1.4 Assessing Safety Culture and Staff Awareness

Assessing Safety Culture and Staff Awareness

CPHRM Domain 1 (Clinical/Patient Safety) carries 25 of the 100 scored items, and its first official task statement is to assess the current state of patient safety and staff awareness within the organization. The operative verb is assess, not assert. Items in this area punish the risk manager who characterizes culture from personal impression, executive rounding anecdotes, or a quiet quarter in the claims file, and reward the one who produces a validated, benchmarked, unit-level measurement that leadership can act on.

Safety Culture Versus Safety Climate

• Patient safety culture is the deep, durable layer. The Joint Commission's Patient Safety Systems (PS) chapter defines it as the product of individual and group beliefs, values, attitudes, perceptions, competencies, and patterns of behavior that determine an organization's commitment to quality and patient safety. It shows up in what people actually do when safety competes with speed, cost, or hierarchy.
• Safety climate is the shallow, measurable layer: staff perceptions of that culture at a single point in time.

A survey measures climate. Climate is the best available proxy for culture, but it is a snapshot of perception, not proof that culture changed. Two administrations separated by time, analyzed at the unit level, are what demonstrate movement. (The Just Culture behavioral taxonomy is treated separately in this guide; it governs how an organization responds to an individual act, while culture assessment measures what the workforce collectively believes.)

The AHRQ SOPS Instruments

The Agency for Healthcare Research and Quality (AHRQ) Surveys on Patient Safety Culture (SOPS) are the default United States measurement tools. AHRQ released the original Hospital Survey on Patient Safety Culture in 2004 and the SOPS Hospital Survey 2.0 in 2019; companion instruments exist for medical offices, nursing homes, ambulatory surgery centers, community pharmacies, and home health. Version 1.0 remains available, but AHRQ encourages 2.0.

The SOPS Hospital Survey 2.0 has 40 total items and takes about 10 to 15 minutes:

• 32 items grouped into 10 composite measures of safety culture.
• 8 single-item measures: one asking how many patient safety events the respondent personally reported, one asking for an overall patient safety rating of the respondent's unit or work area, and six background items (staff position, unit/work area, hospital tenure, unit tenure, work hours, and interaction with patients).

Results are expressed as percent positive — the share of respondents choosing "Strongly agree/Agree" or "Most of the time/Always" on positively worded items (reverse-scored for negatively worded items). There is no AHRQ-published passing score for safety culture; percent positive is interpreted against your own baseline and against the SOPS Database, never against an invented threshold.

Composite Measures and What a Low Score Means Operationally

Version 1.0 to 2.0 Naming (a common exam confuser)

Version 1.0 had 12 dimensions; 2.0 has 10 composites, and several names changed. Nonpunitive Response to Error became Response to Error. Frequency of Events Reported became Reporting Patient Safety Events. Handoffs and Transitions became Handoffs and Information Exchange. Staffing became Staffing and Work Pace. Teamwork Within Units and Teamwork Across Units were consolidated into a single Teamwork composite, and Overall Perceptions of Patient Safety is no longer a composite — an overall patient safety rating survives as a single item. Because the item sets differ, 1.0 and 2.0 scores are not directly comparable; do not trend across the version change as if it were one line.

Administering the Survey So the Results Are Usable

Bad administration produces a number no one can defend. AHRQ's User's Guide sets the mechanics:

• Census or sample. Hospitals with roughly 500 or fewer providers and staff should survey everyone. Larger organizations may sample, using AHRQ's minimum sample-size table, which is built on the assumption of about a 50 percent response.
• Expect 30 to 50 percent. AHRQ tells users to expect completed surveys from roughly 30 to 50 percent of the sample. If the rate is still below 50 percent near the planned close-out, the guidance is to extend data collection and send additional reminders rather than close the survey and report a thin result.
• Low response equals nonresponse bias. With a low rate, the large group that did not respond may hold very different views, and you cannot generalize to the workforce. This is the single most common technical flaw a risk manager must flag before a score reaches the board.
• Confidential versus anonymous. A confidential survey uses identifiers so reminders can be targeted at nonrespondents; an anonymous survey cannot track anyone. Either design requires that only aggregated results are reported, and that no individual response is ever traced back to a person.
• Unit-level minimums. Unit results should only be reported when enough people in that unit responded to prevent identification of individuals; small units are suppressed or rolled up into a larger reporting group. Announce this rule before administration — staff decide whether to answer honestly based on whether they believe it.
• Cadence. The Joint Commission's Sentinel Event Alert 57 (March 2017) recommends repeating an organizational assessment of safety culture every 18 to 24 months, drilling the assessment down to unit levels, and making it part of the strategic measures reported to the board. Separately, the CMS Patient Safety Structural Measure (PSSM) requires an attesting hospital to conduct a hospital-wide culture of safety survey annually, or every two years provided a shorter pulse survey is administered to target units in the off year. Align the calendar to whichever obligation applies to your organization.

Benchmarking Through the SOPS Database

The AHRQ SOPS Hospital Database is a free, voluntary central repository for United States hospitals that administered the survey without modification. Each submitting site receives a customized feedback report comparing its own results against database results, and AHRQ publishes aggregated database reports and infographics. The 2022 SOPS Hospital 2.0 Database Report drew on 400 hospitals and 206,410 provider and staff respondents; the 2024 report included 445 hospitals.

