Patient Safety Culture and Hospital Performance

DECISION SCIENCES INSTITUTE An Empirical Analysis of Using the AHRQ Survey on Patient Safety Culture (Full Paper Submission) Gregory N. Stock, Ph.D. College of Business University of Colorado Colorado Springs 1420 Austin Bluffs Parkway Colorado Springs, CO 80918 719-255-3359 gstock@uccs.edu Kathleen L., McFadden, Ph.D. Department of Operations Management and Information Systems College of Business Northern Illinois University DeKalb, IL 60115 Phone: 815-753-6374 kmcfadden@niu.edu ABSTRACT This paper empirically tests the relationship between patient safety culture and several different dimensions of performance in a sample of 154 hospitals. We use data from the Hospital Survey on Patient Safety Culture User Comparative Database, which is managed by the Agency for Healthcare Research and Quality (AHRQ), to measure patient safety culture. We measure hospital performance using a set of objective patient safety and quality performance metrics collected by the Center for Medicare and Medicaid Services. Our findings indicate that stronger patient safety culture is associated with better hospital performance. KEYWORDS: Health care, patient safety, safety culture, Survey on Patient Safety Culture, Agency for Healthcare Research and Quality INTRODUCTION One of the challenges faced by the health care sector is a growing focus on quality and patient safety. Heightened concerns about healthcare quality and safety grew out of findings from the publication of a report by the Institute of Medicine entitled To Err is Human (Institute of Medicine, 2000). This report claimed that medical errors are responsible for about 98,000 deaths annually, but a more recent study now suggests that the number of deaths due to medical errors is almost four times higher than those original estimates (James, 2013). Medical errors are a persistent and costly problem in terms of not only human lives but also greatly contribute to soaring medical costs. Consequently, the healthcare industry is facing extreme pressure to improve patient safety, and also reduce costs. In addition, changes in the reimbursement system now penalize hospitals for certain medical errors while offering incentives for showing improvements on some process and quality metrics. Because of this

change, hospitals are now incentivized to focus on a broad range of hospital performance measures such as efficiency, process quality, patient safety, and patient satisfaction. Health care organizations are increasingly becoming aware of the importance of embracing a safety culture as a means of improving overall hospital performance (Nieva and Sorra, 2003). As Nieva and Sorra (2003) point out, the general concept of safety culture can be easily adapted to the healthcare context. It has been argued that many hospital intervention strategies have failed to provide the desired benefits because they have overlooked the primary source of the problem, which is a weak organizational safety culture (Singer and Vogus, 2013). There have been many empirical studies investigating the relationship between culture and patient safety performance (DiCuccio, 2014). However, many of these studies suffer from potential methodological shortcomings, including common-method bias, single respondents, poorly validated culture measures, or perceptual outcome measures. In an effort to remedy some of these problems, previous studies have used more robust patient safety culture measurement instruments, such as the Hospital Survey on Patient Safety Culture (HSOPS) survey, developed by the Agency for Healthcare Research and Quality (AHRQ), but used perceptual outcome measures (e.g., Richter et al., 2014a; Richter 2014b). Other studies have used various objective performance measures collected by the Centers for Medicare and Medicaid Services (CMS), but had only a small number of respondents within each hospital to measure safety culture (e.g., Boyer et al, 2012; McFadden et al, 2014). This study addresses these methodological concerns. We use the validated HSOPS measures with a sample of hospitals from the AHRQ HSOPS Comparative Database averaging 836 responses per hospital, to measure safety culture. We also use a robust set of objective patient safety and operational performance outcome measures collected and compiled by CMS. Some prior research has considered the relationship between the safety culture (using either HSOPS measures or other culture measures) and objective outcome measures (e.g., Mardon et al, 2010; Singer et al, 2009). However, to our knowledge, this is the first study to examine the link between safety culture, measured using the comprehensive HSOPS dataset collected by the AHRQ, and a broad set of objective patient safety and operational performance outcome measures, collected by CMS. The strength of this study, and its principal contribution, is in its use of a rigorous methodology to test this potentially critical relationship. The findings from this study should therefore provide hospital administrators with empirical support for the importance of a strong safety culture and a road map for improvement in hospital performance outcomes. CONCEPTUAL FRAMEWORK A theoretical explanation for the connection between culture and patient safety has its origins in the more general High Reliability Organization (HRO) theory. One of the seminal events associated with early development of HRO theory was the Chernobyl accident, which was the first incident to highlight the importance of a safety culture in improving performance outcomes (Flin et al. 2000). Research soon emerged that focused on high reliability organizations (HROs) and the role that safety culture plays in accident prevention (Sorra and Dyer, 2010). HROs are organizations that function in complex, tightly coupled, and often hazardprone environments and yet consistently achieve nearly error-free performance (Roberts 1990; McFadden et al. 2009). HRO theory supports the notion that organizational culture should be related to patient safety measures (Bierly and Spender 1995; Roberts et al. 2001; Bigley and Roberts 2001). Prior studies have demonstrated the importance of leadership as an enabler of safety practices and procedures. Vogus et al. (2010) found that leaders in high reliability organizations possess a strong commitment to safety, enable safety practices and procedures among their subordinates, and place safety as a number one priority. McFadden et al. (2009)

