Measuring Caregiver Retention in Nursing Homes

Abstract

Background and Objectives

Retention of nursing home caregivers is examined. This represents the concept of continuously employing the same caregivers in the same facility for a defined period of time. In this research, several measures of caregiver retention are examined and the utility of these measures for practitioners and policy makers is discussed.

Research Design and Methods

A survey of nursing home administrators conducted in 2016 was used to collect staffing data from 2,898 facilities. This was matched with Nursing Home Compare and the Certification and Survey Provider Enhanced Reporting data. The association of four measures of retention for each of three types of caregivers with six quality indicators was examined.

Results

The descriptive statistics show rates of retention at 5 years for nurse aides (NAs), registered nurses (RNs), and licensed practical nurses to be low. The regression estimates show some support for the relationship that high caregiver retention is associated with better overall quality. The relationship was strongest for NAs and RNs. Support was also found for the notion that different measures of retention were more/less associated with quality. The 3- and 5-year retention measures had the strongest associations with the quality indicators.

Discussion and Implications

The findings presented provide some evidence that caregiver retention may be an important metric that can be used as a means of improving quality of care in nursing homes. However, the findings also show practitioners and policy makers should be more nuanced in the use of caregiver retention metrics.

Nursing homes, Quality, Retention, Stability, Staffing

Retention of nursing home caregivers very generally represents the concept of continuously employing the same caregivers in the same facility for a defined period of time. However, measures of caregiver retention vary. For example, which caregivers are included and the periods of time used in the measures currently vary substantially (Barry et al., 2005; Thomas et al., 2013). In this research, several measures of caregiver retention are examined along with their association with quality indicators, and the utility of these measures for practitioners and policy makers is discussed.

Many nursing homes are known to have staffing issues that affect quality of care. For example, for decades research has shown staffing levels are associated with quality (Harrington & Edelman, 2018; Hyer et al., 2009). A review of this literature is provided by Backhaus et al. (2014). Research on nursing home staffing has expanded beyond just staffing levels to include multiple other staffing issues of concern. For example, turnover of staff has been described as high and also influencing quality of care (Donoghue, 2010). Agency staff (i.e., temporary staff) have been shown to be sometimes problematic and high levels influence quality of care (Bourbonniere et al., 2007). These multiple staffing issues have helped inform practice decisions and broaden the policy debate in this area. Retention may be a further important staffing issue for nursing homes. As we describe below, higher levels of caregiver retention may be advantageous, ultimately influencing quality of care.

The belief that caregiver retention can affect the quality of care of nursing homes is evident from both government and provider groups. The Centers for Medicare and Medicaid Services (CMS; https://www.cms.gov/Outreach-and-Education/American-Indian-Alaska-Native/AIAN/LTSS-TA-Center/program-examples/videos) and provider organizations such as the American Health Care Association (AHCA)/National Center for Assisted Living (NCAL; https://educate.ahcancal.org/products/key-strategies-to-retain-new-hires-and-reduce-employee-turnover) and LeadingAge (https://www.leadingage.org/sites/default/files/Direct%20Care%20Workers%20Report%20%20FINAL%20(2).pdf) have actively encouraged nursing homes to implement caregiver retention activities. However, as described below, little empirical evidence exists that caregiver retention is associated with quality of care.

Existing Literature

Because I did not identify a prior published review of caregiver retention, I review the relevant research below. In this literature review, existing definitions of caregiver retention, specifications of caregiver retention, measures used, levels of caregiver retention reported, and potential associations with quality of care were specifically examined. These factors were examined because they were most germane to the empirical analyses examining measures of caregiver retention and their association with quality indicators for different caregivers (described below). A summary of the findings of the literature review is given in Table 1.

Table 1.

Summary of Studies Including Caregiver Retention and the Association With Quality Indicators

