Investigating Turnover and Satisfaction

Five studies of predictive models of turnover and satisfaction among U.S. nursing assistants and home care workers carried out between 2007 and 2011 are summarized in Table 1. They contribute to defining three dimensions of the present study. First, their primary concern is with certified nursing assistants (CNAs) and home care workers rather than registered nurses; we focus on equivalent categories of non-nursing workers, namely personal care assistants (PCAs) in residential care and community care workers (CCWs) in home- and community-based care. Second, a number of common design features in the studies reduce the effects of methodological differences and enhance the robustness of common findings. The 2004 NNHS and NHAS and NILS 2007 used similar items and measures of satisfaction, providing equivalent accounts of the U.S. and Australian aged-care workforces. Regression analysis was used in two cross-sectional studies to identify significant predictors of ITL (CitationWiener, Squillace, Anderson, & Khatutsky, 2009; CitationDecker, Harris-Kojetin, & Bercovitz, 2009) and, in three longitudinal studies, predictors of actual behavior (CitationCastle, Engberg, Anderson, & Men, 2007; Rosen, Stiehl, Mittal, Leana, 2011; CitationMorris, 2009). Two regression models are used in our study to examine the effects of satisfaction and other variables on ITL. These studies provide comprehensive reviews of research into turnover and satisfaction in aged care.

ITL, Satisfaction, and Actual Behavior

Most of the relationships among ITL, satisfaction, and actual behavior reported in the reviewed studies were found to be significant. In a longitudinal study over 2 years, CitationRosen et al. (2011) distinguished among stayers, switchers, and leavers, finding significant associations among ITL and satisfaction, tenure at baseline, and actual turnover, but satisfaction was not significantly related to actual behavior. Similarly, CitationMorris (2009) reported significant associations among ITL and a number of satisfaction indicators, tenure at baseline and actual turnover at 2 years among home care workers in Maine. While CitationCastle et al. (2007) reported an inverse association among ITL at each stage of progression from thinking about leaving, thinking about job search, searching for a job, and leaving, ITL was not significantly associated with actual turnover at 2 years; satisfaction was significantly associated with ITL and actual turnover.

The two cross-sectional studies reported a range of significant associations with ITL, with tenure providing a proxy measure of behavior over time. CitationDecker et al. (2009) found total experience as a CNA and current job tenure were significantly and inversely associated with ITL. Similarly, CitationWiener et al. (2009) reported the group with tenure of more than 1 year averaged 54.2 months tenure, more than double the average for the total sample.

The consistency in these findings compared to the more varied results of earlier studies may be attributable in part to similarities in the measures and methodologies used in the five studies, and the literature reviews in these studies all note that such differences in previous research in the United States confound interpretation and comparison of findings.

In a previous investigation of stability in the Australian aged-care workforce using NILS 2007, we adopted the dual driver model that CitationMittal, Rosen, & Leana (2009) applied to the long-term care workforce and defined four “indicative stability” groups (CitationHowe et al., 2012). Fully two-thirds of all workers were “committed” (tenure > 1year, expected to be in same job in 1 year), and fewer than 5% were “uncommitted” (tenure ≤ 1 year, did not expect to stay). The remainder was divided almost equally between 18% who were “leavers” (tenure ≥ 1 year, expected to leave) and 15% who were “stayers” (tenure ≤ 1 year, expected to stay). Further differences between nurses and other direct care workers, and between those working in residential and community care, showed that turnover was far from uniform across the workforce and that drivers of retention and intention were asymmetrical rather than reciprocal, in line with the dual-driver model.

Dimensions of Overall, Intrinsic, and Extrinsic Satisfaction

A range of theoretical frameworks developed in labor market studies, industrial psychology, and management have been drawn on in setting out conceptual models of relationships between ITL and satisfaction in the long-term care workforce. Empirical applications of these models have in turn confirmed satisfaction as a predictor of ITL and actual behavior. Three main areas are summarized here.

First, distinctions have commonly been made between overall satisfaction and satisfaction with extrinsic and intrinsic rewards. All five studies reviewed included measures of intrinsic satisfaction reporting workers’ subjective assessment of, or attitudes toward, different aspects of their work as well as extrinsic or objective characteristics; thus, satisfaction with pay and hours worked is reported as well as actual pay and hours. In two studies using similar arrays of items from the NNAS, CitationWiener et al. (2009) found extrinsic rewards were more predictive of tenure, while CitationBishop et al. (2009) found intrinsic rewards were more predictive of satisfaction; neither ITL nor actual behavior were investigated in the latter study.

