Doi:10.1093/geront/gnt129 The Role Of Organizational Culture In .

(Davies et al., 2007; Shortell et al., 2004; Zazzaliet al., 2007).Organizational culture reflects value systemsthat become embedded in managerial practices,work processes and ultimately patient care, and thekey values associated with a particular culture mayresonate or not with the clinical staff in a facility.For example, values key to a group culture includea focus on staff cohesion, internal communication,and individual development; these values may beparticularly important for clinical staff who managedifficult problems in patient care and depend upona team of clinical staff in order to provide excellentcare (Scott-Cawiezell et al., 2005). We subsequentlydevelop several hypotheses, given that professionalnursing norms fit with flexible work environmentsand, furthermore, that clinical professionals willreact positively to work environments that encourage their participation and engagement (Anderson,Corazzini & McDaniel, 2004; Brannon, Zinn, Mor& Davis, 2002; Donoghue & Castle, 2009). It maybe that participatory environments also make itmore difficult to fire employees who do not fit orwho have problems, and subsequently we expectthis relationship to hold regardless of whetherturnover is voluntary or involuntary (Donoghue &Castle, 2007). Subsequently, we expectHypothesis 1: Facilities with stronger group ordevelopmental cultural values will have lowerturnover rates.Furthermore, we expect that facilities withorganizational cultures emphasizing a focus onmeeting performance targets and cost controls willbe less likely to retain staff and may even encouragestaff turnover through downsizing and involuntaryfirings when staff make mistakes. Nursing staff mayquestion whether organizational performance is inthe best interest of residents and may also feel lesscommitted to organizations with relatively rigidcultural values and thus may be more likely to leavethe organization. Thus, we suggest thatHypothesis 2: Facilities with stronger hierarchical or market cultural values will have higherturnover rates.Again, we expect this relationship to holdregardless of whether turnover is voluntary orinvoluntary.MethodsFigure 1. Competing values framework dimensions and cultural types.The nursing homes included in this studyhad been surveyed previously in 2005 and 2006Page 3 of 11Downloaded from http://gerontologist.oxfordjournals.org/ at University of Michigan on March 21, 2014new opportunities). The stability versus flexibilitydimension indicates whether managers enforceadherence to the status quo (such as enforcingexisting rules or processes) or whether they provide flexibility in adapting work to immediateneeds and the changing environment (Zuckerman,Dowling & Richardson, 2000).Juxtaposing the two value dimensions withinthe CVF results in four competing or diametricallyopposed cultural types: a group culture, a developmental culture, a hierarchical culture, and a marketculture. Group cultures (high internal focus andflexibility) have participative leaders, members motivated by membership and attachment to the group,and emphasize member development and commitment. Developmental cultures (high external focusand flexibility) have entrepreneurial leaders, members motivated by growth and creativity needs, andemphasize growth and resource acquisition. Marketcultures (high external focus and stability) havedirective leaders and members who are motivatedby competition, and emphasize productivity and efficiency. Hierarchical cultures (high internal focus andstability) have conservative leaders, members motivated to follow rules and maintain order, and a focuson control, and efficiency as markers of effectiveness.Figure 1 shows the juxtaposition of the CVF’stwo dimensions and four cultural types. A focus on“competing” values implies that managers grapplewith conflicting priorities although scholars nowrecognize that high performing organizations maypromote multiple and sometimes conflicting values

facility characteristics affected response rates forthis survey round using available OSCAR data.We found that respondents and nonrespondentsdiffered on a few variables (Banaszak-Holl et al.,2013), with respondent facilities more likely tocome from geographic areas with lower per capitaincomes, somewhat less likely to be for profit andreported slightly higher occupancy rates. In addition, Appendix A (Supplementary Material) compares our responding facilities to the U.S. nursinghome population in 2009, which shows somestatistically significant differences between thesegroups: responding facilities were more likely to benonprofit and had more beds and higher occupancyand fewer Medicare residents than U.S. nursinghomes overall. These differences limit the generalizability of our study as we discuss in limitations.Variable ConstructsThis study uses turnover measures comparableto those developed in the National Nursing HomeTurnover Study or NNHTS (Castle, Engberg,Anderson, & Men, 2007). Traditionally, much ofthe research on staff turnover has used small samples of facilities; the NNHTS is one exception