Author(s)	Definition of retention	Specification	Retention measure	Rate of retention	Sample, sample size, and data source	Findings
Quality indicators
Barry et al., 2005	Not reported	Not reported	Proportion of NAs employed at the facility for 2 years or longer	Median NA retention was 52% for 2 years or longer	156 facilities in Maine, Mississippi, New York, and Ohio. 1995 MDS and OSCAR data	High retention facilities associated with a lower incidence of pressure ulcers (p < .01)
Castle & Engberg, 2007	Not reported	Not reported	Percent of RNs with 5 or more years tenure at the facility. Percent of LPNs with 5 or more years tenure at the facility. Percent of NAs with 5 or more years tenure at the facilitya	RNs 16.5% retention; LPNs 30.4% retention; NAs 14.9% retention	Survey of 1,071 facilities. 2003 Nursing Home Compare and OSCAR data	Quality index negatively associated with RN stability (p = .045); positively associated with NA stability (p = .058)
Castle & Engberg, 2008	Not reported	Not reported	Percent of RNs with 5 or more years tenure at the facility. Percent of LPNs with 5 or more years tenure at the facility. Percent of NAs with 5 or more years tenure at the facility	RNs 19.9% retention; LPNs 17.9% retention; NAs 14.3% retention	Survey of 6,005 facilities. 2004 Nursing Home Compare and OSCAR data	NA retention associated with five of six quality indicators (p < .05). LPN retention associated with zero of six quality indicators (p < .05). RN retention associated with two of six quality indicators (p < .05)
Castle et al., 2020	The proportion of staff continuously employed in the same facility for a defined period of time	Full-time NAs	NAs consistently employed in the facility for 1 year or more, 2 years or more, and 3 years or more (these were reported for full-time NAs)	53.2% for 1 year; 41.4% for 2 years; and 36.1% for 3 years	Survey of 5,000 facilities. 2016 CASPER data	NA 1-year, 2-year, and 3-year retention associated with all deficiency citations. NA 2-year and 3-year retention associated with quality of care deficiency citations. NA 3-year retention associated with J, K, L, deficiency citations
Donoghue, 2010	Employing the same caregivers in the same facility for a defined period of time	Not reported	Percent of FTE nurses employed for more than 1 year	RNs 67.3% retention; LPNs 68.4% retention; NAs 62.5% retention	2004 National Nursing Home Survey	RN retention, LPN retention, and NA retention associated (p < .001) with the number of months the DON was employed
Thomas et al., 2013	Nursing home stayers	Licensed nurse hours per resident day	Average licensed nurses employed for at least a year in each quarter for the year/average number of licensed nurses employed during each quarter for the year	64.4% retention (licensed nurses)	681 Florida facilities from 2002 to 2009. LTCFocUs data	Licensed nurse (i.e., RNs and LPNs) retention associated (p = .04) with 30-day rehospitalization rate
Other outcomes
Garland et al., 1988	Nursing home stayers	Not reported	NAs working in the same job for the same organization for 10 years	Not reported	79 NAs (stayers) compared with 35 NAs (leavers)	NAs staying at a facility had a higher value of job security (p = .009)
Rosen et al., 2011	Nursing home stayers	Not reported	NAs working in the same job for the same organization for 12 months	85.8% retention	620 NAs in PA responding to a survey	NAs staying at a facility had higher job satisfaction (p = .041)
Berridge et al., 2018	Not reported	Not reported	NAs worked at the facility for at least 12 months	Not reported	2,034 NHAs surveyed in 2009/2010	Staff empowerment associated (p < .01) with higher retention

Note: LTCFocUs = long-term care: facts on care in the US; MDS = minimum data set; NA = nurse aide; OSCAR = The Online Survey Certification and Reporting; RN = registered nurse; LPN = licensed practical nurse; FTE = full-time equivalent; DON = Director of Nursing; NHA = nursing home administrator; CASPER = Certification and Survey Provider Enhanced Reporting.

^aStability was used in this study to be synonymous with retention.

Definitions of Retention

Many publications have used “retention” as a generic term and have not provided much specificity as to what is meant by the concept. It is evident that the concept is used to mean “keeping people,” but more detail is seldom provided. Only two studies defined retention. Castle et al. (2020) defined retention as “the proportion of staff continuously employed in the same facility for a defined period of time” (p. 1), and Donoghue (2010) defined retention as “employing the same caregivers in the same facility for a defined period of time” (p. 27).

Specification of Retention

Studies providing specifications of retention tended to follow other staffing variables (such as turnover) as most have one or more of three different types of staff included (i.e., nurse aides [NAs], registered nurses [RNs], and licensed practical nurses [LPNs]). For example, Barry et al. (2005) examined the retention of NAs. However, less detail is provided in other areas. For example, no detail is provided on how to account for full- or part-time status and if the staff member is counted if they have changed positions within the same facility (given in column 2 of Table 1).

Measures of Retention

The empirical research articles reviewed used different measures of retention (see column 2 of Table 1). These included caregivers employed at the facility for 1 or more years (Donoghue, 2010), 2 years or longer (Barry et al., 2005), and 5 or more years (Castle & Engberg, 2007, 2008). A more recent study has examined caregiver retention at 1 or more years, 2 or more years, and 3 years or longer (Castle et al., 2020). It would seem that no single accepted measure exists in this area and the measures used have been structured around available data.

