The Safety Action Feedback and Engagement (SAFE) Loop: Initial Testing and Refinement of a Novel Intervention to Enhance Hospital Incident Reporting and Patient Safety

Abstract

Objectives:

Voluntary incident reporting has improved safety in many high-risk industries, but barriers limit its effectiveness in hospitals. To overcome these, we designed the Safety Action Feedback and Engagement (SAFE) Loop to obtain input from nurses about and prioritize patient safety problems, teach nurses to write more informative reports and report priority incidents, standardize investigations, and provide feedback to nurses about mitigation plans. The present work sought to iteratively test and refine the nascent intervention.

Methods:

Set at a large U.S. academic hospital, this work had three phases: (A) proof-of-concept testing on two nursing units, exploring potential feasibility and acceptability; (B) iterative refinements to intervention design and implementation plans; and (C) structured pilot testing on one nursing unit to produce descriptive and qualitative data related to the feasibility and acceptability of the refined intervention, and feasibility of future outcome measures.

Results:

Proof-of-concept testing demonstrated the feasibility of intervention elements and nurses’ enthusiasm. Incorporating iterative refinements, the structured pilot testing reaffirmed the feasibility of implementing the full SAFE Loop intervention and of collecting outcome data. Qualitative interviews after the pilot reaffirmed intervention acceptability.

Conclusions:

The SAFE Loop shows promise as an approach to enhancing hospital incident reporting systems. The present work illustrates how patient safety leaders can build from an initial intervention design toward a refined design and implementation plan, with demonstrated acceptability to stakeholders and feasible implementation.

Introduction

Medical errors continue to harm hundreds of thousands of hospitalized patients in the U.S. annually, and voluntary incident reporting is a widely discussed technique for improving patient safety, based in its successful use in other industries.^1–5 The technique relies on two premises: first, a harmful safety event is often preceded by numerous near-miss incidents with similar causes;^6–8 and second, first-hand witnesses can describe specific conditions leading up to an incident.⁹ By analyzing detailed narratives and conducting follow-up investigations, organizations can learn how work-system factors contribute to safety incidents.^4,5

Since 1965, U.S. hospitals have used incident reporting to both improve safety and alert risk managers of potential legal concerns.¹⁰ Nurses see reporting as a responsibility and file most reports,¹¹ but they experience barriers to reporting.^12,13 Nurses are often uncertain about which events to report and what information to include.¹⁴ They fear punitive responses and doubt whether reports will improve systems of care.^14–18 Although hospitals typically receive thousands of reports annually,^19,20 many reports describe similar low-risk problems and lack information about contributing work-system factors.^11,21–24

To address these barriers and enhance hospital incident reporting systems as tools for improving patient safety, we designed the Safety Action Feedback and Engagement (SAFE) Loop, a novel intervention with six complementary elements. The present work sought to iteratively test and refine the nascent intervention, in preparation for a subsequent randomized controlled trial. While the SAFE Loop could apply to any safety problem, we focused on medications because errors are common,^25–28 and preventing them is central to nurse practice.²⁹

Methods

We report this work in accordance with applicable guidelines for pilot and feasibility trials.^30,31

Study Design:

In the development of a novel intervention, Stage I includes generating the basic design (completed previously) and then iteratively testing and refining the design to maximize the acceptability to stakeholders and feasibility of implementation (present work), prior to fully-powered efficacy testing in Stage II (randomized trial).^32–35 Here, our iterative testing and refinement process included three well-recognized steps:^32–35

(A) Proof-of-concept testing in a small select population, using qualitative and descriptive data, can shed initial light on the intervention’s feasibility and acceptability to stakeholders;

(B) Iterative refinement can improve intervention design and support planning for full implementation; and

(C) Structured pilot testing can assess feasibility and acceptability of the refined intervention, and gather preliminary data on efficacy.^32,33

Setting:

This study occurred at a 915-bed major academic hospital in Los Angeles with 1,980 nurses. The hospital employs several medication safety-related systems and technologies, including BD Pyxis^™ automated dispensing machines (BD and Company, Franklin Lakes, NJ).

To support incident reporting and analysis, the hospital uses RL6: Risk (RL Solutions, Cambridge, MA). At baseline, after an incident report is filed, the unit leadership conducts a brief review that may include speaking with involved parties. Additionally, the hospital medication safety team reviews all incidents, discusses risks, assigns levels of harm, conducts investigations, and plans improvements.