found that enabling strategies were associated with enacting practices that ultimately lead to better patient safety outcomes. In addition, they found an indirect relationship between patient safety culture and perceptual patient safety outcomes in hospitals. In another study, patient safety culture was also found to be directly and negatively related to hospital acquired condition rates, an objective measure of patient safety outcomes, and patient safety culture was indirectly related to hospital acquired condition rates via continuous quality improvement initiatives (McFadden et al. 2014). Therefore, prior literature supports the following hypothesis: H1: Stronger patient safety culture will be positively associated with patient safety performance. HRO theory provides support for the proposition that organizational culture should be related to better operational performance. In a hospital context, operational performance can include both quality and efficiency. Quality is a broad performance category, and is itself composed of different dimensions. For example, process quality, which is referred to as process of care, should be related to a safety culture (Bierly and Spender 1995; Roberts et al. 2001; Bigley and Roberts 2001). Process of care refers to the extent to which hospitals follow best clinical practices. HRO theory would support the use of standard operating procedures, checklists, and the determination of best practices. Empirical studies would also support this linkage. For instance, using the enabling-enacting-outcome model, Boyer et al. (2012) did not find direct effects of safety climate on process of care measures, but they did find positive indirect effects through quality practices. Similarly, McFadden et al. (2014) also did not find direct effects of patient safety climate to process quality scores but indirect effects via continuous quality improvement initiatives. A second dimension of quality in health care is patient satisfaction, which is also referred to as patient experience of care. HRO theory also provides support for the idea that organizational culture should be related to improve patient satisfaction (Bierly and Spender 1995; Roberts et al. 2001; Bigley and Roberts 2001). HROs embrace the notion of a safety culture and have fewer errors, thus customers will be more satisfied due to their improved reliability (Roberts, 1990; Ruchlin et al. 2004). Ancarani et al. 2011 found empirical evidence that employees perception of organizational climate mediates the relationship between the hospital managers climate orientation and patient satisfaction. A third dimension of quality would be clinical patient outcomes. Patient outcomes include both medical care outcomes, such as mortality, and patient safety outcomes, such as infections or surgical errors. We have already suggested in H1 that patient safety culture should be related to better patient safety performance, but we would also expect that patient safety culture would be related to better overall clinical outcomes as well. HRO theory also supports the idea that culture would be related to efficiency. Weick (1987) argues that strong organizational cultures provide a centralized and focused cognitive system within which complex and hazard-prone systems can function efficiently and more effectively overall. We expect that patient safety culture would be associated with better quality and efficiency, we therefore hypothesize the following relationship: H2: Stronger patient safety culture will be positively associated with operational performance. METHODS Data This study employed a mixed methodology, with different methods of data collection and different types of measures. The hospital is the unit of analysis. There were three different