Retention Rates

Levels of retention rates reported are given in column 3 of Table 1. The reported levels of retention varied considerably and were different for different caregivers. For NAs, national figures show a 62.5% retention rate that was reported for 1 year, 68.4.4% for LPNs, and 67.3% for RNs (Donoghue, 2010). This information is now dated (i.e., from 2004) and no more recent national estimates were identified. However, other more current data (using a smaller sample size) show a 1-year retention rate of 85.8% for NAs (Rosen et al., 2011).

Association of Retention With Quality

Of the eight studies identified in this area, caregiver retention rates were associated with multiple quality indicators including the incidence of pressure ulcers (Barry et al., 2005) and rehospitalization rates (Thomas et al., 2013). However, the relationships did seem to vary by the caregiver. For example, Castle and Engberg (2008) found NA retention to be associated with five of six quality indicators, LPN retention to be associated with zero of six quality indicators, and RN retention to be associated with two of six quality indicators (see column 6 of Table 1).

Areas of Imprecision

As part of the literature review, two areas of imprecision were identified with respect to caregiver retention. First, caregiver “stability” was sometimes used imprecisely when retention was actually the construct under investigation (Castle & Engberg, 2007). This would seem to reflect the general imprecision in the literature in this area. This also possibly reflects the potential familial relationship between stability, retention, and turnover. Caregiver stability is a broader concept and is a composite of both retention and turnover. That is, caregiver stability is likely influenced by both caregiver retention rates and caregiver turnover rates.

A second area of imprecision is that turnover is often used as the reciprocal of retention. In fact, retention and turnover are different types of measures, and they are not necessarily reciprocal. For example, a facility can have both high retention and high turnover. This occurs when many staff remain, but other positions are regularly vacated. Van Der Merwe and Miller (1971) provide an empirical elaboration of this relationship; Thomas et al. (2013) specifically discuss this in the nursing home context.

Summary of Existing Literature

The existing literature addressing caregiver retention in nursing homes was found to be sparse. Definitions of caregiver retention were found to be broad and retention was generally poorly specified. The measures used varied, and possibly as a result levels of caregiver retention reported also varied. With the caveat that only six empirical studies were identified examining quality and a limited number of quality indicators were used, a potential association with quality would appear to be evident.

Thus, the utility of retention as a characteristic for nursing homes and policy makers to examine and improve would appear to be currently limited. One limitation occurs because measures of caregiver retention vary. If some facilities track 1-year retention rates and others track 2-year retention rates then quality improvement tools such as benchmarking cannot effectively be developed. A second limitation is that the validity of the various measures is unknown. The retention measures seem to have different associations with quality indicators (Castle et al., 2020). A single retention measure with the highest association with quality may prove beneficial. A third limitation is that retention is seldom examined for different caregivers. The influence of retention on quality of care may be different for different caregivers.

The empirical analyses presented here address these three limitations. The empirical analyses include NAs, RNs, and LPNs. The association of retention of these three types of caregivers with six quality measures (QMs) is examined. Four caregiver retention measures are used for each caregiver, that is: (a) continuously employed for 1 or more years, (b) continuously employed for 2 or more years, (c) continuously employed for 3 or more years, and (d) continuously employed for 5 or more years.

Theoretical Framework and Hypotheses

The focus of this study was on understanding the measurement of caregiver retention in nursing homes. As described above, the empirical analyses examine multiple measures of caregiver retention and the association of these different measures with quality indicators. Thus, the investigation was guided by classical measurement theory (Campbell et al., 1954). A similar classical measurement theory approach was used by others in clarifying other staffing measures in nursing homes (Castle & Engberg, 2008). As previously noted, “the concept of retention traditionally has been more difficult to quantify than turnover” (Benedict et al., 1989, p. 78). And, as the literature review above identified significant differences in specifications of caregiver retention and measures used exist. This study addresses facets of construct validity, including associations with criterion measures of quality and comparative validity of different approaches to measuring retention.

Two hypotheses (H) are examined. The first hypothesis is that high caregiver retention will be associated with better overall quality (H1). This follows the findings of other studies presented in the literature review and adds six additional quality indicators to expand our understanding in this area. The second hypothesis is that different measures of retention will be more/less associated with quality (H2). This also follows the findings of other studies presented in the literature review that no single accepted measure exists in this area, but that the measure used may vary in the association with quality.

