Unionization, Ownership Status, and Direct Care Worker Turnover

Geoffrey M Gusoff; Heeeun Jang; Daniel Spertus; Kiran Abraham-Aggarwal; Ariel Avgar; Madeline Sterling

doi:10.1001/jamanetworkopen.2026.4636

. 2026 Apr 2;9(4):e264636. doi: 10.1001/jamanetworkopen.2026.4636

Unionization, Ownership Status, and Direct Care Worker Turnover

Geoffrey M Gusoff ^1,^✉, Heeeun Jang ², Daniel Spertus ³, Kiran Abraham-Aggarwal ², Ariel Avgar ², Madeline Sterling ³

¹Department of Family Medicine, UCLA David Geffen School of Medicine, Los Angeles, California

²School of Industrial and Labor Relations, Cornell University, Ithaca, New York

³Department of Medicine, Weill Cornell Medicine, New York, New York

Accepted for Publication: February 8, 2026.

Published: April 2, 2026. doi:10.1001/jamanetworkopen.2026.4636

Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License, which does not permit alteration or commercial use, including those for text and data mining, AI training, and similar technologies. © 2026 Gusoff GM et al. JAMA Network Open.

^✉

Corresponding Author: Geoffrey M. Gusoff, MD, MBA, MS, Department of Family Medicine, UCLA David Geffen School of Medicine, 10833 Le Conte Ave, 50-080 CHS, Los Angeles, CA 90095 (ggusoff@mednet.ucla.edu).

Author Contributions: Drs Gusoff and Jang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Gusoff and Jang are co–first authors.

Concept and design: Gusoff, Jang, Spertus, Avgar, Sterling.

Acquisition, analysis, or interpretation of data: Gusoff, Jang, Abraham-Aggarwal, Sterling.

Drafting of the manuscript: Gusoff, Jang, Spertus, Abraham-Aggarwal.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Gusoff, Jang.

Obtained funding: Gusoff, Sterling.

Administrative, technical, or material support: Spertus, Sterling.

Supervision: Gusoff, Avgar, Sterling.

Conflict of Interest Disclosures: Dr Sterling reported receiving grants from the National Heart, Lung, and Blood Institute during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was supported by Career Development Awards from the National Institute on Aging (grant K01AG088782 [Dr Gusoff]).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2.

^✉

Corresponding author.

PMCID: PMC13047459 PMID: 41926123

Key Points

Question

How is workforce turnover among direct care workers (DCWs) associated with unionization and the ownership status of their employers?

Findings

In this cross-sectional study of 18 175 DCWs from the 2009-2024 Current Population Survey, unionization and public ownership were associated with significantly lower rates of DCW turnover, with similar unionization-associated turnover differences at nonprofit and for-profit care organizations.

Meaning

This study suggests that structural interventions such as DCW unionization and public ownership may help retain DCWs in the workforce to meet growing demand.

Abstract

Importance

Direct care workers (DCWs) provide essential support for millions of older individuals in the US, but high workforce turnover—the rate at which workers leave the DCW workforce—undermines care access and quality. Structural factors associated with DCW working conditions, such as unionization and employer ownership status, may play an important role in DCW retention, but their association with DCW workforce turnover is not known.

Objective

To assess the association of unionization and ownership with workforce turnover among DCWs.

Design, Setting, and Participants

Data on DCW turnover for this cross-sectional study were obtained from the Outgoing Rotation Groups of the Current Population Survey, an annual survey of 60 000 US households, from January 1, 2009, to December 31, 2024. The study population included individuals aged 15 years or older who were employed as DCWs.

Exposures

Unionization (whether the respondent was covered by a union through their DCW role) and employer ownership status (whether their employer was for profit, nonprofit, or publicly owned).

Main Outcomes and Measures

The main outcome of interest was workforce turnover at 1 year, defined as reporting a non-DCW occupation or no occupation 1 year after the initial survey. Bivariate analyses and logistic regression were used to compare DCW turnover rates across union status and employer ownership type and test whether ownership type moderates the association between unionization and turnover.

Results

The overall sample included 18 175 DCWs (mean [SD] age, 44.0 [14.7] years; 15 860 female DCWs [86.5%]). In the fully adjusted models, the estimated probability of turnover was significantly lower among unionized DCWs than nonunionized DCWs overall (37.4% vs 45.0%; odds ratio [OR], 0.72 [95% CI, 0.64-0.81]), at nonprofit organizations (33.6% vs 47.1%; OR, 0.56 [95% CI, 0.39-0.80]) and at for-profit organizations (35.2% vs 45.9%; OR, 0.63 [95% CI, 0.54-0.75]), but not at public employers (39.8% vs 41.0%; OR, 0.95 [95% CI, 0.78-1.16]). Public ownership was also directly associated with lower turnover (39.1% vs 41.8%; OR, 0.89 [95% CI, 0.80-1.00]) compared with for-profit ownership.

Conclusions and Relevance

In this cross-sectional study of DCWs, employer ownership status and unionization were independently and jointly associated with DCW workforce turnover rates, suggesting that these structural factors may play an important role in DCW retention. State and federal policies that facilitate DCW unionization or public employment of DCWs may significantly improve DCW retention.

This cross-sectional study uses data from the Current Population Survey to evaluate the association of unionization and employer ownership status with workforce turnover among direct care workers.