categories of data that were needed to investigate the relationship between patient safety culture and hospital performance. The first category included data measuring patient safety culture; the second included data providing organizational measures to be used as control variables in the analysis; and the third included data providing performance measures for each hospital. The data for patient safety culture were collected by the Agency for Healthcare Research and Quality (AHRQ) (http://www.ahrq.gov/professionals/quality-patientsafety/patientsafetyculture/hospital/) through their biannual Hospital Survey on Patient Safety Culture (HSOPS). Data for other organizational variables were obtained from the Provider of Service (POS) database maintained by the Centers for Medicare and Medicaid Services (CMS) in the United States Department of Health and Human Services. Data for hospital performance were obtained from the Hospital Compare database, which is also maintained by CMS. The data from these three sources were merged into a single data set used in the analysis. We provide more detail about each data source and how they were merged below. The Agency for Healthcare Research and Quality (AHRQ) in 2004 developed and first administered the Hospital Survey on Patient Safety Culture (HSOPS) (http://www.ahrq.gov/professionals/quality-patient-safety/patientsafetyculture/hospital/). It has been administered on a regular basis, at first annually, and now biannually, since 2004. The HSOPS instrument is based on a thorough review of the literature. This measurement instrument has been used extensively in prior research (e.g., Aiken et al., 2012; Jones et al., 2013), and it has been validated in many different settings (e.g., Sorra and Dyer, 2010; Blegen et al., 2009). The survey has 42 items comprising 12 composite dimensions. Of these twelve composites, seven are considered to be dimensions of unit-level patient safety culture, three to be dimensions of hospital-level patient safety culture, and two to be patient safety outcomes. The Hospital Survey on Patient Safety Culture 2014 User Comparative Database Report (Sorra et al, 2014), published by AHRQ provides a description of the HSOPS composites. This report also includes the questionnaire items used in the HSOPS. The HSOPS has been used in prior studies in two primary ways. In one, researchers use the survey instrument to collect primary data from the target sample. The AHRQ also uses the survey to collect data from several hundred hospitals on a biannual basis and publishes a report summarizing the data collected. Hospitals may use the report to compare their own patient safety culture to those of other hospitals. In the most recent administration of the survey by the AHRQ, 653 hospitals participated. The average hospital had 621 completed surveys and a 54% response rate. A summary report of the data collected from this full sample is available from AHRQ (Sorra et al, 2014). This report also provides a complete set of the questionnaire items from the survey. For this study, we made a formal request for identifiable data to enable matching of the culture data from the HSOPS Comparative Database to the performance data from CMS. After a review to ensure the soundness of the research plan, AHRQ provided data from a subset of hospitals that had granted permission for use of their identifiable data. These data were aggregated at the hospital level. This dataset of 240 hospitals was used as the source of patient safety culture data in our study. A second data source was the Provider of Service (POS) file compiled by the CMS (http://www.cms.gov/research-statistics-data-and-systems/downloadable-public-use- Files/Provider-of-Services/index.html). The POS file includes several hundred organizational variables for each hospital providing Medicare services. We used several of these variables (e.g., number of beds) as control variables in the analysis. The third source of data was the Hospital Compare data base, also maintained by CMS (https://data.medicare.gov/data/hospital-compare). Hospital Compare includes a collection of publicly available data reported by hospitals across several dimensions of performance. The performance dimensions include patient satisfaction, process quality, patient safety, clinical

outcomes, and efficiency. The measures included in the Hospital Compare database are particularly relevant because CMS will begin to adjust a hospital's Medicare reimbursement payments upward or downward based on its performance on these measures (Centers for Medicare and Medicaid Services, 2014a). To assemble the final data set used in the analysis, we had to merge data from these three sources. All hospitals in the three databases have been assigned a Medicare identification number (Medicare ID). This number is unique to each hospital and is the same in each database. We were therefore able to use this Medicare ID to merge the different data sets. The Hospital Compare performance measures were collected only for short-term acute care hospitals. As a result, when the HSOPS data were merged with the Hospital Compare data, the sample size was reduced from 240 to 154. The average number of respondents for the HSOPS per hospital in this final sample was 836, with an average response rate of 58% per hospital. In total, there were 128,712 respondents from all hospitals. The POS data were then merged with this reduced sample to arrive at the final data sample. Dependent Variables The dependent variables used in the analysis include a patient safety outcome measure and an overall operational performance measure comprised of four components: process quality, patient satisfaction, patient outcomes, and efficiency. Hospitals must report these measures to the Centers for Medicare and Medicaid Services. The overall performance score can be used by Medicare to reward high performing hospitals with incentive payments. The separate patient safety score can also be used by Medicare to penalize poor performers by payment reductions. Because a hospital's financial position can be directly affected by its achievement on these measures, they are especially relevant when considering the effects of patient safety culture. We selected these measures explicitly because of their relevance to hospital managers. The patient safety performance measure used in our analysis to evaluate H1 is the Total Hospital Acquired Condition (HAC) score reported to Medicare. The Total HAC Score is reported as part of Medicare s Hospital Acquired Condition (HAC) Reduction Program. An HAC is a condition acquired while in a hospital that would not reasonably be expected to be acquired during the course of normal medical care. An example of an HAC would be a postoperative pulmonary embolism. The HAC Reduction Program scores also can affect a hospital financially, as hospitals that achieve a low level of performance may have Medicare reimbursement payments reduced. The Total HAC Score is the weighted average of two different patient safety domain scores. One domain is based on a set of patient safety indicators (PSIs) developed by AHRQ, and the other domain is based on a set of two patient safety measures developed by the Centers for Disease Control (CDC). The Total HAC Score is normalized to a 1 10 point scale, where lower scores indicate better performance. Some of the metrics used to calculate the Total HAC Score are also used in the calculation of the Patient Outcomes component score for the Value Based Purchasing Program described above. A more detailed description of how this measure is calculated is provided by Medicare.gov (Department of Health and Human Services, 2015) and Medicare s contractor, Quality Net (Quality Net, 2015). The overall operational performance measure is taken from Medicare's Value-Based Purchasing Program (VBPP). The VBPP provides incentive payments to hospitals that achieve high levels of performance on a composite of several different quality, patient safety, and efficiency measures (Centers for Medicare and Medicaid Services, 2014b). This composite score, referred to as the Total Performance score, is calculated as a weighted average of four components:

Process of care (20%): A composite of 12 measures indicating the extent to which the hospital follows best clinical practices in the treatment of several common diseases Patient experience of care (30%): A composite of 8 measures related to various dimensions of the patient s experience of care Patient outcomes (30%): A composite of 5 measures indicating performance on 2 patient safety indicators and mortality rates for 3 different diseases Efficiency (20%): Medicare spending per beneficiary A detailed description of how the Total Performance Score and individual domain scores are calculated is provided in the document available through the CMS web site (Centers for Medicare and Medicaid Services, 2011). Each component score is normalized on a 0 100 scale, and the Total Performance Score therefore has a 0 100 scale as well. When evaluating the relationship between patient safety culture and operational performance, we first estimated a regression model with the Total Performance Score as the dependent variable. Although the Total Performance Score provides an overall picture of the hospital's operational performance, examining the relationship between patient safety culture and each of the performance components provides even more insight. Therefore, we also estimated four additional regression models. Each of these models used a separate operational performance component score as the dependent variable. Independent Variables This study focuses on the relationship between patient safety culture and hospital performance. To measure safety culture, we use data from the HSOPS survey developed and administered by AHRQ. The HSOPS instrument has 42 questionnaire items grouped into twelve composite measures. These twelve composites include seven dimensions of unit-level patient safety culture, three dimensions of hospital-level safety culture, and two outcome measures. Table 1 lists the twelve dimensions of patient safety culture measured by the HSOPS. This survey instrument has been extensively used in prior research and its psychometric properties have been empirically validated many times and in many settings (e.g., Sorra and Dyer, 2010; Blegen et al., 2009). Therefore, our choice of the HSOPS to measure patient safety culture at the hospital level is appropriate. The data in our sample collected by AHRQ has the advantage of having a large number of respondents across a wide range of organizational roles for each hospital. Although the 12 composite measures in the HSOPS instrument have been found to be distinct in prior research, they have tended to be relatively highly correlated at the hospital level, with values in one study as high as 0.81 and averaging 0.56 (e.g., Sorra and Dyer, 2010). In our sample, a similar pattern emerged. However, the correlations between the HSOPS composite dimension scores in our sample are even higher, ranging from 0.46 to 0.94 with a mean of 0.75. Because they are so highly correlated, including more than one of the composite measures in the regression models described below would have resulted in severe multicollinearity and inaccurate results. In the development of the HSOPS, seven unit-level and three hospital-level dimensions of patient safety culture were identified. Therefore, we chose to aggregate the mean composite scores for unit-level and hospital-level culture respectively. However, even these composite scores were highly correlated and resulted in multicollinearity in the regression analysis. In the results of the analysis discussed below, we therefore included only the unit-level patient safety culture aggregated score to avoid the unstable estimates that would result from this multicollinearity. This score was computed by averaging the mean scores

for the seven unit-level culture dimensions. This scale had a Cronbach s alpha of 0.94, so it has a high level of internal consistency. In addition to the explanatory patient safety culture variable, we included several control variables to account for possible organizational and contextual effects on hospital performance. The first control variable we included is hospital size, which is measured as the natural logarithm of the number of beds in the hospital. The natural log transformation was used to reduce possible skew of this variable. We also included three other dichotomous organizational variables. One is whether the hospital is government-owned; another is the teaching status of the hospital; and finally, we included a variable to indicate whether the hospital is located in an urban area. Finally, to account for possible differences among different geographic regions, we included dummy variables for eight different geographic regions defined by the American Hospital Association. Table 1 provides a list and descriptions of all variables used in the analysis. Dependent Variables Total performance Process of care Patient experience of care Patient outcomes Efficiency Total HAC score Description Table 1. Variables. Medicare Value-Based Purchasing Program (VBPP) total performance VBPP Process of care component score VBPP Patient experience of care component score VBPP Patient outcomes component score VBPP Efficiency component score Overall score from Medicare's Hospital Acquired Condition (HAC) Reduction Program Independent Variables Patient safety culture Hospital size Government-owned Teaching Urban Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7 Description Mean of composite scores from the Hospital Survey on Patient Safety Culture measuring seven dimensions of unit-level patient safety culture Ln (total beds) 1 if hospital is government-owned, 0 otherwise 1 if hospital is a teaching hospital, 0 otherwise 1 if hospital is located in an urban area, 0 otherwise 1 if hospital is located in the New England region, 0 otherwise 1 if hospital is located in the Mid-Atlantic region, 0 otherwise 1 if hospital is located in the South Atlantic region, 0 otherwise 1 if hospital is located in the East South Central region, 0 otherwise 1 if hospital is located in the East North Central region, 0 otherwise 1 if hospital is located in the West North Central region, 0 otherwise 1 if hospital is located in the West South Central region, 0 otherwise

