The Gerontologist Vol. 42, No. 2, 159–168 Copyright 2002 by The Gerontological Society of America An Exploration of Job, Organizational, and Environmental Factors Associated With High and Low Nursing Assistant Turnover Diane Brannon, PhD,1 Jacqueline S. Zinn, PhD,2 Vincent Mor, PhD,3 and Jullet Davis, PhD4 Purpose: This article examines factors that distinguish nursing facilities with very high and very low nursing assistant turnover rates from a middle referent group, exploring the possibility that high and low turnover are discrete phenomena with different antecedents. Design and Methods: Data from a stratified sample of facilities in eight states, with directors of nursing as respondents (N 5 288), were merged with facility-level indicators from the On-Line Survey Certification of Automated Records and county-level data from the Area Resource File. Multinominal logistic regression was used to identify factors associated with low (less than 6.6% in 6 months) and high (more than 64% in 6 months) turnover rates. Results: With the exception of registered nurse turnover rate, low turnover and high turnover were not associated with the same factors. Implications: Future studies of facility turnover should avoid modeling turnover as a linear function of a single set of predictors in order to provide clearer recommendations for practice. Cohen-Mansfield, 1997; Helmer, Olson, & Heim, 1993; Mesirow, Klopp, & Olson, 1998). This trend, coupled with increased regulatory and market pressures to improve quality, has prompted industry observers to list maintaining a high-quality nonprofessional staff as a major strategic challenge as the demand for various levels of long-term care continues to rise (Kodner, 1993). Nursing assistant wages and benefits continue to reflect the marginality of chronic caregiving work within the health care system and within society at large (Crown, 1994), and the turnover problem appears to be as salient as it has ever been. In the United States, interventions to improve the quality of the nonprofessional long-term care workforce have focused on training. Since 1991, certification training has been in place, presumably raising the minimum standard of care in nursing homes. It is also possible that an unintended consequence of this certification has been enhanced occupational definition that extends beyond the organizational boundaries of the employer, making these certified staff attractive to other, better paying health care employers. More recently, specialized training (Grant, Kane, Potthoff, & Ryden, 1996; McCallion, Toseland, Lacey, & Banks, 1999), improved preemployment screening (Kettlitz, Zbib, & Motwani, 1997), and a primary-care job design approach (Teresi et al., 1993) have shown promising results. Overall, however, the accumulated findings of research on staff turnover in nursing homes have produced little by way of empirically reliable guidance on remedying the problem. The findings reported in this article are from a multistate study of organizational and environmental factors that are associated with aggregate resident outcomes in a representative sample of 308 nursing facilities. Although not designed to explain individual turnover decisions, the data do allow for a novel, categorical exploration of very high and very low facility turnover, drawing on factors identified in prior work to be correlated with facility turnover rates. The purpose of the analysis was to explore factors from previously established lists of turnover correlates that Key Words: Nursing assistant turnover, Staff turnover Despite two decades of industry and policy research focus on them, persistent staff shortages and turnover rates several times the national average continued to be reported throughout the 1990s (Anderson, Issel, & McDaniel, 1996; Banaszak-Holl & Hines, 1996; Bayer, 1994; Caudill & Patrick, 1991; This work was supported by National Institute on Aging Grants AG11624 and AG00048. Address correspondence to Diane Brannon, Department of Health Policy and Administration, The Pennsylvania State University, 116 Henderson Building, University Park, PA 16802. E-mail: [email protected] 1Department of Health Policy and Administration, The Pennsylvania State University, University Park. 2Department of Risk, Insurance and Healthcare Management, Temple University, Philadelphia, Pennsylvania. 3Bio Med Gerontology Health, Brown University, Providence, Rhode Island. 4Department of Marketing and Management, University of Alabama, Tuscaloosa. Vol. 42, No. 2, 2002 159 mix, and the ratio of administrative costs to clinical costs. Both of these studies and the previous studies summarized by Cohen-Mansfield (1997) indicated that in addition to larger labor market contexts, organizational and job factors, some of which are controllable by management, are influential vis á vis nonprofessional staff turnover. Similarly, research on turnover in general and in the health professions in particular has identified three broad categories of factors that influence turnover (Cotton & Tuttle, 1986)—environmental, job and organizational, and personal factors. Environmental factors include the availability of alternative employment, the local unemployment rate, and general economic conditions. Job and organizational factors include pay (although this appears to be more influential among managerial/professional employees than with others), job stress, role clarity and job design, and leadership effectiveness, all of which influence turnover through job satisfaction. Personal factors shown to be reliably related to turnover behavior include age, education, marital status, and number of dependents supported. distinguish low and high facility turnover levels from a middle referent group, an approach that may open new avenues of inquiry regarding potential solutions to this persistent problem. Prior Research Although extensive research has been conducted in testing both individual-level, behavioral models of the intent to leave and turnover (cf. Mobley, Griffith, Hand, & Meglino, 1979) and aggregate industrylevel models reflecting labor economic approaches (cf. Terborg & Lee, 1984), the role of the organization itself and its immediate environment in influencing turnover has been explored only more recently. Sheridan, White, and Fairchild (1992), for example, estimated that differences in organizational culture explained more than $6 million difference in turnover-related costs among six accounting firms. The application of these models in pink-collar industries—that is, predominantly female and nonprofessional—is rare even though turnover rates tend to be higher among women (see Cotton & Tuttle, 1986, for a meta-analytic review). One notable exception is the study by Schaefer and Moos (1996), which examined the role of work stressors and environment on longterm care staff. Although there have been several studies of staff turnover in nursing homes (Caudill & Patrick, 1991; Wagnild, 1988; Waxman, Carner, & Berkenstock, 1984), two recent ones are noteworthy for their sample sizes and for examining this phenomenon using multivariate analyses to test theoretically derived hypotheses. Banaszak-Holl and Hines (1996) analyzed survey data from 250 facilities in 10 states for the purpose