Culture: Diagnosing and Changing How Safety Is Practiced
The Culture domain tests whether you can define safety culture, distinguish it from safety climate, diagnose it with valid instruments, and connect findings to behavior change. Items typically probe just culture, psychological safety, reporting and learning, and how leaders interpret survey data.
Core rules and concepts
- Safety culture is the shared values, beliefs, attitudes, and behaviors that determine an organization's commitment to safety; safety climate is the measurable snapshot of staff perceptions at one point in time. Climate is what you survey; culture is what you change.
- Reason's four subcultures - reporting, just, flexible, learning - produce safety only when all four are present. A reporting culture without a learning culture generates data nobody acts on.
- Just culture separates behavior from outcome: human error is consoled, at-risk behavior is coached and its incentives removed, reckless behavior is sanctioned. A severe outcome does not automatically mean blameworthy conduct.
- Psychological safety (Edmondson) - the shared belief that speaking up will not bring punishment or humiliation - is the precondition for reporting and for stop-the-line behavior.
- AHRQ SOPS composites include teamwork within units, supervisor expectations, organizational learning, management support, overall perceptions of safety, frequency of events reported, communication openness, feedback about error, nonpunitive response to error, staffing, handoffs and transitions, and teamwork across units.
- Culture is a leading indicator; harm rates are lagging indicators.
Worked example
A hospital's safety culture survey shows event reporting at the 80th percentile but nonpunitive response to error at the 10th percentile, with weak staffing scores in two units. Celebrating great reporting is the wrong move. High reporting paired with low nonpunitive scores usually means staff report because they are required to, not because they trust the system. The correct response is leader-led: publish and apply a just culture algorithm, close the loop by telling staff what changed because of their reports, and address staffing on the two units. Re-measure in 12–18 months, because culture scores move slowly and small fluctuations are noise.
Traps
- Treating a rising event count as proof of a worsening culture - it may be proof of improving trust.
- Confusing just culture with blame-free culture; accountability does not disappear.
- Presenting only the hospital-wide survey average and missing unit-level and role-level subcultures.
- Assuming a written nonpunitive policy equals a nonpunitive lived experience.