Similar variables were included in investigations of satisfaction in the residential aged-care workforce and the residents they cared for in Western Australia (CitationBoldy, Chou, & Lee, 2004). Different effects found for staff in nursing homes and hostels (akin to assisted living facilities) lead these authors to stress the role of management in enhancing satisfaction as a means to reducing turnover, the need for facility-level strategies tailored to staff roles, and the potential for increasing resident satisfaction.

Second, satisfaction has been cast as an intervening variable that mediates the effects of other factors on ITL and actual behavior in models proposed by CitationCastle et al. (2007), CitationDecker et al. (2009), and CitationRosen et al. (2011). The number of significant predictors of satisfaction, and the effects of satisfaction on ITL identified in these studies, demonstrate that inclusion of measures of satisfaction has substantially extended the explanatory scope of multivariate analyses. Intervening effects of intrinsic rewards on ITL are also reported by CitationMorris (2009) but did not translate into significant effects on actual turnover, which was influenced instead by ITL and extrinsic rewards.

The third advance in conceptualization comes from the integration of findings into macro-level explanations for workers’ intentions and behavior. Met expectation theory was adopted and modified by CitationCastle et al. (2007), while CitationMorris (2009) invoked “job-lock” to explain why low-skilled workers remained in their LTC jobs.

Structure of the Australian Long-Term Care Workforce

Compared to the United States, the not-for-profit sector is more dominant in long-term care in Australia. It accounts for 63% of all residential care homes, accommodating 5.3% of the population aged 65 and older (CitationAustralian Institute of Health and Welfare, 2009). Not-for-profit providers are even more dominant in community care where the private sector is negligible; some 25% of the population aged 65 and older received care services in their homes and in community-based settings during a 12-month period (CitationHome and Community Care Program, 2009).

The non-nurse workforce is divided into two groups, similar to CNAs and home health and care workers in the United States. PCAs accounted for 63% of a total residential care workforce of 80,000 (equivalent full time [EFT]), and CCWs accounted for 78% of a total community care workforce of 46,000 (EFT; CitationMartin & King, 2008). While providing mainly personal care, PCAs undertake tasks such as administering medication when they have undertaken relevant training. Trained CCWs undertake personal care and basic health care tasks similar to the work of PCAs but also deliver a substantial amount of homemaker work done by domestic staff in residential care. Women account for over 90% of the total aged-care workforce, and this study includes only female PCAs and CCWs.

Turning to the workplace context, the dominant role of the federal government in both aged-care funding and workplace relations means that employment conditions are relatively uniform across the aged-care workforce, and many conditions are in common with the wider workforce. Australia's universal health insurance scheme provides free public hospital care and free or near-free medical care to the whole population, regardless of workforce status, and while individuals may take out additional private health insurance, employers do not provide this benefit. Access to pension benefits is provided through a superannuation guarantee charge levied on all workers earning above a threshold.

National minimum wage rates are set by a federal body for all sectors of employment. In January 2012, the award rate for certificate-qualified PCAs and CCWs was $18.06 per hour (as of August 2013, $1A = ˜ $.90US). The majority of aged-care workers are employed on a permanent part-time basis and so are covered by national awards that provide paid leave for sickness, holidays, and long service (paid after 10 years of service). Casual workers are partly compensated by higher hourly rates, but they face more uncertainty with regard to rostered hours of work. These much more standard conditions in access to benefits and other workplace conditions are likely to moderate the impact of these factors on variations in turnover and satisfaction in Australia compared to the less regulated employment environment in the United States.

Comparison of Residential and Community Care Workers

Comparison of the residential and community care workforces presented in Table 2 (section A) expands on the similarities and differences already noted. A higher proportion of PCAs than CCWs reported ITL, while the difference of around a year in average tenure reflects the higher proportion of CCWs who were new employees (in job ≤ 1 year). Both groups rated their health as good on average, and half had financial dependents.

In terms of skills, the majority of PCAs and CCWs had some post–high school qualification, but marginally more CCWs had no such qualifications. Substantial minorities of both groups had language and cultural skills; the higher proportion speaking a second language than were born in non–English-speaking countries is explained by second generation migrants who retain their familial language.

