Organizational Factors Associated With Retention of Direct Care Workers: A Comparison of Nursing Homes and Assisted Living Facilities

Abstract

Background and Objectives

Low retention of direct care workers (DCWs), either certified nursing assistants in nursing homes (NHs) or personal care assistants in assisted living (AL), continues to be an unresolved problem. While numerous studies have examined predictors of DCW retention in NHs, little attention has been paid to differences between settings of long-term care. This study compares the predictors of DCW retention rates across both settings.

Research Design and Methods

The 2017 Ohio Biennial Survey of Long-Term Care Facilities provides facility-level information from NHs and ALs (NHs = 739; ALs = 465). We compare the factors that predict retention rates of DCWs utilizing regression analysis. The factors are structural, financial, resident conditions, staffing, and management characteristics, as well as retention strategies.

Results

Average DCW retention rates were 66% and 61% in ALs and NHs, respectively. Not-for-profit status was significantly associated with higher retention rates across settings. While the percent of residents with dementia and less administrator turnover were associated with significantly higher DCW retention in NHs, these were not significant for ALs. However, in the AL context, a higher county unemployment rate and DCWs’ participation in resident care planning meetings were positively related to DCW retention after controlling for all other covariates, while DCW cross-training was negatively associated.

Discussion and Implications

Retention strategies for DCWs may need to differ by setting, as a result of differing working environments, resources, and regulations.

Staffing, Quality, Turnover, Long-term care, Nursing home, Assisted living

The increased media and societal focus on care in our nation’s long-term care settings during the coronavirus disease 2019 (COVID-19) pandemic has pointed out the shortcomings as well as the strengths of the care we provide to vulnerable older adults. One aspect of care often mentioned is the heroic workers on the front lines of caregiving. These frontline or direct care workers (DCWs) provide the bulk of care to residents in long-term care settings (Scales, 2020; Sengupta et al., 2016). DCWs assist every day with daily activities such as bathing and dressing, and their relationships with the residents they care for are essential to good resident quality of life (Ball et al., 2010; Morley, 2014; Shin et al., 2014). Their skills keep residents safe and healthy, and their caring helps residents feel less isolated and lonely.

There is a growing body of evidence that retaining DCWs is linked to long-term care quality. Past work has found nursing homes (NHs) that are better at retaining DCWs have fewer regulatory deficiencies (Castle et al., 2020), have residents with better clinical outcomes (Kimmey & Stearns, 2015), and have lower administrative costs (Pillemer et al., 2008). More importantly, retaining DCWs allows providers to engage in an important staffing strategy—consistent assignment. Consistent assignment is the practice of assigning the same DCWs to the same residents. The practice helps staff get to know the preferences of the residents they care for, identify impending health issues, and to reap the intrinsic rewards of these caregiving jobs (Ball et al., 2010; Mittal et al., 2009; Morgan et al., 2013; Sung et al., 2005). But consistent assignment is difficult to achieve when these essential staff members do not stay in their jobs. National retention rates in NHs and assisted living (AL) are lower compared to other industries. National estimates of 1-year retention rates among DCWs in NHs were just over 50% in 2016 (Castle et al., 2020), and 66% in ALs in 2011 (NCAL, 2012).

While there are a number of studies that have examined factors associated with retention in the NH setting (Berridge et al., 2018, 2020; Donoghue, 2010; Howe et al., 2012; Mittal et al., 2009; Rosen et al., 2011; Sung et al., 2005; Wiener et al., 2009), less is known about retention in AL settings and whether the factors associated with retention in NHs also apply. Additionally, the few studies that have examined both NH and AL settings (Brannon et al., 2007; Kemper et al., 2008) relied on a measure of “intent to quit’’ among current employees or investigated the aspects of DCWs’ jobs that needed most improvement rather than DCW retention. In this paper, we examined retention rates of DCWs in two long-term care settings—NHs and ALs, which are referred to with an array of terms across the nation, including residential care, domiciliary care, and board and care homes. In Ohio, the setting for this study, the licensure name for ALs is residential care facilities.

