Identifying Organizational Cultures That Promote Patient Safety

300 October–December 2009

Identifying organizational cultures that promote patient safety

Sara J. Singer Alyson Falwell David M. Gaba Mark Meterko Amy Rosen Christine W. Hartmann Laurence Baker

Background: Safety climate refers to shared perceptions of what an organization is like with regard to safety, whereas safety culture refers to employees’ fundamental ideology and orientation and explains why safety is pursued in the manner exhibited within a particular organization. Although research has sought to identify opportunities for improving safety outcomes by studying patterns of variation in safety climate, few empirical studies have examined the impact of organizational characteristics such as culture on hospital safety climate. Purpose: This study explored how aspects of general organizational culture relate to hospital patient safety climate. Methodology: In a stratified sample of 92 U.S. hospitals, we sampled 100% of senior managers and physicians and 10% of other hospital workers. The Patient Safety Climate in Healthcare Organizations and the Zammuto and Krakower organizational culture surveys measured safety climate and group, entrepreneurial, hierarchical, and production orientation of hospitals’ culture, respectively. We administered safety climate surveys to 18,361 personnel and organizational culture surveys to a 5,894 random subsample between March 2004 and May 2005. Secondary data came from the 2004 American Hospital Association Annual Hospital Survey and Dun & Bradstreet. Hierarchical linear regressions assessed relationships between organizational culture and safety climate measures. Findings: Aspects of general organizational culture were strongly related to safety climate. A higher level of group culture correlated with a higher level of safety climate, but more hierarchical culture was associated with lower safety climate. Aspects of organizational culture accounted for more than threefold improvement in measures of model fit compared with models with controls alone. A mix of culture types, emphasizing group culture, seemed optimal for safety climate. Practice Implications: Safety climate and organizational culture are positively related. Results support strategies that promote group orientation and reduced hierarchy, including use of multidisciplinary team training, continuous quality improvement tools, and human resource practices and policies.

Key words: hospital characteristics, organizational culture, patient safety, safety climate, survey

Sara J. Singer, MBA, PhD, is Assistant Professor of Health Care Management and Policy, Harvard School of Public Health, Boston, MA. E-mail: [email protected]. Alyson Falwell, MPH, is Director of Operations, The Altos Group, Oakland, CA. David M. Gaba, MD, is Director, Patient Safety Culture Institute, VA Palo Alto Health Care System, CA. Mark Meterko, PhD, is Investigator, Center for Organization, Leadership, and Management Research, VA Boston Healthcare System, MA. Amy Rosen, PhD, is VA Research Career Scientist, Center for Health Quality, Outcomes and Economic Research, Bedford VAMC, MA. Christine W. Hartmann, PhD, is Research Scientist, Center for Health Quality, Outcomes, and Economic Research, Bedford VAMC, MA. Laurence Baker, PhD, is Professor, Department of Health Research and Policy, Stanford University, CA. Health Care Manage Rev, 2009, 34(4), 300-311 Copyright A 2009 Wolters Kluwer Health | Lippincott Williams & Wilkins

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Organizational Cultures That Promote Patient Safety 301

ublication of startling statistics nearly a decade ago about the extent of preventable medical errors Figure 1 Phas directed attention to the ‘‘culture of safety’’ in Competing values framework, adapted health care organizations (Institute of Medicine, 2000). from Quinn 1984 Subsequent research has sought to identify opportunities for improving safety outcomes by studying patterns of variation in what has been termed ‘‘safety climate’’ (Flin, Burns, Mearns, Yule, & Robertson, 2006). Climate and culture are not synonymous, although they are often used interchangeably (Zohar, 1980). Climate has been ‘‘defined as a perceptually based description of what the organization is like in terms of practices, policies, procedures, and routines, while culture helps define the underlying reasons and mechanisms for why these things occur in an organization based on fundamental ideolo- gies, assumptions, values, and artifacts’’ (Ostroff, Kinicki, & Tamkins, 2003). Thus, climate refers to shared perceptions related to a given, specific area of interest, such as safety (Schneider & Bowen, 1993), whereas culture refers to employees’ fundamental ideology and orientation (Trice & Beyer, 1993) and explains why an objec- tive like safety is pursued in the manner exhibited within a particular organization (Schein, 1992). and participation and distributes rewards equally among Limited research has examined organizational charac- members. An entrepreneurial culture is characterized by teristics that may affect safety climate in hospitals (Zohar innovation, by risk taking, by focus on growth, and by & Luria, 2004). In particular, we know little about how rewards for individual initiative. It should be noted that the organizational context could be modified to improve ‘‘risk taking’’ in this context does not imply taking risks safety climate. A better understanding of this relation- with respect to patient safety but rather with patient care ship could reveal, for example, whether aspects of gen- processes, often with the explicit intent of improving eral organizational culture predispose some hospitals to some aspect of patient safety. A hierarchical culture better safety climate. values predictable operations, which it achieves through structure, rules, policies, and procedures; rewards are allocated according to rank. Finally, a production-oriented Theory and Conceptual Framework culture is rational, with a focus on rewards for goal accomplishment. According to the CVF, the overall cul- Drawing on expanding evidence linking safety outcomes ture of an organization will reflect some particular mix of to climate measures in health care and other indus- these four dimensions. Thus, the CVF framework allows tries (e.g., Clarke, 2006; Hofmann & Mark, 2006), we for the simultaneous consideration of potentially com- assumed that a strong safety climate is beneficial for peting organizational characteristics and provides a means patient safety. We conceived of safety climate both as to encapsulate concisely the complex nature of organi- a property of organizations as a whole and as varying zational culture. among groups within organizations (Gaba, Singer, & Working within the CVF framework, we considered Rosen, 2007). Following organizational psychology the- likely relationships between levels of organizational cul- ory which suggests that general organizational context ture and safety climate. Organizations with higher levels shapes specific aspects of climate (Wallace, Popp, & of group culture were hypothesized to have higher levels Mondore, 2006), we examined organizational culture and of safety climate (Hypothesis 1). We based this predic- its relationship to patient safety climate in 92 U.S. hos- tion on previous findings that strong groups are char- pitals. In this way, we explored the potential for basic acterized by the psychological safety that encourages the assumptions, values, and beliefs to play a role in safety openness about mistakes and concerns necessary for or- climate. ganizations to learn and to improve (Edmondson, 1999). In this study, organizational culture was operational- The participation, the inclusion, and the shared decision ized using the competing values framework (CVF; Quinn making inherent in group culture have promoted col- & Rohrbaugh, 1983; Figure 1). The CVF framework laboration and learning from peers (Cameron & Quinn, posits four organizational culture dimensions. A group 1999; Shortell et al., 1995). Collegiality has also facili- culture emphasizes teamwork, cohesiveness, mentorship, tated the implementation of processes related to improved