Introduction

Direct care workers (DCWs)—a broad workforce that includes home health aides, certified nursing assistants, and personal care aides—provide essential support for older adults and individuals with disabilities in the home, nursing homes, and hospital settings.¹ The demand for DCWs in the US is at a historic high as the population ages, but the workforce supply has not kept pace, leading to critical workforce shortages.^2,3 These shortages are compounded by high turnover rates among DCWs.^4,5

Organizational turnover—the rate at which workers leave a particular job—has been found to exceed 75% per year among DCWs in some settings.^5,6 High DCW organizational turnover and the related disruption in care continuity have been associated with lower nursing home care quality, worse home care outcomes, increased recruitment and training costs, and staffing shortages that strain remaining workers.^{5,7,8,9,10,11,12,13}

Although the harms of organizational turnover among DCWs are significant, the effects of high DCW workforce turnover—the rate at which DCWs leave the direct care workforce entirely—are even more concerning. Workforce turnover not only undermines care continuity and quality but also exacerbates industry-wide DCW shortages.¹⁴ Although workforce turnover could theoretically represent a beneficial change to a non-DCW role (ie, licensed vocational nurse), these types of career advancement opportunities are relatively uncommon among DCWs.^15,16 As a result, workforce turnover generally leads to significant losses in human capital and training investments, as experienced DCWs must be replaced.^14,17

Several studies have suggested that high rates of DCW turnover are significantly associated with modifiable job characteristics such as wages, supervisor support, and worker empowerment.^{14,18,19,20,21,22,23} These job characteristics are themselves shaped by structural factors that influence worker power and organizational incentives, such as unionization and ownership type.^{23,24,25,26,27,28}

Unionization provides a means for workers to advocate for better wages, benefits, and workplace protections.²⁵ In several studies, unionization has been associated with higher wages, safer conditions, and lower organizational turnover among DCWs.^{22,25,29,30,31,32,33}

The ownership of care organizations (ie, public, private, nonprofit) also appears to influence DCW workforce outcomes, given their distinct financial incentives and regulatory constraints.^24,27,28 Nonprofit status has been associated with higher employee well-being, training investments, and worker retention at nursing homes and lower organizational turnover at assisted living facilities compared with for-profit status.^28,34,35,36

Overall, these findings suggest that unionization and employer ownership may be associated with DCW workforce turnover. Furthermore, studies on unionization and DCW outcomes suggest that ownership may also moderate the association between unionization and turnover (ie, unionization may have a stronger association with turnover at for-profit care organizations than nonprofit care organizations).^37,38 However, to our knowledge, no prior study has examined the association between unionization or ownership and DCW workforce turnover or the potential moderating role of ownership in the unionization-turnover association.

Therefore, we aimed to examine whether unionization and ownership are associated with DCW workforce turnover and whether ownership moderates the unionization-turnover association. A better understanding of the role of structural factors such as unionization and ownership type in DCW workforce turnover can inform efforts to address the DCW workforce crisis.

Methods

Data Source and Study Population

To assess workforce turnover within the DCW workforce, we used data from the Outgoing Rotation Groups within the Current Population Survey (CPS) from January 1, 2009, to December 31, 2024. The study population included individuals aged 15 years or older who had an occupational code indicating employment as a DCW, including nursing aides, psychiatric aides, home health aides, and personal care aides. CPS obtained verbal consent from participants to participate in the study. This study was exempt from review by the Cornell University institutional review board because it uses publicly available and deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

The CPS surveys approximately 60 000 households annually following a specific rotation pattern: households are surveyed for 4 consecutive months followed by an 8-month break and are then surveyed again for another 4 months. To identify annual DCW turnover, we linked respondent data in the fourth and eighth (final) survey months using their personal IDs. Because personal IDs are occasionally applied to different participants within the dataset, we excluded all linked observations for which there were discrepancies in respondents’ reported race and ethnicity, sex, or age across study periods, consistent with matching methods used in previous studies.^39,40 Excluding respondents who did not complete the follow-up survey and those with demographic discrepancies in linked data resulted in an exclusion of 31.7% of observations, largely consistent with exclusion rates observed in other studies using this data linking methodology.⁴¹

Measures

Exposure Variables

Unionization was defined to include both formal union membership and union coverage without membership. The CPS Outgoing Rotation Groups asked about union status through 2 questions: “On this job, are you a member of a labor union or of an employee association similar to a union?” “On this job, are you covered by a union or employee association contract?” We classified DCWs answering yes to either question as unionized.

To measure ownership, we used the “class of workers” variable from the CPS, which categorizes workers based on the ownership of the employing organization. We consolidated CPS’s 5 ownership categories of employers into 3 major categories: (1) public, for local, state, or federal government entities; (2) for-profit, for private-sector entities; and (3) nonprofit, for charities and religious, civic, or other nonprofit organizations.

Outcome Variable

Workforce turnover was defined as whether respondents left the DCW workforce within 1 year, determined by comparing occupational codes from year t with year t + 1. If DCWs were no longer employed in a DCW occupation after 1 year, they were recorded as “turnover.” This includes respondents who reported a non-DCW occupation and those who left the labor force entirely. If they continued working as DCWs, they were recorded as “stay.” Given our focus on workforce-level turnover, changes between 2 different DCW occupational codes (eg, from “personal care aide” to “home health aide”) were not classified as turnover.

Covariates

We included several sociodemographic and employment-related factors that could be associated with both the exposures and the outcome. These factors included age, sex, self-reported race and ethnicity, educational level, geography, employment status, and care setting. Self-reported race and ethnicity were categorized into 5 groups: Asian, Hispanic, non-Hispanic Black, non-Hispanic White, and other (Hawaiian or Pacific Islander only and all multiracial combinations). Race and ethnicity were included as covariates given the association between racialization and DCW turnover intent.⁴² Educational attainment was divided into 4 levels: high school diploma or less, some college without a degree, associate degree, and bachelor’s degree or higher. Geography included metropolitan status and Census Bureau region (northeast, south, midwest, and west). Care settings were classified into 4 categories: (1) home- and community-based care, (2) nursing and residential care, (3) hospitals, and (4) other settings. Employment status was dichotomized, with full-time work (>35 hours per week) coded as 1 and part-time work coded as 0.