Benchmarking is what converts a bare number into a decision. "62 percent positive on Response to Error" means nothing to a board. "62 percent positive on Response to Error, against a database average, with our surgical service at 34 percent" is a resource-allocation argument.

Related Requirements and Alternative Tools

LD.03.01.01 in The Joint Commission's Leadership chapter states that leaders create and maintain a culture of safety and quality throughout the organization. Its elements of performance require leaders to regularly evaluate the culture of safety and quality using valid and reliable tools (EP 1), to develop a code of conduct defining acceptable behavior and behaviors that undermine a culture of safety (EP 4), and to create and implement a process for managing behaviors that undermine a culture of safety (EP 5). Note who is accountable: leaders. The risk manager recommends the instrument, coordinates administration, analyzes and escalates the findings, and documents the process — the risk manager does not own the culture standard and cannot compel a clinical department to change.

Validated alternatives include the Safety Attitudes Questionnaire (SAQ), which measures six domains — teamwork climate, safety climate, perceptions of management, job satisfaction, working conditions, and stress recognition — and The Joint Commission Center for Transforming Healthcare's Oro 2.0 high reliability organizational assessment, which is a leadership self-assessment used alongside, not instead of, a frontline perception survey.

Unit-Level Analysis: Why the Aggregate Lies

Safety culture is local. It is produced far more by the unit manager, the medical director, and the team on the floor than by the executive suite, and it varies enormously between units in the same building. An organization reporting 68 percent positive overall can contain an intensive care unit at 30 percent on Response to Error, and the aggregate will never show it.

Sentinel Event Alert 57 is explicit on this point: analysis of safety culture survey results must drill down to local unit levels so that unit-specific solutions can be developed and implemented, and the results should be shared both with frontline staff and with governing bodies, including the board.

Practical rules for analysis:

1. Report the distribution, not the mean. Show the spread of unit scores and flag the bottom decile by composite. Pair every flagged unit with a named leader.
2. Segment by role and shift. The background items capture staff position, tenure, work hours, and patient contact. Night shift, float and travel staff, and physicians frequently diverge sharply from the house average on handoffs, staffing, and communication openness.
3. Compare each unit to the database and to itself over time. A unit at 55 percent that was at 40 percent is a success story; a unit at 70 percent that was at 85 percent is an emerging problem.
4. Read the open-ended comments after deidentification. They tell you why a composite is low; the number alone never does.

Proxy Indicators and the Inverse Relationship

Risk managers routinely watch event-report volume as a culture indicator. The direction of that indicator is counterintuitive and heavily tested. A rising event-report rate — especially rising near-miss, good-catch, and unsafe-condition reports — usually signals an improving reporting culture, not deteriorating safety. A unit filing almost no reports and showing no harm is far more likely to be silent than safe.

The correct method is to read report volume against a harm denominator:

Other useful proxies: the ratio of near-miss to harm-event reports, the share of reports submitted by physicians (chronically low is a red flag), time from report to feedback, the anonymous-versus-identified reporting mix, staff turnover and exit-interview themes, and grievance and complaint themes. None of these replaces a validated survey; they triangulate it between administrations.

Scenario: A risk manager at a 400-bed community hospital receives SOPS Hospital Survey 2.0 results: 74 percent positive overall, Response to Error at 41 percent, and one surgical unit at 29 percent on Communication Openness against a 68 percent house average. That same surgical unit filed four event reports in the past year, the fewest in the hospital, and two unplanned returns to the operating room were flagged through peer review. The defensible sequence is to present the unit-level breakout — not just the aggregate — to the patient safety committee and the board quality committee; to interpret the near-zero report volume as suppressed reporting rather than superior performance; to build a unit-based action plan with the unit's medical director and nurse manager rather than for them; and to tell that unit's staff what changed because of their answers before the next survey goes out. What the risk manager does not do is identify who wrote a critical comment, counsel individual staff on their responses, or announce a hospital-wide re-education campaign in place of a unit-specific fix.

Closing the Loop, or Destroying Trust

The most consequential failure in this task area is surveying and then doing nothing visible. Staff who see no action conclude the survey is theater; the next round's response rate falls, its scores fall, and the organization has actively damaged the trust it was trying to measure. AHRQ pairs the surveys with an Action Planning Tool for exactly this reason, and The Joint Commission's Center for Transforming Healthcare found that team members simply stopped raising safety issues when they received no feedback from leaders.

Minimum loop closure: share unit-level results within a defined window, select a small number of prioritized composites rather than all ten, name owners and dates, publicize concretely what changed, and re-measure on the 18-to-24-month cycle.

Exam Traps

• Declaring the culture strong because claims or lawsuits are down. Claims volume is a lagging, heavily filtered financial indicator with a multi-year tail, not a measure of staff perception.
• Comparing your raw percent positive to a number a peer hospital mentioned. Only the SOPS Database comparison is apples-to-apples.
• Reporting only the enterprise aggregate to the board. Sentinel Event Alert 57 expects unit-level drill-down in board-reported strategic measures.
• Treating a rise in event reports as evidence that safety is worsening.
• Modifying, shortening, or reordering the instrument to boost completion.
• Using survey data punitively or attempting to identify a respondent or comment author.
• Surveying more often than the organization can act, producing fatigue with no visible change.
• Assuming the risk manager owns culture. LD.03.01.01 assigns that duty to leaders; the exam-correct risk manager recommends a valid tool, coordinates, analyzes, documents, escalates, and reports.