of examining which job design and other organizational factors affected turnover, controlling for local economic conditions. They observed that turnover was higher in for-profit homes, confirming earlier findings (Waxman et al., 1984), and that facilities that involved nursing assistants in resident care planning, a core clinical organizing task in nursing homes, experienced lower turnover rates. Training and workload issues, case mix severity of the residents, payor source mix, and facility size were not significantly related to nursing assistant turnover, but labor market conditions were significant predictors. Anderson and colleagues (1996) reported on a study of differential predictors of registered nurse (RN), licensed practical nurse (LPN), and nursing assistant turnover in 467 Texas facilities. Consistent with findings from studies in service industries (Cotton & Tuttle, 1986), including Halbur’s work in nursing homes (Halbur & Fears, 1986), although in contrast to Banaszak-Holl and Hines (1996), nursing assistant turnover was not significantly affected by labor market conditions, whereas turnover of professional nurses was. Rather, the nonprofessional staff turnover rates were positively related to organizational factors including for-profit status, facility occupancy rate, and workload, and negatively related to factors including profit margin, nursing staff skill Costs and Benefits of Turnover The disadvantages of turnover from the organizational perspective include the following: replacement costs (including training), lost productivity, compromised quality, and lowered morale. In nursing homes, these costs may well be reflected in the quality of care that residents receive (Burnfeind & O’Connor, 1992; Cohen-Mansfield, 1997; Wagnild, 1988; Waxman et al., 1984). Burnfeind and O’Connor (1992) reported that facilities with high turnover rates had more code violations resulting from the state survey process. It is argued that providing consistent care in a setting that demands intense levels of staff resident interaction places a premium on staff retention (Banaszak-Holl & Hines, 1996; Cohen-Mansfield, 1997). Although excessive turnover is clearly a problem that has continued to plague the long-term care sector, another view found in the literature, although rarely cited in studying this problem, is that not all turnover is bad. Accordingly, a 1% increase in turnover in a facility with very few staff terminations is likely not the same phenomenon—in terms of either antecedents or consequences—as a 1% increase in a high-turnover facility. Abelson (1986) and Iglehart (1990) have argued for health and human services managers to seek their optimal level of turnover rather than assuming all turnover is equally dysfunctional. They have each argued that organizations have optimal aggregate turnover rates that reflect necessary involuntary as well as voluntary terminations and retention costs as well as turnover costs. They further argued that the optimal rate for a given organization or type of organization is almost always greater than zero, given imperfect employee selection practices. Selection ratios and practices, Shaw, Delery, Jenkins, and Gupta (1998) argued, have a disproportionate impact on involuntary 160 The Gerontologist Methods turnover, whereas human resources practices have more influence on voluntary turnover. Given that most nursing facilities recruit from very small pools of applicants, creating unfavorable selection ratios, some turnover is likely needed to weed out hiring mistakes. From the point of view of organizational performance in general, functional aspects of turnover can include the opportunity to reduce costs by replacing higher paid workers with lower paid workers, renewing the workforce through increased demographic diversity, weeding out poor performers, and avoiding retention costs related to promotion, transfers, and so forth. The financial benefits to the nursing facility can quickly be discounted from the perspective of continuity and quality of care, however. Frail and often disoriented elders need consistent care. Furthermore, it has been argued that there are “group learning curves” (Argote et al., 1995) that are adversely affected by turnover. Stryker (1982), however, argued two decades ago that quality of care in nursing facilities would likely not be well served by zero or nearzero rates of turnover. She viewed workforce renewal as an investment in good care. In addition, Halbur and Fears (1986) found that RN turnover positively predicted nursing home resident discharge rates, and they questioned the presumed negative consequences of personnel turnover. Kettlitz and colleagues (1997) observed that poor selection practices are a significant source of nurse’s aide turnover. Logically, there are individuals hired by facilities as nursing assistants, either before or after completion of the mandated training, who are not well suited to this demanding work. In these cases, turnover may be positive in terms of quality. If it takes a facility 10 hires from a limited applicant pool to find a committed and skilled caregiver, does that make this facility less effective than one that retains the first available “warm body”? Thus, the question arises, is the workforce stability implied by very low turnover rates necessarily a desirable goal for nursing facility management? Can a functional range of turnover rates be identified for industry benchmarking purposes that will enable managers to plan for what Haveman (1995) referred to as a “demographic metabolism of organizations” (p. 586) that balances industry dynamics, turnover, and tenure distributions? This question has particular salience in today’s nursing home industry. After several decades of stable funding and relatively little environmental turbulence, clinical and administrative innovation and management of complexity are now survival skills for nursing homes. Personnel who are wedded to established, more custodial practices may be a liability as nursing facilities must respond to new clinical demands and prove the cost effectiveness of their services. Ignoring these questions about the need for workforce renewal and quality through maintenance of some level of attrition or, in research terms, assuming that turnover rate is a linear function of a given set of predictors, may be clouding our understanding of how to manage this problem. Vol. 42, No. 2, 2002 Sample In this article, we report data from a project (National Institute on Aging Grant AG11624-01A1) designed to examine how organizational structure and administrative and clinical processes affect nursing home outcomes. A sample of 360 facilities, stratified by ownership, size, and urban location, was drawn from eight states that were universal in their computerization of the mandated nursing home Resident Assessment Instrument/Minimum Data Set during 1995–1996. An 80% response rate was achieved. Data Sources Three sources of information, including primary and secondary data, were merged to form the analytic database for this analysis: The 1995 On-Line Survey Certification of Automated Records (OSCAR) file maintained by the Health Care Financing Administration, the most closely matched county-level Area Resource File (ARF), and a cross-sectional survey of directors of nursing (DON) of the sample facilities. The OSCAR data and the ARF were used to obtain background characteristics of participating homes. The OSCAR contains facility and aggregated resident data routinely collected through the nursing home certification process and includes data on staffing, resident mix, sex, ownership, and average census. The ARF describes the health-related resources in the environment in which the facility is located, summarizing a large array of census, health, and social resource data. DON in all homes were interviewed by telephone during late 1995 and early 1996 to characterize the internal management structure, participation patterns, and responses to changes in the environment as well as to characterize the nursing care processes in place to guide staff regarding care planning and service delivery. Those job, organizational, and environmental factors in the data set that had been shown to influence turnover in the previous literature were examined in bivariate analyses, and those that showed some correlation (p 5 .20) with turnover rate were included in the multivariate analyses. The Low/High Turnover Model As the overall study was not designed to study turnover, no one theoretical model could be applied. Rather, we examined job, organizational, and environmental factors associated with both low and high turnover rates among nonprofessional nursing assistants at the facility level in a multistate sample. The Cotton and Tuttle (1986) organizing framework of individual, job, and organizational and environmental factors that are associated with turnover is invoked. No individual-level data were included in this facility-level data set, therefore personal, demographic, and wage rate factors were not included. Table 1 describes the variables listed in conceptual configurations under the broad categories of job, organizational, 161 Table 1. Variables, Data Sources, and Their Measurement and Means Variable Turnover Categories Low (DON) Referent (DON) High (DON) Job Factors Leadership Close supervision (DON) Trained supervisors (DON) Performance rewards (DON) Integration Involvement in assessment (DON) Involvement in care planning (DON) Rotating assignments (DON) Workload No. aides/100 beds (OSCAR) Organizational Factors Clinical resources NP (DON) No. aides/RNs & LPNs (OSCAR) RN turnover rate (DON) DON tenure (DON) Nurse aide training (DON) Administrative resources Hospital relationship (DON) Training site (DON) Investor-owned vs. nonprofit (OSCAR) Chain membership (OSCAR) % Medicaid (OSCAR) Administrator span of control (DON) Nurse’s aide union (DON) Recent owner change (OSCAR) Facility size (OSCAR) Investor-owned by nurse’s aide union Investor-owned by chain membership Clinical demands % bladder incontinent (OSCAR) % ambulatory (OSCAR) Emergency room admit rate (OSCAR) Market Environment Factors Log co. unemployment rate (1.8–12.9) (ARF) Co. per capita income (ARF) Nursing facility market concentration (ARF) Measures 0–10th percentile 11th–75th percentile 75th–100th percentile Charge nurse disciplines, 1 5 yes Any management training, 1 5 yes Bonus, 1 5 yes M (SD) #6.6%a 7.0–63%a 64–300%a .73 (.44) .74 (.44) .41 (.49) Nine domains (3-point scale)—3–27 (low) Yes 5 1 Same resident ,1 month to 51 month 19.33 (3.18) .40 (.49) .43 (.49) Aide FTEs/100 beds 41.10 (33.0) NP on site, 1 5 yes Aide FTEs/RN FTEs 1 LPN FTEs No. terminations (6 months)/No. current RN positions ,2 years 5 1 Log hours .08 (.28) 2.49 (1.3) 21.7 (32.9) .31 (.46) 2.37 (.92) Formal relationship 5 1 A clinical training site 5 1 Investor owned 5 1 Chain 5 1 % residents Medicaid No. departments reporting to administration Any aides union 5 1 Since last survey 5 1 Four categories (60, 90, 120, 120) .59 (.49) .64 (.47) .59 (.49) .44 (.49) 66.23 (21.61) 7.08 (2.13) .19 (.39) .18 (.38) 2.63 (1.12) % current residents % current residents No. resident emergency room admits/100 residents 50.69 (16.92) 45.99 (20.45) 9.77 (9.3) 1995 % 1995 $ (10 intervals: $10,653–$49,197) Herfindahl index 6.78 (2.11) 18,740 (45.05) .18 (.22) Note: DON 5 directors of nursing; OSCAR 5 On-Line Survey Certification of Automated Records; FTE 5 full-time equivalent; NP 5 nurse practitioner; RN 5 registered nurse; LPN 5 licensed practical nurse; ARF 5 area resource file. aIndicates range rather than mean. differ from those associated with very high turnover, ordinary least squares regression would obscure such differences. Consequently, we opted to treat turnover as a three-category discrete variable with low and high categories compared against a middle referent category. This approach is exploratory in that there is no prior body of work to guide the development of directional hypotheses for very low versus very high turnover. Here we assumed that neither very high nor very low turnover is desirable and used multinomial logit analysis to identify which predictor variables from a set of six configurations distinguish very low and very high turnover facilities from those in the middle referent category. and market environment factors. These configurations were not intended to represent underlying constructs (i.e., leadership); rather, they served as an organizing framework for this exploratory analysis. Although this work builds on previous conceptualizations in terms of using conceptual configurations of independent variables (Shaw et al., 1998) and on recent work in the nursing home industry by Banaszak-Holl and Hines (1996) and Anderson and associates (1996), the primary new contribution of this study is in the way turnover is operationalized and modeled. If both very low turnover and very high turnover are potentially problematic, then linear regression models can be uninformative. Similarly, if the factors associated with very low turnover 162 The Gerontologist Job Factors.—A limited number of job factors known to influence turnover were available in this data set. These were configured as leadership, integration, and workload. Leader behavior has been consistently shown to affect turnover (Cotton & Tuttle, 1986). As no direct measures of perceived leadership attributes were available, we chose several indicators from the DON questionnaire that we believed distinguished facilities that had a well-articulated structure for supervision of aides from those that did not. The lack of middle management capacity in nursing has been previously noted (Smyer, Brannon, & Cohn, 1992). Lines of authority are often unclear, and poorly managed reward systems can lead to perceptions of inequity and dissatisfaction (Brannon & Streit, 1991). The indicators chosen in this study were whether the charge nurses (the closest supervisory personnel to the aides) applied discipline when it was warranted, whether any of supervising nursing staff had received any management training, and whether performance-based bonus rewards were given to individuals. Again, it is useful to note that these were indicators of managerial structure rather than of leadership effectiveness per se. administrative resources on aggregate facility turnover rates, with clinical resources being more influential in predicting nurses’ turnover than that of nurse’s aides. Several items were included here in