Statistical Analysis

First, we calculated descriptive statistics for all variables and conducted bivariate comparisons across unionization and ownership statuses using design-adjusted χ² tests and survey-weighted Wald tests (α = .05). Then, we used logistic regression to analyze the associations of unionization and ownership, our exposures, with turnover, our binary outcome. We then examined whether ownership status moderated the association between DCW unionization and turnover by adding an interaction term (ownership × unionization) and through logistic regressions stratified by ownership type. Ownership-specific union-nonunion differences were subsequently estimated using postestimation marginal contrasts derived from these interaction models. Statistical significance of differences in the unionization-turnover association across ownership types was evaluated using Wald tests of pairwise contrasts. All regression models accounted for the sociodemographic and employment-related covariates described as well as year fixed effects to better isolate the specific associations of union membership and ownership with turnover. Survey weights were applied in all models to ensure accurate variance estimation.

Sensitivity analyses were conducted to ensure robustness. Although a prior CPS study on DCW unionization found no substantial attrition-related associations with DCW outcomes, we conducted 2 tests to assess for potential attrition-related bias in our analysis.⁴¹ First, we evaluated turnover at 9 months instead of 12 months (matching data from the fourth and fifth in-survey months) to assess whether lower attrition rates in this shorter interval were associated with our findings. Second, we applied inverse probability weighting, combining weights derived from a logistic model estimating sample attrition with the original CPS survey weights, and the main models were reestimated for comparison. We conducted another analysis restricting the unionization variable to union membership, excluding DCWs covered by labor unions but who are not members, to account for the potentially distinct associations of union membership and union coverage. Finally, we conducted an analysis excluding DCWs who left the labor market (restricting turnover to job-to-job transitions) and another analysis restricted to the pre–COVID-19 period, to assess the robustness of our findings to type of workforce transition and pandemic-related changes.

We reported regression results for the exposure variables (unionization and ownership) as well as the marginal associations of unionization with turnover stratified across ownership types. All statistical analyses were conducted with R, version 4.2.1 (R Project for Statistical Computing). All P values were from 2-sided tests and results were deemed statistically significant at P < .05.

Results

Descriptive Statistics

Table 1 displays descriptive statistics of the sample by union status, and Table 2 displays descriptive statistics of the sample by ownership status. The overall sample included 18 175 DCWs (mean [SD] age, 44.0 [14.7] years; 86.5% female DCWs and 13.5% male DCWs; 6.8% Asian, 18.4% Hispanic, 28.4% non-Hispanic Black, 43.6% non-Hispanic White, and 2.8% other race and ethnicity). Overall, 10.9% of DCWs were unionized. Most DCWs were employed by for-profit organizations (72.0%), while 15.6% worked for public organizations and 12.4% worked for nonprofit organizations.

Table 1. Descriptive Statistics of the Study Sample by Unionization Status^a.

Category	Total, No. (%) (N = 18 175)	Unionization status, No. (%)		P value
Category	Total, No. (%) (N = 18 175)	Yes (n = 1972)	No (n = 16 203)	P value
Sex
Male	2315 (13.5)	307 (15.8)	2008 (13.2)	.005
Female	15 860 (86.5)	1665 (84.2)	14 195 (86.8)	.005
Age, mean (SD), y	44.0 (14.7)	46.4 (13.4)	43.7 (14.9)	<.001
Race and ethnicity
Asian	1088 (6.8)	171 (8.1)	917 (6.6)	<.001
Hispanic	2737 (18.4)	349 (20.8)	2388 (18.1)
Non-Hispanic Black	4384 (28.4)	550 (32.5)	3834 (27.8)
Non-Hispanic White	9365 (43.6)	836 (35.7)	8529 (44.7)
Other^b	601 (2.8)	66 (2.9)	535 (2.8)
Educational level
≤High school degree	9313 (51.6)	990 (50.6)	8323 (51.8)	.39
Some college	4597 (24.7)	474 (24.0)	4123 (24.8)
Associate degree	2125 (11.6)	246 (12.4)	1879 (11.5)
≥4-y College degree	2140 (12.1)	262 (13.0)	1878 (12.0)
Residence
Metropolitan	13 861 (83.1)	1702 (90.4)	12 159 (82.1)	<.001
Nonmetropolitan	4314 (16.9)	270 (9.6)	4044 (17.9)	<.001
Region
Northeast	4216 (24.1)	730 (39.2)	3486 (22.0)	<.001
Midwest	4084 (21.9)	330 (17.0)	3754 (22.6)
South	5242 (30.4)	158 (6.3)	5084 (33.8)
West	4633 (23.6)	754 (37.5)	3879 (21.6)
Employment status
Full time	12 082 (66.2)	1427 (72.1)	10 655 (65.4)	<.001
Part time	6093 (33.8)	545 (27.9)	5548 (34.6)	<.001
Care setting
Home-based and community-based care	7318 (42.4)	659 (36.9)	6659 (43.2)	<.001
Nursing and residential care	5137 (26.6)	435 (20.9)	4702 (27.4)
Hospitals	3143 (16.5)	478 (21.4)	2665 (15.8)
Other^c	2577 (14.5)	400 (20.8)	2177 (13.6)
Ownership
Public	2703 (15.6)	796 (39.8)	1907 (12.1)	<.001
For profit	12 723 (72.0)	952 (50.2)	11 771 (75.1)
Nonprofit	2749 (12.4)	224 (10.0)	2525 (12.8)
Direct care worker turnover
Yes	7222 (40.0)	632 (31.9)	6590 (41.2)	<.001
No	10 953 (60.0)	1340 (68.1)	9613 (58.8)	<.001

Open in a new tab

^{^a}

Source: Current Population Survey Outgoing Rotation Groups, 2009-2024.

^{^b}

In the Current Population Survey dataset, “Other” includes Hawaiian or Pacific Islander only and all multiracial combinations.

^{^c}

“Other” care setting includes industries not classified into the preceding categories, with the most prevalent being other health care services (Current Population Survey dataset code 8180) and administration of human resource programs (Current Population Survey dataset code 9480), the latter primarily reflecting public-sector care service.

Table 2. Descriptive Statistics of the Study Sample by Ownership Status^a.