Participants:

On three nursing units, leaders and front-line nurses participated in the iterative testing and refinement process.

Intervention:

The SAFE Loop adds six key elements to established hospital incident reporting systems (Table 1). A critical element is that nursing units select a specific ongoing safety concern, which we call a Target Event (see Table), around which to focus investigative and mitigation activities. This approach promotes nurses’ engagement throughout the SAFE Loop by valuing their insights into safety threats in their work environment. A prior publication describes another element, an online training program, in greater detail.³⁶

Table 1:

Limitations of Current Hospital Incident Reporting Practices, Conceptual Rationale, and Key Elements of the Safety Action Feedback and Engagement (SAFE) Loop

Limitations of Current Reporting Practices	Conceptual Rationale	SAFE Loop Key Elements and Definitions
Lack of engagement: Nurses sometimes doubt that reports are used to improve safety, which deters reporting.15–17,47	Nurses are more likely to report events and adopt changes for problems that they find to be particularly common, severe, or preventable and/or impactful to the quality and efficiency of nursing practice.43–45 Engagement is needed between nurses and safety leaders throughout the process of reporting, investigation, and improvement.	1. Input from Nurses: The SAFE Loop team solicits input from unit leaders and front-line nurses on which patient safety problem to prioritize for reporting, investigation, and mitigation.
Lack of prioritization: Hospitals passively gather thousands of reports on low-priority events each year, making it hard to identify important threats that warrant follow-up and action.11,19,20	Emphasizing one specific problem with a narrow scope may enable safety leaders to make the best use of investigative resources and obtain enough reports to understand the problem. Common, preventable, and potentially severe events should be a high priority.	2. Emphasis on a Selected “Target Event”: The SAFE Loop focuses on each nursing unit’s unique Target Medication Event. These are errors and adverse events related to a selected, defined problem involving a particular drug class (e.g., opioids), population (e.g., kidney disease), type of error (e.g., duplicate therapy), and/or work-system factor (e.g., conflicting protocols).
Lack of guidance about what and how to report: This deters reporting and limits reports’ usefulness.15–17,47 Gathering rich qualitative insights into contributing factors is central to successful incident reporting but hospital reports provide little detail on contributing factors.4–6,10,15–17,19,20,23,47–49 Effective reporting systems prioritize near misses. When reporting events involving actual harm, nurses are often concerned about being blamed or disciplined. Reports that address harm usually contain little information about contributing work-system factors.22	If nurses receive guidance on what and how to report, they would be more likely to submit reports. Report narratives that include more details about contributing work-system factors may be more informative and actionable for efforts to redesign systems and improve safety. A mnemonic may help nurses to remember which types of details to include in incident reports. If reporting emphasizes near misses, this may reduce the risk that reporters might be blamed for the event.	3. Training in Enhanced Reporting: The SAFE Loop involves teaching nurses to write more informative incident reports that include details about contributing work-system factors. The SAFE Loop teaches nurses to organize the narrative using the Situation, Background, Assessment, Recommendations (SBAR) format.
Hospital incident reporting systems typically function as passive receptacles for incident reports. Nurses are often concerned about the time burden of reporting.15–17,47	Obtaining multiple related reports may provide a more thorough understanding of work-system factors. Asking each nurse to submit one report is a modest request. Having a deadline facilitates progress on a project that involves distinct phases.	4. Stimulated Time-limited Reporting: The SAFE Loop prompts nurses to report at least one Target Medication Event, if possible, during a designated period (in addition to any reports that nurses would usually submit).
Lack of standardized investigative procedures: There are no established procedures for conducting follow-up investigations of reported incidents.50,51 Few safety problems are unique to a particular nursing unit or hospital, and there may be opportunities to leverage published or publicly available information learned from other institutions. Certain AHRQ Patient Safety Organizations systematically examine thousands of incident reports and generate syntheses of their findings.	When interviewed, front-line nurses and other relevant parties may share additional insights about work-system factors. Leveraging publicly available information on the causes of similar events at other hospitals and solutions that have been tested at other hospitals would reduce the work involved and enhance the success of investigative activities and mitigation plans.	5. Investigation and Integration of Information: The SAFE Loop integrates information from incident reports, follow-up investigations, interviews with front-line nurses and other relevant parties, and internal and external sources (including published literature and AHRQ Patient Safety Organizations). Standard procedures, including mini-Root Cause Analyses (i.e., root-cause analyses of abbreviated scope) and Ishikawa diagrams, can facilitate investigation and integration and help design a mitigation plan.
Lack of feedback: Nurses seldom receive feedback after filing a report, and a lack of feedback is a major barrier to reporting.15–17,47	If nurses receive feedback and observe improvements in systems of care in response to incident reports, they may have more favorable perceptions of reporting, creating a positive feedback loop that promotes safer care.47	6. Feedback to Nurses: The SAFE Loop Team provides feedback about the causes of safety problems and mitigation plans.