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 302 Health Care Management REVIEW October–December 2009

patient safety (Kaissi, Kralewski, Dowd, & Heaton, 2007). and routines designed to take pressure off employees and Thus, we expected that group-oriented culture would be to reinforce the importance of safety, such as time-outs reflected in the policies, procedures, practices, and rou- for groups to reflect before and after procedures. Con- tines related to safety—the safety climate—in an orga- versely, production pressure has been shown to impede nization. More group-oriented cultures have also been safety climate and improvement in the operating room positively associated with a variety of better outcomes (Gaba, Howard, & Jump, 1994). More production- (Cameron & Quinn, 1999; Carman et al., 1996; Meterko, oriented cultures have also been negatively related to per- Mohr, & Young, 2004; Shortell et al., 1995, 2000; Zazzali, formance measures (Cameron & Quinn, 1999; Carman Alexander, Shortell, & Burns, 2007). et al., 1996; Meterko et al., 2004; Shortell et al., 1995, Higher levels of entrepreneurial culture were also 2000; Zazzali et al., 2007). hypothesized to be associated with higher levels of safety The previous hypotheses emphasize the influences climate (Hypothesis 2). The dynamic and complex na- of each of the organizational culture dimensions indi- ture of health care requires hospitals to design innovative vidually. In practice, however, hospitals will have mix- solutions to problems, to incorporate new technologies, tures of these dimensions. Some mixes of culture types and to adopt new routines safely and effectively. Em- might confer advantages that could lead to better safety ployees in hospitals with a culture that emphasizes ini- climate and outcomes. For example, although safety tiative, adaptability, and resilience—particularly when can be enhanced by open communication and shared balanced by some degree of hierarchical control—may decision making, these values may be strengthened by have expanded opportunities to observe experimentation the presence of some hierarchy, in the form of rules and aimed at improving safety and speaking up when an guidelines, which could beneficially address organiza- action or rule could compromise the safety of a patient. tions’ need for authority and accountability and mini- More entrepreneurial cultures have also been positively mize unproductive and erratic variation in practice. To related to better outcomes (Cameron & Quinn, 1999; explore this, we investigated the mix of organizational Carman et al., 1996; Meterko et al., 2004; Shortell et al., culture associated with the highest safety climate. 1995, 2000; Zazzali et al., 2007). By contrast, higher levels of hierarchical culture were Methods hypothesized to be associated with lower levels of safety climate (Hypothesis 3). Although the standardization that comes with strong hierarchy can be beneficial Sample (Adler, Goldoftas, & Levine, 1999), strong hierarchy can impede incorporation of frontline expertise into de- Sample hospitals came from a stratified random sam- cisions and hamper sharing of valuable information, ple of 92 U.S. hospitals, representing three size cate- stifling safety awareness and learning (Carroll, Rudolph, gories and all four U.S. census regions. They resembled & Hatakenaka, 2002). One key feature of other orga- U.S. hospitals except as dictated by our recruitment nizations with strong safety records (high-reliability strategy, which sought equal representation of small organizations [HROs]) is an ability to ‘‘flatten the hi- (0–99 beds), medium (100–249 beds), and large (250+ erarchy’’ and to promote information flow (Roberts, beds) hospitals. Thus, sample hospitals were larger 1990). Without explicit practices and policies that than the U.S. average. In addition, hospitals from the decentralize power, employees in hospitals that empha- West were overrepresented in the sample. However, size rules and standard procedures may also feel less em- sample hospitals had similar rates of safety events, powered to speak up or to take action when confronted measured by the AHRQ Patient Safety Indicators, as with a safety issue. Finally, organizations with higher all U.S. hospitals. levels of hierarchical culture have been most negatively Questionnaires assessing perceptions of safety climate related to performance measures (Cameron & Quinn, and organizational culture were administered to 35,340 1999; Carman et al., 1996; Meterko et al., 2004; Shortell individuals, including all senior managers, all physicians, et al., 1995, 2000; Zazzali et al., 2007). and a 10% random sample of all other hospital staff at Like hierarchical cultures, we hypothesized that study hospitals between March 2004 and May 2005. higher levels of production-oriented culture will be asso- Survey administration and follow-up procedures are re- ciated with lower levels of safety climate (Hypothesis 4). ported elsewhere (Singer, Falwell, Gaba, & Baker, 2008). In health care, a prime challenge to achieving safe All questionnaires included measures of safety climate; a climates is overt or covert pressure to put production and random one third of these also included measures of efficiency ahead of safety. Managing production pressure organizational culture. We received 18,361 safety cli- has also been important for achieving safe climates mate surveys (52%) and 5,637 organizational culture sur- in other industries (Reason, 1990). Hospitals with less veys (42%). The overall response varied among hospitals production-oriented cultures are likely to use practices and work groups (Singer et al., 2008, 2009). However,

