BACKGROUND AND OBJECTIVE: Safety I error elimination concepts are focused on retrospectively investigating what went wrong and redesigning system processes and individual behaviors to prevent similar future occurrences. The Safety II approach recognizes complex systems and unpredictable circumstances, mandating flexibility and resilience within systems and among individuals to avoid errors. We hypothesized that in our high-complexity and high-risk PICU, Safety II concepts contribute to its remarkably low adverse drug event rate. Our goal was to identify how this microsystem enacts Safety II.
METHODS: We conducted multidisciplinary focus group sessions with PICU members using nonleading, open-ended questions to elicit free-form conversation regarding how safety occurs in their unit. Qualitatively analyzing transcripts identified system characteristics and behaviors potentially contributing to low adverse drug event rates in PICU. Researchers skilled in qualitative methodologies coded transcripts to identify key domains and common themes.
RESULTS: Four domains were identified: (1) individual characteristics, (2) relationships and interactions, (3) structural and environmental characteristics, and (4) innovation approaches. The themes identified in the first 3 domains are typically associated with Safety I and adapted for Safety II. Themes in the last domain (innovation approaches) were specific to Safety II, which were layered on Safety I to improve results under unusual situations.
CONCLUSIONS: Safety II behavior in this unit was based on strong Safety I behaviors adapted to the Safety II environment plus innovation behaviors specific to Safety II situations. We believe these behaviors can be taught and learned. We intend to spread these concepts throughout the organization.
- ADE —
- adverse drug event
What’s Known on This Subject:
Safety II, a novel approach to patient safety, is focused on why processes perform correctly in high-performing units as opposed to why processes fail (Safety I). Specific factors characterizing Safety II behaviors have not been previously identified in health care.
What This Study Adds:
In this analysis of a high-performing microsystem, we identified factors in a hospital system and individuals that lead to more reliable performance and that support increased resilience in delivering patient care (Safety II).
The usual approach to optimizing safety outcomes in health care is often referred to as Safety I.1 This approach incorporates retrospective investigation after an error to determine the root cause(s) of system or individual failures, and thus leads to system redesign to eliminate similar future occurrences. Safety I system modifications generally constrain human behavior2 (eg, new protocol implementation, clinical practice guidelines, and electronic ordering restrictions). Thus, Safety I behaviors are those that prevent individuals from repeating the same mistakes. Although these methods have led to significant improvements,3–10 they can also be problematic for the reasons listed below. A new approach, Safety II, is focused on what has gone right in the system (the 9999 out of 10 000 events that did not fail),11,12 recognizing that systems are complex and viewing human behavior as a source of creativity as opposed to a dangerous source of variation that requires elimination.13 Thus, Safety II behaviors will be those that keep patients safe under circumstances that may not have been previously encountered.
A problem with the Safety I approach is that as errors decline, there are fewer opportunities to learn. In our institution, the adverse drug event (ADE) rate has decreased by 76% since our safety program (Zero Hero) began.7 Consequently, medication errors are often “one-offs” (rare events that are unlikely to happen again), leading to wasted resources because the system is redesigned in reaction to these singular events. However, because system designs have boundaries, unpredictable conditions or infrequent one-off events will continue to occur.
The term “resilience” in the context of engineering has a different and specific meaning compared with its more common usage. In engineering, resilience is defined as an individual’s or a system’s ability to “adjust its functioning before, during, or after changes and disturbances so that it can sustain required operations under both expected and unexpected conditions.”14,15 Resilience is an essential part of Safety II. Safety II requires an “adjustment to functioning,” which goes beyond “good catches” (situations in which error is avoided by performing an expected task). Resilience in the Safety II context can occur in minor, fleeting, and almost imperceptible ways, such as when a nurse reverses the order of 2 tasks, or in major ways, such as during a large-scale catastrophe.15
Because Safety II contains descriptions of individuals’ and systems’ abilities to adapt in real time to unexpected stress or unusual conditions to prevent errors,1,15 identifying and studying a specific situation in which Safety II occurred may teach us novel ways to cope with future episodes of the identical condition. However, greater Safety II applicability will result from understanding the individual and contextual factors that allow resiliency to manifest itself under these conditions. To our knowledge, systematic identification of Safety II drivers or descriptions of real improvement in a health care microenvironment has not occurred. In addition, authors of a recent review found that the research quality in this domain was low and lacked detail.16 We used a qualitative analysis approach in our study to rigorously identify human and system characteristics that enable Safety II application in a complex operational health care environment.
The staff in our PICU microsystem care for children with the highest severity of illness and administers the greatest number of high-risk medication doses hospital-wide, yet have the lowest ADE rate, approaching 10−5 reliability.7 Their low failure rate and high success rate regarding medication safety led us to hypothesize that this microsystem would be a rich environment to use a qualitative research approach to identify factors present in their microenvironment that would provide evidence of reliance on Safety I and II principles.