Turning to organizational characteristics, CCWs were almost three times as likely to be employed on a casual basis as PCAs. Another marked difference is that while the majority of both groups were employed in the not-for-profit sector, PCAs were otherwise more likely to be working in private-sector residential care homes and CCWs in public-sector agencies. PCAs were more likely to be working for providers that were part of a larger organization, reflecting the operation of more chains in residential care, and were somewhat over-represented in metropolitan areas. Overall, PCAs and CCWs were more similar than different in personal characteristics, with more differences in organizational characteristics.

Satisfaction Scores

Satisfaction scores in Table 2 (section B) show that CCWs reported slightly higher levels of overall satisfaction than PCAs, but both mean scores are comparable to an average of 7.97 reported for the Australian female workforce as a whole (CitationMartin & King, 2008). At the same time, there is considerable variation between PCAs and CCWs on satisfaction with different job conditions and aspects of doing care work.

Total pay stands out as the extrinsic reward with which both PCAs and CCWs reported markedly lower satisfaction. The two groups reported similarly high satisfaction with having needed skills and using them, but PCAs were less satisfied than CCWs in all other areas, most markedly in feeling under more pressure and finding their jobs stressful.

Predictors of ITL

Overall satisfaction was a significant and comparably strong predictor of ITL for both PCAs and CCWs. The marginal effects show that for every one-point increase in overall satisfaction scores, ITL decreased by some 5% (Table 3, section B, model 1). Six worker and organizational variables and 11 satisfaction variables were also significant predictors of ITL (p < .05, Table 3, model 2).

Only one worker and one organizational characteristic were common significant predictors of ITL for PCAs and CCWs with large marginal effects (Table 3, section A). Having financial dependents (compared to not having such dependents) reduced ITL, with a marginal effect close to 5% for both PCAs and CCWs, and employment on a casual basis (rather than permanent) increased ITL by more than 8% for PCAs and by approximately half as much for CCWs. For PCAs, having a second language but not using it on the job had a large marginal effect on ITL; the marginal effect of older age was also significant but small. Among CCWs, better self-rated health and being born in a non–English-speaking country had intermediate effects.

Eleven of the 15 satisfaction variables were significant predictors of ITL in model 2. Four variables were significant predictors for both PCAs and CCWs; among job conditions, dissatisfaction with the work itself had a greater marginal effect for PCAs than for CCWs, while dissatisfaction with opportunities and support had lower and similar effects for both groups. Finding the job more stressful than expected was the only aspect of doing the job that was significant for both PCAs and CCWs.

The other three variables significant for PCAs had greater marginal effects than the remaining four variables significant for CCWs. Having the skills needed to do their jobs has the largest marginal effect on ITL for PCAs, followed by not using these skills and not being respected. Among CCWs, dissatisfaction with hours increased ITL while satisfaction with flexibility reduced it; dissatisfaction with staff-management relations and not having enough time had similar effects.

Job Satisfaction

In Australia, the majority of workers were satisfied overall with their jobs and with working in aged care. Comparable findings about overall job satisfaction were reported in the U.S. studies reviewed above, especially CitationRosen et al. (2011), CitationDecker et al. (2009), and CitationCastle et al. (2007). While dissatisfaction was found to be associated with ITL in both countries, it characterizes a minority of workers, not the majority.

These findings may be affected by stayers in the workforce being likely to be less dissatisfied than those who have left; extrapolation of the relationships found between ITL and satisfaction, and reported elsewhere between these variables and actual turnover, would posit this to be so. However, to the extent that workers also leave because of other factors, such as changes in personal circumstances, there may be limits to how far efforts to enhance satisfaction can reduce turnover, especially among the most dissatisfied workers whose decisions to leave may be influenced by other factors.

Nevertheless, there is room for improvement as an increase of two points in overall satisfaction scores of PCAs would see a 10% reduction in ITL. Our analysis provides insights into the particular aspects of job satisfaction that would have the most impact on reducing ITL. Maximizing satisfaction with the work itself was paramount, and ensuring support is available from the work team and employer, providing opportunities for workers to develop their abilities, and reducing stress are identified as means to this end for both groups.

Differences in other aspects of satisfaction on the part of PCAs and CCWs point to somewhat different avenues for action, with one important exception. Based on research in other health settings, autonomy or discretion has been proposed as a factor in job satisfaction in LTC, but freedom to decide how to do their jobs was not a significant predictor of ITL for either PCAs or CCWs in our study.