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Organizational Cultures That Promote Patient Safety 303

supplementary analysis confirmed that hospitals with low reliable results requires consistent, prosafety perceptions response rates did not represent extremes in our distribu- among workers (Gaba, Singer, Sinaiko, & Bowen, 2003; tion of safety climate or organizational culture measures. Roberts, 1990). We prefer to use PPR rather than mean, In addition, differences in response patterns by survey positive, or standard deviation measures of safety climate wave did not suggest bias in response patterns. We ad- because it highlights areas where safety attitudes are not justed responses to account for sampling and known re- uniformly positive. sponse rate differences by job type (Singer et al., 2003). For each respondent, we calculated the PPR for the eight separate multi-item safety climate dimensions as the average of item PPR. We also computed the average Variables PPR for the 38 survey questions as a summary statistic, which we referred to as safety climate overall. Safety Climate. Workers’ perceptions of safety climate were measured using the 2004 Patient Safety Climate Organizational Culture. Organizational culture was in Healthcare Organizations (PSCHO) survey (available assessed using the CVF as operationalized by Zammuto on request). This instrument was selected from among and Krakower (1991 [Z&K]; available on request). Sur- several instruments (Colla, Bracken, Kinney, & Weeks, vey methods have been questioned as a valid measure of 2005; Flin et al., 2006; Singla, Kitch, Weissman, & organizational culture (Ashkanasy, Broadfoot, & Falkus, Campbell, 2006) because it was designed to assess safety 2000; Schein, 1992; Trice & Beyer, 1993), which tradi- climate among all hospital employees, not just clinical tionally has been studied using ethnographic methods. staff or staff in specific units; its reliability and validity However, although admittedly controversial (Scott, have been established; and it has been used extensively Mannion, Davies, & Marshall, 2003), the Z&K measure elsewhere (Cooper, 2006; Ginsburg, Norton, Casebeer, and the CVF are well established (Ostroff et al., 2003), & Lewis, 2005; Hartmann et al., 2008; Singer et al., have yielded findings similar to qualitative research 2003, 2008). The PSCHO survey consisted of 38 ques- (Zammuto & Krakower, 1991), and have proven useful tions about safety climate topics drawn from literature on for understanding the culture of health care organizations HROs, plus six demographic questions. Each of the safety (Cameron & Quinn, 1999; Carman et al., 1996; Meterko climate items used a 5-point Likert scale with a neutral et al., 2004; Shortell et al., 1995, 2000; Zazzali et al., midpoint. Psychometric analysis of response patterns, 2007). Nevertheless, we acknowledge that the Z&K which assessed both convergent and discriminant valid- survey measured perceptions of dominant organizational ity, supported the construction of eight valid and reliable behavior, which may or may not have measured culture dimensions (Singer et al., 2007). Three organizational per se. Compared with qualitative approaches, a survey dimensions were senior managers’ engagement with pa- also has the advantage of efficient implementation across tient safety, the extent to which organizational resources large numbers of organizations. are perceived as sufficient for safety, and a measure of The Z&K measure is structured around five key or- the hospital’s overall emphasis on safety. Two work–unit ganizational features including, for example, the chief safety climate dimensions were informal unit safety norms basis for distributing rewards. For each of these five and formal recognition and support for safety efforts. features, four characterizations of an organization were Two interpersonal dimensions were fear of shame as- presented, one representing each cultural type of the sociated with needing to ask for help and fear of blame CVF. Respondents distributed 100 points across the cul- for having made a mistake. One additional dimension ture characterizations to indicate the extent to which was a self-reported measure of the incidence of unsafe each description resembled their own organization. care. Cronbach’s alpha reliability coefficients for the Cronbach’s alpha reliability coefficients for the organi- safety climate dimensions ranged from .58 ( fear of shame) zational culture types in our study were .48 for the to .89 (senior managers’ engagement; Table 1). production-oriented culture dimension, .55 for entrepre- We summarized results from the PSCHO survey by neurial, .70 for hierarchical, and .77 for group. We computing the fraction of respondents answering in ways included in our models dimensions with relatively low that indicated low levels of safety climate. We referred to reliability because of their theoretical interest. However, this fraction as the percent problematic response (PPR). results related to production-oriented and entrepreneur- Higher PPR indicates a lower level of safety climate. ial culture should be interpreted cautiously. Debate continues regarding the most appropriate way to measure the strength and the uniformity of safety climate Control Variables. Characteristics of individuals and (Klein & Kozlowski, 2000); calculating PPR was one their jobs that may affect safety climate, including level method (Gaba, Singer, & Rosen, 2007) used in previous of supervisory responsibility, age, gender, time at in- research (Hartmann et al., 2008; Singer et al., 2003, stitution, professional discipline, and work area (emer- 2008). Studies of HROs suggest that achieving highly gency department, intensive care unit, operating room/

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 304 Health Care Management REVIEW October–December 2009

Table 1 Descriptive statisticsa Mean PPR Safety among survey climate Dimension Definition respondentsb SD Cronbach’s

Overall Average of Overall level of safety climate as reflected 17.1 45.2 38 items in the eight dimensions described below Organizational Senior Accuracy of senior managers’ understanding 11.9 73.7 .89 dimensions manager of organizational safety issues; engagement supportiveness of actions taken when necessary, and appreciation of qualifications of frontline personnel to solve safety issues Organizational Adequacy of personnel, time, equipment, 24.4 104.4 .67 resources and other resources necessary to provide for safety safe patient care Overall Overall level of emphasis on patient safety at 11.1 75.6 .65 emphasis a facility and whether the respondent feels on safety that safety is improving there Work–unit Unit safety The immediate work environment is one in 9.6 58.0 .82 dimensions norms which safety issues are proactively assessed and addressed, patient safety is a genuine and pervasive value among staff, and concern for safety defines the norms of socially acceptable behavior Unit The immediate work environment is one in 28.4 104.5 .74 recognition which actions that promote safe patient care and support are explicitly acknowledged and patient for safety safety standards are formally used in efforts training and evaluation of performance Interpersonal Fear of shame Respondents’ comfort admitting to mistakes 11.7 55.3 .58 dimensions and gaps in knowledge and seeking help Fear of blame Respondents’ perception that revealing 31.7 128.6 .61 mistakes would result in discipline and punishment Other Provision of Whether respondents witnessed or were 30.0 110.5 .66 safe care directly involved in the provision of unsafe care

Mean score outof100 Organizational among survey culture Dimension Definition respondentsb SD Cronbach’s

Group culture Culture emphasizes teamwork, cohesiveness, 28.2 56.9 .77 mentorship, and participation and distributes rewards equally among members Entrepreneurial Culture is characterized by innovation, risk 15.7 29.5 culture taking, focus on growth, and rewards for .55 individual initiative Hierarchical Culture values predictable operations, which 31.6 56.8 .70 culture it achieves through structure, rules, policies, and procedures; rewards are allocated according to rank Production- Culture is rational, with a focus on rewards 24.5 39.0 .48 oriented for goal accomplishment culture

aDescriptive statistics for control variables included in regression models are available from the authors on request. bWeighted for sampling differences and item nonresponse.