While ALs and NHs provide long-term care, both settings are different in terms of residents, financing, and facility operations that are important when considering DCW retention. NH residents have more functional impairments and have shorter length of stay than AL residents (Applebaum et al., 2019). NH care is primarily paid for by Medicaid, and NHs provide care to residents needing Medicare-reimbursed rehabilitative services (i.e., post-acute care). Post-acute care is provided to AL residents on an outpatient basis via home health care. The principles of independence, autonomy, and negotiated risk are the cornerstone of ALs, whereas NHs have much more of a medical focus. There are also differing degrees of regulatory requirements. ALs are regulated at the state level and tend to face fewer regulatory requirements when compared to NHs, which are regulated via a mix of federal and state regulations and rely much more heavily on federal funds (Trinkoff et al., 2020).

Using statewide data from Ohio facilities for 2017, we calculated retention rates of DCWs in NHs and ALs. Regression analysis was used to examine if the retention rates of DCWs in NHs and ALs are different, and more importantly whether the factors that predict retention in NHs are similar or different in ALs. This paper contributes to the understanding of retention of DCWs by utilizing data that measure retention in both long-term care settings over the same period of time, examines factors affecting retention that are comparable across settings, and includes information regarding specific strategies and practices each facility engages in to retain DCWs.

Conceptual Framework

Most of the literature on retention of DCWs comes from the NH experience, rather than from studies of ALs. The factors associated with higher retention of DCWs include aspects of facility structure and financing, staffing, management, resident conditions, and market characteristics. There are also retention strategies that can be environmental or financial in nature. The convoy model of care (Kemp et al., 2013) specifies the relationship between formal caregiver outcomes, such as retention, and various factors in the regulatory context, industry context, community level, facility level, and individual level. While our data source does not include information from workers or residents directly, it includes variables from the facility and the community.

Building on the literature on NHs, in terms of facility structure and financing, not-for-profit ownership, smaller facilities, and having a special memory care unit are associated with greater retention of DCWs (Berridge et al., 2020; Donoghue, 2010; Lerner et al., 2017; Wiener et al., 2009). Management also matters, with owner-operated, longer-tenured administrators having higher retention or lower turnover rates of DCWs than facilities with shorter administrator tenure and salaried managers (Lerner et al., 2017; Huang & Bowblis, 2018b). In terms of management, DCWs who had more positive perceptions of their supervisor were less likely to leave (Brannon et al., 2007). Higher resident acuity—such as total dependence in eating, toileting, and transferring, and percent psychiatric illness—is also associated with higher DCW turnover (Kennedy et al., 2020; Sengupta et al., 2016).

An important external factor which influences retention is the local market. While retention is often motivated by internal rewards such as a desire to help others (Ball et al., 2010; Brannon et al., 2007; Mittal et al., 2009), the need to earn an income is a critical motivator (Sung et al., 2005; Wiener et al., 2009). Better job alternatives are associated with higher intent to quit for DCWs in NHs and ALs (Brannon et al., 2007). One way to measure job alternatives is through unemployment rates, as higher unemployment in the local area proxies for job availability. A number of studies have found employment duration in ALs and NHs and retention rates in NHs are generally greater when unemployment rates are higher (Ball et al., 2010; Berridge et al., 2018; Huang & Bowblis, 2018a; Wiener et al., 2009), yet some studies find no relationship between NH DCW retention and area unemployment (Berridge et al., 2020; Kennedy et al., 2020). Less is known about the relationship between unemployment rates and retention of DCWs in ALs.

Environmental and financial retention strategies can also play an important role in retaining DCWs by improving the working conditions and the value DCWs find in the tasks they perform. Practices that empower DCWs such as participation in resident care planning and consistent assignment of staff to residents have also been shown to affect retention rates (Berridge et al., 2018, 2020). Overwhelmingly, DCWs report that better compensation and supervision are needed to improve their jobs (Bishop et al., 2008; Kemper et al., 2008; Sung et al., 2005). Extrinsic benefits such as pensions, higher wages, and health insurance have a positive effect on retention rates and intent to quit (Bishop et al., 2008; Brannon et al., 2007, p. 823; Dill et al., 2013, 2014; Morgan et al., 2013; Wiener et al., 2009). Growth opportunities, such as career ladders and tuition reimbursement provided to DCWs in both types of settings, are also associated with more stable staffing (Ball et al., 2010; Bishop et al., 2008; Dill et al., 2014; Pillemer et al., 2008). However, this research was framed around the turnover issue and did not explicitly distinguish from retention in terminology or measures (Castle et al., 2020).