Region 8 1 if hospital is located in the Mountain region, 0 otherwise Region 9 (the Pacific region) is the reference region, where all region dummy variables are equal to 0. RESULTS Table 2 reports descriptive statistics for our study variables. Because submission of HSOPS data to AHRQ and allowing the use of the data are voluntary, the possibility of bias in the sample is a concern. We were able to compare the hospitals in our sample to the non- HSOPS hospitals on size, urban location, and geographic region. In addition, we include descriptive statistics for the performance measures for the hospitals in the Hospital Compare database that were not HSOSPS respondents. The study sample differed significantly from the full Hospital Compare dataset for only the Process of Care component of the Total Performance score, size, and geographic distribution. Table 3 shows correlations between the study variables. Table 2. Descriptive statistics. Study Sample Non-HSOPS Respondent Hospitals Mean S.D. Mean S.D. T-statistic Total HAC Score 5.32 2.01 5.41 2.04 0.55 Total Performance 43.33 12.09 41.62 12.57 1.71 Process of Care 59.93 17.89 55.32 20.47 3.06** Patient Experience 44.04 18.65 41.86 20.75 1.40 Patient Outcomes 46.69 17.77 45.07 18.53 1.07 Efficiency 18.42 23.56 20.94 25.97 1.28 Ln(Beds) 5.36 0.91 5.12 0.93 3.25*** Unit Safety Culture 3.71 0.14 Count Proportion Count Proportion 2 (df) Govt. Ownership 13 0.084 Teaching Hospital 61 0.394 Urban Location 39 0.253 1008 0.326 3.58 (1) Region 1 2 0.013 136 0.044 146.2 (8)*** Region 2 20 0.130 360 0.117 Region 3 14 0.091 512 0.166 Region 4 80 0.519 490 0.159 Region 5 9 0.058 276 0.089 Region 6 13 0.084 246 0.080 Region 7 3 0.019 472 0.153 Region 8 6 0.039 213 0.069 Region 9 7 0.045 384 0.124 * < 0.05, ** p < 0.01, *** p < 0.001

Table 3. Correlations. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 1. Total Performance 1 2. Process of Care.41 1 3. Patient Experience.72.14 1 4. Patient Outcomes.54 -.04.00 1 5. Efficiency.66.07.38 -.04 1 6. Total HAC Score -.21 -.04 -.20 -.21 -.03 1 7. Unit Safety Culture.36.18.43 -.06.22 -.25 1 8. Govt. Ownership.01 -.07.13 -.06.01 -.04.13 1 9. Teaching Hospital -.26 -.20 -.29 -.01 -.12.31 -.22 -.01 1 10. Size -.42 -.18 -.46.13 -.42.32 -.47 -.05.50 1 11. Urban location -.19 -.04 -.22.12 -.34.19 -.18 -.20.25.51 1 N = 154 Correlations with absolute value greater than or equal to 0.18 are significant at p < 0.05. Correlations with absolute value greater than or equal to 0.21 are significant at p < 0.01. We used ordinary least squares linear regression to test the hypothesized relationships. Tables 5 and 6 show the results of the regression analysis. In all models, standardized coefficients are shown to enable easier interpretation of the coefficients. The results were very interesting and in general supported our hypotheses. In particular, Model 1 from Table 4 showed that patient safety culture was negatively related Total HAC score, which indicates that stronger patient safety culture is related to better patient safety performance. Therefore, H1 is supported. Model 2 shows that patient safety culture is positively related to the Value-Based Purchasing Total Performance Score, so H2 was supported. To examine the effects of patient safety culture on operational performance in more depth, we extended the analysis to consider each Value Based Purchasing performance component score in a separate regression model, shown in Table 5. In Model 3, the dependent variable is the Process of Care measure, which shows the extent to which hospitals follow best practices in the treatment of a collection of common diseases. Patient safety culture was positively related to better process of care performance. Similarly, as shown in Model 4, patient safety culture was also positively related to patient satisfaction. On the other hand, Model 5 shows that patient safety culture was not significantly related to efficiency. In addition, patient safety culture was not significantly related to the patient outcomes measure, as shown in Model 6. This result was somewhat surprising. In Model 1, where Total HAC Score is the dependent variable, we saw a significant relationship between patient safety culture and better patient safety performance. We therefore might infer that patient safety culture is associated with better patient safety but not better mortality. We discuss these findings in more detail below.