Casual Employment

The strength of association between ITL and employment on a casual basis puts action in this area squarely with employers in the Australian aged-care system as the main means by which they could reduce ITL and thereby reduce turnover. As well as creating uncertainty about having a continuing job, fluctuating hours of work, and level of income, casual employment has other negative consequences for workers, such as lack of access to paid leave for sick days and holidays, that are not fully compensated for by higher hourly rates. While casual employment per se has not been reported in U.S. studies, transposed effects by way of lack of benefits appear widespread (CitationBishop, Squillace, Meagher, Anderson, & Wiener, 2009; CitationDecker et al., 2009; CitationRosen et al., 2011; CitationWiener et al., 2009;).

Casual employment had a larger impact on ITL for PCAs than CCWs. Further investigation into how casual employment is managed in both settings is warranted to establish, for example, how far providers use casual contracts to cover transition periods. There seems to be little about the structure of care work that requires high levels of casual labor, and it is relatively common for shift workers in other industries to be on permanent contracts. The higher levels of dissatisfaction with casual work in residential care are likely to stem from comparisons with other health care workers in institutional settings, the majority of whom are on permanent contracts (CitationMartin, 2007), and with the majority of PCAs who are on permanent contracts. In contrast, workers in community care may be more likely to compare themselves to workers in other service sectors or the broader community care workforce in which casual contracts are increasingly the norm.

While some level of casual employment may be desired by workers and employers, maintaining a large casual workforce may be counterproductive if it means frequent hiring of new and inexperienced workers and a lack of continuity in contact between care workers and clients, with consequences for quality of care (CitationNetten, Jones, & Sandhu, 2007).

Job-Lock

The third issue that our analysis highlights is that the phenomenon of job-lock described by CitationMorris (2009) is evident in Australia. The finding that job security was not significant for either workforce group may reflect low expectations about tenure due to personal or contextual circumstances or a lack of concern about being able to get another job in aged care or a similar service area in a strong labor market.

Where workers are dissatisfied with their work but unable to move, they may feel trapped in their jobs. While this may decrease ITL, it is not necessarily a positive outcome. To the extent that lower ITL indicates job-lock, our findings suggest that workers with financial dependents, older workers, workers who speak a language other than English and use it in their jobs, and those in rural areas may find themselves in this position. The question is whether job-lock can be turned into a positive experience, so that while workers may be stuck in jobs, they are satisfied to be there.

Having financial dependents and workers’ ages cannot be changed. These workers may, however, respond positively to benefits that improve their financial situation and work conditions. For example, retaining an older workforce may require offering transition to retirement programs that are responsive to preferred hours, or the flexible use of leave entitlements and safeguarding occupational health and safety.

Factors associated with ethnicity show mixed effects on job-lock, possibly reflecting the diversity of cultural backgrounds among workers born overseas, who include not only long-standing and more recent migrants but student nurses intent on remaining in Australia when they qualify. As well as optimizing use of language and other skills on the part of workers from similar backgrounds to clients, there is a need to develop English proficiency, cultural competency, and aged-care skills among others to make them effective members of the mainstream workforce.

Limitations

Many of the findings of this study are in common with U.S. research on predictors of ITL, and where marked differences were evident, explanations can be found largely in contextual differences, but three limitations need to be noted.

First, there are some differences between the number of respondents in the total NILS 2007 surveys and the subsamples in our study, which excluded respondents with missing data on any of the items analysed. The number of PCAs in the subsample was lower by 16% and the number of CCWs by 32%. No systematic selection bias is apparent, however, because the subsamples closely matched the distribution of the total samples on all variables.

While the marginal effects method was an advance in being able to establish the magnitude of the effect of different variables on ITL, a second limitation is that we have not been able to assess the strength and direction of interactions among variables. The adoption of the moderated mediation model developed by CitationAvgar, Pandey, and Kwon (2012) would enable clarification of the nature of such interactions. A further three areas for methodological development were identified: the need for more nuanced analyses of job satisfaction in relation to specific worker and workplace characteristics; separating dissatisfaction with a particular job from dissatisfaction with working in aged care more generally; and including measures of quality of care, as assessed by workers and care recipients, to enable exploration of associations between worker satisfaction and client outcomes.

The cross-sectional nature of our study is a third limitation. We have not been able to link ITL to behavior, and an important contribution of longitudinal work would be to distinguish workers who switched between jobs but remained part of the LTC workforce from those who stayed or left their jobs. Following CitationRosen et al. (2011), identification of switchers could be useful in separating the effects of dissatisfaction with some aspects of a particular job from dissatisfaction with care work in general. A longitudinal panel design would address this limitation since panel estimation would be able to control for time invariant heterogeneity and enable some inferences to be drawn about causal effects between variables.