measuring clinical resources. These included whether the facility had a nurse practitioner. This clinical resource is viewed as needed support for maintaining role clarity, setting standards, and providing guidance for nurse’s aides (Anderson et al., 1996; Monahan & McCarthy, 1992). Also, following Anderson and colleagues’ (1996) logic, we examined the skill mix (the ratio of aides to RNs and LPNs). Professional staff turnover has been linked to nonprofessional staff turnover in nursing facilities (Anderson et al., 1996; Halbur & Fears, 1986); consequently, we included the turnover rate for RNs for the facility and whether the DON had been hired within the past 2 years. Finally, we included the measure of the number of inservice training programs offered by the facility for its nursing assistants during the previous year. Hypothesis 4: Facilities with a nurse practitioner, with a lower aide-to-professional-staff ratio, with lower RN turnover, with a longer tenured director of nursing, or with more inservice offering for aides will be less likely to report very high or very low turnover. Hypothesis 1: Facilities reporting close supervision of nursing assistants, management training for supervisors, or performance-based rewards will be less likely to report very high or very low turnover. Administrative resources served as proxy for the pay, benefits, and organizational culture variables that are influential in turnover but were not measured in this study. Larger organizations have been reported to have higher turnover rates (Cotton & Tuttle, 1986), as have investor-owned nursing facilities (Anderson et al., 1996; Banaszak-Holl & Hines, 1996). Union contracts, probably through their effects on pay, benefits, and job security, tend to be associated with lower turnover (Cotton & Tuttle, 1986). Cotton and Tuttle also noted, however, that the union effect was moderated by the nature of the firm, with service organizations’ turnover rates being less consistently influenced by the presence of a union. Whether the facility has a formal relationship with a hospital or a multifacility chain and whether it is a clinical training site speak to its reputation and legitimacy, factors also shown to influence individual turnover decisions (Cotton & Tuttle, 1986). Banaszak-Holl and Hines found that the proportion of residents whose care was paid for at least in part by Medicaid was a factor in predicting aggregate turnover rates, presumably due to financial constraints imposed by lower per resident reimbursement. Finally, we viewed administrative span of control, or the number of departments reporting directly to the top administrator, as an inverse indicator of managerial resources. The assumption here was that the traditional flat bureaucracy that once typified the industry may be inadequate for managing such organizations in the contemporary environment. Because human resource management, for example, falls within the purview of the administrator in most nursing facilities, this function has a better chance of being reasonably well developed when the administrator is not directly overseeing all or most departments. Finally, we were Integration into decision making has been shown to affect turnover (Cotton & Tuttle, 1986) and has been reported as a reason why nurse’s aides quit (Caudill & Patrick, 1991). Job designs that promote nurse’s aide integration or participation in clinical decision making have been suggested for years as a potentially powerful way to improve the quality of their work (Smyer et al., 1992; Waxman et al., 1984). Banaszak-Holl and Hines (1996), however, found that only one of four aspects of job design, participation in care planning, was a significant predictor of turnover for nurse’s aides. Hypothesis 2: Facilities that involve nursing assistants in resident assessments and care planning or provide stable resident assignments (not rotating) will be less likely to report very high or low turnover. Workload was the third job factor included in these analyses. Staffing ratios have become a common process-of-care measure in nursing homes. We presumed that caregiving staff are more susceptible to emotional and physical burnout when they have higher patient loads (Banaszak-Holl & Hines, 1996; Wagnild, 1988). Hypothesis 3: Facilities that maintain higher nursing assistant-to-bed ratios will be less likely to report very high or very low turnover. Organizational Factors.—Organizational factors were configured as clinical and administrative resources and clinical demands. Anderson and associates (1996) found differential effects of clinical and Vol. 42, No. 2, 2002 163 Results interested in whether a change of ownership would influence turnover because ownership turnover has been considerable in this industry and may result in disruptions in collective bargaining agreements and other personnel policies. A wide range of turnover rates was reported by this sample. Extreme cases, those of more than 100% for the 6-month period, were discussed with the DON in the telephone interviews and were assumed to be accurate. The mean for the sample was 51% for a 6-month period, with a standard deviation of almost 54%. This mean rate is higher than the mean 6month rate of 32% reported by Banaszak-Holl and Hines (1996), but not as high as the annual mean rate of 179% observed by Anderson and associates (1996) in Texas facilities. The distribution of facility turnover rates was skewed, with half the facilities reporting 33% or less turnover in a 6-month period. Twenty-five percent of the homes reported 15% or less turnover among their nursing assistants during the previous 6-month period, and 18% reported 10% or less. At the high end, 25% of the facilities experienced more than 63% turnover, and 15% reported more than 90% in 6 months. As no guidelines for establishing cut-off points for dysfunctional aggregate turnover rates were found in the literature or in consultation with industry trade associations, the creation of categories of high and low turnover with a middle referent category was necessarily arbitrary. As the argument for very low turnover being dysfunctional is somewhat novel and thus exploratory, we used the relatively extreme cut-off point of the 10th percentile of the sample turnover rates to create the low category. The result was that this category included 30 facilities with a range of reported turnover rates from 0% to 6.6% during the previous 6 months. Thus, although the group was arbitrarily defined, it was unequivocally composed of low-turnover nursing facilities. Because the distribution was considerably more spread out at the high end, however, we opted for a more inclusive cut-off point and used the 75th percentile. This resulted in a grouping of those 74 facilities that reported greater than 63% turnover in the previous 6 months. The range of turnover rates for these facilities was 64% to 300%. As with the low group, although the cut-off point was arbitrary and reasonable arguments could be made for setting it lower than the 75th percentile, the selected group clearly represented facilities with very high turnover. The multinominal logit analysis then compared the low and high groups with the remaining middle group, which consisted of 189 facilities