having the same size of the combined vectors. This operator can be particularly useful when the features are aligned and represent the same spatial locations or attributes.

The Late fusion strategy processes separately input data provided by each modality through distinct deep neural models, and their outputs are combined at the later stage, usually at the classification stage. In our study, the Late fusion strategy, illustrated in Fig. 9 , uses an architecture with two identical, parallel encoder and decoder paths that take as input $\mathbf {X_{S1}}$ and $\mathbf {X_{S2}}$ , respectively. The outputs returned by the two decoders are stacked into a single hypercube and the Sigmoid activation function is employed in the final layer. Final considerations concern the expected behaviour of the three data fusion schemes. According to the discussion reported in Zhang et al. ( 2021 ), the Early fusion strategy is expected to better leverage cross-modal information interaction as early as possible in the learning stage. On the other hand, the Late fusion strategy is considered flexible, but it may lack sufficient cross-modal correlation. Finally, the Middle fusion strategy is expected to find a trade-off between Early fusion and Late fusion , with possible advantages in terms of final performances.

5 Empirical evaluation and discussion

5.1 implementation details.

We implemented DIAMANTE in Python 3.0. The source code is available online. Footnote 7 In this study, we consider a U-Net architecture optimized for satellite images implemented using the Keras 2.15 and TensorFlow as back-end Footnote 8 . Both encoder and decoder components of the different variants of U-Net architectures tested in this study are composed of five main blocks. In the encoder part, each block consists of 3 blocks containing a Batch Normalization layer and a 2D Convolutional layer, followed by a $2 \times 2$ Max-Pooling operation or downsampling. The stride of the Max-Pooling operation was set equal to 2. In the decoder part, each main block consists of a transposed Convolutional layer (for upsampling) followed by 3 blocks containing a Batch Normalization layer and a 2D Convolutional layer. The kernel size of each Convolutional layer was set equal to $3 \times 3$ . In all hidden layers the Rectified Linear Unit function (ReLU) was used as the activation function, while the Sigmoid activation function was used in the final semantic segmentation layer. The SUM operator was implemented using the Add layer available in TensorFlow. Footnote 9 The training of the U-Net architectures was performed using imagery tiles of size $32\times 32$ extracted from the imagery scenes by using tiler library. Footnote 10 Both Sentinel-1 and Sentinel-2 data were scaled between 0 and 1 using the Min-Max scaler (as it is implemented in the Scikit-learn 0.22.2 library) In addition, we considered a tile augmentation strategy to improve the performance of the U-Net architecture by using the Albumentations library Footnote 11 . Specifically, we quintupled the number of training imagery tiles by creating new tiles applying traditional computer vision augmentation operators (i.e., Horizontal Flip, Vertical Flip, Random Rotate, Transpose and Grid Distortion). We used the tree-structured Parzen estimator algorithm to optimize hyper-parameters of U-Net architectures (i.e., mini-batch size in { $2^2$ , $2^3$ , $2^4$ , $2^5$ , $2^6$ }, learning rate between 0.0001 and 0.01 and image augmentation in {True, False}), by using 20% of the training set as the validation set. In particular, the hyper-parameter configuration that achieves the highest F1 score on the minority class (“damaged”) in the validation set was automatically selected as the best semantic segmentation model. We performed the gradient-based optimisation using the Adam update rule. Finally, each U-Net model was trained with a maximum number of epochs equal to 150, using an early stopping approach to retain the best semantic segmentation model.

5.2 Metrics

To evaluate the accuracy of the semantic segmentation masks, we measured the following metrics: F1 score ( F1 ) computed for the two opposite classes, Macro F1 score ( Macro F1 ) averaged on the two opposite classes and Intersection-over-Union ( IoU ). Specifically, the F1 score measures the harmonic mean of Precision and Recall . The Precision = $\frac{TP}{TP+FP}$ is the fraction of pixels correctly classified in a specific class ( TP ) among pixels of the considered class ( $TP+FP$ ). The Recall = $\frac{TP}{TP+FN}$ is the fraction of pixels correctly classified in a specific class ( TP ) among pixels classified in the considered class ( $TP+FN$ ). In this study, we computed the F1 score for the two opposite classes of both case studies: “healthy” ( F1(h) ) and “damaged” ( F1(d) ). Macro F1 measures the average of each F1 score value per class, that is, Macro F1 = $\frac{F1(h) + F1(d)}{2}$ . The IoU score is the ratio of the intersected area to the combined area of prediction and ground truth, that is, IoU = $\frac{TP}{TP+FP+FN}$ . This is commonly used to evaluate the accuracy of models trained in both semantic segmentation and object detection problems. All metrics are reported in percentages and computed on the images collected for the testing scenes. For each metric, the higher the value, the better the performance of the semantic segmentation masks predicted.

5.3 Results

The illustration of results is organised as follows. Section 5.3.1 presents the results achieved by processing the multisensor imagery dataset collected in the study area in October 2018. This analysis is done to evaluate the performance of the data fusion strategies at the same time the ground truth masks of the study scenes were created. Section 5.3.2 presents a temporal study where we explore the performance of the models trained and evaluated considering satellite images acquired in July, August and September 2018. This analysis is done to explore the ability of the considered data fusion strategies to learn a model capable to perform early detection of tree dieback phenomena. Finally, Section 5.3.3 illustrates the results achieved by considering multisensor semantic segmentation models trained from satellite images acquired in October 2018 to predict the mask of tree dieback caused by a bark beetle infestation in a new scene located in the Northeast of France, but monitored in March 2020. This analysis explores the transferability over time of a semantic segmentation model.

5.3.1 Performance Analysis

In this Section, we analyse the performance of the semantic segmentation masks produced for the testing scenes of the Northeast France study by using the multisensor semantic segmentation models trained via the three data fusion schemes illustrated in Section 4 . As baselines, we consider the single-sensor semantic segmentation models trained with a traditional U-Net by processing either the Sentinel-1 images ( S1 U-Net ) or the Sentinel-2 images ( S2 U-Net ) alone. With regard to the Middle fusion strategy, we report the results achieved with the two fusion operators: SUM and CONC . This evaluation was conducted by processing the dataset of images acquired in October 2018 for both the training and evaluation stages. The accuracy metrics measured on the semantic segmentation masks produced for the images of the testing scenes are reported in Table 3 .