The conceptual foundation for the SAFE Loop is a long-standing technique in quality improvement, the Plan-Do-Study-Act (PDSA) cycle.³⁷ Elements in the SAFE Loop design also drew from institutional knowledge and literature on the limitations of current reporting systems,^{11,21–23,38,39} and evidence-based strategies for improving quality/safety.^40–49

Step (A) Proof-of-concept Testing:

The Chief Patient Safety Officer (ES) oversaw a preliminary test on two nursing units from January to March 2019, including a neurosurgery/spine floor unit with 78 nurses, and an adult medicine floor unit with 103 nurses. We selected these units based on having medical-surgical patient populations, engaged nursing staff, and a leader to serve as the intervention champion.

Implementation:

First, we met with unit leaders and nursing staff to prioritize a Target Event at routine shift-change “huddles,” where nurses hear key announcements. Next, the safety leadership team encouraged nurses to submit examples of the Target Event, emphasizing near misses. We met weekly with unit managers and nursing staff during huddles to discuss findings from incident reports and mini-root-cause analyses (see below).⁵⁰ Finally, we worked with unit managers to consider potential system changes, obtain feedback, implement tests, and disseminate information.

Evaluation:

From February 5 through March 25, 2019, we examined the numbers of reports submitted per week on these nursing units as well as proportions that were near misses (National Coordinating Council for Medication Error Reporting [NCC MERP] categories A-B).^51,52 Before and after implementation, we surveyed nurses to generate a Culture of Safety composite measure, drawn from the AHRQ Hospital Survey on Patient Safety.^53–55 Lastly, we asked nursing unit leaders and front-line nurses open-ended questions to obtain feedback.

Step (B) Iterative Refinement:

Building on proof-of-concept testing and guided by the updated Consolidated Framework for Implementation Research, a widely used implementation framework,⁵⁶ we refined and operationalized the SAFE Loop and planned implementation procedures. We broke key elements down into component steps, proposed a timeline, defined roles in the SAFE Loop (Table 2), we obtained input and feedback from stakeholders, and we established how to engage personnel on each nursing unit.

Table 2:

Roles of Personnel in the Existing Reporting System and the SAFE Loop Intervention

	Position	Roles in Existing Reporting System	ADDITIONAL Roles in SAFE Loop Intervention
	Nurses on Study Nursing Units	Spontaneously submit incident reports.	Provide input on the Target Events, undergo training in reporting, submit reports, receive feedback, and implement mitigation strategies.
	Nursing Unit Leaders	Administratively responsible for overseeing efforts to improve the safety of care on each nursing unit. Participate in analyzing reports, conduct any investigations, design and implement any solutions.	Select the Target Events and work with the SAFE Loop Team throughout implementation.
SAFE Loop Team	Incident Report Clinical Reviewer	Classifies events, assigns harm severity, identifies source nursing unit, and triages high-risk events to relevant departmental and unit leaders.
	Patient Safety Leaders (Hospital, Pharmacy, Nursing)	Assist nursing units with investigations, obtain resources, engage leaders of relevant departments, and overcome barriers. Oversee root-cause analyses of harmful events.	Part of the SAFE Loop team. Facilitate investigation of Target Events.
	Target Event Champion	Not applicable.	Obtains input from and trains nurses, identifies and extracts information from reports on Target Events, assists with investigations, integrates information from reports and other sources, drafts feedback to nurses.
	AHRQ Patient Safety Organization	Not applicable.	Serves as a resource to the SAFE Loop Team, providing information on events at other hospitals.

We report iterative refinements and implementation under Step (B) and evaluation methods under Step (C), even though these steps were intertwined. Refinement and pilot testing occurred from March 15, 2022 to November 21, 2023 with a floor unit for adults with neurological conditions. We chose this nursing unit based on the leadership team’s willingness to play an active role in all aspects of testing and refining the intervention, particularly in providing the research team with detailed input and feedback. The number of nurses on the unit was initially 26 and increased to 59 following a reorganization.