Many of the factors associated with retention in NHs may also be relevant for the AL setting, but to a different extent. First, the DCW characteristics in NHs and ALs are different, with ALs employing more men, workers with some college education (Kelly et al., 2020; Scales, 2020), and stronger intent to leave (Brannon et al., 2007; Kemper et al., 2008). Second, DCWs in NHs were found to be more likely to respond to financial retention strategies than DCWs in ALs. Third, NH DCWs perceive their workloads as higher and have lower job satisfaction compared to other settings such as AL and home care (Brannon et al., 2007; Kemper et al., 2008). Finally, AL workers seem to be more affected by intrinsic factors such as job advancement, job challenges, and collegiality in the workplace.

Given previous research we propose three hypotheses. Our first hypothesis (H1) is that factors that are associated with retention among NH DCWs are expected to be associated with retention in ALs in the same direction. For example, not-for-profits when compared to for-profit facilities would have better retention rates for NH and AL settings. Our second hypothesis (H2) is that factors that predict retention in ALs and NHs are expected to have different magnitudes because NH and AL settings have different residents, financing, and operational characteristics. To illustrate, past work found that AL DCWs have stronger intent to leave than NH DCWs (Brannon et al., 2007). This would mean that while higher unemployment rates would increase retention rates in NHs and ALs, the effect of higher unemployment rates would be larger for AL DCWs. Our third hypothesis (H3) is if NH DCWs perceive their workloads as being higher, which is consistent with past literature, retention rates are expected to be higher in ALs holding other factors such as facility, market, and retention strategies constant.

Design and Methods

Data Sources

This study leveraged unique facility-level data collected from Ohio NHs and ALs from 2017. The main source of data was the 2017 Ohio Biennial Survey of Long-Term Care Facilities, an ongoing survey with 12 previous rounds of data collection. The survey is completed by facility administrators, is required by the state, and has an overall response rate among NHs and ALs of 91% and 88%, respectively (Applebaum et al., 2019, 2020). The NH version of the Ohio Biennial Survey of Long-term Care Facilities was merged with Certification and Survey Provider Enhanced Reports (CASPER) data, which contain facility-level information collected as part of federally mandated recertification surveys. We also merged the NH data with the Payroll-Based Journaling data to obtain resident census information. Finally, we merged all data sets with the Rural–Urban Commuting Area 2004 ZIP Data (Version 2.0) to obtain rurality at the zip-code level (Rural Health Research Center, n.d.) and the Area Health Resource File to obtain unemployment rates.

In 2017, there were 963 licensed NHs and 721 licensed ALs that were eligible to complete the Ohio Biennial Survey of Long-term Care Facilities. To assure a uniform sample across NHs and ALs, we restricted the sample to facilities that completed the survey, did not change ownership during 2017, were not government-owned, and among NHs, facilities that were not hospital-based. This resulted in a potential sample of 824 NHs and 616 ALs. After further restricting the sample to facilities that had reported information to calculate DCW retention and covariates, our analytic sample included 1,204 long-term care facilities, of which 739 were NHs and 465 were ALs.

Retention Rate

The dependent variable in our analysis was the retention rate of DCWs. The retention rate is the proportion of DCWs employed in the first payroll period of 2017 that remained employed in the last payroll period of 2017. The retention rate can range from a minimum of 0, indicating no employees employed at the beginning of the year remained at the end of the year, to 100, which indicates that all employees employed at the start of the year were retained throughout the year. For NHs, DCWs were certified nurse aides (CNAs) and in ALs, DCWs were those personal care assistants that provide direct care to residents. DCWs in ALs in Ohio do not need to be CNAs, although many do have state certification.