RGB of the Sentinel-2 image acquired in October 2018 for a testing scene of the study area in the Northeast of France ( 10 a). Inventory masks of tree dieback areas caused by bark beetle hotspots in this scene as they are predicted by S1 U-Net ( 10 b), S2 U-Net ( 10 c), Early fusion U-Net ( 10 d), ( 10 g), Middle fusion U-Net with operators SUM ( 10 e) and CONC ( 10 f) and Late fusion U-Net trained on the imagery set acquired in October 2018 for the training scenes of the study area

As we expected, the output of the stand-alone use of Sentinel-1 images is unsatisfactory for this inventory task. In fact, the configuration S1 U-Net achieves the lowest performance in all accuracy metrics. Better performance can be achieved by processing Sentinel-2 images in place of Sentinel-1 images. However, this evaluation study shows that the data fusion of Sentinel-1 and Sentinel-2 images can help us to improve the performance of the semantic segmentation model regardless of the type of data fusion strategy employed. In fact, the Early fusion U-Net , Late fusion U-Net and Middle fusion U-Net all achieve better performance than S2 U-Net that considers Sentinel-2 images only. More in detail, the best configuration in terms of F1(d) , IoU and Macro F1 is achieved with the Middle fusion schemes having Middle fusion (CONC) U-Net as runner-up of Middle fusion (SUM) U-Net . These conclusions are consistent with the observations on the expected behaviour of the data fusion schemes reported in Section 4 . Figure 10 b-g show the semantic segmentation masks of a sample testing scene predicted by the compared models, while Fig. 10 a shows the RGB image of this sample scene. The masks highlight how the use of a data fusion strategy helps us to reduce the number of false alarms in this case. Specifically, the bark beetle infestation masks predicted using the multisensor U-net trained with both Early fusion and Middle fusion schemes show only one false infested patch, while the U-Net trained from Sentinel-1 data shows large extensions of false infested areas and the U-Net trained from Sentinel-2 data shows two false infested patches. Notably, the multisensor U-Net trained with Late fusion strategy removes one of the false patches discovered by S2 U-Net , but, at the same time, it alerts a new false patch that is undetected in the other masks. We note that the Late fusion strategy is the worst-performing fusion strategy of this experiment. This result suggests that although the Late fusion strategy may allow us to correct some false patches detected processing Sentinel-2 data only, it may also produce some artefacts at the decision level, which may cause false alarms unseen in the remaining configurations. Finally, the masks of this example show that the use of SUM operator performs better than the CONC operator in delineating the large damaged patch located on the left side of the scene.

5.3.2 Temporal analysis

To complete this investigation, we illustrate the results of a temporal study conducted to explore the accuracy performance of the semantic segmentation maps produced when the Sentinel-1 and Sentinel-2 images were acquired in the middle of summer (i.e., July 2018) and the late summer (i.e., August 2018 and September 2018), while the ground truth map of the tree dieback was observed in early autumn (October 2018). This analysis is done to explore the performance of the presented data fusion strategies in the early detection of areas where bark beetle infestation disturbance events are likely to cause (near-)future tree dieback. The temporal snapshots of this experiment were selected according to the recent achievements of the analysis on the spectral separability between the healthy and bark beetle attacked trees illustrated in Dalponte et al. ( 2023 ). In particular, this study shows that bark beetle attacks commonly occur in the summer, while the spectral separability between the two opposite classes (“Healthy” and “Damaged”) increases moving from July to October. In addition, it highlights that a time span of approximately one month commonly occurs between the attack of the beetles to a tree and the development of the first symptoms (green-attack) in the tree. Hence, based on the conclusions drawn in this study, the green attack detection stage can reasonably arise in the summer period spanned from July to August. Based on these premises, the accuracy metrics measured on the semantic segmentation maps produced for the testing scenes of this study in each month between July and October 2018 are reported in Table 4 .

These results show that the data fusion of Sentinel-1 and Sentinel-2 continues to help us to gain accuracy also when the multisensor semantic segmentation model is trained to forecast tree dieback areas caused by the bark beetle infestation. Notably, Middle fusion (SUM) U-Net achieves the highest F1(d) , IoU and Macro F1 in segmentation maps produced in experiments performed in July 2018, August 2018 and October 2018. The only exception is observed in the segmentation maps produced for the evaluation in September 2018. However, also in the experiment conducted in September 2018, the Middle fusion (SUM) U-Net still achieves good performance by ranking as the runner-up of the Late fusion U-Net . To draw conclusive conclusions on the better data fusion strategy, we perform the Friedman-Nemenyi test to compare the Macro F1 measured for S1 U-Net , S2 U-Net , Early fusion U-Net , Middle fusion (SUM) U-Net , Middle fusion (CONC) U-Net and Late fusion U-Net on the multiple segmentation maps produced for the testing data of the multisensor datasets of this temporal analysis. This non-parametric test ranks the model configurations compared for each dataset separately, so the best-performing model is given a rank of 1, the second-best rank of 2 and so on. The results of the Friedman-Nemenyi test reported in Fig. 11 shows that the test groups the configurations adopting a multisensor data fusion strategy as statistically different from the configurations that consider either Sentinel-1 data only ( S1 U-Net ) or Sentinel-2 data only ( S2 U-Net ). In addition, the Middle fusion (SUM) U-Net achieves the highest rank by having the Middle fusion (CONC) U-Net as runner-up. Notably, these results of the comparative test support the conclusions already drawn in 5.3.1 and 5.3.3 on the superior performance of a Middle fusion strategy to combine Sentinel-1 and Sentinel-2 data for bark beetle infestation detection.