with reported rates from 6.7% to 62.5%. It should be noted that this middle group cannot be claimed to have productive or even moderate turnover. The categories are arbitrary and the analyses exploratory. Given the arbitrary definitions of the categories, we assessed whether observed findings were robust across alternative—higher and lower— turnover category cutoff points by running several models using alternative cut-off points and groupings for the dependent variable. Although variation in the relative significance of different factors was observed, no substantive or directional changes in the results were observed Hypothesis 5: Facilities with formal relationships with hospitals, those that serve as training sites, are nonprofit, are part of multifacility chains, are smaller in size, are less reliant on Medicaid funding, have smaller administrator span of control, have a union contract, or have no recent ownership change will be less likely to report very high or very low turnover. Few of the resident status variables, or clinical demands, showed any notable correlations with turnover in preliminary analyses. To provide case mix control variables, we chose three variables that reflected somewhat different potentially stressful aspects of caregiving work. Bladder incontinence requires considerable patience and may be distasteful. Monitoring residents who are ambulatory, although less physically demanding, requires vigilance. Emergency room admissions represent negative, unplanned events that disrupt workflow and increase uncertainty and stress. Hypothesis 6: Facilities with fewer proportions of residents who are bladder incontinent, that have fewer ambulatory residents, or that experience fewer emergency room admissions will be less likely to report very high or very low turnover. Market Environment Factors.—Local economic conditions have been shown to be associated with turnover, reflecting the perception of alternative employment opportunities as employees periodically evaluate their current situation (Abelson, 1986). Generally, higher per capita income and lower unemployment rates suggest the availability of job alternatives and, hence, higher turnover rates, although Cotton and Tuttle (1986) pointed to evidence that these economic factors may be less important for service workers than for those in manufacturing jobs. Recently, these two variables have been shown to be significant predictors of nursing home turnover (Banaszak-Holl & Hines, 1996), and thus we included them in these analyses. Banaszak-Holl and Hines also used the number of nursing facility beds in the county as a measure of competition and found that, indeed, turnover was lower in counties with fewer beds. We used a Herfindahl index of market concentration. This index was calculated as the sum of the squared market shares of the nursing homes in the county. Lower market concentration suggests more facilities competing for staff, which in turn likely raises wages in the long-term care sector and reduces overall turnover of nursing assistants. Hypothesis 7: Facilities located in counties with relatively high unemployment and low per capita income and those in less concentrated nursing home markets will be less likely to report very high or very low turnover. 164 The Gerontologist Table 2. Weighted Multinomial Logit Results for Low- and High-Turnover Groups Compared with Moderate-Turnover Group: Relative Risks as a result of changing the high and low category cutoff points. The sample was stratified to be nationally representative in terms of size, ownership, and rural versus urban location. Staffing and participation patterns and case mix variables were similar to those reported elsewhere (Anderson et al., 1996; Banaszak-Holl & Hines, 1996). With regard to the variables not examined in these prior studies, 19% of facilities reported union representation for nursing assistants, and 41% used a bonus reward system for performance. That on average seven departments reported directly to the top administrator (administrative span of control) suggests that nursing facilities continue to operate as relatively flat bureaucracies, with minimal managerial capacity. Mean per capita income and unemployment rates in the sample facilities’ counties were similar to those observed by Banaszak-Holl and Hines. The relative risk associated with each predictor for the facility’s being in the low or high turnover groups (using the middle category as the referent) is shown in Table 2. The analyses were conducted using STATA 5.0 (StataCorp, 1997). Variable Job Factors Leadership Close supervision (0,1) Trained supervisors (0,1) Performance rewards (0,1) Integration Involvement in assessment (3–27) Involvement in care planning (0,1) Rotating assignments (0,1) Workload No. aides/100 beds Organizational Factors Clinical resources Nurse practitioner (0,1) No. Aides/RNs and LPNs RN turnover rate (%) DON tenure (0,1) Nurse’s aide training Administrative resources Hospital relationship (0,1) Training site (0,1) Investor-owned vs. nonprofit (0,1) Chain membership (0,1) % Medicaid (0–100) Administrator span of control (2–14) Nurse’s aide union (0,1) Recent owner change (0,1) Facility size (4 categories) Investor-owned by nurse’s aide union (1,1) Investor-owned by chain membership (1,1) Clinical demands % bladder incontinent (0–100) % ambulatory (0–100) Emergency room admit rate (0–48) Job Factors Table 2 shows that the job factors, for the most part, bore little relationship to the likelihood of being either a high- or a low-turnover facility when other factors in the model were controlled for. The exception was that one of the leadership variables (Hypothesis 1), having supervisors trained in management, was associated with lower likelihood of being in the low-turnover group of facilities. That is, facilities with supervisors trained in management were more likely to be in the referent group than in the very low turnover group. This makes sense if one assumes that much of what is missing in very low turnover facilities is productive, possibly involuntary, turnover. Extreme staff stability may reflect an absence of functional management and resulting low performance expectations. Trained supervisors, by contrast, are more likely to set expectations and provide feedback to staff about performance, thereby promoting desirable attrition among poor performers. It is also possible that facilities with trained supervisors also have more effective hiring and human resources practices and thus experience very low turnover. Neither nurse’s aides per 100 beds (Hypothesis 3) nor any of the job integration variables (Hypothesis 2) were significantly associated with either high or low turnover, consistent with Banaszak-Holl and Hines’s (1996) observation that only one of four job design variables affected turnover rate. Market Environment Factors Log co. unemployment rate Co. per capita income Nursing facility market concentration High 0.53 (33) 1.69 (.20) 0.25 (.05) 1.09 (.84) 0.87 (.83) 1.36 (.39) 1.10 (.48) 1.07 (.36) 1.34 (.74) 1.07 (.89) 1.43 (.59) 0.86 (.69) 1.01 (.79) 1.00 (.83) 0.21 (.37) 0.59 (.25) 0.94 (.03) 0.41 (.24) 1.17 (.69) 2.31 (.16) 0.89 (.57) 1.02 (.01) 0.66 (.27) 0.92 (.71) 2.02 (.28) 1.29 (.70) 1.15 (.88) 0.75 (.82) 