Obtaining Input from Nurses and Selecting a Target Event:

Choosing the Target Event was an iterative process that involved several rounds of discussions between the research team, champion, and nursing unit leaders, based on perceptions of frequency of risk, clarity of event definition for nurses, likelihood of mitigation, among others. The research team and champion visited nursing huddles to introduce the SAFE Loop and discuss potential Target Events. Once the unit leaders chose the Target Event, we revisited the huddles to encourage reporting over a three-month period.

Training in Enhanced Reporting:

As reported previously,³⁶ our team created an online training program that introduced the SAFE Loop and taught nurses to report near misses, describe contributing work-system factors in the narrative section, and use the Situation, Background, Assessment, Recommendation (SBAR) format to organize the narrative.⁵⁷

Prompting Incident Reporting:

We periodically revisited huddles to remind nurses to report, monitored incidents being reported, and let nurses know that we received and appreciated relevant reports. We also checked in with the unit leaders about nurses’ perspectives on the training, Target Event, and other aspects of the SAFE Loop.

Investigative Procedures:

The next element is an abbreviated (“mini”) root cause analysis (RCA) of the Target Event. The mini-RCAs are derived from traditional RCAs, which involve comprehensive efforts to reveal all active and latent errors that lead to adverse outcomes.^50,58 However, the mini-RCAs are shorter and more focused because they seek to reveal work-system factors that contribute to a specific type of error.

First, the champion reviewed the medication-related incident reports submitted during the six months preceding and during the pilot test. They considered whether each incident might be relevant to the Target Event, identified contributing work-system factors in the narrative, and used the Human Factors Analysis and Classification System for Healthcare (HFACS-Healthcare) to categorize them. HFACS-Healthcare is a tool for analyzing safety events in healthcare by identifying and categorizing the work-system factors that lead to errors, to help standardize the analysis of causes and develop effective mitigation strategies. HFACS-Healthcare categorizes causes of errors into four levels: unsafe acts, preconditions for unsafe acts, unsafe supervision, and organizational influences.⁵⁹

Concurrently, the champion and research team interviewed nursing staff on the unit over two 90-minute sessions. During each session, unit leaders directed one-to-three nurses to a conference room, to explore their thoughts about the Target Event and contributing factors. Each nurse participated for about 10–15 minutes. The champion and research team briefly interviewed hospital personnel who could offer additional insights, such as pharmacists, physicians, and other administrative leaders.

At the same time, we discussed the Target Event history at the institution with the hospital’s Pharmacist Manager for Medication Safety and the Chief Patient Safety Officer, who shared findings from prior investigations, relevant work-system factors, and potential mitigation strategies and barriers. The champion also sought external information through a brief, directed (non-systematic) literature review and requested relevant information from the study’s partners at two AHRQ-designated Patient Safety Organizations: the Hospital Quality Institute in California and the Institute for Safe Medication Practices.

To conclude, the champion created an Ishikawa Diagram for the Target Event, which listed the specific contributing factors identified via incident reports and interviews.

Producing a Mitigation Plan and Providing Feedback:

We met with the nursing unit leaders, the Pharmacist Manager for Medication Safety, the Chief Patient Safety Officer, hospital nursing leaders, and leaders of other relevant departments as needed. We discussed the Target Event and the contributing factors, including those that were reported most often, appeared most preventable, and/or posed the most significant risk. We discussed relevant institutional history and information from external sources.

Nursing leaders ultimately decided what work-system factors to prioritize for mitigation and implemented the mitigation plan based on what changes were feasible. Once the strategy was selected, the research team drafted a mitigation plan, reiterating the Target Event, describing the contributing factors, steps we took in the work, any outside information identified, and the unit’s final mitigation plan. After nursing unit director approval, we emailed the mitigation plan to nurses on the unit. The nursing unit director also communicated the mitigation plan to nurses via routine daily huddles, in-service activities, and day-to-day management on the nursing unit.

Step (C) Structured Pilot Testing:

Here, the focus was on assessing feasibility and acceptability to nursing unit leaders and nursing staff, and gathering initial data on efficacy measures to refine assessment methods.