Data and Methods

Primary and secondary data were combined for this analysis. A survey of nursing home administrators (NHAs) was the source of primary data collected by the authors and was used to identify several staffing measures, including retention. Nursing Home Compare (NHC) and the Certification and Survey Provider Enhanced Reporting (CASPER) data were used as sources of secondary data. The QMs and facility characteristics of the nursing homes, respectively, came from these data sources.

NHA Survey

The NHA survey was conducted by mail in 2016. Previously, the same survey items had been used in national samples to examine staffing in nursing homes. This included staffing levels, turnover, and agency staff use (Castle et al., 2008). The 2016 survey added additional items addressing retention.

All nursing homes in the contiguous U.S. states were included in the sampling frame, but this excluded small (i.e., <30 beds), large (i.e., >800 beds), and hospital-based facilities. From this sampling frame, 5,000 nursing homes were randomly selected to receive the mail survey. Address information came from the CASPER data (described below), and the survey was sent to the NHA. Following recommended survey practices (Dillman, 1991), two repeat survey mailings were used.

The survey items included retention, turnover, agency staff use, and staffing levels. Most of these survey items have been described elsewhere (Castle & Engberg, 2008). The overall questionnaire psychometrics have also been described elsewhere (Castle & Engberg, 2008).

Four questions were used addressing retention: Staff consistently employed in the facility for (a) 1 year or more, (b) 2 years or more, (c) 3 years or more, and (d) 5 years or more. Respondents were asked to give the percent of staff for each of these retention questions, reported separately for full- and part-time staff, and by staff type (i.e., NAs, LPNs, and RNs). This followed the same approach used for the other staffing items on the questionnaire.

Nursing Home Compare

NHC is a web-based source of information for all Medicare and/or Medicaid-certified nursing facilities (www.Medicare.gov/NHCompare/home.asp). It provides descriptive information on nursing homes along with numerous quality indicators. Several of the quality indicators reported on the NHC website were used in this initiative.

The quality indicators in NHC are termed QMs. They are derived from the Minimum Data Set and are updated quarterly (General Accounting Office [GAO], 2002; Perraillon et al., 2019). The development of NHC and the QMs is extensively described in a technical report (Abt, 2004).

Given the large number of QMs, and given that little prior work exists examining caregiver retention, not every QM was examined. Studies in the hospital literature have examined quality indicators termed “nursing quality indicators (NQIs)” (Needleman et al., 2001). That is, these quality indicators are more sensitive to care processes provided by staff than other measures. Using a similar approach, the QMs used here were chosen to be sensitive to care processes provided by staff. These QMs are restraints, catheter use, inadequate pain management, and pressure ulcers and are all listed as potential NQIs (Montalvo, 2007). The definitions of these QMs are provided in Table 2 and include information from 2016.

Table 2.

Dependent and Independent Variables Used in Analyses

Variable	Definition	Mean or percent (SD)
Dependent variablesa
Pain (long stay)	Percent with moderate to severe pain (long-stay residents)	4.9 (4.1)
Pressure sores (low risk)	Percent low-risk residents with pressure sores (long-stay residents)	5.5 (3.3)
Pressure sores (high risk)	Percent high-risk residents with pressure sores (long-stay residents)	5.9 (5.0)
Physical restraint	Percent physical restraint use (long-stay residents)	6.1 (3.0)
Catheterized	Percent had a catheter inserted and left in the bladder (long-stay residents)	9.5 (6.1)
Pain (short stay)	Percent with moderate to severe pain (short-stay residents)	16.5 (7.4)
Independent variables
Staffing characteristicsb
RN staffing	FTE RNs per 100 residents (including full-time and part-time workers)	10.6 (6.2)
LPN staffing	FTE LPNs per 100 residents (including full-time and part-time workers)	12.5 (6.6)
NA staffing	FTE NAs per 100 residents (including full-time and part-time workers)	32.1 (8.4)
Professional staff mix	Ratio of professional staffing levels (i.e., RNs + LPNs) to other caregiver staffing levels (i.e., NAs)	0.25 (0.4)
RN turnover	The total number of RN staff who left employment during 2016 divided by the total number of RN staff who were employed during this period	42.3% (7.5)
LPN turnover	The total number of LPN staff who left employment during 2016 divided by the total number of LPN staff who were employed during this period	47.1% (14.1)
NA turnover	The total number of NA staff who left employment during 2016 divided by the total number of NA staff who were employed during this period	58.4% (16.5)
RN agency	Percent of FTE positions filled by agency RNs in 2016	8.9% (2.1)
LPN agency	Percent of FTE positions filled by agency LPNs in 2016	11.4% (4.6)
NA agency	Percent of FTE positions filled by agency NAs in 2016	12.2% (4.9)
Facility characteristicsb
Organizational size	Number of beds	129 (80)
Ownership	For-profit	68%
Chain membership	Member of a nursing home chain	59%
Occupancy	Average daily occupancy rate	88% (12)
Medicaid occupancy	Average percent of residents in the facility with Medicaid as the payor	61% (20)
Resident case-mix	The average score for three ADLs (eating, toileting, and transferring). Constructed by giving a score of 1 for low assistance, 2 for moderate assistance, and 3 for high need for assistance summed for each ADL	1.9 (0.9)
Market characteristics
Competitiona	Herfindahl index. The sum of each facility’s squared percentage share of beds in the county for all facilities in the county (0–1)	0.2 (0.2)
Medicaid reimbursementb	State average daily (payment rate for Medicaid residents ($)	95.3