1.01 (.55) 0.55 (.09) 3.06 (.01) 6.31 (.00) 3.88 (.09) 1.00 (.70) 1.41 (.03) 10.68 (.02) 0.47 (.43) 0.79 (.49) 1.50 (.76) 1.13 (.14) 3.54 (.12) 1.40 (.43) 1.01 (.97) 0.18 (.07) 0.54 (.67) 0.16 (.04) 1.00 (.89) 1.00 (.67) 1.50 (.76) 1.01 (.21) 1.02 (.56) 1.04 (.09) 0.32 (.28) 1.37 (.61) 1.00 (.17) 1.00 (.99) 0.75 (.30) 0.79 (.13) Note: p levels are given in parentheses. DON 5 directors of nursing; RN 5 registered nurse; LPN 5 licensed practical nurse. n 5 257; x2(58) 5 104.51, p , .001; Psuedo R2 5 .23. of being in the very low aide turnover group and higher risk of being in the high aide turnover group (Hypothesis 4). None of the other clinical resource variables was significant. Administrative Resources Organizational Factors As for the administrative resources identified in Hypothesis 5, only the administrative span of control and the nurse’s aides having a union contract were associated with being in the low-turnover group. The likelihood of being in the high-turnover group rather than the referent group was associated with being a training site, for-profit ownership status, and although Consistent with prior work, the logit analysis supported a positive linear association between RN turnover and nurse’s aide turnover in that the likelihood of being in the low- or the high-turnover group was significant in the expected directions. That is, higher RN turnover was associated with lower risk Vol. 42, No. 2, 2002 Low 165 in the high-turnover category. The excessive turnover rates observed in the for-profit sector in this sample, then, were disproportionately attributable to the independent or freestanding proprietary nursing facilities. not significant at the standard .5 level, being part of a multifacility firm. The wider the administrator’s span of control— that is, the greater the number of people who reported directly to the administrator—the more likely the facility was to be in the low-turnover group. Thus, flatter organizations with less middle management were more likely to have very low turnover. This is consistent with the earlier finding that the likelihood of being in the very low turnover group was reduced by having supervisors trained in management. Having a union contract was associated with a 10fold increase in the likelihood that a facility would be in the very low turnover category. The union impact on the risk of being in the high-turnover group was not significant, although positive, when controlling for other factors in the model. Previous studies from other industries (Cotton & Tuttle, 1986) have suggested that the presence of the union reduces turnover by improving aspects of the job that we did not measure here, such as pay, benefits, and job security. As the logit analysis shows, however, the union presence was significant only in distinguishing facilities with very low turnover; it did not reduce the likelihood of being in the high-turnover group. In fact, there was a trend (relative risk ratio 5 3.53, p 5 .12) for the union factor to be associated with increased risk of very high turnover. We sought additional insight regarding the effect of unionization by examining the interaction between having a union and ownership status. Although not significant at the .05 level, the interaction term suggests that the investor-owned facilities with union contracts were less likely to be in the very high turnover group than were the nonprofit facilities with union contracts. Having a formal relationship with a hospital, financial constraints as measured by the percentage of Medicaid reimbursement, and facility size were not significantly associated with the high- and low-turnover categories defined here. Facilities that served as clinical training sites, however, were more than three times as likely to be in the high-turnover group than in the referent group. It seems likely that these facilities were hiring, providing certification training, and then losing to other health care facilities a disproportionate number of nursing assistants. What is not known, of course, is whether these training sites were more likely to lose their strongest or their weakest trainees and whether the turnover was voluntary or involuntary. At any rate, it likely represents a substantial financial burden. It is also possible, given the cross-sectional nature of the data, that facilities initiated training programs in response to high turnover. As shown in previous studies, investor-owned facilities were significantly more vulnerable to very high staff turnover than were nonprofits. An interaction term analysis indicated, however, that not all investor-owned facilities were equally vulnerable. The interaction of investor-owned status and chainmembership showed that the investor-owned chain facilities were more likely than investor-owned independent facilities to be in the referent category than Clinical Demands Clinical demands, as identified in Hypothesis 6, were not predictive of the high- and low-turnover groups. Neither of the “bed and body” clinical demands, the percentage of residents incontinent and the percentage ambulatory, was significantly associated with different turnover categories. The emergency room admission rate for the facility showed some (nonsignificant) effect on increased likelihood of high turnover, however. What cannot be shown from these cross-sectional analyses, however, is whether clinical acuity and stress led to turnover in these facilities or vice versa. Market Environment Factors None of the variables stated in Hypothesis 7 was significant in distinguishing either high- or low-turnover facilities. Neither the unemployment rate nor the per capita income in the county in which the facility was located appeared to be a factor in distinguishing extreme turnover rates among nursing assistants. Examination of the distribution of unemployment rate by turnover category, however, showed that the variances in unemployment were not similar across the three categories, making it difficult to compare effects. Market concentration of nursing home beds in the county, another typical indication of competition for employees, was not a significant factor in this analysis. In sum, the factors that significantly distinguished low-turnover facilities from the referent group of facilities, with the exception of RN turnover rate, were not the same factors associated with very high turnover. Further, RN turnover rate was the only factor that suggested evidence of a somewhat linear and monotonic relationship with turnover in that it was both negatively associated with being in the lowturnover group and positively associated with being in the high-turnover group. Summary and Discussion One limitation of this research is that although it drew on previous inquiries regarding the correlates of staff turnover, it did not test a single theoretical perspective, and the overall predictive validity of the model tested should not be used for comparison purposes. Rather than testing a well-defined model of turnover, our intent was to explore an alternative empirical approach to viewing turnover in light of known predictors. Although these known predictors are organized in conceptual configurations and stated as hypotheses for conceptual clarity, we view the findings