Comparison of the configurations: Macro F1 measured for S1 U-Net , S2 U-Net , Early fusion U-Net , Middle fusion (SUM) U-Net , Middle fusion (CONC) U-Net and Late fusion U-Net , performed with the Friedman-Nemenyi test run on Macro F1 measured in the temporal analysis performed from July 2018 to October 2018 ( computed $pvalue=0.013$ )

5.3.3 Transferability analysis

In this Section, we examine the accuracy of the semantic segmentation models learned in October 2018 when used to detect the tree dieback events caused by bark beetle infestations in March 2020. The accuracy metrics measured in this experiment are reported in Table 5 . These results show that also in this evaluation scenario, the data fusion of Sentinel-1 and Sentinel-2 may help us to improve the performance of a semantic segmentation model even when it was trained on past images and used for mapping the bark beetle infestation in future images. The only exception is observed for the Late fusion strategy that achieves lower performance than S2 U-Net . In general, the highest F1(d) , IoU and Macro F1 are achieved with the Middle fusion (CONC) U-Net schema having Middle fusion (SUM) U-Net as runner-up. This confirms the conclusions on the better performance of the Middle fusion strategy already drawn in Section 5.3.1 . Finally, Fig. 12 b-g show the semantic segmentation masks predicted for the scene under evaluation. The RGB image of the scene in March 2020 is shown in Fig. 12 a. The extracts show that the data fusion schemes, except for Late fusion , allow us to reduce the extension of the false alarm areas detected. In both Early fusion and Middle fusion (SUM) schemes, the higher precision is achieved at the cost of a lower recall. Both data fusion configurations allow us to map correctly a percentage of the infested area that is lower than the one mapped processing Sentinel-2 data only. Instead, the use of the Middle fusion (CONC) strategy allows us to achieve the best trade-off between precision and recall in detecting the tree dieback areas caused by the bark beetle infestation. In general, these maps confirm the idea that also when the semantic segmentation model is trained on historical data, the main contribution to the correct detection of the bark beetle infestation is given by Sentinel-2 data, while Sentinel-1 data can aid in reducing false alarms and better delimiting infested areas.

RGB of the Sentinel-2 image acquired in March 2020 ( 12 a). Inventory masks of tree dieback areas caused by bark beetle hotspots in this scene as they are predicted by S1 U-Net ( 12 b), S2 U-Net ( 12 c), Early fusion U-Net ( 12 d), Middle fusion U-Net with operators SUM ( 12 e) and CONC ( 12 f) and Late fusion U-Net ( 12 g) trained on the imagery set acquired in October 2018 for the training scenes of the study area

5.4 Considerations and findings

The experimental assessment highlights the general advantages of using multisensor data over a single data source in various scenarios of bark beetle detection, including early disease detection and out-of-year temporal transfer. While Sentinel-1 alone is not suitable for the considered downstream mapping task, using Sentinel-2 alone yields satisfactory results. However, the combined use of these two publicly available and freely accessible remote sensing data sources provides the best overall results.

More specifically, the joint use of Sentinel-1 and Sentinel-2 data significantly reduces false alarms and improves the delineation of infested areas in the resulting binary maps. Regarding the early detection of bark beetle attacks (Section 5.3.2 ), signs of the attack can be detected with reasonable accuracy one month before the acquisition of ground truth data (September 2018). However, the disease’s early stages (before July 2018) are weakly detectable via satellite imagery.

An additional challenge is represented by the out-of-year transfer of the model trained on 2018 data to 2020 data. Recent studies in the domain of remote sensing analysis have highlighted that spatial and temporal distribution shifts can hinder the direct deployment of a model trained on a particular area or time period to a different area or time period (Capliez et al., 2023 ; Nyborg et al., 2022 ). The results obtained in Section 5.3.3 confirm this point, indicating that there is still room for research activities in the way historical data can be leveraged in order to improve current mapping results. Finally, the comparison of the different approaches indicates that all fusion strategies are statistically significant compared to single source analysis, with the Middle fusion (SUM) U-Net model exhibiting the best average performance. This finding underscores once more the importance of combining multisensor satellite data for mapping tree dieback induced by bark beetle infestation.

6 Conclusion

In this study, we investigate the effectiveness of a data-centric semantic segmentation approach to map forest tree dieback areas caused by bark beetle hotspots. First, we define a data-centric pipeline to collect and prepare images acquired from both the SAR Sentinel-1 sensor and the optical Sentinel-2 sensor. Then, we explore the accuracy performance of several data fusion strategies, namely Early fusion , Middle fusion and Late fusion adopted for the development of a U-Net-like model combining both Sentinel-1 and Sentinel-2 images acquired in the Northeast of France. Finally, we investigate the performance of the proposed strategies in multisensor imagery data acquired in Northeast of France with the map of bark beetle infestation available in October 2018. We conducted the evaluation with imagery data prepared according to the data curation pipeline presented in this study. The experimental results show that multisensor data can actually help us to improve the ability of the U-Net model to detect tree dieback areas caused bark beetle infestations. The evaluation also explores the transferability of the output of the model development step, as well as the performance of the proposed approach in early detection of infestations that will cause tree dieback.

As future work, we plan to continue the investigation of multisensor data fusion strategies in combination with ecological and weather data, as well temporal data trend information. In addition, we plan to extend the investigation of the transferability of the semantic segmentation model, trained with the described multisensor data fusion techniques to unseen data settings. In particular, we intend to start a systematic exploration of some transfer learning approaches to obtain the transferability of a “general” semantic segmentation model trained for a specific disturbance agent to different disturbance agents. For example, we intend to investigate the transferability of a semantic segmentation model trained for mapping forest tree die-back hotspots caused by bark beetle infestation to perform the inventory of tree die-back hotspots caused by different families of fungal forest pathogens. In addition, we hope to be able to acquire large-scale data within the experimental phase of the EU project SWIFTT to be able of investigating, on large scale, the transferability of a semantic segmentation model trained in a geographic area to a different geographic area, in addition to a future time.