Covariates

When constructing covariates, we attempted to measure each variable consistently across both settings, although the data sources were not always consistent. The covariates included the following categories: facility structure and financing, staffing and management factors, aggregated resident characteristics, market characteristics, and environmental and financial retention strategies.

Facility structural measures included number of beds, ownership, chain affiliation, presence of a memory care unit, and whether the facility is part of a continuing care retirement community (CCRC). Because NHs and ALs are generally of different size, we identified small, medium, and large facilities. Small facilities were within the first quartile of number of beds within each setting, large facilities in the fourth quartile, and medium were within the second and third quartiles. We also identified whether the building and the facility were owned by the same company or were different. Financial characteristics included the occupancy rate and the percentage of the facility’s residents paid for by Medicaid. In ALs, Medicaid includes residents on the Medicaid AL Waiver and integrated Medicare/Medicaid plans (i.e., MyCare). Occupancy rates for NHs came from CASPER and for ALs were calculated as the number of units occupied each day to available unit-days in the year as reported on the Biennial Survey. We did not include the percent of resident days paid for by Medicare, because Medicare does not cover basic AL care.

We included three sets of staffing variables. First, we wanted to include staffing levels for DCWs. However, because the level of need of residents is different between both settings, we first calculated the number of DCWs per 100 residents based on the number of DCWs in the last payroll period and number of residents in the facility on December 31, 2017. We then categorized each DCW staffing level into quartiles relative to each setting. This identified whether the facility was lower or higher staffed compared to their peers in the state. For the few facilities that did not report information on the number of residents or reported no staff in the last payroll period, we included an indicator variable for unknown quartile. The second staffing variable we calculated was the quartile of staffing level for licensed nurses (i.e., registered nurses and licensed practical nurses). These quartiles were calculated in the same manner as the DCWs. The final staffing variable was the starting hourly wage of DCWs on a daytime shift.

Management covariates considered were whether the administrator knows all the names of the DCWs and whether the facility had two or more administrators in the last 2 years. To capture resident characteristics, we included the percentage of residents with Alzheimer’s Disease and Related Dementias (ADRD) and with serious mental illness. Market characteristics included an indicator for whether the facility was in a rural or urban area, the total number of NHs and ALs in the county, and the county unemployment rate.

For environmental retention strategies, we included whether the facility engaged in consistent assignment (for NHs defined as 85% of “residents have 12 or fewer DCWs in a 30-day period”; for ALs, defined as a yes response to “direct care workers are consistently assigned to the same group of workers”), empowered DCWs, cross-trained DCWs for various tasks (e.g., DCWs trained to provide activities), DCWs participated in resident care planning meetings, and whether DCWs care for residents they choose or are assigned. To determine whether a facility empowered DCWs, we calculated an empowerment scale developed by Abbott et al. (2019) and defined an empowered facility as having a score of 3 or higher. This empowerment scale includes the following five, equally weighted empowerment practices: DCWs involved in interviews of new staff; scheduling managed by a staff team; DCWs participate in quality improvement teams; DCWs work together to cover shifts; and facility uses a career ladder to retain staff. Financial retention strategies included indicators for whether the facility offered any wage or longevity bonuses, health insurance, a retirement account (i.e., 401k, pension), any paid time off, and tuition reimbursement.

Analytic Approach

In order to align our analytic approach with our three hypotheses, we examined three samples. We examined a sample of all long-term care facilities, which included both NHs and ALs. This sample was used to examine whether NHs and ALs have different retention rates. Next, we conducted analysis on NHs and ALs, separately. This allowed us to examine which factors were associated with retention in NHs and ALs, and to determine if our hypotheses that compared NHs to ALs were true. These hypotheses were whether the factors associated with retention for NHs are similar to ALs, and whether there are differences in the importance of those factors (i.e., the magnitude of the effects is different).

While not shown, we started by comparing summary statistics for observations included and excluded from the analytic sample to ensure they did not affect our results. Excluded facilities were more likely to have unknown staffing levels and did not report information on retention strategies because both sets of variables were from the same module and nonresponses were generally for the entire module. To assure this did not affect our results, we compared the analytic and excluded sample and found few differences for other covariates. We believe that the exclusion of these facilities did not affect our general conclusions.