Category	Total, No. (%) (N = 18 175)	Ownership, No. (%)			P value
Category	Total, No. (%) (N = 18 175)	Public (n = 2703)	For profit (n = 12 723)	Nonprofit (n = 2749)	P value
Sex
Male	2315 (13.5)	458 (18.3)	1410 (11.9)	447 (16.9)	<.001
Female	15 860 (86.5)	2245 (81.7)	11 313 (88.1)	2302 (83.1)	<.001
Age, mean (SD), y	44.0 (14.7)	46.4 (14.0)	43.7 (14.8)	42.7 (15.2)	<.001
Race and ethnicity
Asian	1088 (6.8)	221 (9.5)	746 (6.4)	121 (5.3)	<.001
Hispanic	2737 (18.4)	498 (22.6)	2005 (18.8)	234 (10.9)
Non-Hispanic Black	4384 (28.4)	557 (23.7)	3319 (29.9)	508 (25.6)
Non-Hispanic White	9365 (43.6)	1323 (40.8)	6270 (42.2)	1772 (54.7)
Other^b	601 (2.8)	104 (3.3)	383 (2.6)	114 (3.6)
Educational level
≤High school degree	9313 (51.6)	1422 (52.8)	6665 (52.7)	1226 (44.1)	<.001
Some college	4597 (24.7)	638 (23.0)	3205 (24.5)	754 (27.4)
Associate degree	2125 (11.6)	313 (11.5)	1461 (11.4)	351 (12.8)
≥4-y College degree	2140 (12.1)	330 (12.6)	1392 (11.4)	418 (15.7)
Residence
Metropolitan	13 861 (83.1)	2021 (82.4)	9901 (83.7)	1939 (80.7)	.002
Nonmetropolitan	4314 (16.9)	682 (17.6)	2822 (16.3)	810 (19.3)	.002
Region
Northeast	4216 (24.1)	392 (14.0)	2971 (25.1)	853 (31.2)	<.001
Midwest	4084 (21.9)	514 (16.9)	2745 (21.8)	825 (29.4)
South	5242 (30.4)	552 (19.3)	4158 (34.2)	532 (22.2)
West	4633 (23.6)	1245 (49.9)	2849 (19.0)	539 (17.2)
Employment status
Full time	12 082 (66.2)	1771 (64.1)	8390 (66.1)	1921 (69.4)	.002
Part time	6093 (33.8)	932 (35.9)	4333 (33.9)	828 (30.6)	.002
Care setting
Home-based and community-based care	7318 (42.4)	879 (35.2)	5643 (46.3)	796 (29.0)	<.001
Nursing and residential care	5137 (26.6)	310 (9.7)	3991 (29.8)	836 (29.3)
Hospitals	3143 (16.5)	417 (13.9)	1952 (14.9)	774 (29.2)
Other^c	2577 (14.5)	1097 (41.2)	1137 (9.0)	343 (12.5)
Unionization
Unionized	1972 (10.9)	796 (31.7)	952 (8.6)	224 (9.9)	<.001
Nonunionized	16 203 (89.1)	1907 (68.3)	11 771 (91.4)	2525 (90.1)	<.001
Direct care worker turnover
Yes	7222 (40.0)	1040 (38.1)	5038 (39.9)	1144 (43.2)	.005
No	10 953 (60.0)	1663 (61.9)	7685 (60.1)	1605 (56.8)	.005

Open in a new tab

^{^a}

Source: Current Population Survey Outgoing Rotation Groups, 2009-2024.

^{^b}

In the Current Population Survey dataset, “Other “ includes Hawaiian or Pacific Islander only and all multiracial combinations.

^{^c}

Statistically significant differences by union status were observed across all sociodemographic and employment-related variables, except for educational attainment (Table 1). Compared with nonunionized DCWs, unionized DCWs were more likely to be male (15.8% vs 13.2%) and older (mean [SD] age, 46.4 [13.4] vs 43.7 [14.9] years). Unionized DCWs were also more likely than nonunionized DCWs to be non-Hispanic Black (32.5% vs 27.8%), live in metropolitan areas (90.4% vs 82.1%), and live in the northeast (39.2% vs 22.0%) or west (37.5% vs 21.6%).

Significant differences by employer ownership type were also observed across all the DCW variables (Table 2). Compared with nonprofit and publicly employed DCWs, DCWs at for-profit organizations were disproportionately female (for profit, 88.1% vs nonprofit, 83.1% and public, 81.7%), non-Hispanic Black (for profit, 29.9% vs nonprofit, 25.6% and public, 23.7%), and based in the south (for profit, 34.2% vs nonprofit, 22.2% and public, 19.3%). Compared with DCWs at for-profit and public organizations, DCWs at nonprofit organizations were disproportionately non-Hispanic White (nonprofit, 54.7% vs for profit, 42.2% and public 40.8%), employed full time (nonprofit, 69.4% vs for profit, 66.1% and public, 64.1%), based in the northeast (nonprofit, 31.2% vs for profit, 25.1% and public, 14.0%), and more likely to have at least some college education (nonprofit, 55.9% vs for profit, 47.3% and public, 47.2%). Publicly employed DCWs were much more likely to be unionized than DCWs employed at for-profit organizations or nonprofit organizations (public, 31.7% vs for profit, 8.6% vs nonprofit, 9.9%).

Associations of Unionization and Ownership With DCW Workforce Turnover

The unadjusted workforce turnover rate for all DCWs in the sample was 40.0% (Table 1). Workforce turnover rates were significantly lower among unionized DCWs compared with their nonunionized counterparts (31.9% vs 41.2%; P < .001) and were significantly higher among nonprofit employers compared with for-profit and public employers (nonprofit, 43.2% vs for profit, 39.9% and public, 38.1%; P = .005) (Table 2).

In the fully adjusted model, unionized DCWs demonstrated a significantly lower workforce turnover probability of 37.4% (95% CI, 34.3%-40.4%) compared with 45.0% (95% CI, 43.0%-47.1%) among nonunionized counterparts (Table 3), representing a 7.6% absolute decrease in workforce turnover. The adjusted odds ratio (OR) for turnover among unionized DCWs was 0.72 (95% CI, 0.64-0.81; P < .001), indicating unionized workers had 28% lower odds of experiencing workforce turnover than their nonunionized counterparts.