Data, Material, and/or Code Availability

The source code is available at https://github.com/gsndr/DIAMANTE

https://planetarycomputer.microsoft.com/

https://planetarycomputer.microsoft.com/dataset/sentinel-1-rtc

https://planetarycomputer.microsoft.com/dataset/sentinel-2-l2a

https://planetarycomputer.microsoft.com/docs/quickstarts/reading-stac/

https://catalyst.earth/

https://macarte.ign.fr/carte/3bd52aa2b6422a3a58b5086576f91080/Foyers+de+scolytes+dans+les+pessi%C3%A8res+et+les+sapini%C3%A8res+du+Nord-Est+de+la+France,+automne+2018-printemps+2019

https://github.com/gsndr/DIAMANTE

https://github.com/karolzak/keras-unet/tree/master

https://www.tensorflow.org/api_docs/python/tf/keras/layers/Add

https://github.com/the-lay/tiler

https://albumentations.ai/

Abdullah, H., Skidmore, A.K., Darvishzadeh, R., Heurich, M. (2019).Sentinel-2 accurately maps green-attack stage of European spruce bark beetle (ips typographus, l.) compared with landsat-8. Remote Sensing in Ecology and Conservation , 5(1), 87–106.

Alshayef, M.S., Musthafa, M. (2021). Identification of bark beetle infestation in part of bohemian forest using sentinel-1 time series insar. In: 2021 IEEE International India Geoscience and Remote Sensing Symposium (InGARSS), pp. 5–9.

Altarez, R. D. D., Apan, A., & Maraseni, T. (2023). Deep learning u-net classification of sentinel-1 and 2 fusions effectively demarcates tropical montane forest’s deforestation. Remote Sensing Applications: Society and Environment, 29 , 1–21.

Google Scholar

Andresini, G., Appice, A., & Malerba, D. (2023). $\sf SILVIA$ : An explainable framework to map bark beetle infestation in Sentinel-2 images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , 16, 10050–10066.

Andresini, G., Appice, A., & Malerba, D. (2024). Leveraging sentinel-2 time series for bark beetle induced forest dieback inventory. In: The 39th ACM/SIGAPP Symposium on Applied Computing, SAC 2024, pp. 875–882. CAM.

Audebert, N., Le Saux, B., & Lefèvre, S. (2018). Beyond rgb: Very high resolution urban remote sensing with multimodal deep networks. ISPRS Journal of Photogrammetry and Remote Sensing, 140 , 20–32.

Article Google Scholar

Bárta, V., Lukeš, P., & Homolová, L. (2021). Early detection of bark beetle infestation in norway spruce forests of central europe using Sentinel-2. International Journal of Applied Earth Observation and Geoinformation, 100 , 102335.

Berger, M., Moreno, J., Johannessen, J. A., Levelt, P. F., & Hanssen, R. F. (2012). Esa’s sentinel missions in support of earth system science. Remote Sensing of Environment, 120 , 84–90.

Candotti, A., De Giglio, M., Dubbini, M., Tomelleri, E. (2022). A Sentinel-2 based multi-temporal monitoring framework for wind and bark beetle detection and damage mapping. Remote Sensing , 14(23).

Capliez, E., Ienco, D., Gaetano, R., Baghdadi, N. N., & Hadj-Salah, A. (2023). Temporal-domain adaptation for satellite image time-series land-cover mapping with adversarial learning and spatially aware self-training. IEEE J. Sel. Top. Appl. Earth Obs. Remote. Sens, 16 , 3645–3675.

Couprie, C., Farabet, C., Najman, L., Lecun, Y. (2013): Indoor Semantic Segmentation using depth information. In: First International Conference on Learning Representations (ICLR 2013), pp. 1–8.

Dalponte, M., Cetto, R., Marinelli, D., Andreatta, D., Salvadori, C., Pirotti, F., Frizzera, L., & Gianelle, D. (2023). Spectral separability of bark beetle infestation stages: A single-tree time-series analysis using planet imagery. Ecological Indicators, 153 , 110349.

Dalponte, M., Solano-Correa, Y.T., Frizzera, L., Gianelle, D. (2022). Mapping a European spruce bark beetle outbreak using Sentinel-2 remote sensing data. Remote Sensing , 14(13).

de Carvalho, O. L. F., de Carvalho Junior, O. A., de Albuquerque, A. O., Orlandi, A. G., Hirata, I., Borges, D. L., Gomes, R. A. T., & Guimarães, R. F. (2023). A data-centric approach for wind plant instance-level segmentation using semantic segmentation and gis. Remote Sensing, 15 (5), 1–23.

Estrada, J.S., Fuentes, A., Reszka, P., Auat Cheein, F. (2023). Machine learning assisted remote forestry health assessment: a comprehensive state of the art review. Frontiers in Plant Science , 14.

Fernandez-Carrillo, A., Patočka, Z., Dobrovolný, L., Franco-Nieto, A., Revilla-Romero, B. (2020). Monitoring bark beetle forest damage in central europe. a remote sensing approach validated with field data. Remote Sensing , 12(21).

Ferreira de Carvalho, O.L., Olino de Albuquerque, A., Luiz, A.S., Henrique Guimarães Ferreira, P., Mou, L., e Silva, D.G., Abílio de Carvalho Junior, O. (2023). A data-centric approach for rapid dataset generation using iterative learning and sparse annotations. In: IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium, pp. 5650–5653.

Gomez, D.F., Riggins, J.J., Cognato, A.I. (2023). Forest Entomology and Pathology: Volume 1: Entomology – Bark Beetles, pp. 299–337. Springer International Publishing, Cham.

Hafner, S., Nascetti, A., Azizpour, H., & Ban, Y. (2022). Sentinel-1 and sentinel-2 data fusion for urban change detection using a dual stream u-net. IEEE Geoscience and Remote Sensing Letters, 19 , 1–5.

Hinton, G.E., Vinyals, O., Dean, J. (2015). Distilling the knowledge in a neural network. CoRR,928 abs/1503.02531

Hoekman, D., Kooij, B., Quiñones, M., Vellekoop, S., Carolita, I., Budhiman, S., Arief, R., & Roswintiarti, O. (2020). Wide-area near-real-time monitoring of tropical forest degradation and deforestation using sentinel-1. Remote Sensing, 12 (19), 1–32.