Incident Reporting Practices:

We measured the number of incident reports submitted on the units from June 2, 2022, to December 12, 2022, and a prior six-month period (December 1, 2021, to June 1, 2022). Dual reviewers examined and reached a consensus on whether each report pertained to the unit’s selected Target Event and included work-system factors; the team adjudicated any ties. We also used severity scores assigned by the reporter to identify near misses (NCC MERP categories A-C).^51,52,60

Surveys of Nurses:

Before and after implementation, we fielded a 10-item survey with front-line nurses by email (Appendix). To promote the survey, we visited the nursing units with snacks and provided encouragement. Additionally, nurses who completed each survey were entered into a lottery for wireless earbuds.

From the AHRQ Hospital Survey on Patient Safety (SOPS) Culture^™, Version 2.0 (English),⁶¹ we derived two composite measures, Communication about Error (3 items) and Reporting Patient Safety Events (2 items), and two individual items, Blame-free Culture and Incident Report Submissions. We also created a novel composite measure, Confidence in Reporting (2 items), and a novel individual measure, Time per Report.

Most individual items were scored on a scale from 1 (never or not at all) to 5 (always or extremely). Composite scores are calculated as the percent of positive responses (most of the time or always). Incident Report Submissions included five categories (0 to 4+ reports). Time per Report included 5-minute increments up to 90 minutes.

Implementation Interviews:

After SAFE Loop implementation, we conducted one-on-one in-person interviews with a convenience sample of three nursing unit leaders and three frontline nurses. We used a semi-structured interview guide with open-ended questions and follow-up probes, guided by the Consolidated Framework for Implementation Research and key publications on information to report about quality improvement interventions.^56,62–64 Questions elicited respondents’ views regarding contextual factors, implementation details, and other information. Pairs of researchers trained in qualitative analysis analyzed interview transcripts in Dedoose© (Sociocultural Research Consultants, LLC, Manhattan Beach, CA) using a combination of content analysis and qualitative inquiry, allowing them to identify key themes nurses emphasized. We developed initial themes a priori based on previous literature and notes from interviews. We then used content coding and team review to iteratively refine themes and develop new themes for emergent concepts using a constant comparative approach.⁶⁵

Human Subjects Protections:

The Institutional Review Board (IRB) reviewed and approved this work. The proof-of-concept testing (Pro00055431) was considered quality improvement and not research. The iterative refinement and pilot testing (STUDY00001025) was also considered quality improvement, while the associated data collection constituted research. Written informed consent was obtained for nurse surveys and interviews. Consent was waived for reuse of existing data.

Results

Step (A) Proof-of-concept Testing:

Feasibility of Implementation:

Unit leaders and front-line nurses were enthusiastic about and able to select a Target Event. The nurses submitted several incident reports relevant to the Target Events. The unit leaders and nursing staff were receptive to discussing findings from the incident reports and mini-root-cause analyses. Finally, it was feasible to work with the nursing units to develop and disseminate mitigation plans.

Reporting Practices:

Before the pilot test on these two nursing units, an average of four incident reports were submitted per week and 6.25% represented near misses. During the pilot test, an average of 10 reports were submitted per week and 47% of reports addressed near misses.

Surveys of Nurses:

Before implementation, the Culture of Safety composite measure was 91% on one unit and 75% on the other. These percentages were 94% and 79%, respectively, after implementation.

Interviews with Nurses:

Feedback from unit leaders was strongly favorable: “SAFE Loop was an effective test of change as it engaged front-line staff to voice input and concerns to prevent near misses. It improved their workflow processes in terms of helping eliminate workarounds or delays. Staff had an increased feeling of empowerment.”

One front-line nurse commented, “We saw the value of these rounds. It made our job easier and more efficient, and we could see the long-term efforts of putting in the near misses now.” Another said, “[The advantages include] more reporting of ‘near misses’ and ensuring all staff receive follow-up reports.”

Step (B) Refined Intervention Design and Implementation:

Obtaining Input from Nurses and Selecting a Target Event:

During SAFE Loop implementation in the structured pilot study, unit directors initially proposed the Target Event as “challenges using the Pyxis^™ safely,” due to recent incident reports on this issue, and consistent with recent literature.⁶⁶ Discussions revealed a wide variety of concerns. Consequently, the study team and nursing unit leaders narrowed the scope and reframed the Target Event as “wrong medication or wrong patient errors involving the Pyxis^™.” Further narrowing occurred during the investigative activities, as noted below.