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Organizational Cultures That Promote Patient Safety 305

postanesthesia care unit, medical/surgical ward, and oth- gression models do not permit direct assessment of the ers), served as controls in all models. These were drawn variance in the dependent variable explained by inde- from individual responses to the PSCHO survey. pendent variables in the way that Ordinary Least Squares We also recognized that structural characteristics of regressions do. Deviance differences allow assessment of hospitals, such as size, tax status, teaching status, region, model fit to the data, albeit in relative terms. urban location, nurse staffing levels, and financial status, To explore what might be considered an ‘‘optimal’’ may affect patient safety. We thus included these vari- mix of general organizational culture types implied by ables as additional controls in our analysis. Most of the our data, we compared the average culture-type scores of hospital characteristics used in this study were derived five hospitals with the highest safety climate overall to from the American Hospital Association 2004 Annual those with the lowest safety climate overall. First, we Survey of Hospitals. For hospital size, we included both aggregated safety climate and organizational culture number of hospital beds and its square in our models responses from the individual level to the hospital level. because visual inspection suggested a curvilinear rela- Aggregation was justified by one-way analysis of variance tionship between size and safety climate. As our measure models, which assessed within-group versus between- of nurse staffing, we calculated the ratio of total full-time group variance for safety climate overall and for each equivalent nurse hours per patient day. For financial sta- organizational culture dimension. The intraclass corre- tus, we relied on the Dun & Bradstreet percentile rank- lation coefficients of .027 for safety climate overall ings of hospital credit scores in 2004. evaluated at the hospital level based on 197 respondents per hospital and of .030 to .122 for the organizational Statistical Analysis culture dimensions based on 61 respondents per hospital were all statistically significant ( p = .000) Next, we Hierarchical linear models examined the relationship graphically compared average organizational culture between hospitals’ organizational culture and patient scores for the two groups of hospitals using a quadrant safety climate, using individual-level data (Snijders & map that has been applied previously to the CVF mea- Bosker, 1999). Consistent with our conceptualization of sures (Zazzali et al., 2007). Then, we compared the mean safety climate as a property of both groups and orga- scores for each culture type between the five highest and nizations, we performed three-level random intercept lowest safety climate hospitals using a t statistic to test analysis, which accounted for the nesting of individuals for significance of the differences. within work areas within hospitals while allowing for variation both within hospitals between work areas and between hospitals. A chi-square test comparing two- and Findings three-level models with linear regressions found significant differences among both work areas and hospitals and confirmed appropriateness of using three-level models, Sample Characteristics 2 (2) = 630.86, p < .001. Respondents’ scored their hospitals as having a hierar- We estimated hierarchical linear models in which chical organizational culture the most (average = 31.6 safety climate and its eight dimensions were the depen- points) and as having an entrepreneurial culture the dent variables and measures of organizational culture least (average = 15.7 points; Table 1). Thus, the average were the independent variables of interest. Due to the respondent worked in a hospital with a culture that ipsative nature of the organizational culture measures, emphasized rules but also encouraged participation and only one culture type was included in any given model as teamwork and attended less to productivity and least to in prior research with this instrument (Zazzali et al., risk taking. The average PPR for safety climate overall 2007). Therefore, for each of the nine dependent was 17.1%. Additional characteristics of respondents are variables, we first estimated a baseline model including available from the authors on request. only the control variables using our full sample of respondents. We then added each of the four culture types in turn to this basic model, applying these to the Relationship of Organizational Culture subsample of individuals who responded to the ques- and Safety Climate tionnaire version including both the safety climate and the organizational culture components. Thus, we exam- Five regression models predicting safety climate overall ined 45 models (5 models for each of the nine dependent are presented. Results for the remaining 40 regression variables). Analyses were performed using STATA-10. models, providing more detail regarding the relationship Next, we compared deviance differences between of organizational culture dimensions with specific as- models to approximate information provided by an ad- pects of safety climate, are available from the authors on justed R2 (Snijders & Bosker, 1999). Hierarchical re- request.

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 306 Health Care Management REVIEW October–December 2009