Hollaus, M., & Vreugdenhil, M. (2019). Radar satellite imagery for detecting bark beetle outbreaks in forests. Current Forestry Reports, 5 , 240–250.

Hong, D., Gao, L., Yokoya, N., Yao, J., Chanussot, J., Du, Q., Zhang, B. (2020). More diverse means better: Multimodal deep learning meets remote-sensing imagery classification. IEEE Transactions on Geoscience and Remote Sensing, pp. 1–15.

Hu, J., Mou, L., Schmitt, A., Zhu, X.X. (2017). Fusionet: A two-stream convolutional neural network for urban scene classification using polsar and hyperspectral data. In: 2017 Joint Urban Remote Sensing Event (JURSE), pp. 1–4.

Huo, L., Persson, H. J., & Lindberg, E. (2021). Early detection of forest stress from European spruce bark beetle attack, and a new vegetation index: Normalized distance red & swir (ndrs). Remote Sensing of Environment, 255 , 112240.

Jamali, S., Olsson, P. O., Ghorbanian, A., & Müller, M. (2023). Examining the potential for early detection of spruce bark beetle attacks using multi-temporal sentinel-2 and harvester data. ISPRS Journal of Photogrammetry and Remote Sensing, 205 , 352–366.

Kautz, M., Peter, F., Harms, L., Kammen, S., & Delb, H. (2022). Patterns, drivers and detectability of infestation symptoms following attacks by the European spruce bark beetle. Journal of Pest Science, 96 , 403–414.

Konig, S., Thonfeld, F., Forster, M., Dubovyk, O., & Heurich, M. (2023). Assessing combinations of landsat, Sentinel-2 and Sentinel-1 time series for detecting bark beetle infestations. GIScience & Remote Sensing, 60 (1), 2226515.

Li, J., Hong, D., Gao, L., Yao, J., Zheng, K., Zhang, B., & Chanussot, J. (2022). Deep learning in multimodal remote sensing data fusion: A comprehensive review. Int. J. Appl. Earth Obs. Geoinformation, 112 , 102926.

Louis, J., Debaecker, V., Pflug, B., Main-Knorn, M., Bieniarz, J., Mueller-Wilm, U., Cadau, E., Gascon, F. (2016). Sentinel-2 sen2cor: L2a processor for users. In: Proceedings of the Living Planet Symposium 2016, pp. 1–8. Spacebooks Onlin.

Muszynski, M., Hölzer, T., Weiss, J., Fraccaro, P., Zortea, M., Brunschwiler, T. (2022). Flood event detection from sentinel 1 and sentinel 2 data: Does land use matter for performance of u-net based flood segmenters? In: 2022 IEEE International Conference on Big Data (Big Data), pp. 4860–4867.

Nyborg, J., Pelletier, C., Lefèvre, S., & Assent, I. (2022). Timematch: Unsupervised cross-region adaptation by temporal shift estimation. ISPRS Journal of Photogrammetry and Remote Sensing, 188 , 301–313.

Park, S.J., Hong, K.S., Lee, S. (2017). Rdfnet: Rgb-d multi-level residual feature fusion for indoor semantic segmentation. In: The IEEE International Conference on Computer Vision (ICCV).

Patil, M. (2018). Interpolation techniques in image resampling. International Journal of Engineering and Technology, 7 , 567–570.

Phillips, J., Zhang, C., Williams, B., & Jarvis, S. (2022). Data-driven sentinel-2 based deep feature extraction to improve insect species distribution models. EGU General Assembly, 2022 , 1–1.

Qian, X., Lin, X., Yu, L., & Zhou, W. (2023). Fasflnet: feature adaptive selection and fusion lightweight network for rgb-d indoor scene parsing. Optics Express, 31 (5), 8029–8041. https://doi.org/10.1364/OE.480252

Ronneberger, O., Fischer, P., Brox, T. (2015). U-Net: Convolutional networks for biomedical image segmentation. In: Medical Image Computing and Computer-Assisted Intervention, MICCAI 2015, pp. 234–241. Springer.

Roscher, R., Rußwurm, M., Gevaert, C., Kampffmeyer, M., dos Santos, J.A., Vakalopoulou, M., Hänsch, R., Hansen, S., Nogueira, K., Prexl, J., Tuia, D. (2023). Data-centric machine learning for geospatial remote sensing data. arXiv:2312.05327

Sainte Fare Garnot, V., Landrieu, L., Chehata, N. (2022). Multi-modal temporal attention models for crop mapping from satellite time series. ISPRS Journal of Photogrammetry and Remote Sensing , 187, 294–305.

Schmarje, L., Santarossa, M., Schröder, S. M., Zelenka, C., Kiko, R., Stracke, J., Volkmann, N., & Koch, R. (2022). A data-centric approach for improving ambiguous labels with combined semi-supervised classification and clustering. In S. Avidan, G. Brostow, M. Cissé, G. M. Farinella, & T. Hassner (Eds.), Computer Vision - ECCV 2022 (pp. 363–380). Cham: Springer Nature Switzerland.

Chapter Google Scholar

Solórzano, J. V., Mas, J. F., Gao, Y., & Gallardo-Cruz, J. A. (2021). Land use land cover classification with u-net: Advantages of combining sentinel-1 and sentinel-2 imagery. Remote Sensing, 13 (18), 1–23.

Wu, J., Zhang, Y., Wang, K., & Tang, X. (2019). Skip connection u-net for white matter hyperintensities segmentation from mri. IEEE Access, 7 , 155194–155202.

Zha, D., Bhat, Z.P., Lai, K., Yang, F., Jiang, Z., Zhong, S., Hu, X. (2023). Data-centric artificial intelligence: A survey. arXiv:2303.10158

Zha, D., Lai, K., Yang, F., Zou, N., Gao, H., Hu, X. (2023). Data-centric AI: techniques and future perspectives. In: A.K. Singh, Y. Sun, L. Akoglu, D. Gunopulos, X. Yan, R. Kumar, F. Ozcan, J. Ye (eds.) Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2023, pp. 5839–5840. ACM.