presented here as exploratory. The approach of 166 The Gerontologist manage the long-term care staff retention problem. The findings presented here offer a starting point. Furthermore, human resources professionals from the nursing home industry should address with researchers the measurement of turnover and provide guidance regarding productive and nonproductive levels of turnover, both high and low. These findings should be interpreted in light of the fact than an optimal turnover rate for nursing homes has not been established and the categorical outcome variable used in the logit analysis was arbitrarily defined. Nevertheless, the logit analysis, in contrast with the traditional ordinary least squares modeling approach, is instructive. Specifically, it gives some credence to the argument that simple analysis of turnover rates expressed as percentages may not be desirable. Accordingly, a goal of reducing turnover by 10% where turnover is very high would be a very different challenge than reducing turnover by 10% where it is already very low. Managing voluntary and involuntary terminations of chronically poor performers requires a different approach than stemming a significant outflow of good staff to better paying organizations. Facilities that have a collective bargaining agreement may struggle to avoid both high- and very low-turnover situations. Having a union, again, does not prevent high turnover. Further, it may promote very low turnover in some facilities. This is a question that warrants further research. Having a collective bargaining agreement generally promotes a more stable staff, but it also reduces management’s discretion, affects organizational culture, and may prevent workforce renewal through job security measures. Given the cross-sectional nature of the data, it should be noted, union presence may be endogenous with high turnover in that both high turnover itself and the conditions that lead to it to may increase the likelihood of a union contract. Unionization, on the other hand, would not likely be caused by low turnover. Some of the reports of very high turnover are apparently from facilities that have their own training programs. Some of these may be serving as entry points to the occupation for nurse’s aides who leave after becoming certified for employment in other nursing facilities or hospitals. Reimbursement incentives for providing nurse’s aide certification may need to be examined. If nurse’s aides are, in fact, the backbone of long-term care, then policy-making bodies which have historically addressed shortages of health personnel through education financing and other strategies urgently need to work with the industry to increase the supply of these caregiving staff. The finding from the introduction of the Investor Owned 3 Chain Membership interaction also suggests that a stronger partnership between public policy makers and industry leaders may be productive. Although each of these separate factors predicted the likelihood of high turnover (although chain membership was not significant at .05 in the model), the interaction term showed that being a for-profit, chain-owned facility significantly reduced the chances of being in either extreme group compared with being a for-profit independent. The expertise of the investor-owned comparing high- and low-turnover facilities against a middle referent group is itself exploratory and not intended to test well-conceived theoretical models. Our intent was to reintroduce to the turnover literature the possibility that not all staff turnover is equally damaging to nursing facilities and their residents and, consequently, to show that turnover should not necessarily be viewed as a linear function of a set of predictors. The findings presented here push this argument one step further in that they suggest that different sets of predictors operate at the low end of the turnover continuum (fewer than 6.6% in 6 months) than at the high end (more than 64% in 6 months). Table 2 shows four predictors to be significant at the .05 level in distinguishing low-turnover facilities from the referent group: not having supervisors trained in management, having lower RN turnover, having a flatter management structure in which many people report directly to the administrator, and having a union contract governing nursing assistant/management relations. As a group, these predictors are somewhat suggestive of the simple, stable nursing home that was once the norm when one-size-fits-all, custodial care was all that was expected. Does this model still work well in some communities and with some resident populations? A question for future research is whether this simple and stable model is truly still identifiable and, if so, whether residents are receiving good care. The likelihood of being in the high-turnover category was increased by three factors significant at the .05 level—having higher RN turnover, being a training site, and being investor owned rather than nonprofit. Interaction analysis revealed that investorowned chain facilities were less likely to be in the high-turnover group than were investor-owned independents. This set of predictors is suggestive of a facility type with quite permeable borders and much inward and outward movement of employees between the facility and its environment. Future research should address the question of whether there are circumstances where this instability works well to produce good nursing home care or whether this level of instability is categorically incompatible with good long-term care. Beyond these statistically significant findings, the reader is encouraged to examine the relative risk ratios and significance levels throughout Table 2 in building future models because, as has been discussed, the direction of the findings was stable across alternative groupings of the dependent variable, although the significance levels of some predictors did vary depending on the cut-off points for the dependent variable categories. Researchers attempting to model the causes of high turnover, we now argue, may produce more clear and precise findings if they omit very low turnover facilities from their analyses. Studies are needed that evaluate models predicting turnover in low- to moderateturnover facilities and those that evaluate separate, distinct models to predict turnover in moderate- to high-turnover facilities. This next step of inquiry is needed to move forward our understanding of how to Vol. 42, No. 2, 2002 167 chains in hiring and retaining staff should not be underestimated in the search for replicable solutions. These findings suggest that practicing managers in nonunionized facilities may want to consider how they can provide alternative support structures to fill the role a labor contract plays without the loss of renewing turnover that may accompany the collective bargaining agreement. On the other hand, the nursing home industry has been a major focus of unionization efforts within the health care system for some time, and the commitment of the unions to quality-of-worklife issues in long-term care has been substantial. RN turnover is closely tied to turnover among nonprofessional staff, suggesting the need for progressive human resource management at all