Relationships between organizational culture and production-oriented (22.4), and entrepreneurial (13.9) safety climate were generally strong (Table 2, columns dimensions. Differences between levels of group and 2–5). Higher group culture was associated with higher hierarchical culture present in high and low safety safety climate overall and for each safety climate climate hospitals were statistically significant at p < .05. dimension individually (Hypothesis 1). Overall, a one- point-higher score for group culture correlated with a .24 percentage point lower PPR. Group culture was most Practice Implications highly correlated with organizational resources and work–unit support for safety, with an effect size of .42 We found that patient safety climate was better when percentage points for both. Higher entrepreneurial individuals also perceived that their hospital emphasized culture was also related to higher safety climate overall more group participation and less hierarchy. The and for all but the two interpersonal dimensions (fear of association between higher group culture and better safety shame and fear of blame; Hypothesis 2). A one-point- climate was not surprising, given that teamwork is neces- higher score for entrepreneurial culture related to .28 sary in patient care and problem-solving activities to percentage point lower PPR overall and .63 percentage maintain a safe environment. That more entrepreneurial point lower PPR for work–unit support for safety (the culture may also relate positively to safety climate implies strongest effect for an individual safety climate dimen- that some innovation and adaptability may be necessary sion). Higher hierarchical culture related to lower safety for organizational learning and process change. Our find- climate overall and for all dimensions except fear of ing that better safety climate was related to lower hier- shame (Hypothesis 3). A one-point-higher score for archical culture suggests that high levels of bureaucracy hierarchical culture related to .30 percentage point may have a dampening effect on communication and higher PPR overall and .62 percentage point higher PPR information flow, which can be an important impediment for work–unit support for safety. Higher production- to safety climate. In addition, better safety climate may oriented culture was related to lower safety climate relate to a lower production orientation, implying that overall, for the three organizational dimensions and the demand for efficiency—at least to the extent that it comes provision of safe care, but not the work–unit or at the expense of safety—could impede safety climate. interpersonal safety climate dimensions (Hypothesis 4). Results are relatively consistent across safety climate dimensions. However, organizational culture is most Comparison of Model Fit. Statistics assessing the fit consistently related to features of safety climate that of our models to the data suggested substantial im- pertain to organizations and least consistently related to provement in fit when the organizational culture vari- interpersonal dimensions of safety climate. The latter ables were included relative to the model including only may be more a function of individual differences relative controls (Table 2). The organizational culture variables to institutional characteristics in comparison with other accounted for more than threefold improvement in de- safety climate dimensions. viance measures of model fit compared with models with The graphical comparison of the distribution of controls alone. cultural-type scores in hospitals with the highest and lowest levels of safety climate suggests that the rela- Optimal Mix of Organizational Culture Type. To tionship between culture types and safety climate may explore the role of culture-type mix, we examined five be complicated. The optimal mix probably weights hospitals in our sample with the highest and five with group orientation most heavily. Relatively high levels of the lowest safety climate scores overall. Mean PPR for hierarchical culture observed among hospitals with the safety climate overall was 11.5 among the highest safety highest safety climate suggest that it may be impor- climate hospitals and 24.6 among the lowest safety tant to couple strong teamwork with entrepreneurial climate hospitals ( p = .000). Figure 2 is a graphical initiatives and adaptability with sufficient standard depiction of the average culture-type mixtures in these operating procedures and managerial control to achieve two groups. The results suggest that the optimal safety appropriate uniformity and performance of clinical climate includes a mix of general organizational culture processes. types. The five hospitals with the highest levels of safety These results have implications for the design and climate scored highest on group culture (40.1), followed effectiveness of hospitals. Our comparison of organiza- by hierarchical culture (24.6), production-oriented cul- tional culture in hospitals with different levels of safety ture (20.0), and entrepreneurial culture (15.3). The climate and related findings in other industries (Zohar, pattern among the five hospitals with the lowest levels 2002) suggests that promoting group-oriented cultures of safety climate was also consistent with regression and reducing hierarchical ones may be desirable. Reduc- results, with the highest score on hierarchical culture ing production pressure by improving workload predic- (36.7) and considerably lower scores on group (26.9), tion and planning and through policies and procedures

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Organizational Cultures That Promote Patient Safety 307

Table 2 Relationship of organizational characteristics with patient safety climate overall Fixed effects Safety climate overalla (omitted variable) Coefficient 1 2 3 4 5

Organizational Group culture À0.241 (0.011)** culture Entrepreneurial À0.279 (0.022)** (average culture points Hierarchical 0.300 (0.011)** allocated culture of 100) Production- 0.0666 oriented (0.017)** culture Hospital size No. beds 0.00861 À0.00165 0.0126 0.00251 0.0101 (0.0056) (0.0051) (0.0047) (0.0056)* (0.0044) No. beds 0.0000113 À0.0000016 0.0000126 0.00000335 0.0000114 squared (0.000005)* (0.000004) (0.000005)* (0.000004) (0.000005)* Nurse staffing Full-time À0.0455 À0.0448 0.0117 À0.00491 À0.00636 equivalent (0.60) (0.55) (0.67) (0.051) (0.066) registered nurse hours per patient day Financial Credit score À0.0223 (0.01)* À0.0136 (0.01) À0.0187 (0.01) À0.0126 (0.01) À0.0198 (0.01) status percentile Region Northeast 1.753 (0.98)* 0.989 (0.89) 1.031 (1.08) 0.681 (0.83) 1.261 (1.06) (Midwest) South 1.373 (0.92) 0.856 (0.84) 0.727 (1.02) 0.743 (0.79) 0.91 (1.01) West 2.322 (0.93)* 1.505 (0.85) 2.184 (1.04)* 1.625 (0.80)* 2.212 (1.02)* Tax status Government À0.17 (1.66) 1.511 (1.64) À0.0385 (1.93) À0.21 (1.55) 0.805 (1.91) (for-profit) Nonprofit 0.0196 (1.48) 1.32 (1.50) 0.365 (1.75) 0.498 (1.41) 0.795 (1.73) Teaching Other teaching 0.847 (1.18) 1.567 (1.00) 0.768 (1.25) 1.425 (0.93) 0.934 (1.23) status Nonteaching À1.114 (1.40) 0.0555 (1.22) 0.00306 (1.51) 0.419 (1.14) À0.223 (1.48) (major teaching)b Location Urban À1.278 (0.81) À1.187 (0.79) À2.177 (0.93)* À1.540 (0.74)* À1.980 (0.92)* (nonurban) Job level Supervisor 2.897 (0.38)** 1.698 (0.64)** 2.687 (0.66)** 1.169 (0.63) 2.870 (0.67)** (senior Frontline 4.317 (0.30)** 2.201 (0.51)** 3.309 (0.52)** 1.308 (0.50)** 3.743 (0.53)** manager)c worker Age (18–25 26–30 years 0.3 (0.73) À0.594 (1.24) À0.347 (1.28) À0.634 (1.21) À0.29 (1.30) years) 31–40 years 0.0481 (0.62) À0.541 (1.04) À0.983 (1.07) À1.206 (1.01) À0.577 (1.09) 41–50 years 0.0109 (0.62) 0.251 (1.03) À0.394 (1.07) À0.391 (1.01) 0.0394 (1.09) 51–60 years À0.825 (0.64) À0.0115 (1.06) À0.61 (1.09) À0.54 (1.03) À0.309 (1.11) >60 years À3.084 (0.72)** À2.224 (1.19) À3.203 (1.23)**À2.624 (1.16)* À3.015 (1.25)* Gender Male 0.122 (0.27) À0.655 (0.46) À0.628 (0.47) À0.727 (0.44) À0.622 (0.48) (female) Time at 6 months to 0.543 (0.80) 0.909 (1.32) 0.959 (1.36) 0.361 (1.28) 1.276 (1.38) institution 1 year (0–6 1–3 years 1.943 (0.72)** 1.986 (1.16) 1.85 (1.19) 1.168 (1.13) 2.273 (1.21) months) 3–5 years 3.627 (0.73)** 4.170 (1.18)** 4.391 (1.21)** 3.233 (1.15)** 4.926 (1.23)** 5–10 years 3.538 (0.72)** 3.708 (1.16)** 3.934 (1.20)** 2.575 (1.14)* 4.503 (1.22)** >10 years 3.048 (0.71)** 3.433 (1.15)** 3.433 (1.18)** 2.570 (1.12)* 3.906 (1.20)** Professional Physician 0.039 (0.37) 0.932 (0.61) 0.536 (0.63) 0.293 (0.59) 1.05 (0.64) discipline Resident À3.335 (0.99)** À5.232 (1.97)** À5.479 (2.05)**À5.457 (1.91)** À5.056 (2.08)* (nurse)d Other clinician À0.356 (0.44) 0.0871 (0.75) 0.363 (0.77) 0.449 (0.73) 0.248 (0.78) Nonclinician À2.489 (0.30)** À2.832 (0.51)** À2.189 (0.52)**À2.753 (0.49)** À2.399 (0.53)**