Variable Table 4. Primary regression results. Model 1 Model 2 Patient Safety Performance (Total HAC Score) Size 0.12-0.28** Government 0.00 0.04 Teaching 0.17-0.10 Urban 0.07 0.04 Region 1 0.03-0.08 Region 2-0.19 0.02 Region 3-0.33** -0.13 Region 4-0.31-0.08 Region 5-0.08-0.15 Region 6-0.12 0.14 Region 7 0.00-0.04 Region 8 0.02-0.12 Patient safety culture -0.21* 0.24** Operational Performance (Total Performance Score) F 3.15*** 4.44*** R 2 0.23 0.30 * p < 0.05, ** p < 0.01, *** p < 0.001

Table 5. Operational performance component regression results. Model 3 Model 4 Model 5 Model 6 Variable Process of care Patient satisfaction Efficiency Patient Outcomes Size -0.19* -0.27** -0.40*** 0.14 Government -0.01 0.12 0.06-0.08 Teaching -0.02-0.08 0.06-0.08 Urban 0.07 0.00-0.13 0.06 Region 1-0.06 0.03-0.09-0.07 Region 2 0.24-0.03 0.00-0.07 Region 3-0.11-0.01-0.12-0.05 Region 4-0.02 0.16-0.21-0.05 Region 5-0.11-0.08-0.18 0.04 Region 6 0.06 0.10 0.19-0.02 Region 7-0.01 0.06-0.14 0.03 Region 8-0.05-0.13-0.10 0.07 Patient safety culture 0.20* 0.30*** 0.04-0.03 F 2.14* 5.86*** 5.72*** 0.50 R 2 0.17 0.36 0.35 0.05 * p < 0.05, ** p < 0.01, *** p < 0.001 DISCUSSION AND CONCLUSIONS In this study, we found significant relationships between patient safety culture and several dimensions of hospital performance. These findings are particularly important because of the data and measures used in the analysis. Prior research has found similar relationships between safety culture and performance (e.g. Aiken et al., 2012; Singer et al., 2009), but this study makes a significant contribution in its use of a unique dataset drawn from several separate sources. In particular, this study is the first to test relationships between the AHRQ Hospital Survey on Patient Safety data collected by AHRQ and comprehensive performance data from Medicare s Value Based Purchasing and Hospital Acquired Conditions Reduction Programs. The data used for independent variables, including the primary culture measures, were collected independently from the data used to measure the dependent variables, which eliminates common method bias as a concern. Moreover, the HSOPS research instrument has been extensively validated in prior research (e.g., Sorra and Dyer, 2010). Therefore, we have confidence that our measure of patient safety culture is reliable and valid. The performance measures and data used for the dependent variables are equally as important. These measures, as noted above, include performance metrics for the Value Based

Purchasing and Hospital Acquired Condition Reduction Programs. They assess performance across a wide range of operational dimensions, including process quality, patient satisfaction, patient outcomes, and efficiency, so they are valid for understanding how culture affects organizational performance. However, what is particularly important is that Medicare will use a hospital s performance on these measures to adjust payments to hospitals treating Medicare patients. The results of this study therefore go beyond interesting academic research findings to have tangible implications for a hospital s bottom line. Our results suggest that improving the patient safety culture of a hospital can pay off financially as well as improve the care of its patients. Although our general expectations were supported by the results, there some aspects of the findings that were somewhat surprising. An unexpected result was the lack of a relationship between culture and patient outcomes. However, as we noted above, the patient outcomes measure included measures of mortality rate as well as patient safety outcomes, so it is likely that safety culture has an effect on patient safety outcomes such as adverse events and medical errors rather than on mortality rate. We also did not see a significant relationship between patient safety culture and efficiency. It might be that strengthening a hospital s patient safety culture results in higher costs. It might also be a result of the efficiency measure used. A simple average of the Medicare spending per beneficiary, which we used because it was part of the Medicare VBPP Total Performance score, may not provide a true assessment of operational efficiency in a hospital. Our findings were interesting and provide significant contributions to research and practice, but there are some limitations to this study. First, in our sample, the dimensions of patient safety culture measured by HSOPS were highly correlated, which led to an analysis where we were able to test the effects of only one dimension. Part of the reason for this result might lie in the purpose for which this survey instrument was conceived and developed. The AHRQ originally developed the HSOPS as a diagnostic tool to be used at the hospital or unit level to assess the patient safety culture of the organization. AHRQ maintains a database and issues a biannual report that allows hospitals to compare themselves to other hospitals with respect to patient safety culture. Hospitals are encouraged to use the data and reports as part of their efforts to improve patient safety (Sorra et al., 2014). HSOPS was not developed with the intention that it be used as a set of measures for scholarly research. However, using a single, highly reliable, dimension, as we have done, does provide us with a validated measure of patient safety culture. Our findings showed a significant relationship between patient safety culture and hospital performance, but we were not able to investigate the organizational mechanisms through which stronger culture results in better performance Future research might focus on the development of conceptual frameworks and measures that would examine these organizational mechanisms. A possible starting point for future research in this vein might be the enablingenacting-elaborating framework introduced by Vogus et al (2010) and further developed by Singer and Vogus (2013). In this framework, safety culture has been described as a function of a series of interrelated organizational factors of enabling (activities that help shape perceptions of safety climate), enacting (activities by frontline employees intended to improve safety) and elaborating (learning practices that reinforce safety behavior) (Vogus et al. 2010). Future research could develop and validate a survey instrument that would measure the enabling, enacting, and elaborating constructs and then link them with the HSOPS culture composite dimensions. Richter et al (2014b) mapped the HSOPS composites as sub-dimensions of the Enabling, Enacting, and Elaborating constructs of this framework, so a first step in this direction has already been taken. Figure 1 shows a possible conceptual model that might be used in future research.

Figure 1. Possible conceptual model for future research. ENABLING Management support Supervisor expectations Communication openness Nonpunititive response to errors Staffing PATIENT SAFETY CULTURE ENACTING Teamwork across units Teamwork within units Handoffs and transitions ELABORATING Error feedback Organizational learning PATIENT SAFETY PERFORMANCE OPERATIONAL PERFORMANCE HOSPITAL PERFORMANCE An additional limitation is the size of the sample. Of the 653 hospitals that completed the HSOPS, only 240 agreed to allow their responses to be used in this research. Of those 240 hospitals, only 154 met the sample requirements of being a general acute care hospital that had reported outcome data to the CMS. Future research might collect data from a larger sample using both the HSOPS and an additional survey instrument as discussed above. In this study, we have provided compelling evidence of the importance and impact of patient safety culture. Our findings show that a stronger patient safety culture is beneficial to both the hospital and its patients. We believe that our study is useful for both scholars and practitioners in aiding efforts to improve patient safety. REFERENCES Aiken, L.H., Sermeus, W., Van Den Heede, et al. (2012). Patient safety, satisfaction, and quality of hospital care: cross sectional surveys of nurses and patients in 12 countries in Europe and the United States. BMJ, 344:e1717. Ancarani, A., Di Mauro, C., Giammanco, M. D. (2011). Patient satisfaction, managers' climate orientation and organizational climate. International Journal of Operations & Production Management, 31(3), 224-250.

Bierly III, P. E., & Spender, J. C. (1995). Culture and high reliability organizations: The case of the nuclear submarine. Journal of Management, 21(4), 639-656 Bigley, G. A., & Roberts, K. H. (2001). The incident command system: high-reliability organizing for complex and volatile task environments. Academy of Management Journal, 44(6), 1281-1299. Blegen, M.A., Gearhart, S., O Brien, R., Sehgal, N.L., and Alldredge, B.K. (2009). AHRQ s Hospital Survey on Patient Safety Culture: Psychometric analyses. Journal of Patient Safety, 5 (3), 139-144. Boyer, K. K., Gardner, J. W., Schweikhart, S. (2012). Process quality improvement: an examination of general vs. outcome-specific climate and practices in hospitals. Journal of Operations Management, 30(4), 325-339. Centers for Medicare and Medicaid Services (2011). http://www.cms.gov/outreach-and- Education/Medicare-Learning-Network- MLN/MLNProducts/downloads/Hospital_VBPurchasing_Fact_Sheet_ICN907664.pdf. Centers for Medicare and Medicaid Services (2014a). http://www.cms.gov/newsroom/mediareleasedatabase/fact-sheets/2014-fact-sheetsitems/2014-12-18-2.html. Centers for Medicare and Medicaid Services (2014b). https://www.cms.gov/medicare/quality- Initiatives-Patient-Assessment-Instruments/hospital-value-basedpurchasing/index.html?redirect=/hospital-value-based-purchasing/. Flin, R., Mearns, K., O'Conner, P. & Bryden, R. (2000) Measuring safety climate: identifying the common features. Safety Science 34, 177 192 Institute of Medicine. (2000). To Err is Human: Building a Safer Health System. Washington DC: National Academies Press. Jones, F., Podila, P., and Powers, C. (2013). Creating a culture of safety in the emergency department. Journal of Nursing Administration, 43 (4), 194-200. McFadden, K.L., Henagan, S.C., and Gowen III, C.R. (2009). The Patient Safety Chain: Transformational Leadership s Effect on Patient Safety Culture, Initiatives, and Outcomes. Journal of Operations Management, 27(5), 390-404. McFadden, K.L., Stock, G.N., Gowen III, C. R. (2014). Leadership, safety climate and CQI: impact on process quality and patient safety. Health Care Management Review, forthcoming. Nieva, V. F., & Sorra, J. (2003). Safety culture assessment: a tool for improving patient safety in healthcare organizations. Quality and Safety in Health Care, 12(suppl 2), ii17-ii23. Quality Net (2015). https://www.qualitynet.org/dcs/contentserver?c=page&pagename=qnetpublic%2fpage%2fq nettier2&cid=1228774189166.

Richter, J. P., McAlearney, A. S., & Pennell, M. L. (2014a). The influence of organizational factors on patient safety: Examining successful handoffs in health care. Health Care Manage Review, Richter, J. P., McAlearney, A. S., & Pennell, M. L. (2014b). Evaluating the Effect of Safety Culture on Error Reporting A Comparison of Managerial and Staff Perspectives. American Journal of Medical Quality, 1062860614544469. Roberts, K. H., (1990). Some characteristics of one type of high reliability organizations. Organization Science, 1(2), 160-177. Roberts, K. H., Bea, R., and Bartles, D. L. (2001). Must accidents happen? Lessons from highreliability organizations, Executive commentary. The Academy of Management Executives, 15(3), 70-79. Ruchlin, H. S., Dubbs, N. L., & Callahan, M. A. (2004). The role of leadership in instilling a culture of safety: Lessons from the literature. Journal of Healthcare Management, 49(1), 47-59. Singer, S. J., Gaba, D. M., Geppert, J. J., Sinaiko, A. D., Howard, S. K., & Park, K. C. (2003). The culture of safety: results of an organization-wide survey in 15 California hospitals. Quality and Safety in Health Care, 12(2), 112-118. Singer, S. J., A. Falwell, A., Gaba, D. M., Meterko, M., Rosen, A., Hartmann, C. W., and Baker, C. L. (2009). Identifying organizational cultures that promote patient safety. Health Care Management Review, 34(4), 300-311. Singer, S. J., Vogus, T.J. (2013a). Reducing hospital errors: interventions that build safety culture. Annual Review of Public Health, 34: 373-396. Singer, S. J., Vogus, T.J. (2013b). Safety climate research: taking stock and looking forward. Quality and Safety in Health Care, 22:1-4. Sorra, J. S., & Dyer, N. (2010). Multilevel psychometric properties of the AHRQ hospital survey on patient safety culture. BMC Health Services Research, 10(1), 199. Sorra J, Famolaro T, Yount ND, et al. (2014). Hospital Survey on Patient Safety Culture 2014 User Comparative Database Report. (Prepared by Westat, Rockville, MD, under Contract No. HHSA 290201300003C). Rockville, MD: Agency for Healthcare Research and Quality. AHRQ Publication No. 14-0019-EF. Sorra, J.S. and Nieva V.F. (2004). Hospital Survey on Patient Safety Culture. (Prepared by Westat, under Contract No. 290-96-0004). AHRQ Publication No. 04-0041. Rockville, MD: Agency for Healthcare Research and Quality. Sorra, J.S. and Dyer, N. (2010). Multilevel psychometric properties of the AHRQ hospital survey on patient safety culture. BMC Health Services Research, 10: 199-211.

Stock, G.N., McFadden, K.L., and Gowen III, C.R. (2007). Organizational culture, critical success factors, and the reduction of hospital rrrors, International Journal of Production Economics 106 (2), 368-392. United States Department of Health and Human Services (2015). http://www.medicare.gov/hospitalcompare/hac-reduction-program.html Vogus, T. J., Sutcliffe, K. M., Weick, K. E., (2010). Doing no harm: enabling, enacting, and elaborating a culture of safety in health care. Academy of Management Perspectives 24(4), 60-77. Weick, K. E. (1987). Organizational culture as a source of high reliability. California Management Review, 29, 112-127.