levels. Professional leadership instability may indicate a lack of commitment to the facility that trickles down to nursing assistants. Further analysis may reveal that local labor market conditions other than per capita income do affect nursing assistant turnover indirectly through their impact on professional staff turnover. The insignificant role of market factors evidenced here is somewhat puzzling, although it has been noted elsewhere that these environmental factors tend to be less influential in service employment turnover than in manufacturing (Cotton & Tuttle, 1986). Further, although it is not possible to establish causal order between emergency room admissions and staff turnover from these cross-sectional data, a relationship between resident and caregiver instability is suggested. From an analytic perspective, these findings raise questions about the use of linear regression as the best approach to modeling nursing facility turnover. Both in terms of the antecedents and presumed consequences of turnover, very low turnover cannot be assumed to be less of a problem than, or even assumed to be less of the same substantive problem as, very high turnover. In addition, although aggregate, organizational turnover is important to understand regardless of the individual events, quits and terminations are different and combining them does create “noise” from a conceptual perspective. This study, like many others, does not distinguish between voluntary and involuntary turnover. Future analysis of these data will enable explorations of the shape of the relationship between levels of staff turnover and resident outcomes. What this analysis has accomplished is to deconstruct a series of relationships that may influence turnover. Future research should identify a range of facility turnover rates that is neither too high nor too low in terms of quality of care and financial outcomes. Banaszak-Holl, J., & Hines, M. A. (1996). Factors associated with nursing home staff turnover. The Gerontologist, 36, 512–517. Bayer, E. J. (1994). Innovative work-force initiatives for long-term-care paraprofessionals: A resource directory. Generations, 18, 79–84. Brannon, D., & Streit, A. (1991, November). The effects of a merit-based reward system on performance in nursing homes. Paper presented at the 44th Annual Scientific Meeting of The Gerontological Society of America, San Francisco, CA. Burnfeind, J. D., & O’Connor, S. J. (1992, August). Employee turnover and retention rates as predictors of nursing home code violations. Paper presented at the annual meeting of the Academy of Management, Las Vegas, NV. Caudill, M., & Patrick, M. (1991). Turnover among nursing assistants: Why they leave and why they stay. Journal of Long Term Care Administration, 19, 29–32. Cohen-Mansfield, J. (1997). Turnover among nursing home staff: A review. Nursing Management, 28, 59–62. Cotton, J., & Tuttle, J. (1986). Employee turnover: A meta-analysis and review with implications for research. Academy of Management Review, 11, 55–70. Crown, W. H. (1994). A national profile of homecare, nursing home, and hospital aides. Generations, 18, 29–33. Grant, L., Kane, R., Potthoff, S., & Ryden, M. (1996). Staff training and turnover in Alzheimer special care units: Comparisons with non-special care units. Geriatric Nursing, 17, 278–282. Halbur, B. T., & Fears, N. (1986). Nursing personnel turnover rates turned over: Potential positive effects on resident outcomes in nursing homes. The Gerontologist, 26, 70–76. Haveman, H. A. (1995). The demographic metabolism of organizations: Industry dynamics, turnover and tenure distributions. Administrative Science Quarterly, 40(4), 586–590. Helmer, F. T., Olson, S. F., & Heim, R. I. (1993). Strategies for nurse aide job satisfaction. Journal of Long-Term Care Administration, 21, 10–14. Iglehart, A. P. (1990). Turnover in the social services: Turning over to the benefits. Social Service Review, December, 649–657. Kettlitz, G. R., Zbib, I., & Motwani, J. (1997). Reducing nurse aide turnover through the use of weighted applications bland procedure. Health Care Supervisors, 16, 41–47. Kodner, D. L. (1993). Long-term care 2010: Speculations and implications. Journal of Long-Term Care Administration, 21, 82–86. McCallion, P., Toseland, R. W., Lacey, D., & Banks, S. (1999). Educating nursing assistants to communicate more effectively with nursing home residents with dementia. The Gerontologist, 39, 546–558. Mesirow, K. M., Klopp, A., & Olson, L. L. (1998). Improving certified nurse aide retention. Journal of Nursing Administration, 28, 56–61. Mobley, W. H., Griffith, R. W., Hand, H. H., & Meglino, B. M. (1979). Review and conceptual analysis of the employee turnover process. Psychological Bulletin, 86, 493–522. Monahan, R. S., & McCarthy, M. S. (1992). Nursing home employment: The nursing assistant’s perspective. Journal of Gerontological Nursing, 18, 13–16. Schaefer, J. A., & Moos, R. H. (1996). Effects of work stressors and work climate on long-term care staff’s job morale and functioning. Research in Nursing and Health, 19, 63–73. Shaw, J. D., Delery, J. E., Jenkins, G. D., & Gupta, N. (1998). An organization-level analysis of voluntary and involuntary turnover. Academy of Management Journal, 41(5), 511–525. Sheridan, J. E., White, J., & Fairchild, T. J. (1992). Ineffective staff, ineffective supervision, or ineffective administration? Why some nursing homes fail to provide adequate care. The Gerontologist, 32, 334–341. Smyer, M., Brannon, D., & Cohn, M. (1992). Improving nursing home care through training and job redesign. The Gerontologist, 32, 327–333. StataCorp. (1997). Stata statistical software (Release 5.0) [Computer software]. College Station, TX: Stata. Stryker, R. (1982). The effect of managerial interventions on high personnel turnover in nursing homes. Journal of Long Term Care Administration, 10, 21–33. Terborg, J. R., & Lee, T. W. (1984). A predictive study of organizational turnover rates. Academy of Management Journal, 27, 793–810. Teresi, J., Holmes, D., Beneson, E., Monaco, C., Barrett, V., Ramirez, M., & Koren, M. J. (1993). A primary care nursing model in long-term facilities: Evaluation of impact on affect, behavior, and socialization. The Gerontologist, 33, 667–674. Wagnild, G. (1988). A descriptive study of nurse’s aide turnover in long-term care facilities. Journal of Long Term Care Administration, 16, 19–23. Waxman, H. M., Carner, E. A., & Berkenstock, G. (1984). Job turnover and job satisfaction among nursing home aides. The Gerontologist, 24, 503–509. References Abelson, M. A. (1986). Strategic management of turnover: A model for the health service administrator. Health Care Management Review, 11, 1–71. Anderson, R. A., Issel, L. M., & McDaniel, R. R., Jr. (1996, August). Nursing staff turnover in nursing homes: A new look. Paper presented at the annual meeting of the Academy of Management, Health Care Administration Division, Cincinnati, OH. Argote, L., Insko, C., Yovetich, N., & Romero, O. (1995). The effects of turnover and task complexity on group performance. Journal of Applied Social Psychology, 25(6), 512–529. Received December 5, 2000 Accepted October 12, 2001 Decision Editor: Laurence G. Branch, PhD 168 The Gerontologist
© Copyright 2025 Paperzz