Chapter 6. Comparing Your Results
Hospital Survey on Patient Safety Culture: 2010 User Comparative Database
To compare your hospital's survey results with the results from the database hospitals, you will need to calculate your hospital's percent positive response on the survey's 42 items and 12 composites (plus the two questions on patient safety grade and number of events reported). Refer to the Notes section at the end of this report for a description of how to calculate these percent positive scores. You will then be able to compare your hospital's results with the database averages and examine the percentile scores to place your hospital's results relative to the distribution of database hospitals.
When comparing your hospital's results with results from the database, keep in mind that the database only provides relative comparisons. Even though your hospital's survey results may be better than the database statistics, you may still believe there is room for improvement in a particular area within your hospital in an absolute sense. As shown in the database results, there are some patient safety composites that even the highest scoring hospitals could improve on. Therefore, the comparative data provided in this report should be used to supplement your hospital's own efforts to identify areas of strength and areas on which to focus efforts to improve patient safety culture.
Description of Comparative Statistics
In addition to the average percent positive scores presented in the charts in Chapter 5, a number of statistics are provided in this report to facilitate comparisons with the database hospitals. A description of each statistic shown in the comparative results tables in this chapter is provided next.
Average Percent Positive
The average percent positive scores for each of the 12 patient safety culture composites and for the survey's 42 items (plus the two questions on patient safety grade and number of events reported) are provided in the comparative results tables in this chapter. These average percent positive scores were calculated by averaging composite-level percent positive scores across all hospitals in the database, as well as averaging item-level percent positive scores across hospitals. Since the percent positive is displayed as an overall average, scores from each hospital are weighted equally in their contribution to the calculation of the average.3
The standard deviation (s.d.), a measure of the spread or variability of hospital scores around the average, is also displayed. The standard deviation tells you the extent to which hospitals' scores differ from the average:
- If scores from all hospitals were exactly the same, the average would represent all their scores perfectly and the standard deviation would be zero.
- If scores from all hospitals were very close to the average, the standard deviation would be small and close to 0.
- If scores from many hospitals were very different from the average, the standard deviation would be a large number.
When the distribution of hospital scores follows a normal, bell-shaped curve (where most of the scores fall in the middle of the distribution, with fewer scores at the lower and higher ends of the distribution), the average, plus or minus the standard deviation, will include about 68 percent of all hospital scores. For example, if an average percent positive score across the database hospitals were 70 percent with a standard deviation of 10 percent and scores were normally distributed, about 68 percent of all the database hospitals would have scores between 60 and 80 percent.
Statistically "significant" differences between scores. You may be interested in determining the statistical significance of differences between your scores and the averages in the database, or between scores in various breakout categories (hospital bed size, teaching status, etc). Statistical significance is greatly influenced by samples size, so as the number of observations in comparison groups gets larger, small differences in scores will be statistically significant. While a 1 percent difference between percent positive scores might be "statistically" significant (that is, not due to chance), the difference is not likely to be meaningful or "practically" significant. Keep in mind that statistically significant differences are not always important, and nonsignificant differences are not always trivial. Therefore, we recommend the following guideline:
- Use a 5 percentage point difference as a rule of thumb when comparing your hospital's results to the database averages. Your hospital's percent positive score should be at least 5 percentage points greater than the database average to be considered "better" and should be at least 5 percentage points less to be considered "worse" than the database average. A 5 percentage point difference is likely to be statistically significant for most hospitals given the number of responses per hospital and is also a meaningful difference to consider.
Minimum and Maximum Scores
The minimum and maximum percent positive scores are presented for each composite and item. These scores provide information about the range of percent positive scores obtained by hospitals in the database and are actual scores from the lowest and highest scoring hospitals. When comparing with the minimum and maximum scores, keep in mind that these scores may represent hospitals that are extreme outliers (indicated by large differences between the minimum and the 10th percentile score, or between the 90th percentile score and the maximum).
The 10th, 25th, 50th (or median), 75th, and 90th percentile scores are displayed for the survey composites and items. Percentiles provide information about the distribution of hospital scores. To calculate percentile scores, all hospital percent positive scores were ranked in order from low to high. A specific percentile score shows the percentage of hospitals that scored at or below a particular score. For example, the 50th percentile, or median, is the percent positive score where 50 percent of the hospitals scored the same or lower and 50 percent of the hospitals scored higher. When the distribution of hospital scores follows a normal bell-shaped curve, the 50th percentile, or median, will be very similar to the average score. Interpret the percentile scores as shown in Table 6-1.
To compare with the database percentiles, compare your hospital's percent positive scores with the percentile scores for each composite and item. Look for the highest percentile where your hospital's score is higher than that percentile. For example: On survey item 1 in Table 6-2, the 75th percentile score is 49 percent positive, and the 90th percentile score is 62 percent positive.
- If your hospital's score is 55 percent positive, it falls above the 75th percentile (but below the 90th), meaning that your hospital scored higher than at least 75 percent of the hospitals in the database.
- If your hospital's score is 65 percent positive, it falls above the 90th percentile, meaning your hospital scored higher than at least 90 percent of the hospitals in the database.
Composite and Item-Level Comparative Tables
Table 6-3 presents comparative statistics (average percent positive and standard deviation, minimum and maximum scores, and percentiles) for each of the 12 patient safety culture composites. The patient safety culture composites are shown in order from the highest average percent positive response to the lowest.
Table 6-4 presents comparative statistics for each of the 42 survey items. The survey items are grouped by the patient safety culture composite they are intended to measure. Within each composite, the items are presented in the order in which they appear in the survey.