Notes: NA = nurse aide; RN = registered nurse; LPN = licensed practical nurse; FTE = full-time equivalent; ADL = activities of daily living. Figures from 2016 are shown. N = 2,898 nursing facilities.

^aVariables were taken from Certification and Survey Provider Enhanced Reporting data.

^bVariables were from primary data collection.

CASPER Data

CASPER data are collected as part of the Medicare and Medicaid certification process. Because most nursing homes participate in Medicare and/or Medicaid, the data include information from almost all facilities in the U.S. Certification that occurs approximately yearly (Office of Inspector General, 2003).

The CASPER data are extensive and include multiple items. One group of items includes facility characteristics (e.g., the number of beds and occupancy level). A second group of items includes resident characteristics (e.g., the number of residents with limitations in activities of daily living). A third group of items includes quality characteristics (e.g., deficiency citations). The State Operations Manual (2010) comprehensively lists all of the data elements available in the CASPER. The items used in this research are provided in Table 2 and include information from 2016.

Model Specification

Definitions of the variables and descriptive statistics are provided in Table 2. This includes the quality, staffing, and facility variables used in the analyses. All of the variables included in the model specification have well-established associations with quality indicators, especially when examining staffing characteristics (Barry et al., 2005; Hyer et al., 2009; Smith et al., 2019).

Statistical Methods

Descriptive statistics for all of the variables are presented. This consists of means, percentages, and standard deviations. Multicollinearity and collinearity levels among the variables were examined (SAS Institute, 1999) prior to conducting the multivariate analyses.

Missing data and improbable values did occur for the retention variables collected as part of the survey. Facilities that returned the survey but did not provide any retention information were excluded from the analyses (N = 267). The four retention measures should have declining or at least equal values. For example, a 2-year retention rate should not be higher than a 1-year retention rate. Facilities that did not report declining or equal values were excluded from the analyses (N = 74).

For the four retention measures, missing data also occurred: (a) 1 year or more (N = 155), (b) 2 years or more (N = 159), (c) 3 years or more (N = 171), and (d) 5 years or more (N = 192). In these cases, multiple methods of imputation were used (details available from the authors). Possibly because of the small number of cases, imputation did not influence the results in any way.

Missing data also occurred for the turnover (N = 121), agency staff use (N = 194), and staffing level (N = 66) variables. These facilities were excluded from the analyses. Based on previously reported statistics, some (N = 109) levels of turnover, agency staff use, and staffing levels appeared improbable. In these cases, multiple methods of imputation were used; however, imputation or the method of imputation did not influence the results reported.

Least squares regression was used in the analyses because all of the dependent variables were approximately normally distributed. Standard errors were estimated using a Huber–White sandwich estimator that accounts for unmodeled correlation of quality within counties (Hastie et al., 2001).

Results

The survey response rate was 71% giving a sample of 3,550 facilities, and after the exclusions described above the analytic sample consisted of 2,898 facilities. The facilities were located in 829 counties (i.e., markets). Characteristics such as ownership and size of these facilities were examined for representativeness excluding those with less than 30 beds, more than 800 beds, and hospital-based facilities (analysis not shown). The analytic sample was representative of nursing facilities in the United States (p < .05).

Table 2 presents descriptive statistics for the variables used in the analysis (excluding the retention measures). The scores for the dependent variables were all similar to national averages (p < .05).