Table 3. Association Between Direct Care Worker Turnover Rates and Unionization and Ownership Status in Fully Adjusted Model, 2009-2023^a.

Characteristic	Turnover probability, % (95% CI)^b	Odds ratio (95% CI)	P value
Unionization
Nonunionized	45.0 (43.0-47.1)	1 [Reference]	NA
Unionized	37.4 (34.3-40.4)	0.72 (0.64-0.81)	<.001
Ownership status
For profit	41.8 (39.4-44.1)	1 [Reference]	NA
Public	39.1 (36.3-41.9)	0.89 (0.80-1.00)	.04
Nonprofit	42.7 (39.7-45.8)	1.04 (0.94-1.16)	.46

Open in a new tab

Abbreviation: NA, not applicable.

^{^a}

All models adjusted for sex, race and ethnicity, age, educational level, metropolitan or nonmetropolitan residence, region, employment status, and survey year. The 95% CIs were calculated based on robust standard errors to address heteroscedasticity. Source: Current Population Survey Outgoing Rotation Group, 2009-2024. Complete regression results provided in eTable 1 in Supplement 1.

^{^b}

Turnover rate values represent the percentage of direct care workers leaving the field annually.

The association between ownership status and workforce turnover in the fully adjusted model was also statistically significant (Table 3). Public ownership was directly associated with lower turnover compared with for-profit ownership (39.1% vs 41.8%; OR, 0.89 [95% CI, 0.80-1.00]; P = .04), but nonprofit ownership was not (42.7% vs 41.8%; OR, 1.04 [95% CI, 0.94-1.16]; P = .46).

Although unionized DCWs had lower workforce turnover rates across all ownership types, the interaction analysis (eTable 2 in Supplement 1) and postestimation analyses stratified by ownership status (Figure) indicate that the magnitude of these turnover differences varied significantly across ownership type. As shown in the Figure, the magnitude of the difference in workforce turnover rates between unionized and nonunionized HCWs was greatest in the nonprofit sector (33.6% vs 47.1%; OR, 0.56 [95% CI, 0.39-0.80]; P = .001), somewhat smaller but still statistically significant in the for-profit sector (35.2% vs 45.9%; OR, 0.63 [95% CI, 0.54-0.75]; P < .001), and smallest and not statistically significant in the public sector (39.8% vs 41.0%; OR, 0.95 [95% CI, 0.78-1.16]; P = .63). Formal tests of interaction indicated that the magnitude of the union-nonunion difference did not differ significantly between nonprofit and for profit (OR, 0.88 [95% CI, 0.59-1.31]; P = .53), although the union-nonunion difference in the public sector differed significantly from the nonprofit setting (OR, 0.58 [95% CI, 0.39-0.87]; P = .01) and for-profit setting (OR, 0.66 [95% CI, 0.51-0.86]; P = .002). Similar findings were observed in sensitivity analyses (eTables 2-10 in Supplement 1) with the exception of the pre–COVID-19 analysis, in which only unionization was significantly associated with turnover (eTables 7 and 8 in Supplement 1).

Figure. — Error bars indicate 95% CIs. Source: Current Population Survey Outgoing Rotation Group, 2009 to 2024.

Discussion

Overall, we found that unionization and public ownership were associated with significantly lower rates of DCW workforce turnover and that ownership status moderated the unionization-turnover association. The sensitivity analyses suggest that these findings are robust to shorter turnover intervals and narrower definitions of unionization and turnover, and there was no evidence of systematic attrition-related bias. Only the unionization-turnover association was observed when assessing just the pre–COVID-19 period.

Although previous studies have found lower rates of organizational turnover among unionized DCWs within subsets of the DCW workforce,^18,19,21,22 our findings suggest unionization is also associated with lower workforce turnover across DCW settings. This association may be explained by improved DCW compensation, workplace participation, and other working conditions that have been associated with both unionization and DCW retention.^{18,19,21,22,26,30,38} However, further studies are needed to more directly assess the causal role of unions in reducing DCW workforce turnover.

The magnitude of the unionization-turnover association that we found among DCWs is also notable, representing a 7.6% absolute decrease in workforce turnover. Given inflation-adjusted estimates of DCW turnover costs of $3400 to $4960 per employee, if these decreases in turnover were realized across the DCW workforce, it could generate an estimated $1.04 billion to $1.51 billion in reduced direct turnover costs annually (eAppendix in Supplement 1).^{1,17,43,44,45} Estimating the overall economic impact of broad unionization in the DCW sector would require additional analyses to assess the causal contribution of unionization to direct turnover costs, labor costs (ie, worker salaries and benefits), and health care savings (ie, related to care quality and outcomes).

In addition to unionization, we found that ownership status was also directly associated with DCW turnover, with public employers demonstrating significantly lower turnover rates than nonprofit and for-profit employers. This association was smaller in magnitude than the unionization-turnover association and less consistent over time but still significant. Public DCW employers are relatively understudied in the US context, with analyses tending to focus on nonprofit and for-profit employers.^28,34,35,36 Spillover effects, in which relatively high union density in the public sector can also improve working conditions for nonunionized public workers, may help explain lower turnover in the public sector, but further DCW workforce studies are needed to assess the unique public sector features associated with DCW turnover.^38,46

Finally, we found a significant interaction between unionization and ownership, with relatively strong associations between unionization and turnover at for-profit organizations and nonprofit organizations but no significant association in public settings. Spillover effects from higher union density in the public sector may help explain the negligible unionization-associated turnover differences observed in public settings.^38,46

Nonprofit organizations, which in theory lack the profit-maximizing and labor cost–reducing imperatives of for-profit organizations, demonstrated similar unionization-associated turnover differences as for-profit organizations.⁴⁷ This finding is consistent with several studies that have found that, in practice, nonprofit organizations often have similar or worse working conditions compared with for-profit organizations.^47,48,49,50 These findings suggest that the role of unionization in addressing DCW turnover may be at least as important at nonprofit organizations as at for-profit organizations, but further analyses permitting causal inference are needed to better understand the role of unionization in reducing DCW turnover across ownership types.