inwhich researchers collected turnover data from alarge number of facilities across multiple time points(Castle, 2008). The NNHTS and other large surveyshave revealed that collecting turnover data can bedifficult; although, NHAs track turnover data routinely through administrative processes. We interviewed several NHAs when designing the survey,who reported tracking the percent staff turnoveradministratively. Subsequently, we asked NHAs toidentify the percent of staff who leave the organization within 12- and 6-month periods for staffingcategories of Registered Nurses (RNs), LicensedPractice Nurses (LPNs), and nurse aides (NAs)separately. These questions differ slightly from theoriginal NNHTS questions in which NHAs wereasked to report the number of staff experiencingturnover. There are no published results comparing reporting methods using either the number orpercent of staff who leave, and we recognize thatcomparisons to the NNHTS are limited.The NNHTS developed a method of requesting turnover data separately for RNs, LPNs, andNAs in several forms, including 6- and 12-monthreports and overall as well as within categories offull and part-time staff and for contract staff separately. These details provide some indication ofhow turnover rates vary within as well as acrossPage 4 of 11The GerontologistDownloaded from http://gerontologist.oxfordjournals.org/ at University of Michigan on March 21, 2014regarding their turnover rates (Castle, 2008). Weselected a random sample of the nursing homes previously surveyed, and surveys were sent to them inthe period from late fall 2008 through early 2009.The original sample was nationally representative of U.S. nursing home facilities based upon acomparison to the Online Survey Certificationand Reporting (OSCAR) database. In this round,surveys were mailed to both the nursing homeadministrators (NHAs) and Directors of Nursing(DONs) in facilities. The NHA and DON weresurveyed regarding organizational culture becausethey play a primary role in the creation of organizational values and managing human resources(Davies et al., 2007; Banaszak-Holl et al., 2010).NHAs were also asked to report staff turnover fortheir facilities. We limited survey length in orderto minimize response burden because previousresearch has shown that response rates amongnursing home staff are not high. Human-subjectinvolvement was approved by the InstitutionalReview Board Committees at the Universities ofMichigan and Pittsburgh.Our follow-up and incentive payment methods were based on Dillman’s (2000) recommendations that token payments, representations ofgoodwill, and rigorous follow-up methods are themost effective tools for reducing respondent loss(Dillman, 2000). We included a 5 gift card as atoken of appreciation with all mailed surveys. Ofthe contacted facilities, 37 had closed since the lastcontact and were thus dropped from the study.Telephone call reminders were made to nonrespondents in the four months following the initialsurvey mailing and a second mailing was sent toall nonrespondents. Excluding the closed facilities, our final sample came from 1,056 facilities or2,112 contacted individuals. Among individuals,38.7% or 817 responded to the survey, including419 NHAs and 398 DONs. Results are reportedhere for the 419 facilities with NHA responses.We matched survey data to the 2009 OSCARdata, which includes all nursing home facilitiesthat are Medicaid and Medicare certified. OSCARis the most comprehensive source of facility-levelinformation on nursing homes and from thesedata, we drew information on organizational characteristics including facility size and ownershipand aggregate resident mix. Past research basedon the earlier sample found that despite lower participation among poorer quality facilities, nonresponse did not jeopardize generalizability (Castle& Engberg, 2006). We further examined whether

adjust for voluntary turnover (Castle, 2006). Wereplaced turnover rates with the estimated voluntary turnover rates when available and found somechanges in the relationship between culture andturnover, as reported and discussed in sensitivityanalyses.Culture Measures.—Reliable and valid CVF survey instruments use a set of value statements for eachcultural type to which respondents assign priority orimportance (Yeung et al., 1991; Scott et al., 2003). TheCVF measures have been demonstrated to have reliability of the order of .70 or higher using Cronbach’salpha and the validity of these measures has beenestablished in other industries using multitrait, multimethod analyses (Spreitzer & Quinn, 1991). Each of thefour cultural values in the CVF is identified using fourLikert-scaled questions. In these questions, respondents were asked to rate on a scale from 1 (strongly disagree) to 7 (strongly agree) whether “possible values areoperating and emphasized in your nursing home as awhole.” Subsequently, a nursing home may score highon more than one cultural type. For example, a nursing home administrator may evaluate values of botha group and developmental culture as strong withintheir facility.For a group culture, respondents