To test each hypothesis, we calculated summary statistics. Next, we estimated linear regression models where the dependent variable was the DCW retention rate and all covariates except retention strategies were included in the model. We did this because some facilities did not complete the survey section related to retention strategies and limiting the sample to only those that did would reduce the sample size. Then, we estimated a second set of regressions which included all of the covariates, including retention strategies. In all regression analyses, we adjusted for standard errors for heteroscedasticity. This research utilized SAS for data management, Stata for all statistical analyses, and received institutional review board approval.

Results

Sample Characteristics

Table 1 depicts the characteristics of our analytic samples as combined together (NH and AL) and separately. NHs in the sample retained about 61% of their DCWs in 2017 compared to 66% in ALs. In both settings, most facilities are for-profit organizations and affiliated with a chain. About one third of both NHs and ALs had a special care memory unit. Occupancy rates were higher in ALs, and NHs were more reliant on Medicaid. Administrators reported knowing more of their DCWs names in ALs, and administrative turnover in ALs was lower than NHs. The average facility had over 40% of residents with ADRD, and NHs were more likely to have any resident with a serious mental illness. The market characteristics of facilities in both settings were rather similar.

Table 1.

Summary Statistics for All Facilities, NHs, and ALs

Variables	All facilities (n = 1,204)	NHs (n = 739)	ALs (n = 465)
DCW retention rate	62.748 (20.918)	60.704 (18.641)	65.997 (23.761)
Facility structure
Facility is a NH (%)	61.4	100	0
Number of beds	87.5	93.3	78.1
Small facility (1st quartile) (%)	25.1	25.0	25.2
Large facility (4th quartile) (%)	25.2	25.2	25.2
Nonprofit ownership (%)	23.3	20.3	28.0
Chain-affiliated (%)	66.2	68.9	61.9
Memory care special care unit (%)	34.5	33.8	35.5
CCRC (%)	19.0	16.5	23.0
Different building owner and operator (%)	14.3	16.1	11.4
Financial characteristics
Occupancy rate	85.295 (69.424)	82.860 (13.070)	89.165 (110.451)
Medicaid payer-mix	43.392 (30.593)	58.888 (20.444)	18.764 (27.777)
Staffing
DCW staffing level (2nd quartile) (%)	24.1	23.5	24.9
DCW staffing level (3rd quartile) (%)	23.6	23.0	24.5
DCW staffing level (4th quartile) (%)	23.5	23.8	23.0
DCW staffing level (unknown) (%)	6.2	6.6	5.6
Licensed nurse staff level (2nd quartile) (%)	23.9	22.9	25.6
Licensed nurse staff level (3rd quartile) (%)	23.6	23.4	23.9
Licensed nurse staff level (4th quartile) (%)	23.1	23.0	23.2
Licensed nurse staff level (unknown) (%)	6.1	7.8	3.2
DCW starting hourly wage	$10.805 (1.197)	$11.186 (1.085)	$10.199 (1.115)
Management
Administrator knows all DCW names (%)	43.5	32.1	61.7
Two or more administrators in 2 years (%)	33.4	37.6	26.7
Resident characteristics
Residents with ADRD	42.954 (20.675)	45.030 (15.651)	39.655 (26.473)
Residents with serious mental illness	27.876 (24.906)	41.389 (19.469)	6.400 (15.887)
Market
Rural (%)	29.2	29.5	28.8
Number of AL/NHs in county	57.042 (54.373)	56.464 (54.231)	57.959 (54.643)
County unemployment rate	5.142 (0.974)	5.172 (0.968)	5.094 (0.984)

Notes: Summary statistics include mean and standard deviation for continuous variables. Percentages are reported for binary variables. ADRD = Alzheimer’s Disease and Related Dementias; AL = assisted living; CCRC = continuing care retirement community; DCW = direct care worker; NH = nursing home.

Factors Associated With Retention

Table 2 reports the regression results when all covariates were included except retention strategies. The first column reports the regression results when all facilities are included in the regression but NHs and ALs are differentiated by including an indicator for being a NH. The other two columns report regression results when the sample is restricted only to NHs and ALs.