Zhang, J., Cong, S., Zhang, G., Ma, Y., Zhang, Y., Huang, J. (2022). Detecting pest-infested forest damage through multispectral satellite imagery and improved UNet++. Sensors , 22(19).

Zhang, Y., Sidibé, D., Morel, O., & Mériaudeau, F. (2021). Deep multimodal fusion for semantic image segmentation: A survey. Image and Vision Computing, 105 , 1–17.

Zhou, W., Yue, Y., Fang, M., Qian, X., Yang, R., & Yu, L. (2023). Bcinet: Bilateral cross-modal interaction network for indoor scene understanding in rgb-d images. Information Fusion, 94 , 32–42.

Download references

Acknowledgements

Annalisa Appice acknowledges support from the SWIFTT project, funded by the European Union under Grant Agreement 101082732. Dino Ienco acknowledges support from the Eco2Adapt project, funded by the European Union under Grant Agreement 101059498. Giuseppina Andresini and Vito Recchia are supported by the project FAIR - Future AI Research (PE00000013), Spoke 6 - Symbiotic AI, under the NRRP MUR program funded by the NextGenerationEU. The authors wish to thank the remote sensing company WildSense for preparing the ground truth masks of the evaluation study.

Open access funding provided by Università degli Studi di Bari Aldo Moro within the CRUI-CARE Agreement.

Author information

Authors and affiliations.

Department of Informatics, Università degli Studi di Bari Aldo Moro, via Orabona, 4-70125, Bari, Italy

Giuseppina Andresini, Annalisa Appice & Vito Recchia

Consorzio Interuniversitario Nazionale per l’Informatica - CINI, Bari, Italy

Giuseppina Andresini & Annalisa Appice

INRAE, UMR TETIS, University of Montpellier, Montpellier, France

INRIA, University of Montpellier, Montpellier, France

You can also search for this author in PubMed Google Scholar

Contributions

Giuseppina Andresini : Conceptualization, Methodology, Software, Validation, Investigation, Supervision, Writing - original draft, Writing - review & editing Annalisa Appice : Conceptualization, Methodology, Validation, Visualization, Investigation, Writing - original draft, Writing - review & editing, Supervision, Project administration. Dino Ienco : Conceptualization, Methodology, Investigation, Writing - original draft, Writing - review & editing. Vito Recchia : Conceptualization, Methodology, Data curation, Software, Validation, Visualization, Investigation, Writing - review & editing

Corresponding author

Correspondence to Giuseppina Andresini .

Ethics declarations

Conflict of interests.

The authors declare that they have no conflict of interest.

Competing Interests

No competing interests.

Ethics Approval

We declare that this submission follows the policies as outlined in the Guide for Authors. The current research involves no Human Participants and/or Animals.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Andresini, G., Appice, A., Ienco, D. et al. DIAMANTE: A data-centric semantic segmentation approach to map tree dieback induced by bark beetle infestations via satellite images. J Intell Inf Syst (2024). https://doi.org/10.1007/s10844-024-00877-6

Download citation

Received : 27 February 2024

Revised : 24 May 2024

Accepted : 01 August 2024

Published : 04 September 2024

DOI : https://doi.org/10.1007/s10844-024-00877-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Data-centric remote sensing
Sentinel-1 data
Sentinel-2 data
Data fusion
Data imbalance
Semantic segmentation
Bark beetle detection

Find a journal
Publish with us
Track your research

An official website of the Department of Health & Human Services

Search All AHRQ Sites
Email Updates

1. Use quotes to search for an exact match of a phrase.

2. Put a minus sign just before words you don't want.

3. Enter any important keywords in any order to find entries where all these terms appear.

The PSNet Collection
All Content
Perspectives
Current Weekly Issue
Past Weekly Issues
Curated Libraries
Clinical Areas
Patient Safety 101
The Fundamentals
Training and Education
Continuing Education
WebM&M: Case Studies
Training Catalog
Submit a Case
Improvement Resources
Innovations
Submit an Innovation
About PSNet
Editorial Team
Technical Expert Panel

The next step in learning from sentinel events in healthcare.

Bos K, Dongelmans DA, Greuters S, et al. The next step in learning from sentinel events in healthcare. BMJ Open Qual. 2020;9(1):e000739. doi:10.1136/bmjoq-2019-000739.

To support the development of a national approach to systems learning from sentinel events (SEs) in the Netherlands, this study evaluated how SEs are handled and leveraged for learning across eight academic hospitals. These hospitals use different approaches to analyze events, rarely use external experts in the process, and lack predetermined criteria for selecting recommendations for implementation. Given that five of eight included hospitals reported recurrence of similar SEs, the authors highlight the need for a transparent approach to assessing events and implementing recommendations successfully.

Grading recommendations for enhanced patient safety in sentinel event analysis: the recommendation improvement matrix. June 12, 2024

Performance variability in perioperative sentinel events: report on a nationwide data set. April 20, 2022

Prioritising recommendations following analyses of adverse events in healthcare: a systematic review. November 4, 2020

The effect of computerised decision support alerts tailored to intensive care on the administration of high-risk drug combinations, and their monitoring: a cluster randomised stepped-wedge trial. February 14, 2024

Adverse drug events caused by three high-risk drug-drug interactions in patients admitted to intensive care units: a multicentre retrospective observational study. October 18, 2023

Improving patient safety in the ICU by prospective identification of missing safety barriers using the Bow-Tie prospective risk analysis model. October 2, 2013

A prospective study of paediatric cardiac surgical microsystems: assessing the relationships between non-routine events, teamwork and patient outcomes. July 6, 2011

Dedicated teams to optimize quality and safety of surgery: a systematic review. November 16, 2022

Effect of cognitive aids on adherence to best practice in the treatment of deteriorating surgical patients: a randomized clinical trial in a simulation setting. January 8, 2020

Where are my instruments? Hazards in delivery of surgical instruments. August 10, 2016