were askedabout values emphasizing: (a) human relations,teamwork, and cohesion; (b) employee concernsand ideas; (c) participation and open discussion;and (d) morale. For a developmental culture, values included: (a) innovation and change, (b) newideas, (c) creative problem solving, and (d) decentralization. For a market culture, values included:(a) outcome excellence and quality, (b) getting thejob done, (c) goal achievement and (d) doing one’sbest. For a hierarchical culture, values included: (a)order, (b) stability and continuity, (c) dependabilityand reliability, and (d) predictable outcomes.Relative strength of culture, and whether havinga predominantly stronger culture in one domainaffects turnover. Reports of the average values forfacilities on the cultural items have been reportedelsewhere (Banaszak-Holl et al., 2013).Controls.—Control variables include those fromour survey, OSCAR data, and American Health CareAssociation (AHCA) reports of the average 2007 turnover rates by state. OSCAR measures include whetherthe facility is chain owned or for profit (both codedas dummy variables), whether the facility is hospitalbased, number of beds in the facility, occupancy ratePage 5 of 11Downloaded from http://gerontologist.oxfordjournals.org/ at University of Michigan on March 21, 2014facilities. In this study, we found NHAs reluctantto provide detailed turnover information across allthe subcategories of staffing, and subsequently, wereport analyses using overall 12-month turnovermeasures. For example, administrators were muchless likely to provide 6-month turnover rates than12-month turnover rates in our survey. In examining the impact of culture, we look at the effectson the estimated count of staff leaving during the12-month window, by multiplying reports of percent turnover by reported Full-time Equivalents(FTEs) within staff categories.We have compared the turnover rates from ourstudy with a number of earlier studies over the lastdecade, including those reported by Donoghue(2010), Castle (2008), and Donoghue and Castle(2009). Our turnover rates are consistent withreports in these studies, which all used nationalsampling frameworks. For example, our estimatedturnover rates are within 10% of those from 2007(Donoghue and Castle, 2009), 2 years previous.Overall, comparisons across these studies appearto indicate a slight downward trend in staff turnover across the decade; although, a full meta-analysis is needed to make conclusive statements abouttrends.We did find that NHAs from facilities with few(i.e., 5) staff in any particular nursing category(either RN, LPN, or NA) were more likely to reportrelatively small percentages of staff turnover (e.g.,in a facility with only one RN, the 12-month turnover rate may be reported as low as 1%). To adjustfor potential underreporting of turnover within thefacilities with the fewest staff, we adjusted reportsof percent turnover by a factor of 10 when staffing in a category fell below 5 employees. In doingthis, we sought to realistically adjust the turnoverreports upwards while still using count models topredict the impact of culture. We have tried alternative methods for analyzing these data. For example, turnover was categorized by quartile, into thelowest, second lowest, second highest, and highestquartiles in order to minimize assumptions aboutthe data distribution, and then quartile membership was predicted using an ordinal logistic model;the effects of culture were similar (these resultsare available from the authors upon request). Theadjusted counts are reported because they providea meaningful way to discuss the impact of cultureon differences in turnover.Although only 22% of responding NHAs provided separate estimates for voluntary turnover,past research has argued that it is important to

Methods of Analysis.—Predictors of the numberof staff leaving in a 12-month window are modeledusing negative binomial count models (Cameron &Trivedi, 1998). Models were estimated using STATA’snbreg procedure (see Long & Freese, 2003), and inresults, we report several measures of model goodness of fit, including the overdispersion coefficient andthe Log Ratio Chi-square test. Parameter effects fromthe multivariate models are reported as incidence riskratio (IRRs). Sample size is fewer than the 419 original respondents because several NHAs did not reportstaff FTEs or other measures; in the tables, sample sizeis reported. We also ran multivariate models using theoriginal questions on percent staff turnover to reducemissing data, but the general significance and directionof parameter estimates did not change. Here, we reportmodels predicting numbers of turnovers.FindingsTable 1 reports descriptive statistics on the turnover measures and adjustments to these measures asused within this study. We highlight the differenceshere because (a) multiple definitions of turnoverhave been used in past studies and (b) sensitivityanalyses show our results are fairly robust whenwe adjust turnover measures for reporting issues.Additional descriptive information on the independent variables from our model is