Table 2.

Regression Results for Factors Associated With Retention Excluding Retention Strategies

Variables	All facilities (n = 1,204)		NHs (n = 739)		ALs (n = 465)
	Coeff	SE	Coeff	SE	Coeff	SE
Facility structure
Facility is a NH	−8.068***	2.175
Small facility	−1.104	1.590	−2.259	1.886	1.560	2.885
Large facility	−1.198	1.418	0.697	1.629	−3.931	2.650
Nonprofit ownership	7.984***	1.624	3.389*	2.023	12.169***	2.666
Chain-affiliated	−1.843	1.283	−2.251	1.478	−1.331	2.338
Memory care special care unit	−0.146	1.388	0.439	1.565	−0.548	2.758
CCRC	0.857	1.649	0.139	2.055	0.659	2.789
Different building owner and operator	0.014	1.808	−0.190	2.083	1.743	3.572
Financial characteristics
Occupancy rate (%)	0.009	0.006	−0.080	0.067	0.008	0.006
Medicaid payer-mix (%)	0.023	0.030	0.003	0.046	−0.018	0.043
Staffing
DCW staffing level (2nd quartile)	0.492	1.778	2.067	1.996	−1.119	3.391
DCW staffing level (3rd quartile)	0.456	1.817	1.408	2.061	−1.161	3.463
DCW staffing level (4th quartile)	0.740	2.036	4.054*	2.321	−3.040	3.934
DCW staffing level (unknown)	−1.272	4.586	−6.068	7.838	0.468	5.701
Licensed nurse staff level (2nd quartile)	−1.991	1.726	−1.666	2.038	−2.013	3.178
Licensed nurse staff level (3rd quartile)	−2.584	1.823	−2.349	2.168	−3.648	3.276
Licensed nurse staff level (4th quartile)	−2.171	1.986	−2.930	2.338	−1.834	3.700
Licensed nurse staff level (unknown)	3.371	4.716	6.589	7.365	6.708	7.316
DCW starting hourly wage	0.470	0.551	0.005	0.655	0.791	0.965
Management
Administrator knows all DCW names	0.804	1.289	0.999	1.606	0.478	2.224
Two or more administrators in 2 years	−2.472*	1.336	−3.299**	1.533	−1.829	2.567
Resident characteristics
Residents with ADRD (%)	−0.004	0.034	0.108**	0.052	−0.034	0.048
Residents with serious mental illness (%)	0.068*	0.040	0.045	0.044	0.071	0.087
Market
Rural	2.828*	1.526	1.277	1.781	5.010*	2.786
Number of AL/NHs in county	−0.007	0.013	−0.011	0.016	−0.009	0.023
County unemployment rate	1.891***	0.634	0.088	0.732	4.256***	1.128
Constant	50.356***	7.744	61.927***	11.026	36.716***	13.126
R2	0.072		0.049		0.131

Notes: The table reports the coefficient estimates (Coeff) and standard errors (SEs) for a linear regression with a dependent variable of DCW retention rate. SEs are adjusted heteroscedasticity. ADRD = Alzheimer’s Disease and Related Dementias; AL = assisted living; CCRC = continuing care retirement community; DCWs = direct care workers; NH = nursing home.

***p < .01. **p < .05. *p < .1.

Consistent with our third hypothesis (H3), the regression using all facilities finds that NHs have a DCW retention rate that is 8.0 percentage points lower than ALs. In other words, holding other factors constant, retention is lower in NH settings. Also, consistent with H1, when we examine the coefficient estimates across all three regression models, we find that the coefficient estimates are generally in the same direction. This means that factors that predict retention in NHs also generally apply to ALs.

Even though the directional effects of factors are typically similar in both settings, variables associated with retention (i.e., magnitude and statistical significance) do differ by setting, supporting H2. Among facility structures, not-for-profits have higher retention rates than for-profits, but the effect is nearly 4 times larger in ALs. All other facility structure covariates are not statistically significant. None of the financial characteristics are found to be statistically significant at conventional levels in either setting and the same is true for most of the staffing factors. NHs with a greater proportion of residents with dementia have higher DCW retention rates. Facilities with higher administrator turnover had lower retention, but this was only significant for NHs and the sample overall. ALs in rural areas had higher retention rates at the marginal significance level (10% level). Finally, facilities in counties with higher unemployment rates have higher retention, but the effects are found only to be statistically significant in the all facility and AL samples.