We conducted three 1-hour multidisciplinary focus group sessions with PICU members who participated voluntarily. Each session had between 2 and 6 participants, including physicians, nurses, pharmacists, nurse practitioners, and respiratory therapists. Participants were told they were being audio and video recorded. When each session concluded, participants were asked to suggest other practitioners who they expected would best be able to describe the safety environment, and we invited those suggested individuals to subsequent sessions.
We used 5 open-ended questions (Table 1) intended to elicit participants’ thoughts about safety behaviors in the PICU. One of the questions was designed to suggest that we were open to hearing about how staff handled unexpected conditions; however, we did not lead the participants into describing any particular safety methodologies. Participants were not educated about the differences between Safety I and Safety II. A transcriptionist generated text from each session’s audio, and these were analyzed and coded by an independent team supervised by another author who has substantive qualitative research experience. Conventional content analysis using a constant comparative approach was used,17–19 and general principles to assure quality in qualitative research20 were incorporated. Three coding team members read transcripts first to obtain a holistic view of each session, then to identify and group thought units, generate and modify potential codes (a word or short phrase that assigns a summative, salient, essence-capturing, and/or evocative attribute for a portion of language-based data), and create notes for discussion. A summary report was generated outlining domains, codes, and quotes.
This work was deemed quality improvement by our institutional review board and is exempt from further review.
Domains, Themes, and Supporting Quotations
In the qualitative analysis, we identified 19 themes describing the PICU’s safety culture, which are clustered into 4 domains (Table 2). Although themes in Domains I to III (individual characteristics, relationships and interactions, and structural and environmental factors) describe attributes usually associated with Safety I, we discovered that participants had also applied these same themes in Safety II situations, as described below. Domain IV (innovation approaches) specifically describes attributes that increase individual or system flexibility, adaptability, and resilient behaviors (consistent with Safety II principles). Quotations outlined in the Results section are referred to by number as listed in Table 2.
Domain I: Individual Characteristics
Six of the 19 themes that participants associate with increased safety relate to characteristics possessed by individuals.
1. Staying Calm and Maintaining Focus
Participants said errors were less likely with staff who can handle chaos (quote 1) as well as those who maintain focus even when situations are mundane (quote 2). People have their own systems for staying organized (quote 3).
2. Taking a Global Perspective
Individuals who make fewer mistakes think about a situation from the perspective of others, creating a situational awareness (an ability to be more predictive and proactive about what is about to happen with a patient). They “think ahead” to increase system capacity to handle possible future contingencies (quote 4).
3. Experience and Expertise
Participants believe more experienced nurses who do not float out of the unit make fewer errors. Senior nurses model safe behavior for newer nurses. Clinical leaders’ participation in quality improvement creates a culture of awareness regarding potential errors and how to address them. Experience was also mentioned as contributing to flexibility and adaptability when standard procedures were not working. Experience provides a safety net when seasoned nurses double-check or question physician and/or advanced practice nurse orders or point out proper procedures for less experienced nurses. Interviewees recognized that individuals have different and often complementary forms of expertise, and those with less expertise in a particular task seek out those with more experience (quote 5).
4. Attention to Detail
This was seen as an important contributor: ensuring that individuals reliably work through the expected standard approach (Safety I) before flexing to other approaches (Safety II). Meticulous individuals frequently have individualized approaches allowing them to track their tasks and timeliness (quotes 6–9).
5. Taking Control
Interviewees described an ability to assume responsibility and act with authority as a highly valuable trait in the PICU (quotes 10, 11).
6. Appreciating the Consequences of Mistakes
Individuals maintain a healthy fear of mistakes and are not overconfident (quotes 12, 13). Recognizing the inevitability of error is viewed as an important check on overconfidence (quote 14).
Domain II: Relationships and Interactions
7. Personal Relationships
Knowledge of co-workers’ personal lives facilitates safety by improving the quality of communication, reducing the risk of misunderstandings, and allowing for better interpretation of the meaning behind things that are said or not said (quote 15).
Emphasizing collaboration enhances patient care. It is easy to ask for help (quote 16), and participants noted that they often preidentify who they want to help them during a shift (quote 17). The ready availability of and commitment to provide help was mentioned consistently (quote 18).
9. Culture of Questioning
Questions are valued and expected across disciplines and experience levels, working their way up a chain of command if necessary. Anyone can ask for a “hard stop” to question impending actions and devise alternate plans. Individuals feel responsible to accept a challenge and respond when they are questioned (quote 19).
Open and clear communication emerged as a driver of both Safety I (policy compliance) and Safety II (adaptation of everyday work). The team reliably uses written communication for orders, handoffs, and changes in procedures or policies (quote 20). Staff deliver messages using multiple modalities and a respectful tone to increase the likelihood that messages are received as intended.
11. Training to Introduce Cultural Values
New team members learn the unit’s cultural values during training with an emphasis on attention to detail and having a questioning attitude. Leaders from multiple disciplines lead training for all types of individuals, causing them to expect cross-discipline communication, respect, and questioning (quote 21).
12. Careful Examination and Feedback After Errors Are Made
After an error occurs, a feedback loop ensures maximal learning and correction of individual or system issues (quote 22), emphasizing an expectation of maximal application of Safety I concepts before using adaptive and resilient methods.
Domain III: Structural and Environmental Factors
13. Familiarity and Proximity
The physical space contributes to building personal relationships. This camaraderie defuses tense situations and encourages debate and exploration (quote 23).
14. Number, Acuity, and Intensity of Patients
Interviewees noted that the unit tended to function better when patient number or acuity was higher. On the surface this is counterintuitive; however, the staff attributed this to having a heightened sense of situational awareness during these times. Importantly, they noted exceptions when extremely high census and acuity persist for extended periods of time, leading to exhaustion and overreliance on shortcuts (quote 24).
15. Shift Resource Availability
The only topic that emerged describing a hindrance to safety on the unit was the less experienced staffing and leadership that occurs during night shifts (quote 25).
Domain IV: Innovation Approaches
In this final domain, how individuals or the microsystem approached innovation, flexibility, and adaptability (resilience) was described. Although it was described how the themes in the other domains were used for both Safety I and Safety II, themes in this domain were only applicable for Safety II - related concepts.
16. Relying on Teamwork if Something Novel Is Considered
When innovative approaches are needed, individuals are reluctant to act independently. They gather people from multiple disciplines in real time to troubleshoot and problem solve. They double-check facts against each other. They are honest and open about explaining why they wish to deviate from protocol or try novel actions. They ask for constructive criticism, ensuring that they are taking the best possible novel approach (quote 26).
17. Teams Responding to Challenging Circumstances
The PICU team triages and prioritizes work. Everyone is expected to “go the extra mile” and work together. A clear sense of “being in this together” is a predominate mind-set in the PICU culture (quote 27).
Some participants articulated that error reduction resulted from a commitment to standard procedures and they would become skeptical if they saw a shortcut being suggested or pursued (quotes 28–30). Even when there is high census and acuity, individuals are skeptical of creating shortcuts or variations that stray from standard or expected practices or processes.
19. Bringing Atypical Approaches From Other Microenvironments
Although a cohesive team was described as being critical for high function in previous themes, there are occasions in which team members come from other microsystems. This was seen as vital for providing enough variety and “outside the box” thinking to the unit (quote 31).
To our knowledge, this is the first article in the health care arena to outline concrete behaviors and activities present in a complex microsystem that may be contributing to remarkably low error rates and that likely exemplify the use of Safety II principles at work.
Themes within the domains of individual characteristics, relationships and interactions, and structural and environmental factors reflect Safety I and Safety II behaviors in the PICU. For example, participants described how personal relationships with co-workers reduces psychological barriers to voicing concerns when an established pathway or protocol is not being followed (Safety I). However, these relationships also allow better interplay for creative thought under unanticipated circumstances (Safety II). In our model (Fig 1), we suggest that the themes outlined within domains I through III are used effectively in this unit to prevent error (Safety I) and to improve success (Safety II). We conclude that a strong Safety I culture is an essential foundation to support a Safety II culture.
In contrast, themes in domain IV, innovative approaches, appear to describe characteristics unique to Safety II implementation. Although the dominant message from our institution’s quality and safety improvement efforts has been to reduce variation, PICU team members spontaneously described techniques they use when they feel that variation is necessary. They have internalized that safety and quality derive from reducing unintended variation, but intended variation on occasion is apparently a hallmark of this Safety II culture (Table 1: quotes 26, 27, and 31). Two of 4 themes in this domain specifically address teamwork in which team members support each other in responding to unanticipated circumstances and act as sounding boards to improve anticipation and response. One theme, bringing atypical approaches from other microenvironments, suggests that having team members who have worked in other microsystems or who can reach outside the microsystem for new ideas reduces narrow thinking or groupthink. The necessity for a tight yet nonrigid team suggests that there may be an optimal turnover level: too high and teamwork fails, too low and novel ideas cannot be elicited. Finally, the skepticism theme suggests that the microsystem understands that guardrails should be in place to allow variation without entering into chaos. In Safety II, deviations from prescribed protocols are intentional and required for safest care, not unintentional disregard of the typical process.
Researchers of previous qualitative research studies have examined other aspects of high-performing microsystems but not Safety II behavior specifically. For example, Santana et al21 defined 5 characteristics of high-performing microsystems including constancy of purpose, investment in improvement, and interdependent multidisciplinary teams among others.21 However, not all identified characteristics were present in the microsystem they examined. The researchers for another study examined leadership issues in the critical care setting and concluded that nurse leaders who provide direct patient care are crucial to success.22 Because Safety I versus Safety II was not the foundational issue underlying the studies, comparison and contrast is not feasible. Nonetheless, it is possible that some Safety II behavior may have been at work in both settings.
Safety II involves applying resilience engineering to health care systems.15 Resilient systems require humans to learn from what goes right and develop adaptations and flexibility to incorporate that learning going forward. Resilient systems and individuals’ resilient behavior are key elements of high-reliability systems as characterized by Weick and Sutcliffe.23 Our PICU staff have described the relationship of stress, adaptation, and safety in their unit by noting that they are safest over a range of stress levels. However, suboptimal performance exists at both extremes (very low and very high stress [Table 1: quotes 24, 25]), supporting the theories of Yerkes and Dodson.24
The findings of a study performed in emergency departments in Brazil and in the United States revealed resilient behavior similar to the observations and themes we identified in our PICU.25 These authors identified several skill categories, including but not limited to collaborative work, communication, recognizing the impact of actions, and context awareness. They suggest education and training frontline staff through simulation, role playing, and other teaching strategies, while focusing on “imperfect” situations which can enhance resilience capacity. Another example of resilience in health care described an implementation of a wireless call system intended to reduce nursing workflow interruptions that led instead to increased interruptions. The difference between “work as imagined” and “work as done” forced nurses to be resilient and change their work as done with the new system to create the desired effect.26
Our study has some potential weaknesses. Participants in the focus groups were not entirely randomly selected. Some individuals included in later focus groups were suggested by individuals in preceding focus groups. To the extent their views coincided with those of their nominators, the qualitative data could be skewed in that direction. A data review did not reveal that was the case. Secondly, our focus groups were small (2–6 participants). This largely reflected the challenges of scheduling in this group of busy people. Correlation with high scores from this unit in our 2017 culture survey (Safety Attitudes Questionnaire)27 on questions regarding staff’s ability to “speak up” and “discuss error” suggested that we had a representative sample of the unit. Finally, our data were collected in 1 unit in 1 hospital and may not extrapolate to other units in our hospital or to other hospitals. We intend to test these data in other high-intensity units (eg, hematology, oncology, or emergency department) in our hospital and in simulation settings before spreading these learnings to the rest of our hospital. We will also test the data in lower performing units to see if the themes and domains were qualitatively different or simply less reliably practiced.
To spread Safety II through our hospital, we will not introduce Safety II into other microsystems and processes until they are performing at ∼10−4 error rates. This is because of a concern that chaos could result if Safety II is not appropriately limited. Spread will require development of additional Safety II “tools” that can be used in addition to the Zero Hero high-reliability tools we have used to encourage eliminating practice variation. We intend to test and then spread such tools to additional microsystems at an appropriate time for each. Our current plan, based on learnings to date, will include the following: (1) creating an environment in which it is acceptable to learn how to “work as done” versus “work as imagined,” (2) capturing resilient behavior from debriefings,28 (3) determining how frontline staff predict and anticipate problems,29 (4) simulating imperfect situations both in situ and off-unit, and (5) providing frontline staff with specific tools that will guide safe improvisation in unanticipated situations.
We identified 19 themes in which individual and group behaviors were described in a unit with low error rates. The low error rates exemplify Safety II activity contributing to ADE prevention. Fifteen of these themes are also characteristic of a strong Safety I culture, which appears necessary, but not sufficient, to attain Safety II status. Four themes in the innovation approaches domain seemed specific to Safety II performance, including resiliency. Similar studies will be conducted in some of our other high-performing units to explore similarities and differences. We believe these behaviors can be spread across our hospital and will contribute to further harm reduction.
- Accepted March 27, 2018.
- Address correspondence to J. Terrance Davis, MD, Nationwide Children’s Hospital, 700 Children’s Dr, Columbus, OH 43205. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funded internally through Nationwide Children’s Hospital.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Deutsch ES; Pennsylvania Patient Safety Authority; ECRI (Organization); Institute for Safe Medication Practices
- Institute of Medicine Committee on Quality of Health Care in America
- Hollnagel E,
- Wears RL,
- Braithwaite J
- Hollnagel E,
- Braithwaite J,
- Wears RL
- Baxter R,
- Taylor N,
- Kellar I,
- Lawton R
- Krueger RA,
- Casey MA
- Mays N,
- Pope C
- Santana C,
- Nunez-Smith M,
- Camp A,
- Ruppe E,
- Berg D,
- Curry L
- Weick KE,
- Suttcliffe KM
- Pascal Metrics
- Jeffcott SA,
- Ibrahim JE,
- Cameron PA
- Cabitza F,
- Simone C
- Copyright © 2018 by the American Academy of Pediatrics