included inAppendix B (Supplementary Material). The percentof nursing aide (NA) staff leaving in a 12-monthperiod was 35.8% compared with 19.5% of LPNsand 19.3% of RNs (column 1 of Table 1). Acrossall staff categories, the standard deviation for thepercent turnover is greater than the mean, indicating significant overdispersion in turnover ratesamong facilities. There are a number of facilitiesthat experience no turnover, the minimum possible; and at the other extreme, the maximum turnover rates are quite high, with the highest for NAsat approximately 180% turnover (over 200 staffat that facility), for LPNs at 120%, and for RNsat 188%. These rates may seem high on averagebut are consistent with prior research, which hasshown that staff turnover rates are quite high inthe industry but also vary considerably (Castle,2008). Adjustments made to turnover rates, forfacilities with the fewest staff present, increasedTable 1. Descriptive Statistics for Reported and Adjusted Staff Turnover RatesEst. number staff inMaxPercent turnoverfrom surveyBased on NHA responseAdj when staff smallAdj when 15.722.40213.1aAdjustment for voluntary turnover in addition to corrections for facilities with staffing 5.Page 6 of 11The GerontologistDownloaded from http://gerontologist.oxfordjournals.org/ at University of Michigan on March 21, 2014defined as number of residents relative to the numberof beds, and the percent of residents within the facility covered by Medicare, Medicaid (private pay is usedas excluded category). We also controlled for whethera facility routinely tracked turnover which preliminary analyses showed was related to higher turnoverrates. In our survey, 86% tracked turnover informationroutinely, with 52% using computer systems and 43%using written records, and some facilities using bothmethods. We originally planned to control for NHAand DON tenure at a facility, as measured by whetherthey had been there 2 years or less, but these variablesdid not significantly affect results. Finally, we includeda control for historical rates of turnover within a facility’s state using the average state turnover rates in 2007as reported by the AHCA (2008).

turnover rates and no cultural factor was predictive in the model of NA turnover.These results partially support our hypothesesregarding the impact of cultural values on turnover rates. In particular, we find that Hypothesis 1proposing that facilities with stronger group anddevelopmental cultures have lower turnover ratesholds only among LPNs and not for RNs or NAs.And, although Hypothesis 2 proposed that facilities with stronger hierarchical cultures would havehigher turnover rates, we found the opposite inmodels of RN turnover, and the relationship wasnot supported for either LPN or NA turnoverrates. Furthermore, hypothesis 2 proposed furtherthat market values would increase turnover rates,which was true for RNs and LPNs only.Among the control variables included in Table 2,the number of beds in the facility, occupancy rate, andthe state average turnover rate in 2007 had statisticallysignificant effects on the turnover rates of LPNs andNAs, whereas the tracking of turnover significantlypredicted NA turnover. The overdispersion coefficient, alpha, is significantly greater than zero and statistically significant in all of our models; furthermore,the χ2 statistic is highly significant indicating that ingeneral our models produce a better fit than one inwhich these factors do not predict turnover rates.Table 2. Negative Binomial Models Predicting Staff Turnover RatesInterceptTracks turnoverChain-ownedFor-profit# BedsOccupancy RateProp. MedicareProp. MedicaidHospital-basedState Avg07 turnoverCultural value scalesGroupDevelopmentalMarketHierarchicalModel statisticsα(Overdispersion coeffic)Likelihood χ2 (12)p-value χ2Observations(1)(2)(3)RN turnoverLPN turnoverNursing aide turnoverRate ratioSERate ratioSERate 8)(.023)(.042)(.030)(.094)***Notes: We use RN for Registered Nurse and LPN for Licensed Practice Nurse in our findings. Tests of significance: *p .10,**p .05, ***p .01Page 7 of 11Downloaded from http://gerontologist.oxfordjournals.org/ at University of Michigan on March 21, 2014estimated numbers of staff leaving and had the biggest impact on RN turnover rates. Adjusting nextto voluntary turnover rates reduces the turnoverrates across categories but has the biggest impacton the estimated NA turnover. This is not surprising. Because, as largely unskilled workers, NAsexperience substantially more involuntary turnover than either RNs or LPNs, and eliminating theseevents from our counts affects the NA turnovermore than LPN or RN turnover rates.Table 2 reports the IRR from multivariate negative binomial models predicting the effects ofculture and facility-level contro

The Role of Organizational Culture Developing organizational culture is a basic managerial tool for improving the work envi ronment by emphasizing core values necessary for individual and organizational effectiveness. Organizational culture is closely related to but should not be confused as equivalent to the con-cept of organizational climate.