Table 3 presents descriptive statistics for the retention variables used in the analysis. An average rate of 53.2% was reported for the 1-year NA retention measure which declined to 17.6% with the 5-year NA retention measure. An average rate of 65.5% was reported for the 1-year LPN retention measure which declined to 22.4% with the 5-year LPN retention measure. An average rate of 65.6% was reported for the 1-year RN retention measure which declined to 17.6% with the 5-year RN retention measure.

Table 3.

Caregiver Retention Variables Used in Analyses

Variablea	Definition	Percent (SD)
NA retention (1 year)	NAs consistently employed in the facility for 1 year or more (full-time NAs)	53.2 (8.1)
NA retention (2 years)	NAs consistently employed in the facility for 2 years or more (full-time NAs)	41.4 (9.6)
NA retention (3 years)	NAs consistently employed in the facility for 3 years or more (full-time NAs)	36.1 (9.9)
NA retention (5 years)	NAs consistently employed in the facility for 5 years or more (full-time NAs)	17.6 (9.2)
LPN retention (1 year)	LPNs consistently employed in the facility for 1 year or more (full-time LPNs)	62.5 (10.0)
LPN retention (2 years)	LPNs consistently employed in the facility for 2 years or more (full-time LPNs)	40.7 (9.8)
LPN retention (3 years)	LPNs consistently employed in the facility for 3 years or more (full-time LPNs)	32.1 (9.5)
LPN retention (5 years)	LPNs consistently employed in the facility for 5 years or more (full-time LPNs)	22.4 (8.3)
RN retention (1 year)	RNs consistently employed in the facility for 1 year or more (full-time RNs)	65.6 (8.1)
RN retention (2 years)	RNs consistently employed in the facility for 2 years or more (full-time RNs)	52.8 (8.7)
RN retention (3 years)	RNs consistently employed in the facility for 3 years or more (full-time RNs)	44.5 (8.9)
RN retention (5 years)	RNs consistently employed in the facility for 5 years or more (full-time RNs)	26.4 (9.3)

Notes: NA = nurse aide; RN = registered nurse; LPN = licensed practical nurse. Figures from 2016 are shown. N = 2,898 nursing facilities.

^aVariables were from primary data collection.

Table 4 presents the coefficient estimates for the regressions of the six QMs. The QMs are presented in the columns and the rows show the estimated coefficients for the different retention measures. Standard errors, adjusted for clustering within the market, are in parentheses below the coefficient estimates. For NA retention at 1 year, one of the six coefficients was significant (p < .05), whereas for NA retention at 5 years, four of the six coefficients were significant (p < .05). For LPN retention at 1 year, zero of the six coefficients were significant (p < .05), whereas for LPN retention at 5 years, two of the six coefficients were significant (p < .05). For RN retention at 1 year, one of the six coefficients was significant (p < .05), whereas for RN retention at 5 years, four of the six coefficients were significant (p < .05).

Table 4.

Results of Least Squares Regressions Examining Caregiver Retention Measures Related to Six Quality Indicators

Caregiver retention measures	(1) Percent physical restraint use (long-stay residents)	(2) Percent with moderate to severe pain (long-stay residents)	(3) Percent low-risk residents with pressure sores (long-stay residents)	(4) Percent high-risk residents with pressure sores (long-stay residents)	(5) Percent had a catheter inserted and left in the bladder (long-stay residents)	(6) Percent with moderate to severe pain (short-stay residents)
NA retention (1 year)	−0.231(0.129)	−0.209 (0.135)	−0.271 (0.190)	−0.131* (0.070)	−0.105 (0.110)	−0.268 (0.169)
NA retention (2 years)	−0.093* (0.035)	−0.237 (0.199)	−0.193 (0.156)	−0.170** (0.062)	−0.167 (0.099)	−0.152 (0.144)
NA retention (3 years)	−0.139** (0.035)	−0.235 (0.300)	−0.182 (0.029)**	−0.177 (0.065)**	−0.156 (0.045)*	−0.159 (0.131)
NA retention (5 years)	−0.178** (0.036)	−0.245 (0.302)	−0.167 (0.025)**	−0.156 (0.061)**	−0.130 (0.036)*	−0.196 (0.224)
LPN retention (1 year)	−0.311 (0.324)	0.145 (0.124)	−0.167 (0.129)	−0.318 (0.142)	0.300 (0.225)	−0.226 (0.162)
LPN retention (2 years)	−0.239 (0.211)	0.137 (0.055)*	0.180 (0.112)	−0.150 (0.120)	−0.265 (0.192)	−0.231 (0.221)
LPN retention (3 years)	−0.222 (0.220)	−0.200 (0.101)*	−0.133 (0.105)	−0.185 (0.152)	−0.251 (0.172)	−0.2180 (0.201)
LPN retention (5 years)	−0.190 (0.154)	−0.189 (0.051)**	−0.119 (0.116)	−0.129 (0.061)*	−0.228 (0.125)	−0.199 (0.169)
RN retention (1 year)	−0.133 (0.035)**	−0.137 (0.150)	−0.111 (0.099)	−0.171 (0.104)	−0.146 (0.135)	−0.124 (0.101)
RN retention (2 years)	−0.140 (0.033)**	−0.140 (0.106)	−0.120 (0.111)	−0.189 (0.049)*	−0.197 (0.124)	−0.143 (0.145)
RN retention (3 years)	−0.117 (0.030)**	−0.201 (0.163)	−0.217 (0.101)*	−0.186 (0.044)*	−0.203 (0.099)*	−0.190 (0.139)
RN retention (5 years)	−0.125 (0.030)**	−0.221 (0.144)	−0.181 (0.039)**	−0.159 (0.039)**	−0.151 (0.069)*	−0.166 (0.122)