Retention Strategies

The summary statistics and regression results for facilities reporting retention strategies are reported in Table 3. Interestingly, NHs and ALs tend to use most retention strategies at similar rates, with the only exceptions being ALs are more likely to use consistent assignment of DCWs to residents (69% vs 47%) and cross-trained DCWs than NHs (65.6% vs 42.5%). However, NHs are more likely to have DCWs participate in care planning (64.8% vs 47.5%). Most retention strategies are not statistically associated with DCW retention rates. The only exception is DCWs in ALs have lower retention rates if they are cross-trained, but have higher retention rates if they participate in care planning.

Table 3.

Summary Statistics and Regression Results for Environmental and Financial Retention Strategies on DCW Retention Rates

	All facilities (n = 1,065)			NHs (n = 600)			ALs (n = 465)
Variables	(%)	Coeff	SE	(%)	Coeff	SE	(%)	Coeff	SE
Environmental retention strategies
Consistent assignment	56.6	0.711	1.338	47.0	2.273	1.577	69.0	−1.464	2.553
Utilizes 3+ empowerment practices	35.5	1.463	1.436	36.5	0.845	1.744	34.2	1.766	2.424
DCW are cross-trained	52.6	−1.761	1.333	42.5	0.427	1.637	65.6	−5.064**	2.359
DCWs participate in care planning	57.3	2.821**	1.404	64.8	−1.763	1.657	47.5	7.559***	2.335
DCWs choose resident cared for	7.4	−1.980	2.688	9.2	−4.361	2.950	5.2	3.717	5.292
Financial retention strategies
Offers wage increases/bonuses	92.9	−1.845	2.449	95.0	−0.019	3.213	90.1	−2.323	3.444
Offers health insurance	92.9	−2.533	3.365	96.2	2.172	4.092	88.6	−6.158	4.576
Offers retirement account	84.5	0.956	2.251	85.8	−0.891	2.748	82.8	1.480	3.782
Offers paid time off	97.9	−2.137	5.186	98.0	−2.545	7.200	97.8	−1.976	7.165
Offers tuition reimbursement	65.7	1.150	1.491	68.3	2.401	1.886	62.4	0.975	2.356
R2		0.076			0.052			0.139

Notes: The table reports the summary statistics as the percentage of facilities with the following strategy and regression results (coefficient estimates [Coeff] and standard errors [SEs]) for a linear regression with a dependent variable of DCW retention rate. All regressions also control for covariates in Table 2. SEs are adjusted for heteroscedasticity. AL = assisted living; DCWs = direct care workers; NH = nursing home.

***p < .01. **p < .05.

Discussion

The results overall supported our hypothesis that factors associated with retention rates would be similar across the NH and AL settings. Our findings also showed that even though the factors associated with retention are similar across settings, the magnitude of the effects can be different (e.g., effect of not-for-profit ownership).

One difference is higher county unemployment being associated with higher retention rates in ALs that could have implications for facility strategy. The local market significantly affected DCWs in ALs, though it did not affect the retention of NH DCWs. This means the development of benefit packages for AL and NH DCWs may need to be different, as benefit packages in ALs may need to be more responsive to the strength of the local economy. It also suggests the differences between the AL and NH workforces may play a role, such that personal contingencies like being the primary breadwinner motivate DCWs to leave a facility in search of a better job (Dill et al., 2013), but DCWs also leave the industry due to physical health concerns (Rosen et al., 2011).

Another difference is higher turnover of administrators in NHs is significantly associated with lower retention in NHs but not in ALs. According to a recent study, facilities could enhance their support of NH administrators to prevent job dissatisfaction and intent to quit, which usually concerns regulations, staffing problems, and corporate challenges (Nelson et al., 2020). But this still leaves important practice questions, such as what is it about the role of administrator in the NH setting that can affect retention, and can facilities do anything to either increase administrator longevity or to mitigate the negative effects that occur as a result of administrator turnover?