Training in Enhanced Reporting:

Twenty of 26 full-time nurses (77%) completed the online SAFER training program.

Prompting Incident Reporting:

During the investigative interviews, we learned that nurses might collectively recognize a high-priority safety issue but not spontaneously submit reports about it. In this case, the Target Event occurred so often as to seem a routine part of care rather than a reportable incident. As such, the interviewers reinforced submission of reports on the Target Events.

Investigative Procedures:

During the mini-RCA process, we learned that nurses were generally eager to share their thoughts about the Target Event during the brief interviews. While speaking to individual nurses was helpful, nurses became more engaged when we talked with multiple at a time, with discussions stimulating richer insights. We also learned that interviews were able to quickly reveal a wide range of contributing factors.

The investigation yielded four incident reports directly relevant to the Target Event, 10 interviews (six with nurses on the unit, and four with other informants), 12 relevant peer-reviewed publications, and eight resources shared by the Patient Safety Organizations.

The incident reports and interviews yielded 21 unique contributing factors for this Target Event (Figure 1, Ishikawa Diagram). Of 21 factors, 17 were derived from nursing interviews alone, one from an incident report alone, and three from both nursing interviews and incident reports. The contributing factor discussed most often was distractions or interruptions while using the Pyxis^™, which was in an open area by the nurse station.

Figure 1: Results of Structured Pilot Test of SAFE Loop Intervention: Contributing Work-system Factors Identified by Investigation (Ishikawa Diagram).

Black text signifies contributing factors identified only during interviews, red text signifies contributing factors identified only from incident reports, and orange text signifies contributing factors identified and/or characterized by both data sources.

Producing a Mitigation Plan and Providing Feedback:

Nurses recognized numerous challenges to using Pyxis^™ machines safely, with many diverse contributing work-system factors. As such, nursing unit leaders had to prioritize and they chose to focus on distractions during use.

To address this, the nursing directors selected two mitigation strategies: (1) a new “quiet zone” sign on the Pyxis^™ to alert people to not interrupt nurses retrieving medications, and (2) a new policy to defer all non-urgent calls (i.e., from patients’ families) during morning medication administration periods to reduce interruptions. The nursing unit directors immediately implemented these changes, which were continued to be sustained on the unit as of October 2025, two years later.

Step (C) Structured Pilot Testing:

Reporting Practices:

In the six months before the pilot test, nurses submitted 48 medication-based incidents, including 18 (37.5%) related to the Pyxis^™. Of 28 events for which severity was reported, 22 (79%) were near misses. Across the 48 reports, nurses described 53 contributing work-system factors (Table 3).

Table 3:

Results of Structured Pilot Test of SAFE Loop Intervention: Contributing Work-system Factors Obtained from Incident Report Narratives

HFACS-Healthcare Tier	Contributing Factor (CF) Category	Baseline Reports	Pilot Test Reports
Organizational Influences	Organizational Climate, N (% of total CF)	0 (0%)	0 (0%)
	Operational Process, N (% of total CF)	0 (0%)	0 (0%)
	Resource Management, N (% of total CF)	0 (0%)	0 (0%)
	Subtotals	0 (0%)	0 (0%)
Supervisory Factors	Inadequate Supervision, N (% of total CF)	0 (0%)	1 (2.6%)
	Planned Inappropriate Operations, N (% of total CF)	0 (0%)	0 (0%)
	Failed to Correct Problem, N (% of total CF)	0 (0%)	1 (2.6%)
	Supervisory Violations, N (% of total CF)	0 (0%)	0 (0%)
	Subtotals	0 (0%)	2 (5.1%)
Preconditions for Unsafe Acts	Physical Environment, N (% of total CF)	0 (0%)	0 (0%)
	Tools/Technology, N (% of total CF)	2 (3.8%)	0 (0%)
	Task, N (% of total CF)	1 (1.9%)	3 (7.7%)
	Mental State, N (% of total CF)	1 (1.9%)	0 (0%)
	Physiological State, N (% of total CF)	0 (0%)	0 (0%)
	Fitness for Duty, N (% of total CF)	0 (0%)	0 (0%)
	Communication, N (% of total CF)	7 (13%)	3 (7.7%)
	Coordination, N (% of total CF)	3 (5.7%)	3 (7.7%)
	Leadership, N (% of total CF)	0 (0%)	0 (0%)
	Subtotals	14 (26%)	9 (23%)
Unsafe Acts	Decision errors, N (% of total CF)	6 (11%)	7 (18%)
	Skill-Based Errors, N (% of total CF)	26 (49%)	17 (44%)
	Perceptual Errors, N (% of total CF)	0 (0%)	0 (0%)
	Routine Violations, N (% of total CF)	0 (0%)	3 (7.7%)
	Exceptional Violations, N (% of total CF)	1 (1.9%)	0 (0%)
	Subtotals	33 (62%)	27 (69%)
	Total Contributing Factors	53 (100%)	39 (100%)
	Total Incident Reports	48	41