To err is system: a comparison of methodologies for the investigation of adverse outcomes in healthcare. May 5, 2021

Will the COVID-19 pandemic transform infection prevention and control in surgery? Seeking leverage points for organizational learning. January 27, 2021

Transformational improvement in quality care and health systems: the next decade. November 25, 2020

Operational measurement of diagnostic safety: state of the science. October 7, 2020

Resilience and regulation, an odd couple? Consequences of Safety-II on governmental regulation of healthcare quality. May 13, 2020

Sharing lessons learned to prevent adverse events in anesthesiology nationwide. August 21, 2019

Development of "SWARM" as a model for high reliability, rapid problem solving, and institutional learning. November 4, 2015

Establishing a global learning community for incident-reporting systems. November 10, 2010

Using medical-error reporting to drive patient safety efforts. October 11, 2006

Creating a safer health care system: finding the constraint. December 21, 2005

Connect With Us

Sign up for Email Updates

To sign up for updates or to access your subscriber preferences, please enter your email address below.

Agency for Healthcare Research and Quality

5600 Fishers Lane Rockville, MD 20857 Telephone: (301) 427-1364

Accessibility
Disclaimers
Electronic Policies
HHS Digital Strategy
HHS Nondiscrimination Notice
Inspector General
Plain Writing Act
Privacy Policy
Viewers & Players
U.S. Department of Health & Human Services
The White House
Don't have an account? Sign up to PSNet

Submit Your Innovations

Please select your preferred way to submit an innovation.

Continue as a Guest

Track and save your innovation

in My Innovations

Edit your innovation as a draft

Continue Logged In

Please select your preferred way to submit an innovation. Note that even if you have an account, you can still choose to submit an innovation as a guest.

Continue logged in

New users to the psnet site.

Access to quizzes and start earning

CME, CEU, or Trainee Certification.

Get email alerts when new content

matching your topics of interest

in My Innovations.

COMMENTS

Case Examples
Case Examples. Based on a composite of a number of real sentinel event reports to The Joint Commission, Case Examples can be used for educational purposes to identify lapses in patient safety and missed opportunities for developing a safety culture. This learning resource highlights safety actions and strategies to have a better result.
Sentinel events. In memory of Ben—a case study.
Haas D. Sentinel events. In memory of Ben--a case study. Jt Comm Perspect. 1997;17(2):12-5. Copy Citation. ... Haas D. Sentinel events. In memory of Ben--a case study. Jt Comm Perspect. 1997;17(2):12-5. Copy Citation. Format: Download Citation. Related Resources From the Same Author(s) Exclusion of residents from surgery-intensive care team ...
Sentinel Event
Sentinel Event. Group of medical experts working on wireless technology at doctor's office. A sentinel event is a patient safety event that results in death, permanent harm, or severe temporary harm. Sentinel events are debilitating to both patients and health care providers involved in the event. The Joint Commission works closely with its ...
Sentinel Event
The Joint Commission defines a sentinel event as a patient safety event that results in death, permanent harm, or severe temporary harm. Sentinel events are debilitating to both patients and health care providers involved in the event. The term sentinel refers to a system issue that may result in similar events in the future. The National Quality Forum defined the term serious reportable ...
Sentinel Events: Approaches to Error Reduction and Prevention
Several case studies and examples dem-onstrate successful event investigation and improvement efforts in health care organizations. This excerpt addresses prevention of sentinel events through proactive, risk-reduction approaches.
Team experiences of the root cause analysis process after a sentinel
A case study is a flexible research design that captures holistic and meaningful characteristics of actual life events . Case studies can provide a detailed understanding of what is happening and solid grounds for improvement . Case study research has a strong advantage in examining the relevant process . It can capture the complexity of a case ...
"The Other Side of the Fence": A Geriatric Surgical Case Study of Error
Finally, the surgeon's discovery during re-operation that the "wrong angle" plate had been used in the initial surgery focused attention on the case as a sentinel event. A sentinel event is defined by The Joint Commission as an unexpected occurrence involving death or serious physical or psychologic injury, or the risk thereof. 1 Regarding ...
Clinical nurses' experiences with sentinel events
This study describes nurses' experiences with sentinel events in hospital settings, including intensive care, medical-surgical, long-term care, psychiatric, and Alzheimer units. Figure. Little is known about nurses' perceptions of sentinel events (SEs) and/or the changes needed in the work environment to best support nurses following such events.
Team experiences of the root cause analysis process after a sentinel
Keywords Root cause analysis, Qualitative case study, Sentinel events, Organizational learning, Norway, Childbirth Introduction In the healthcare landscape, the paramount objective
Root Cause Analysis: Responding to a Sentinel Event
Root Cause Analysis. The Joint Commission designates events as sentinel because they require an immediate investigation and response. Accredited organizations are expected to respond to sentinel events with a "thorough and credible root cause analysis [RCA] and action plan" (The Joint Commission, 2013a, p. 12).
Sentinel Event
The Joint Commission defines a sentinel event as a patient safety event that results in death, permanent harm, or severe temporary harm. Sentinel events are debilitating to both patients and health care providers involved in the event. The term sentinel refers to a system issue that may result in similar events in the future.
Sentinel Events
The Human Toll. Such sentinel events are all too common. According to a just-released report, Preventing Medication Errors, prepared by the Institute of Medicine (IOM) at the behest of the Centers for Medicare and Medicaid Services, medication errors harm 1.5 million people yearly in the U.S. and kill thousands.The annual cost: at least $3.5 billion
Sentinel event
A sentinel event is "any unanticipated event in a healthcare setting that results in death or serious physical or psychological injury to a patient, not related to the natural course of the patient's illness". [1] Sentinel events can be caused by major mistakes and negligence on the part of a healthcare provider, and are closely investigated by healthcare regulatory authorities.
Intraoperative Sentinel Events in the Era of Surgical Safety Checklists
The Joint Commission defines sentinel events as unexpected occurrences involving death or serious physical or psychological injury or the risk thereof. 1 Such events are called "sentinel" because they signal the need for immediate investigation and response. Although The Joint Commission allows hospitals to define their own list of additional sentinel events, mandatory reportable surgical ...
Root Cause Analysis: Responding to a Sentinel Event : Home Healthcare Now
Accredited organizations are expected to respond to sentinel events with a "thorough and credible root cause analysis [RCA] and action plan" ( The Joint Commission, 2013a, p. 12). RCA can be defined as "a process for identifying the basic or causal factors that underlie variation in performance ( Anderson et al., 2010, p. 8).
Suicide, A Sentinel Event HESI Case Study Flashcards
Study with Quizlet and memorize flashcards containing terms like As the nurse documents Mr. Fearon's assessment, the nurse is correct to question which activity of a client with DM2?, Mr. Fearon mentions that he feels "blue" lately because his wife died one year ago, and his children live out of state and seldom visit. The nurse knows that the greatest risk for major depression includes which ...
Revised definition of suicide in Sentinel Event Policy
The Joint Commission revised its definition of suicide in the Sentinel Event Policy, effective Jan. 1, 2024. The original definition, developed more than 10 years ago, focused on inpatient and "staffed around-the-clock" care settings or suicides within 72 hours of discharge. Data and evidence-based literature support extending the time ...
Team experiences of the root cause analysis process after a sentinel
Root cause analysis (RCA) is a frequently used, and sometimes mandatory, method to investigate sentinel events. In this study, members of an RCA committee were interviewed before and after an RCA investigation to elicit their experiences and assess compliance with the Norwegian RCA process. Organizational factors and team composition presented challenges, particularly the inclusion of staff ...
Sentinel Event.
Since 1998, The Joint Commission has issued sentinel event alerts in response to unexpected incidents involving death or serious physical or psychological injury (or risk thereof). These events are identified as sentinel due to the gravity of the injury and the need for immediate investigation and response. The goal is often to determine the root causes involved and provide recommendations for ...
Sentinel events. In memory of Ben--a case study
Sentinel events. In memory of Ben--a case study. Sentinel events. In memory of Ben--a case study Jt Comm Perspect. Mar-Apr 1997;17(2):12-5. Author D Haas 1 Affiliation 1 Martin Memorial Health System, USA. PMID: 10177138 No abstract available. Publication types Case Reports ...
The relationships between patient safety culture and sentinel events
Our study confirms that a more positive patient safety culture is associated with lower occurrence of sentinel events. To minimize the fear of sentinel events reporting and to improve overall patient safety a culture change is needed by promoting a blame-free culture and improving teamwork, handoffs, and communication openness.
Sentinel events, serious reportable events, and root cause ...
This commentary describes the importance of performing root cause analyses following sentinel events and never events in order to identify factors that contribute to failure and develop solutions to reduce risks. The authors use ophthalmologic examples to illustrate the elements of a systematic approach to root cause analysis following a never event, along with recommendations for ...
DIAMANTE: A data-centric semantic segmentation approach to ...
The evaluation shows the effectiveness of the proposed approach in a real inventory case study that regards non-overlapping forest scenes from the Northeast of France acquired in October 2018. ... Hölzer, T., Weiss, J., Fraccaro, P., Zortea, M., Brunschwiler, T. (2022). Flood event detection from sentinel 1 and sentinel 2 data: Does land use ...
The next step in learning from sentinel events in healthcare
To support the development of a national approach to systems learning from sentinel events (SEs) in the Netherlands, this study evaluated how SEs are handled and leveraged for learning across eight academic hospitals. These hospitals use different approaches to analyze events, rarely use external experts in the process, and lack predetermined criteria for selecting recommendations for ...

Lippincott ® Nursing Center ®

Article Content

Sentinel Event

Root Cause Analysis

Identifying the RCA Team

Gathering Information

Organizing Information

Contributing Factors

Identifying the Root Cause

Action Plans

RCA Case Study: Retained Foreign Object

Connect with NursingCenter

All Content

The Human Toll

Focus on Medication Errors

Secondary Logo

Root Cause Analysis

Sentinel Event

Identifying the RCA Team

Gathering Information

Organizing Information

Contributing Factors

Identifying the Root Cause

Action Plans

RCA Case Study: Retained Foreign Object

Revised definition of suicide in Sentinel Event Policy

Save citation to file

Add to My Bibliography

Sentinel events. In memory of Ben--a case study

Similar articles

Publication types

LinkOut - more resources

The relationships between patient safety culture and sentinel events among hospitals in Saudi Arabia: a national descriptive study

Yasser A. Alaska

Alia Albaharnah

Nawfal Aljerian

Rabab Alkutbe

Supplementary Information

Study design

Study setting and data sources

Dataset 1: The patient safety culture survey

Dataset 2: The sentinel events reports

Data analysis

Descriptive summary of the two data sources

Linking the patient safety culture and the reported sentinel events

The relationship between the patient safety culture and the reported sentinel events

Patient safety culture domains and contributory factors

Patient safety culture domains and level of harm

The influence of patient safety domains on sentinel events, patient overall safety grade (POSG), and positive events reported (PER)

Conclusions

Acknowledgements

Abbreviations

Authors’ contributions

Availability of data and materials

DIAMANTE: A data-centric semantic segmentation approach to map tree dieback induced by bark beetle infestations via satellite images

Cite this article

Explore related subjects

1 Introduction

2 Related work

2.1 Bark beetle detection in remote sensing

2.2 Data-centric artificial intelligence in remote sensing

3 Study area and data preparation

3.1 Sentinel-1 and Sentinel-2 data collection

3.2 Multisensor data alignment

3.3 Ground truth data, datasets and statistics

4 Semantic segmentation model development

5 Empirical evaluation and discussion

5.2 Metrics

5.3 Results

5.3.1 Performance Analysis

5.3.2 Temporal analysis

5.3.3 Transferability analysis

5.4 Considerations and findings

6 Conclusion

Data, Material, and/or Code Availability

Acknowledgements

Author information

Contributions

Corresponding author

Ethics declarations