Notes: NA = nurse aide; LPN = licensed practical nurse; RN = registered nurse. Please note that the model specification also included staffing, turnover, agency, organizational size, ownership, chain membership, occupancy, Medicaid occupancy, resident case-mix, competition, and Medicaid reimbursement (Table 2).

*p < .05, **p < .01, ***p < .001.

Discussion

In this research, several measures of NA, LPN, and RN caregiver retention and the association with quality indicators are examined. Both the descriptive and multivariate findings provide important information.

The descriptive findings would seem to indicate that caregiver retention in many facilities is problematic. For NAs, it would appear that in many facilities barely half stay 1 year and this plummets to 17% by 5 years. For LPNs and RNs, a slightly different pattern results. More LPNs and RNs stay for at least a year; however, the 5-year retention rates are low.

Further complicating the staffing issues in nursing homes, each issue may have its own etiology. For example, as previously noted, the reasons for high or low turnover may not be applicable to high or low retention (Donoghue, 2010). Very little research has examined how to influence caregiver retention. High staff empowerment practices were found by Berridge et al. (2018) to be associated with higher retention. And, Pillemer et al. (2008) propose using a “retention specialist” in nursing homes.

The importance of retention as a staffing issue is highlighted by the multivariate findings examining H1. Some support was found for high caregiver retention as associated with the better overall quality. The relationship was strongest for NAs and RNs; in both cases, four of the six QMs were significant (at the 3- and 5-year retention levels). However, very little support was found for the relationship with LPNs. With our data, it is not possible to identify why the relationships identified differ for LPNs as opposed to NAs and RNs. We speculate that LPNs are less involved with the specific QMs used in this investigation due to their roles in caregiving which may entail less resident contact.

In H2, the premise that different measures of retention would be more/less associated with quality was examined. Strong support was found for this relationship. However, again the relationships differed for the three caregivers. For LPNs, the 5-year retention measure had the strongest association with the QMs. For both NAs and RNs, the 3- and 5-year retention measures had the strongest associations with the QMs.

In retrospect, it would seem probable that the longer retention measures would have a stronger association with the QMs. Retention implies that caregivers have appropriately learned institutional protocols, they have relationships with other staff and residents, and they provide institutional memory that can help guide ongoing decisions (Thomas et al., 2013). A 3- to 5-year time horizon for this accumulation of knowledge would appear reasonable.

One goal of this research was to identify a measure of retention that could be further examined as a potential industry standard moving forward. Such a single measure was not identified. We do note that such a single measure would need extensive testing beyond this empirical analysis. This testing at the very least would include the suggestions for future investigation provided below.

Nevertheless, two findings are significant. First, it is clear that the measure of retention used is important with respect to the associations with quality indicators. Using a 3- or 5-year measure would be a starting point for standardizing retention measures. Second, it is clear that for all caregivers a 1-year measure is not optimal. All measures longer than this time frame seem to be more sensitive to the association with quality (at least up to 5 years).

These findings may be significant for practitioners and policy makers. Two decades ago, the GAO (2001) identified retention of caregivers in nursing homes as problematic. Since that time CMS has promoted caregiver retention activities for nursing homes with the belief that this may help improve quality. This includes a satisfaction survey aimed at increasing retention (www.mcknights.com/news/cms-announces-tools-to-help-nursing-homes-boost-employee-satisfaction/). However, little specificity is provided by CMS, and improving overall retention is the focus.