The comparative results in Tables 6-3 and 6-4 show considerable variability in the range of hospital scores (lowest to highest) across the 12 patient safety culture composites. The standard deviation around the average percent positive scores ranged from 5.86 percent to 11.66 percent on the composites and ranged from 5.58 percent to 13.39 percent on the items.
Patient safety grades shown in Table 6-5 had a wide range of response, from at least one hospital where none of the respondents (0 percent) provided their unit with a patient safety grade of "A-Excellent," to a hospital where 65 percent did.
Number of events reported also had a wide range of response, as shown in Table 6-6, from a hospital where 82 percent of respondents had not reported a single event over the past 12 months, to a hospital where only 14 percent had not reported an event.
Appendixes A and B: Overall Results by Hospital and Respondent Characteristics
In addition to the overall results on the database hospitals presented, Part II of the report presents data tables showing average percent positive scores on the survey composites and items across database hospitals, broken down by the following hospital and respondent characteristics:
Appendix A: Results by Hospital Characteristics
- Bed size.
- Teaching status.
- Ownership and control.
- Geographic region.
Appendix B: Results by Respondent Characteristics
- Work area/unit.
- Staff position.
- Interaction with patients.
The breakout tables are included as appendixes because there are a large number of them. Highlights of the findings from the breakout tables in these appendixes are provided on the following pages. The appendixes are available on the Web at: http://www.ahrq.gov/professionals/quality-patient-safety/patientsafetyculture/hospital/2010/index.html/..
Note: New to the 2010 database, breakouts by respondent characteristics (go to Appendix B) were only calculated for hospitals that had at least five respondents in the breakout category. If a hospital had fewer than five respondents in a certain category, the hospital is not included in the statistics displayed for that category. (Further explanation is in Notes: Description of Data Cleaning and Calculations.)
Highlights From Appendix A: Overall Results by Hospital Characteristics
Bed Size (Tables A-1, A-3, A-4)
- Smaller hospitals (49 beds or fewer) had the highest average percent positive response on all 12 patient safety culture composites.
- Large hospitals (400-499 beds) scored lowest on the percentage of respondents who gave their work area/unit a patient safety grade of "Excellent" or "Very good" (70 percent positive for 400-499 beds compared with 79 percent positive for 25-49 beds).
- There were no noticeable differences in number of events reported based on bed size (all differences were 2 percentage points or less).
Teaching Status and Ownership and Control (Tables A-5, A-7, A-8)
- Non-teaching hospitals had a higher average percent positive response on Handoffs and Transitions than teaching hospitals (46 percent positive compared with 41 percent positive).
- There were no noticeable differences in the patient safety culture composites based on ownership and control (all differences were 3 percentage points or less).
- There were no noticeable differences in patient safety grade or number of events reported based on teaching status or ownership and control (all differences were 3 percentage points or less).
Geographic Region (Tables A-9, A-11, A-12)
- East South Central hospitals had the highest average percent positive response across the composites (66 percent positive); Mid-Atlantic/New England hospitals had the lowest (60 percent positive).
- West South Central hospitals scored highest on the percentage of respondents who gave their work area/unit a patient safety grade of "Excellent" or "Very Good" (78 percent).
- Pacific hospitals had the highest percentage of respondents who reported one or more events in the past year (53 percent); the lowest percentage of respondents reporting events was in the West South Central region (41 percent).
Highlights From Appendix B: Overall Results by Respondent Characteristics
Work Area/Unit (Tables B-1, B-3, B-4)
- Respondents in, Rehabilitation had the highest average percent positive response across the composites (68 percent positive); Emergency had the lowest (57 percent positive).
- Rehabilitation had the highest percentage of respondents who gave their work area/unit a patient safety grade of "Excellent" or "Very Good" (84 percent); Emergency had the lowest percentage (62 percent).
- ICU (any type) had the highest percentage of respondents reporting one or more events in the past year (65 percent); Anesthesiology had the lowest percentage of respondents reporting events (40 percent).
Staff Position (Tables B-5, B-7, B-8)
- Respondents in Administration/Management had the highest average percent positive response across the composites (73 percent positive); Pharmacists had the lowest (58 percent positive).
- Administration/Management had the highest percentage of respondents who gave their work area/unit a patient safety grade of "Excellent" or "Very Good" (85 percent); Pharmacists had the lowest percentage (65 percent).
- Pharmacists had the highest percentage of respondents reporting one or more events in the past year (72 percent); Unit Assistants/Clerks/Secretaries and Dietitians had the lowest percentage reporting events (19 percent).
Interaction With Patients (Tables B-9, B-11, B-12)
- Respondents with direct patient interaction were 8 percent more positive on Handoffs and Transitions compared with those without direct patient interaction (46 percent positive compared with 38 percent positive).
- Respondents without direct patient interaction were 6 percent more positive about Management Support for Patient Safety than those with direct patient interaction (77 percent positive compared with 71 percent positive).
- Respondents without direct patient interaction had the highest percentage of respondents who gave their work area/unit a patient safety grade of "Excellent" or "Very Good" (79 percent) compared with those with direct patient interaction (74 percent).
- More respondents with direct patient interaction reported one or more events in the past year (52 percent) than respondents without direct patient interaction (31 percent).
3 As described in the Notes section, an alternative method would be to report a straight percentage of positive response across all respondents, but this method would give greater weight to respondents from larger hospitals since they account for almost twice as many responses as those from smaller hospitals.