Limitations

This study has several limitations. First, the use of cross-sectional data limits our ability to make causal inferences. Although there are several plausible mechanisms through which unionization may reduce turnover (ie, higher wages), it is also possible that lower-turnover workplaces are more likely to unionize. Longitudinal, quasi-experimental analyses can help determine the relative contributions of each factor.

Second, reliance on participant-reported data for unionization and ownership introduces the potential for bias. The observed distributions of union and ownership status were generally consistent with other data sources, but future studies should verify these measures with external data where possible.^51,52

Third, there is also a possibility of missing variable bias. Although we controlled for several potential confounders described in the literature, there may be additional unmeasured confounders (ie, local labor laws or market conditions). Controlling for region in our analysis helped capture some of these factors, but future analyses should explore the role of these potential confounders in more detail.

Fourth, this analysis was not powered to detect differences in the observed associations between different care settings or across different types of workforce exit (eg, retirement and layoff), which should be explored in future research. In addition, this analysis was not able to assess turnover differences across specific ownership types (ie, sole proprietorship and private equity), which could provide additional insight into the role of employer ownership in DCW retention.

Conclusions

Our findings in this cross-sectional study suggest that unions and the practices of public employers may play an important role in retaining DCWs in the workforce. State and federal policies that facilitate DCW unionization or public employment of DCWs may significantly improve DCW retention. To the extent that unions and public employers reduce DCW turnover through improved working conditions and workplace participation, additional interventions such as health care worker minimum wage laws and DCW representation on company boards may also improve DCW retention.^53,54 Given the large size of the DCW workforce and the direct and indirect costs of turnover, these interventions have the potential to produce substantial health care savings, even with greater investment in DCW compensation and staffing.^1,43 Workforce interventions to reduce DCW turnover on a system level also have the potential to improve care quality for the millions of individuals in the US who rely on DCWs.^5,7,8,9,10

The high rate of DCWs leaving the workforce is an urgent and growing problem that harms both DCWs and their patients, but it is not immutable. Changes on the structural level that empower workers or change organizational incentives and governance may play an important role in improving retention across the sector. Additional research can help better determine what unionization-related and ownership-related factors may be most conducive to improving DCW retention.

Supplement 1.

eTable 1. Logistic Regression Results Between Unionization Status and Ownership Status on Direct Care Worker Turnover

eTable 2. Association Between Union Status and Direct Care Workers’ Turnover Rates, 2009–2023: Linking 4th and 5th Survey-in-Months

eTable 3. Association Between Ownership Status and Direct Care Workers’ Turnover Rates, 2009–2023: Linking 4th and 5th Survey-in-Months

eTable 4. Association Between Union Membership and Direct Care Workers’ Turnover Rates, 2009–2023: Narrow Definition of Union Status

eTable 5. Association Between Unionization and Direct Care Worker Turnover After Applying Inverse Probability–Weighted Survey Adjustment, 2009–2023

eTable 6. Association Between Ownership Status and Direct Care Worker Turnover After Applying Inverse Probability–Weighted Survey Adjustment, 2009–2023

eTable 7. Association Between Unionization and Direct Care Workers’ Turnover Rates During Pre–COVID-19 Period, 2009–2018

eTable 8. Association Between Ownership Status and Direct Care Workers’ Turnover Rates During Pre–COVID-19 Period, 2009–2018

eTable 9. Association Between Unionization and Direct Care Workers’ Job-to-Job Transition, 2009–2023

eTable 10. Association Between Ownership Status and Direct Care Workers’ Job-to-Job Transition, 2009–2023

eAppendix. Methodology for Estimating Projected Savings From Reduced Turnover Costs

jamanetwopen-e264636-s001.pdf^{(471KB, pdf)}

Supplement 2.