Improving overall caregiver retention may be an important goal, and initiatives such as those by CMS can certainly increase awareness for retention as a potentially important issue. Nevertheless, many nursing homes have both limited resources and multiple staffing issues, so it may help facilities to target retention activities. A potential industry standard to focus upon and a uniform measure would certainly be helpful. Again, unfortunately, our research does not provide a specific measure for this. Our findings are somewhat limited, but promoting caregiver retention beyond 1 year seems warranted.

CMS recently introduced electronic staffing data submission by nursing homes for inclusion in NHC. This Payroll-Based Journal (PBJ) data collect information on staffing levels, turnover, and agency staff (www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/NursingHomeQualityInits/Downloads/PBJ-Policy-Manual-Final-V25-11-19-2018.pdf). An extension of this work with PBJ could include a retention measure (but again would likely require a potential industry standard measure). Caregiver retention measures could be included in NHC. This development may be important, as with most self-reported staffing information from nursing homes (including our study), the validity and reliability of the information can be suspect.

AHCA/NCAL and LeadingAge as provider organizations could also move beyond simple 1-year caregiver retention metrics. Benchmarking information and retention statistics going beyond simple 1-year figures could promote the idea that longer periods of caregiver retention are important.

Suggestions for Future Investigation

Given that little difference was found between the 3- and 5-year retention measures (for NAs and RNs), some further granularity is needed with respect to examining these measures. Clearly, using a 4-year and 6-year measure would be useful in this analysis. However, a continuous metric may be even more advantageous. A continuous score could be used to show the functional form between retention and quality.

The specifications used for measuring retention may be further informed by the turnover literature. Many turnover measures exist; however, work with the PBJ turnover measures mentioned above may be helpful in further standardizing and developing retention measures. Also, with respect to specification, using job tenure may be useful. That is, caregivers’ retention in their current job may be more strongly associated with quality than their retention in the facility (wherein, they may have changed jobs).

Six QMs are examined. Many other QMs could also be used in additional research. Moreover, the QMs are known to provide more of a clinical picture of quality. Quality indicators that include measures coming from residents and families would add to our understanding of caregiver retention. Arguably, measures such as satisfaction should be very sensitive to caregiver retention.

As noted above, the relationship between the different staffing characteristics is complex. Following other work identifying some of these relationships may be important. For example, prior work examining multiple staffing characteristics has included interaction terms among the staffing variables (Castle & Engberg, 2008). Interaction terms help identify if the effect of one characteristic is influenced by the level of a different characteristic. In our case, 30 interactions between the staffing characteristics were possible (i.e., 10 staffing characteristic variables multiplied by three caregiver retention variables). Thus, some further conceptual modeling in this area may be helpful to potentially reduce the number of interactions. Moreover, many of the relationships between staffing variables and quality are also nonlinear and may also need to be taken into account (Castle & Engberg, 2008).

It is also known that the top management of nursing homes can have a profound influence on quality of care. Top management consists of the NHA and the Director of Nursing (and possibly the medical director). Research has shown that high (low) top management turnover is associated with high (low) caregiver turnover (Castle, 2005). The same relationship may also be true for top management retention. A retention measure for top management may be of interest as an additional metric to examine.

Limitations

The analyses only include full-time caregivers. The questionnaire used included items to collect information on part-time staff. Several facilities were not able to provide retention information for all of these part-time staff, therefore they were excluded from the analyses.

Some bias could occur in measuring retention as having been consistently employed in the facility for X years or more. Caregivers could leave a facility then return at a different date. In these cases, the most recent period of employment would be counted. This would exclude the prior period of employment causing some bias; although, we believe this bias is likely minor.

The different measures of retention may be subject to differential bias. For example, the 1-year retention rate may be easier to report than a 5-year or more retention rate. This may also be reflected in the higher level of missing data found for the 5-year rate, compared to the other measures.

The results presented are cross-sectional, and as such, no causal interpretation can be made. Higher levels of caregiver retention may cause better quality; nevertheless, it is also possible that nursing homes with better quality promote higher caregiver retention.

Conclusions

Some support for the association of caregiver retention with quality was identified. However, the association with the QMs was strongest for NAs and RNs. Support was also identified for the use of specific retention measures. The 3-year and 5-year retention measures seemed to be most advantageous. The findings presented provide some evidence that practitioners and policy makers should be more nuanced in the use of caregiver retention as a means of improving the quality of care in nursing homes.

Funding

This study was supported by the Agency for Healthcare Research and Quality (1K18HS021980-01A1).

Conflict of Interest

None declared.