A second important message from this study is that most facility retention strategies did not appear to be associated with retention rates. Our finding differs from prior work (Bishop et al., 2008; Dill et al., 2013; Rosen et al., 2011). These differences could be due to worker-level measures of retention differing from facility-level measures, differences in the quality and definition of retention strategies, or that the strategies can have varying effects depending on the strength of the economy. Our study period is consistent with a strong economy, but it does not take away from the fact that the work of DCWs is difficult, the pay and benefits generally low, the work conditions can be difficult, and the value of the work is not always well recognized by society (Bishop et al., 2008). With limited organizational resources, the retention strategies offered just may not be enough to influence worker retention. Consistent with previous work, retention specialists may need to implement strategies that are tailored to the specific needs of the facility (Pillemer et al., 2008).

Our study found a significant positive effect for involving DCWs in care planning in the AL and total samples, and a negative effect of cross-training in ALs. This suggests a particular sensitivity of the AL DCW workforce to broadening their job responsibilities (Brannon et al., 2007). It is also not surprising that involving DCWs in resident care planning, a management practice that shows respect toward DCWs and values their input, was related to better retention. Involving AL DCWs in care plan meetings on a consistent basis is a practice that practitioners should adopt for the benefit of the workers, residents, and families. What is perhaps surprising from our results is that empowerment practices were not significant in light of work by Berridge et al. (2020). Yet earlier research revealed that job enhancements were not associated with intent to stay and job commitment; lack of uptake of high-performance management practices and high wages in long-term care settings could explain the lack of association between job enhancements and retention or turnover measures (Bishop et al., 2008).

As one of the few studies examining retention across settings, this work offers some important ideas for practice and research, but also has limitations. First, it uses data from one state, and while Ohio is large and generally looks like the nation, it does have some unique system elements. The primary data for the study came from a self-reported survey of facilities completed by the administrator. Their perceptions in some areas, such as staff empowerment, could reflect bias about how they would like their facility to be classified and could be inaccurate. This work also does not include the perspectives of the workers, who need to be heard if researchers are going to really understand retention. Finally, because ALs and NHs are different settings, we were required to construct covariates that could be measured in the same context in both settings. There could be factors that affect retention in one setting that are not valid in another, such as the proportion of post-acute care patients in NHs.

Furthermore, this work also highlights how little we know about retention of DCWs. Study results explain little of the variation in retention rates, as measured by R², in explaining the factors associated with retention. While additional data can be added to the information collected from administrators and from administrative data, it is also clear that putting more resources into primary data collection with DCWs and their supervisors is critical. Until we understand the stressors and rewards of the work and how we can design strategies to better support workers, these issues will remain. Future studies can explore DCW perceptions about financial benefits, such as not just offering, but the cost of health care premiums and the monetary value of bonuses.

In conclusion, the fact that the overall retention rates in both settings indicate that 4 in 10 workers are not retained in a 1-year time period is indicative of a larger industry problem. As noted, the work is hard and not well supported either internally or externally. The financial pressures faced by NHs and ALs are not going to subside under the current funding approaches and increased costs of COVID-19. With Medicaid and out-of-pocket expenditures, the two major sources of revenue, the enormous cost pressures will remain a constant. Because DCWs represent a large component of facility expenditures, continued efforts to limit costs are likely. While not all retention strategies require additional funds, many do, such as retention specialist models that involve investments in training (Maier, 2002) and where lack of funding was a barrier to implementation (Pillemer et al., 2008). Thus, at a time when financial pressures are rising, the likelihood of increased staff investment is low. This suggests that unless changes to the financing and structure of the long-term care industry occur, the ability to affect retention could be increasingly more difficult.

Acknowledgments

We would like to thank Matt Nelson for his assistance with obtaining the Biennial Survey data sets and his helpful comments on this paper.

Funding

None declared.

Conflicts of Interest

J. R. Bowblis provides consulting services to the health care industry, which includes long-term care providers. All other authors report no known conflicts of interest.