Data Sharing Statement

jamanetwopen-e264636-s002.pdf^{(14.5KB, pdf)}

References

1.Direct care workers in the United States: key facts 2024. PHI. September 2, 2024. Accessed September 9, 2024. https://www.phinational.org/resource/direct-care-workers-in-the-united-states-key-facts-2024/
2.Jumabhoy S, Jung HY, Yu J. Characterizing the direct care health workforce in the United States, 2010-2019. J Am Geriatr Soc. 2022;70(2):512-521. doi: 10.1111/jgs.17519 [DOI] [PubMed] [Google Scholar]
3.Scales K. It is time to resolve the direct care workforce crisis in long-term care. Gerontologist. 2021;61(4):497-504. doi: 10.1093/geront/gnaa116 [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Famakinwa J. Home care’s industry-wide turnover rate reaches nearly 80%. Home Health Care News. Updated July 3, 2024. Accessed June 28, 2025. https://homehealthcarenews.com/2024/07/home-cares-industry-wide-turnover-rate-reaches-nearly-80/
5.Gandhi A, Yu H, Grabowski DC. High nursing staff turnover in nursing homes offers important quality information. Health Aff (Millwood). 2021;40(3):384-391. doi: 10.1377/hlthaff.2020.00957 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.2022 HCP benchmarking report. Home Care Pulse. 2022. Accessed June 28, 2025. https://store.activatedinsights.com/product/2022-hcp-benchmarking-report-digital/
7.Shen K, McGarry BE, Gandhi AD. Health care staff turnover and quality of care at nursing homes. JAMA Intern Med. 2023;183(11):1247-1254. doi: 10.1001/jamainternmed.2023.5225 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Lerner NB, Johantgen M, Trinkoff AM, Storr CL, Han K. Are nursing home survey deficiencies higher in facilities with greater staff turnover. J Am Med Dir Assoc. 2014;15(2):102-107. doi: 10.1016/j.jamda.2013.09.003 [DOI] [PubMed] [Google Scholar]
9.Castle NG, Engberg J. Staff turnover and quality of care in nursing homes. Med Care. 2005;43(6):616-626. doi: 10.1097/01.mlr.0000163661.67170.b9 [DOI] [PubMed] [Google Scholar]
10.Russell D, Rosati RJ, Peng TR, et al. Continuity in the provider of home health aide services and the likelihood of patient improvement in activities of daily living. Home Health Care Manag Pract. 2012;25(1):6-12. doi: 10.1177/1084822312453046 [DOI] [Google Scholar]
11.Lapane KL, Hughes CM. Considering the employee point of view: perceptions of job satisfaction and stress among nursing staff in nursing homes. J Am Med Dir Assoc. 2007;8(1):8-13. doi: 10.1016/j.jamda.2006.05.010 [DOI] [PubMed] [Google Scholar]
12.Huang S, Yau SY, Lee YKL, Song J, Guo Y, Dong D. Job characteristics, personal characteristics and well-being of nursing assistants in long-term care facilities: a mixed methods systematic review and narrative synthesis. Int J Nurs Stud. 2025;161:104934. doi: 10.1016/j.ijnurstu.2024.104934 [DOI] [PubMed] [Google Scholar]
13.Larson SA, Tolbize MA, Kim O, et al. Direct Support Professional Turnover Costs in Small Group Homes: A Case Study. University of Minnesota, Research and Training Center on Community Living; 2016:1-13. [Google Scholar]
14.Gleason HP, Miller EA. Maximizing home health aide retention: the impact of control and support on the job. Gerontologist. 2021;61(4):517-529. doi: 10.1093/geront/gnab003 [DOI] [PubMed] [Google Scholar]
15.Baughman RA, Smith KE. Labor mobility of the direct care workforce: implications for the provision of long-term care. Health Econ. 2012;21(12):1402-1415. doi: 10.1002/hec.1798 [DOI] [PubMed] [Google Scholar]
16.Dill JS, Morgan JC. Employability among low-skill workers: organizational expectations and practices in the US health care sector. Hum Relat. 2018;71(7):1001-1022. doi: 10.1177/0018726717734035 [DOI] [Google Scholar]
17.Mukamel DB, Spector WD, Limcangco R, Wang Y, Feng Z, Mor V. The costs of turnover in nursing homes. Med Care. 2009;47(10):1039-1045. doi: 10.1097/MLR.0b013e3181a3cc62 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Stone R, Wilhelm J, Bishop CE, Bryant NS, Hermer L, Squillace MR. Predictors of intent to leave the job among home health workers: analysis of the National Home Health Aide Survey. Gerontologist. 2017;57(5):890-899. doi: 10.1093/geront/gnw075 [DOI] [PubMed] [Google Scholar]
19.Jang Y, Lee AA, Zadrozny M, Bae SH, Kim MT, Marti NC. Determinants of job satisfaction and turnover intent in home health workers: the role of job demands and resources. J Appl Gerontol. 2017;36(1):56-70. doi: 10.1177/0733464815586059 [DOI] [PubMed] [Google Scholar]
20.Butler SS, Brennan-Ing M, Wardamasky S, Ashley A. Determinants of longer job tenure among home care aides: what makes some stay on the job while others leave? J Appl Gerontol. 2014;33(2):164-188. doi: 10.1177/0733464813495958 [DOI] [PubMed] [Google Scholar]
21.Wiener JM, Squillace MR, Anderson WL, Khatutsky G. Why do they stay? job tenure among certified nursing assistants in nursing homes. Gerontologist. 2009;49(2):198-210. doi: 10.1093/geront/gnp027 [DOI] [PubMed] [Google Scholar]
22.Berridge C, Tyler DA, Miller SC. Staff empowerment practices and CNA retention: findings from a nationally representative nursing home culture change survey. J Appl Gerontol. 2018;37(4):419-434. doi: 10.1177/0733464816665204 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Zarska A, Avgar AC, Sterling MR. Relationship between working conditions, worker outcomes, and patient care: a theoretical model for frontline health care workers. Am J Med Qual. 2021;36(6):429-440. doi: 10.1097/01.JMQ.0000735508.08292.73 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Berry D, Bell M. Worker cooperatives: alternative governance for caring and precarious work. Equality, Diversity and Inclusion: An International Journal. 2018;37:376-391. doi: 10.1108/EDI-02-2017-0027 [DOI] [Google Scholar]
25.Abraham-Aggarwal K, Spertus D, Avgar A, et al. The impact of unions on US direct care workers in long-term care settings: a systematic review. J Am Med Dir Assoc. 2024;25(12):105236. doi: 10.1016/j.jamda.2024.105236 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Jang H, Avgar A, Weaver R, et al. What do unions do for direct care workers? assessing employment and economic outcomes. J Am Med Dir Assoc. 2025;26(3):105427. doi: 10.1016/j.jamda.2024.105427 [DOI] [PubMed] [Google Scholar]