(continues)

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 308 Health Care Management REVIEW October–December 2009

Table 2 Continued Fixed effects Safety climate overalla (omitted variable) Coefficient 12345

Work area Intensive care 0.456 (0.59) 1.953 (0.91)* 1.429 (0.95) 1.995 (0.87)* 1.376 (0.97) (operating unit room/ Emergency 2.843 (0.62)** 4.312 (0.97)** 3.819 (1.01)** 4.438 (0.93)** 3.881 (1.03)** postanesthesia department care unit)e Ward 1.016 (0.54) 1.339 (0.79) 0.767 (0.83) 1.557 (0.76)* 0.745 (0.85) Other clinical À1.354 (0.47)** À0.976 (0.67) À1.200 (0.70) À0.773 (0.64) À1.248 (0.72) areas

Constant 12.68 (2.60)** 20.67 (2.79)** 15.62 (3.19)** 5.761 (2.62)* 8.824 (3.17)** Observations 14,257 4,537 4,537 4,537 4,537 No. groups 79 79 79 79 79 2, overall model 682.8 806.8 461.9 1,107 300.3 df,overall 33 34 34 34 34 model Testofmodelfit Log-restricted À56,691 À17,795 À17,938 À17,676 À18,013 likelihood Deviance = À113,382 À35,590 À35,876 À35,352 À36,026 À2 Â log likelihood Improvement 77,792 77,506 78,030 77,356 versus model 1

aResults derived from three-level hierarchical regressions. Dependent variable is overall average PPR. Standard errors in parentheses. bA major teaching hospital is defined as a member of the Council of Teaching Hospitals of the Association of American Medical Colleges. cSenior managers, supervisors, or frontline personnel are designated according to individuals’ response to a survey item asking whether they were ‘‘senior management—department head or above,’’ ‘‘a supervisor, but not a senior manager,’’ or ‘‘not a supervisor. dProfessional discipline is as follows: physician, if respondents indicated on the PSCHO survey that they were ‘‘physician–staff’’; residents, if they indicated they were ‘‘resident,’’ ‘‘intern,’’ or ‘‘fellow’’; nurses, if they indicated they were registered nurse, registered nurse practitioner, or licensed vocational nurse; other clinician, if they indicated they were ‘‘nursing assistant,’’ ‘‘pharmacist,’’ ‘‘physician assistant,’’ ‘‘food technician,’’ ‘‘respiratory therapist,’’ ‘‘audiologist,’’ or ‘‘radiologist.’’ Other respondents were classified as nonclinicians. eWork areas include operating room/postanesthesia care unit, intensive care unit, emergency department, and medical/surgical ward. We classified respondents as working in ‘‘other clinical areas’’ if they indicated on the PSCHO survey that they worked in ambulatory care, laboratory, pharmacy, home care, urgent care, or other clinical. **p < .01, *p < .05. for handling unpredictable workload changes is a known Several strategies could be used to improve teamwork strategy, also consistent with our findings (Gaba et al., and group orientation. For example, teamwork might be 1994). There is also evidence that senior managers do improved through multidisciplinary team training, in- not perceive safety climate in the same way that frontline cluding use of simulation techniques that provide expe- workers do, and this may hamper their ability to ap- riential learning opportunities and allow individuals to preciate and to support the changes required to optimize ‘‘walk in someone else’s shoes’’ to gain understanding and patient safety (Singer et al., 2008). Interventions for appreciation for other roles and perspectives. Similarly, improving senior managers’ awareness of frontline safety greater application of continuous quality improvement hazards may address this concern (Frankel et al., 2008; tools such as plan-do-study-act cycles could be encour- Tucker, Singer, Hayes, & Falwell, 2008). Even in hos- aged. These tools emphasize working effectively in pitals where senior managers are engaged, physician groups, use of effective team–management techniques, independence and focus on individual performance can consensus-driven decision making, and broad participa- pose significant challenges to efforts to promote group- tion of stakeholders, potentially fostering more group oriented behavior. culture (Shortell et al., 1995). Human resource practices

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Organizational Cultures That Promote Patient Safety 309

Figure 2 comparison with structural characteristics such as size or to market conditions as reflected in nurse staffing ra- Culture map for highest and lowest tios and hospital financial status—are strongly related to patient safety climate hospitals safety climate. They also suggest that organizational culture explains more of the variance in safety climate among hospitals than structural characteristics do. Al- though difficult to achieve, hospital leaders can change organizational culture over time (Schein, 1992). Altering the mix of organizational culture seems to be a potentially useful strategy for promoting safety climate and ul- timately patient safety. Our results suggest that hospitals may be able to increase safety climate by cultivating improvement-oriented teamwork and openness to innovation, leavening these with enough hierarchy to keep things under control. The lessons of HROs (which op- erate at high production and tempo while still achieving the most highly reliable results) suggest that a production orientation should be tempered by specific practices targeted to ensure safety. Several limitations of our study are worth noting. First, despite a fairly representative sample of U.S. hospitals, reasonable response rates, compensation for ex- Note. High safety climate hospitals include the five hospitals pected and actual differences in nonresponse by job (average n = 55) with the lowest PPR overall to questions on the category through oversampling and weighting adjust- PSCHO survey. Low safety climate hospitals (average n = 53) ments, and lack of systematic differences in responses include the five hospitals with the highest PPR overall. Average values for each of the organizational culture types were among survey waves, the possibility of selection bias re- calculated for both of these groups and are presented above. mains. Second, obtaining data from a single survey raises P-values compare significance of difference between high and concern regarding endogeneity, that is, that the rela- low safety climate hospitals for each culture-type. tionship found between safety climate and organizational culture was exaggerated. In practice, common method bias could inflate or deflate the apparent relationship addressing selection, orientation, training, performance (Spector, 2006). Also, the simplicity and distinctiveness appraisal, and reward systems could be designed to re- of survey items related to safety climate as compared inforce work in teams. Finally, policies might attack with culture and the observed variation in safety climate antigroup behaviors, such as tolerating abusive conduct results by type of organizational culture mitigate this toward fellow workers and ensuring no blame is attached concern. Third, although we have argued that organiza- to those who report safety or quality problems. tional culture and safety climate are related yet distinct Likewise, organizations with excessively hierarchical features of organizations, we are unable to confirm dis- cultures could develop strategies to ameliorate detrimen- criminant validity between these measures. The ipsative tal effects. HROs, such as nuclear powered aircraft nature of the organizational culture-type scale precludes carriers, have developed practices that allow hierarchy traditional factor analysis. to transform from rigid centralized control in some circumstances to flexible, distributed control in others (Roberts, 1990). Specific techniques include using Conclusion decision-making algorithms and policies to distribute authority, assigning relatively high responsibility and This study provides the first empirical evidence of a link accountability to low-level employees, and explicitly between safety climate and organizational culture. giving safety veto power to those nearest to a process Others have found relationships between both structural regardless of rank. characteristics and organizational culture and other Although this study shares limitations with other outcomes of interest in hospitals and physician orga- cross-sectional analyses, it provides some useful insights nizations (Cameron & Quinn, 1999; Carman et al., for improving patient safety climate. Our findings can be 1996; Meterko et al., 2004; Shortell et al., 1995, 2000; interpreted as good news for hospitals in their efforts to Zazzali et al., 2007). However, none have examined the improve patient safety. They imply that characteristics of link between these characteristics of organizations and organizations that are relatively mutable—at least in safety climate. The results of our study suggest that

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 310 Health Care Management REVIEW October–December 2009

general organizational culture is important to organiza- Flin, R., Burns, C., Mearns, K., Yule, S., & Robertson, E. tions’ climate of safety. Transforming hospital culture (2006). Measuring safety climate in health care. Quality & Safety in Health Care, 15(2), 109–115. may be a powerful tool to advance patient safety. Future Frankel, A., Pratt Grillo, S., Pittman, M., Thomas, E. J., research should seek to establish more persuasive evi- Horowitz, L., Page, M., et al. (2008). Revealing and dence of the link between safety climate and safety per- resolving patient safety defects: The impact of leadership formance and to identify situations in which selected walkrounds on frontline caregiver assessments of patient processes, such as implementation of health information safety. Health Services Research, 43(6), 2050–2066. technology, moderate this relationship. Gaba, D., Howard, S., & Jump, B. (1994). Production pressure in the work environment: California anesthesiologists’ attitudes and experiences. Anesthesiology, 81, 488–500. Gaba, D., Singer, S., Sinaiko, A., & Bowen, J. (2003). Dif- Acknowledgments ferences in safety climate between hospital personnel and navy aviators. Human Factors, 45(2), 173–185. The authors thank Tobias Rathgeb for research as- Gaba, D. M., Singer, S. J., & Rosen, A. K. (2007). Safety sistance and the individuals and hospitals of the Patient culture: Is the ‘‘unit’’ the right ‘‘unit of analysis’’? Critical Safety Consortium who participated in this study. Fi- Care Medicine, 35(1), 314–316. nancial support for this research was provided by the Ginsburg, L., Norton, P. G., Casebeer, A., & Lewis, S. (2005). Agency for Healthcare Research and Quality RO1 An educational intervention to enhance nurse leaders’ per- HSO13920. Preliminary results were presented at the ceptions of patient safety culture. Health Services Research, Academy Health Annual Research Meeting, June 5, 40(4), 997–1020. 2007, Orlando, Florida. Approval from all relevant in- Hartmann, C., Rosen, A., Meterko, M., Shokeen, P., Zhao, S., Singer, S., et al. (2008). An overview of patient safety climate stitutional review boards was granted before conducting in the VA. Health Services Research, 43(4), 1263–1284. this study. Hofmann, D. A., & Mark, B. (2006). An investigation of the relationship between safety climate and medication errors as well as other nurse and patient outcomes. Personnel References Psychology, 59(4), 847–869. Adler, P. S., Goldoftas, B., & Levine, D. I. (1999). Flexibil- Institute of Medicine. (2000). To err is human: Building a safer ity versus efficiency? A case study of model changeovers health system. Washington, DC: National Academy Press. in the Toyota production system. Organization Science, Kaissi, A., Kralewski, J., Dowd, B., & Heaton, A. (2007). The 10(1), 43. effect of the fit between organizational culture and struc- Ashkanasy, N., Broadfoot, L., & Falkus, S. (2000). Question- ture on medication errors in medical group practices. Health naire measures of organizational culture. In N. M. Ashkanasy, Care Management Review, 32(1), 12. C. P. M. W., & M. F. Peterson (Eds.), Handbook of orga- Klein, K. J., & Kozlowski, S. W. (2000). From micro to meso: nizational culture & climate (pp. 131–146). Thousand Oaks, Critical steps in conceptualizing and conducting multilevel CA: Sage. research. Organizational Research Methods, 3(3), 211–236. Cameron, K. S., & Quinn, R. E. (1999). Diagnosing and Meterko, M., Mohr, D., & Young, G. (2004). Teamwork changing organizational culture. Upper Saddle River, NJ: culture and patient satisfaction in hospitals. Medical Care, Prentice Hall. 42(5), 492–498. Carman, J. M., Shortell, S. M., Foster, R. W., Hughes, E. F. X., Ostroff, C., Kinicki, A. J., & Tamkins, M. M. (2003). Orga- Boerstler, H., O’Brien, J., et al. (1996). Keys for successful nizational culture and climate. In Handbook of psychology: implementation of total quality management in hospitals. Vol. 12. Industrial and organizational psychology. Hoboken, Health Care Management Review, 21(1), 48–60. NJ: John Wiley & Sons, Inc. Carroll, J., Rudolph, J., & Hatakenaka, S. (2002). Learning Quinn, R. E., & Rohrbaugh, J. (1983). A spatial model of from experience in high-hazard organizations. In B. Staw effectiveness criteria: Towards a competing values ap- & R. M. Kramer (Eds.), Research in organizational behavior proach to organizational analysis. Management Science, (Vol. 24, pp. 87–137). Oxford: Elsevier/JAI Press. 29(3), 363. Clarke, S. (2006). The relationship between safety climate Reason, J. (1990). The contribution of latent human failures and safety performance: A meta-analytic review. Journal of to the breakdown of complex systems. Philosophical Trans- Occupational Health Psychology, 11(4), 315–327. actions of the Royal Society of London. Series B, Biological Colla, J., Bracken, A., Kinney, L., & Weeks, W. (2005). Sciences, 327(1241), 475–484. Measuring patient safety climate: A review of surveys. Roberts, K. H. (1990). Some characteristics of one type of Quality & Safety in Health Care, 14(5), 364–366. high reliability organization. Organization Science, 1(2), Cooper, J. B., Blum, R. H., Carroll, J. S., Dershwitz, M., 160–177. Feinstein, D. M., Gaba, D. M., et al. (2008). Differences in Schein, E. H. (1992). Organizational culture and leadership safety climate among hospital anesthesia departments and (2nd ed.). San Francisco: Jossey-Bass. the effect of a realistic simulation-based training program. Schneider, B., & Bowen, D. (1993). The service organization: Anesthesia & Analgesia, 106(2), 574–584. Human resource management is crucial. Organizational Dy- Edmondson, A. C. (1999). Psychological safety and learning namics, 21, 39–52. behavior in work teams. Administrative Science Quarterly, Scott, T., Mannion, R., Davies, H., & Marshall, M. (2003). 44(2), 350–383. The quantitative measurement of organizational culture in

Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Organizational Cultures That Promote Patient Safety 311

health care: A review of the available instruments. Health duction to basic and advanced multilevel modeling. Thousand Services Research, 38(3), 923. Oaks, CA: SAGE Publications. Shortell, S. M., Jones, R. H., Rademaker, A. W., Gillies, R. R., Spector, P. E. (2006). Method variance in organizational Dranove, D. S., Hughes, E. F., et al. (2000). Assessing research: Truth or urban legend? Organizational Research the impact of total quality management and organizational Methods, 9(2), 221–232. culture on multiple outcomes of care for coronary artery Trice, H. M., & Beyer, J. M. (1993). The cultures of work bypass graft surgery patients. Medical Care, 38(2), 207–217. organizations. Englewood Cliffs, NJ: Prentice-Hall Inc. Shortell, S. M., O’Brien, J. L., Carman, J. M., Foster, R. W., Tucker, A., Singer, S., Hayes, J., & Falwell, A. (2008). Front- Hughes, E. F., Boerstler, H., et al. (1995). Assessing the line staff perspectives on opportunities for improving the impact of continuous quality improvement/total quality man- safety and efficiency of hospital work systems. Health agement: Concept versus implementation. Health Services Services Research, 43(5 Pt 2), 1807–1829. Research, 30(2), 377–401. Wallace, J. C., Popp, E., & Mondore, S. (2006). Safety climate Singer, S., Falwell, A., Gaba, D., & Baker, L. (2008). Hospital as a mediator between foundation climates and occupa- patient safety climate: Variation by management level. tional accidents: A group-level investigation. Journal of Medical Care, 46(11), 1149–1156. Applied Psychology, 91(3), 681. Singer, S., Gaba, D., Falwell, A., Lin, S., Hayes, J., & Baker, L. Zammuto, R., & Krakower, J. (1991). Quantitative and quali- (2009). Patient safety climate in 92 US hospitals: Differences tative studies of organizational culture. Research in Orga- by work area and discipline. Medical Care, 47(1), 23–31. nizational Change and Development, 5, 83–114. Singer, S., Gaba, D., Geppert, J., Sinaiko, A., Howard, S., & Zazzali, J. L., Alexander, J. A., Shortell, S. M., & Burns, L. R. Park, K. (2003). The culture of safety in California hos- (2007). Organizational culture and physician satisfaction pitals. Quality and Safety in Health Care, 12(2), 112–118. with dimensions of group practice. Health Services Research, Singer, S., Meterko, M., Baker, L., Gaba, D., Falwell, A., & 42, 1150–1176. Rosen, A. (2007). Workforce perceptions of hospital safety Zohar, D. (1980). Safety climate in industrial organizations: culture: Development and validation of the Patient Safety Theoretical and applied implications. Journal of Applied Climate in Healthcare Organizations survey. Health Ser- Psychology, 65(1), 96–102. vices Research, 42(5), 1999–2021. Zohar, D. (2002). Modifying supervisory practices to improve Singla, A. K., Kitch, B. T., Weissman, J. S., & Campbell, E. G. subunit safety: A leadership-based intervention model. Jour- (2006). Assessing patient safety culture: A review and syn- nal of Applied Psychology, 87(1), 156–163. thesis of the measurement tools. Journal of Patient Safety, Zohar, D., & Luria, G. (2004). Climate as a social–cognitive 2(3), 105–115. construction of supervisory safety practices: Scripts as proxy Snijders, T., & Bosker, R. (1999). Multilevel analysis: An intro- of behavior patterns. Journal of Applied Psychology, 89(2), 322.

ERRATUM

Commentary: Critical next steps in developing evidence-based management of health care organization

In the Commentary by Jane Banaszak-Holl, Kai Zheng, and John R. Griffith that was published on page 219 of the July–September 2009 issue, the e-mail address given for Jane Banaszak-Holl was incorrect. Her correct e-mail address is [email protected].

Reference Banaszak-Holl, J., Zheng, K., & Griffith, J. R. (2009). Critical next steps in developing evidence-based management of health care organizations. Health Care Manage Rev, 34, 219–221.