During the pilot, nurses submitted 41 medication-based incidents, including 8 (19.5%) related to the Pyxis^™. Of 30 events that rated severity, 25 (83%) were near misses. Across the 41 reports, nurses described 39 contributing work-system factors.

Surveys of Nurses:

Survey responses were similar before and after SAFE Loop implementation (Table 4), except that the time per report rose from 10 to 15.5 minutes.

Table 4:

Results of Investigation Structured Pilot Test on One Nursing Unit: Attitudes of Front-line Nurses Toward Reporting

	Baseline Survey	Survey after Pilot Test
Response Rate	22 of 54 (40.7%)	29 of 59 (49.2%)
Composite Measures
Confidence in Reporting, % Positive Responses	70.5%	82.7%
Communication About Error, % Positive Responses	74.2%	70.1%
Reporting Patient Safety Events, % Positive Responses	70.5%	69.0%
Individual Items
Blame-free Culture, % Positive Responses	86.4%	82.8%
Incident Report Submissions, 0–4+
None	15 (68.2%)	19 (65.5%)
1	5 (22.7%)	5 (17.2%)
2	1 (4.5%)	2 (6.9%)
3	1 (4.5%)	1 (3.4%)
4 or more	0 (0%)	2 (6.9%)
Time per Report, Mean Minutes (SD)*	10 (5.8)	15.5 (10.0)

^*Excludes missing responses

Implementation Interviews:

Before SAFE Loop implementation, attitudes towards incident reporting were generally favorable. However, nurses were uncertain about what changes were made to improve safety in response to reporting.

After implementation, nurses voiced positive opinions of the potential for reporting to improve medication safety, and nursing leaders reported nurses were highly receptive and invested in reporting the Target Event (Table 5).

Table 5:

Qualitative Interviews with Nurses in Structured Pilot Test: Themes and Excerpts from Interviews

Comments Related to Incident Reporting on the Nursing Unit before SAFE Loop Implementation

All nurses reported it was their responsibility to submit incident reports because: “[Nurses are] the first in line to see [a] reportable event.”

The culture of safety was generally positive: “If we see something then we say something. We can tell the charge nurse or they really encourage us to just report anything, whether it’s medication related or [something else].”

However, nurses’ opinions were mixed about whether they received follow-up information after submitting a report. “I mean it really depends. So far, every time we’ve submitted forms, I’ve gotten the email [confirming that the incident report was received by the reporting system] for sure. And then just depending on the situation, I’ll have management come and say, ‘Hey, just so you know, we got the form and we’re looking into it’ kind of thing.” “Really, once the file has been reported, we don’t really hear about what happens.”

Nurses were less certain about whether submitting an incident report leads to changes in medication safety on the unit: “I’d like to think that leadership sees it and does things about it. And sometimes I feel like they do, sometimes they talk about what had happened in our huddle, but it’s not for everything. So, I really don’t know how effective the incident reporting is.” “I’m not sure because I don’t know if people, or who is reporting medication errors and what gets done about them. So maybe it’s effective, but I don’t see how it’s affected me before.” “And have you found that that med report incident reporting is effective at improving medication safety? Yes. For pharmaceuticals.”

Comments Related to Incident Reporting on the Nursing Unit after SAFE Loop Implementation

Nurses voiced a potential for an increase in incident reporting to improve medication safety, particularly the combination of reporting via the electronic interface and in-person to charge nurses. “[I think] people learn from their mistakes, people learn through mistakes… So, if the errors are recorded properly, then I believe they’ll learn about it. And then, we can make plans on how we can further prevent the same incident from happening, and then also study how this kind of incidents happen, if they have this multiple times.”

One nursing leader reported a high level of receptivity and investment from nurses in reporting the Target Event: “I think they have it down. They have it down. Okay. Yeah, they’re very invested and making sure that when the whole system started, the whole CS-Safe entry process started, they saw it as a way of maybe potentially falling on someone’s desk who was removed from the clinical aspects. We may do nothing with it. So, we weren’t seeing the real meaty, detailed CS-Safe events, but now we’re seeing more and a lot more of the entries being made and capturing. We try to tell them, remove the emotion, just get to the facts. And we’re seeing a lot more of that. So, it’s slowly getting less personal and getting to the core of what the issue was. So, we’re seeing they are more receptive.”

Discussion

The SAFE Loop is a novel multifaceted intervention designed to enhance hospital incident reporting systems as tools for improving patient safety. The present work involved initial proof-of-concept testing, detailed implementation planning, and iterative refinements, followed by a structured pilot test of the full intervention using qualitative and descriptive evaluation methods. The pilot test reaffirmed the feasibility of implementing the six intervention elements together, and qualitative interviews with nurses reaffirmed intervention acceptability. The pilot test also confirmed the feasibility of future efficacy measures for a planned randomized controlled trial.

This work illustrates a common scenario that both patient safety researchers and operational leaders face when developing a novel intervention. Developers may invest effort in understanding local safety problems and incorporate evidence-based improvement strategies—yet this does not assure feasibility, acceptability, or efficacy. Small-scale proof-of-concept testing, like we performed, can identify major barriers quickly and with modest resources, enabling refinements or justifying a decision not to proceed. Codifying intervention design and implementation plans is essential in research, such as our trial planning, but also in operational contexts where tacit knowledge might be lost if a safety leader leaves the role or institution. While assessing an intervention’s feasibility and acceptability can involve multiple descriptive and qualitative analyses, as used here, operational leaders can often leverage relevant data sources that already exist at an institution and obtain invaluable insights from a few conversations with relevant stakeholders. Statistical testing is rarely necessary at this stage because pilot tests have high rates of false positive and negative results relative to fully-powered trials of efficacy.⁶⁷

For the SAFE Loop, our present work informed intervention design and proposed analyses in the subsequent randomized trial. For example, we decided that the SAFE Loop would prioritize input from nurses, over numbers of previously reported incidents, when selecting the Target Event for a nursing unit. Pilot testing revealed that nurses collectively recognized a high-priority safety issue during the Target Event selection process, despite having filed few relevant incident reports because the event occurred so often as to seem a routine part of care.

As a second example, analyses in the randomized trial will now quantify and compare information obtained from interviews and incident reports. In pilot testing, the SAFE Loop detected 21 unique work-system factors that contributed to the Pyxis^™ Target Event. A surprise was that only one unique factor came from incident reporting alone, whereas 20 more factors came from brief investigative interviews with 10 nurses. Three of these factors were also in incident reports, but the interviews provided much richer information given the ability to ask follow-up questions. Thus, the mechanism by which the SAFE Loop improves safety may extend beyond the incident reports themselves.

Such insights have informed our ongoing pragmatic randomized controlled trial on 20 nursing units (1R01HS027455, registration number NCT05381441).^68,69 To our knowledge, no prior randomized trials have studied hospital incident reporting systems. The SAFE Loop addresses more barriers than prior interventions have,⁷⁰ and it includes novel elements including training nurses to recognize and report contributing work-system factors and standardizing follow-up procedures.⁷¹

Despite our efforts to optimize the SAFE Loop in preparation for the trial, it is likely to have limitations. Implementation will likely differ across nursing units due to local priorities and nurse engagement. The SAFE Loop focuses on events prioritized by nurses, who may have different perspectives relative to an objective prioritization process. Nonetheless, nursing units would benefit from new strategies for solving local safety problems. The SAFE Loop is resource intensive, so it would be best used to mitigate higher priority problems. As with many multifaceted interventions, it will be hard to isolate which SAFE Loop elements are most important. Preliminary testing involves non-representative populations, and insufficient sample sizes to draw quantitative conclusions. We tested the SAFE Loop at one hospital with a strong commitment to safety, so adaptation to and testing in other hospitals and contexts may be needed. Sustainability and reproducibility of the SAFE Loop remain to be determined.

Conclusion

The SAFE Loop is a novel patient safety intervention that shows promise as an approach to enhancing hospital incident reporting systems. The present work illustrates how patient safety leaders can build from an initial intervention design toward a refined design and implementation plan, with demonstrated acceptability to stakeholders and feasible implementation.