27.Herrera CA, Rada G, Kuhn-Barrientos L, Barrios X. Does ownership matter? an overview of systematic reviews of the performance of private for-profit, private not-for-profit and public healthcare providers. PLoS One. 2014;9(12):e93456. doi: 10.1371/journal.pone.0093456 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Bos A, Boselie P, Trappenburg M. Financial performance, employee well-being, and client well-being in for-profit and not-for-profit nursing homes: a systematic review. Health Care Manage Rev. 2017;42(4):352-368. doi: 10.1097/HMR.0000000000000121 [DOI] [PubMed] [Google Scholar]
29.Dill J, Tanem J. Gender, race/ethnicity, and unionization in direct care occupations. Am J Public Health. 2022;112(11):1676-1684. doi: 10.2105/AJPH.2022.307022 [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Delp L, Wallace SP, Geiger-Brown J, Muntaner C. Job stress and job satisfaction: home care workers in a consumer-directed model of care. Health Serv Res. 2010;45(4):922-940. doi: 10.1111/j.1475-6773.2010.01112.x [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Temple A, Dobbs D, Andel R. Exploring correlates of turnover among nursing assistants in the National Nursing Home Survey. Health Care Manage Rev. 2009;34(2):182-190. doi: 10.1097/HMR.0b013e31819c8b11 [DOI] [PubMed] [Google Scholar]
32.Dean A, McCallum J, Venkataramani A, Michaels D. Labor unions and staff turnover in US nursing homes. JAMA Netw Open. 2023;6(10):e2337898-e2337898. doi: 10.1001/jamanetworkopen.2023.37898 [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Dean A, McCallum J, Kimmel SD, Venkataramani AS. Resident mortality and worker infection rates from COVID-19 lower in union than nonunion US nursing homes, 2020-21. Health Aff (Millwood). 2022;41(5):751-759. doi: 10.1377/hlthaff.2021.01687 [DOI] [PubMed] [Google Scholar]
34.Kash BA, Castle NG, Naufal GS, Hawes C. Effect of staff turnover on staffing: a closer look at registered nurses, licensed vocational nurses, and certified nursing assistants. Gerontologist. 2006;46(5):609-619. doi: 10.1093/geront/46.5.609 [DOI] [PubMed] [Google Scholar]
35.Castle NG, Engberg J. Organizational characteristics associated with staff turnover in nursing homes. Gerontologist. 2006;46(1):62-73. doi: 10.1093/geront/46.1.62 [DOI] [PubMed] [Google Scholar]
36.Kennedy KA, Applebaum R, Bowblis JR, Straker JK. Organizational factors associated with retention of direct care workers: a comparison of nursing homes and assisted living facilities. Gerontologist. 2021;61(4):530-539. doi: 10.1093/geront/gnaa130 [DOI] [PubMed] [Google Scholar]
37.Allen SG. The effect of unionism on productivity in privately and publicly owned hospitals and nursing homes. J Labor Res. 1986;7(1):59-68. doi: 10.1007/BF02685300 [DOI] [Google Scholar]
38.Sojourner AJ, Frandsen BR, Town RJ, et al. Impacts of unionization on quality and productivity: regression discontinuity evidence from nursing homes. Ind Labor Relat Rev. 2015;68(4):771-806. doi: 10.1177/0019793915586380 [DOI] [Google Scholar]
39.Drew JA, Flood S, Warren JR. Making full use of the longitudinal design of the Current Population Survey: methods for linking records across 16 months. J Econ Soc Meas. 2014;39(3):121-144. doi: 10.3233/JEM-140388 [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Madrian BC, Lefgren L. A Note on Longitudinally Matching Current Population Survey (CPS) Respondents. National Bureau of Economic Research Cambridge; 1999. doi: 10.3386/t0247 [DOI] [Google Scholar]
41.Neumark D, Kawaguchi D. Attrition bias in labor economics research using matched CPS files. J Econ Soc Meas. 2004;29(4):445-472. doi: 10.3233/JEM-2004-0236 [DOI] [Google Scholar]
42.Landes SD, Weng SS. Racial-ethnic differences in turnover intent among home health aides. J Appl Gerontol. 2020;39(5):490-501. doi: 10.1177/0733464819827806 [DOI] [PubMed] [Google Scholar]
43.2017 Home care benchmarking study. Home Care Pulse . 2017. Accessed June 28, 2025. https://activatedinsights.com/resource-library/2017-home-care-benchmarking-study/
44.Caudill ME, Patrick M. Costing nurse turnover in nursing homes. Nurs Manage. 1991;22(11):61-62, 64. doi: 10.1097/00006247-199111000-00022 [DOI] [PubMed] [Google Scholar]
45.CPI inflation calculator. U.S. Bureau of Labor Statistics. Accessed August 7, 2025. https://www.bls.gov/data/inflation_calculator.htm
46.Rosenfeld J, Denice P. What do government unions do? public sector unions and nonunion wages, 1977-2015. Soc Sci Res. 2019;78:41-56. doi: 10.1016/j.ssresearch.2018.10.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Preston AE, Sacks DW. Nonprofit wages: theory and evidence. In: Seaman BA, Young DR, eds. Handbook of Research on Nonprofit Economics and Management. Edward Elgar Publishing; 2010:106-119. [Google Scholar]
48.Hamann DJ, Foster NT. An exploration of job demands, job control, stress, and attitudes in public, nonprofit, and for-profit employees. Rev Public Person Adm. 2014;34(4):332-355. doi: 10.1177/0734371X13491119 [DOI] [Google Scholar]
49.Zhao R, Anasti T, Welch R. Addressing “idealism exploitation” through unionization: causes and outcomes of nonprofit worker labor organizing in the U.S. Affilia. 2025;40(2):265-282. doi: 10.1177/08861099241296470 [DOI] [Google Scholar]
50.Hirsch BT, Macpherson DA, Preston AE. Nonprofit wages: theory and evidence. In: Seaman BA, Young Dr, eds. Handbook of Research on Nonprofit Economics and Management. Edward Elgar Publishing; 2018:146-179. [Google Scholar]
51.Hirsch BT, Macpherson DA. Union membership and coverage files from the Current Population Surveys: note. Ind Labor Relat Rev. 1993;46(3):574-578. doi: 10.1177/001979399304600309 [DOI] [Google Scholar]
52.Nursing home including rehab services: provider information. Centers for Medicare & Medicaid Services . Accessed January 20, 2026. https://data.cms.gov/provider-data/dataset/4pq5-n9py
53.Angell S, Stepick L, Scales K. Direct care workforce state index: mapping workforce policies and outcomes. PHI. July 30, 2024. Accessed September 9, 2024. https://www.phinational.org/resource/direct-care-workforce-state-index-mapping-workforce-policies-and-outcomes/
54.Gusoff GM, Stepick L, Soriano-Versoza A, Kazda K. Centering marginalized care: Home care cooperatives and system change. Health Aff Sch. 2025;3(3):qxae184. doi: 10.1093/haschl/qxae184 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials