Integrating Safety-I and Safety-II conceptual frameworks to enhance safety measurement and management

If one were to ask a safety professional “How do we know a hospital is safe?,” it is very likely that the question would lead to debate. Many would likely agree that despite decades of effort, safety measurement is not as robust as it needs to be. However, we continue to rely on ill-defined metrics and narrow methodologies to evaluate safety. This viewpoint aims to explore the question “How can Safety-I and Safety-II conceptual frameworks be integrated to inform more robust approaches to safety measurement and management?”

Other than agreement about the goal in healthcare of “freedom from preventable harm,” there has been historical disagreement around what constitutes safety. This tension may be due to the complexity of healthcare delivery; the small “n” of documented safety events; and the varied perceptions of the clinical, financial, and emotional effects of preventable harm. ¹ Health systems have attempted to measure and manage safety events and have evolved to detect some instantiations of failure, focus on the most severe harmful events, and increase compliance with policies and procedures aimed to reduce them. Known as the “Safety-I” approach, this traditional measure-your-defects approach led to improvements in harm detection and has increased the appetite to learn from these failures to prevent their recurrence. ² Indeed, it seems insensitive to prioritize resources in any other way after major healthcare harm events occur, given the physical, emotional, and financial implications of the event for patients, family members, and staff. However, even after a rigorous event analysis, it is not uncommon for clinical stakeholders and safety leaders to remain uneasy about the organization's safety. The “root causes” that are identified may feel superficial or inadequate. Studying one event such as a CAUTI, at a time under one set of circumstances appears to only scratch the surface.

But if the staff in Hospital A perceive, via safety culture surveys, that the hospital is safe, but they have a CAUTI every month, does that mean that the hospital is safer than Hospital B, where staff perceive more risks around every corner but have been CAUTI-free for years? How do we develop a measurement strategy that more holistically depicts organizational safety?

These debates underpin two major schools of thought within the safety field: Safety-I and Safety-II. Safety-I primarily views adverse events as deviations from safe operations, whereas Safety-II acknowledges the adverse event as an emergent outcome of a complex adaptive system and emphasizes the importance of understanding the tensions, adaptations, and variability in normal work to enable continued resilience. ³ Critics of Safety-I have critiqued the Safety-I approach of using narrowly defined metrics to represent safety without considering the complexities of frontline work. ⁴ Critics of Safety-II have cited its lack of pragmatism, the considerable time investment required to understand work-as-done, and the ambiguous roadmap to guide application of the theory.^5,6

We suggest that fidelity to one perspective to the extent that it denounces the insights offered by the other may be a significant contributor to our stagnation. Instead, integration of the Safety-I and Safety-II perspectives is essential to inform more robust approaches to safety measurement.

Holistic understanding of the current safety landscape

Like the proverb of the elephant and the blind men, each approach affords a different depiction of a given clinical process. Whereas a Safety-I approach may yield insights into the contributing factors, active and latent conditions, and root causes, it fails to explain why these contributing factors did not come to fruition as a harm event during the other 99.9% of times that this clinical process was performed. ⁷ A Safety-II–framed analysis may unveil tradeoffs, adaptations, constraints, and tensions faced by the frontline when performing this clinical process, but our insights will be cut short without examples of how typically adaptive maneuvers may fail.

The rule to forbid administration of a patient's own home medication in the emergency department (ED) is an example that illustrates the imperative of integrating the two schools of thought for more robust measurement. A patient with a rare condition has tried many medications in their life and has finally found one that works for them in urgent situations. However, the hospital's policy states that home medications are not to be administered if there is a formulary equivalent. When the patient arrives to the ED with their home medication, but it is unable to be used in their ED course of care, the patient experiences a life-threatening delay as the healthcare team aims to remain compliant with rules while providing the best care for the patient. A Safety-I approach may examine this event in isolation and identify factors that contributed to this specific event, such as the hospital pharmacy's lack of an adequate substitute for the patient's own home medication. A Safety-II informed approach, however, would examine the “normal work” associated with the process in question, thereby detecting tradeoffs and adaptations that normally happen without resulting in a safety event. For example, using a Safety-II approach, we may find that providers normally have to carve out time to explain alternatives to patients, that patients often reconsider seeking care in the ED at all if not able to use their home medications, or that declining to administer the patient's home medication has downstream effects once the patient is admitted. When Safety-I and Safety-II conceptual frameworks are applied together, the findings, and worthwhile metrics informed by the findings, are more likely to holistically represent system safety. In this example, the metrics that result may be informed by both perspectives and include, for example, time spent identifying alternatives, patient access to care, patient satisfaction, or availability of substitute medications. Integration of Safety-I and Safety-II conceptual frameworks, in safety investigations for example, is a key step in developing more robust metrics and ultimately better depicting the current state of safety in healthcare organizations.

Providing context for robust interpretation

In addition to more holistically depicting the safety landscape, integration of Safety-I and Safety-II conceptual frameworks can provide context to better interpret important outcome metrics. The principle of equivalence of failures safe and successes means that similar conditions can produce different outcomes in complex environments. When the only examined or reported outcome metrics are instances of failure, blame for noncompliance with policies and procedures becomes contingent on the outcome when the noncompliance was tolerated in all other normal circumstances that did not result in harm. ⁸ Interpreting outcome metrics through the lens of both Safety-I and Safety-II frameworks may better depict the complexity of the system, reveal how upstream activities influence the ability to remain compliant, and expose the system pressure–generated tradeoffs between efficiency and thoroughness. ⁹ Safety practitioners may combine Safety-I and Safety-II approaches in analysis but our experience suggests that recommendations for improvement often lead to Safety-I type interventions like re-education or policy development for example. Synthesis of these approaches in our analyses by examining when to improve compliance with policies and procedures versus when to support the frontline work using other strategies will likely inform more robust interventions.

Embedding Safety-I and Safety-II conceptual frameworks into existing safety measurement

These Safety-I and Safety-II concepts should be integrated within existing safety structures and processes, such as safety investigations. The emergence of Safety-II may have brought with it the perception that its integration can only be made possible after the traditional Safety-I approaches and tools—such as root cause analysis, hierarchical task analysis, and incident reporting—have been marginalized. The one-or-the-other perception is likely one of the drivers of the stagnation of Safety-II use in practice. Our experience with safety investigations across multiple institutions suggests that traditional Safety-I approaches are still heavily embedded in organizational safety work, and the idea of replacing them can seem daunting. However, the Safety-II principles should be embedded in existing work and may not necessarily replace many of our existing practices. For example, parallel use of a Safety-I informed tool, such as Hierarchical Task Analysis and a Safety-II informed tool, such as the Functional Resonance Analysis Method, in the same safety investigation may yield a more complete picture of the clinical context than either alone. Similarly, asking about the pressures that influence decision making related to clinical activities surrounding the event being investigated may reveal compromises or ambiguities that those on the frontline must navigate to perform optimally.

Conclusion

Measurement is helpful for understanding performance, tracking progress, and developing strategies to meet the end goal of improved and consistent safe care. “Good” safety measurement should avoid the temptation to base our assessments of safety only on a few readily available outcomes. Measuring more broadly than the Safety-I approaches to measurement, such as healthcare-acquired conditions and number of serious events, is an imperative next step. Collecting qualitative data to explain the how and why of quantitative data, and dedicating time and attention to understanding what goes well and why, are some actionable steps to move toward “better” safety measurement. However, our traditional approaches to measurement should not be forgotten in favor of new Safety-II frontiers. Instead, there is a need to examine how the two approaches complement each other for the field to advance. Just as we would critique the researcher who studies marital happiness by only studying divorced couples or only through couples celebrating their 50^th anniversary, we would benefit by infusing both Safety-I and Safety-II schools of thought when attempting to answer the question: “How do we know if a hospital is safe?”