Abstract
This study estimates the effect of nursing home closure on occupancy, net profit margin, and operating margin of nearby nursing homes. We use national nursing home data from 2009 to 2019 from Medicare cost reports, Medicare Provider of Services (POS), and LTCfocus.org data. Using the Callaway and Sant’Anna difference-in-differences model, we compare the changes in occupancy, net profit margin, and operating margin between incumbent nursing homes in markets with any closure and nursing homes in markets without a closure, overall, and across rurality. Our findings suggest that nursing home closure improves the occupancy rates of remaining nursing homes in the same market in rural areas but there is little evidence of effects in metropolitan and micropolitan areas. Nursing home regulators and local officials should consider the long-term care market heterogeneity when considering interventions targeted at nursing home closure.
Introduction
Nursing homes in the United States provide postacute and long-term care services to over a million residents every year but many are increasingly facing a risk of closing due to declining occupancy, low Medicaid reimbursements, and staffing shortages (American Health Care Association [AHCA] & National Center for Assisted Living [NCAL], 2022; Healy, 2019; Leys, 2023; Olenski, 2022; Pugh, 2021). About 100 to 150 nursing homes closed annually between 2008 and 2018 (Sharma, Baten, et al., 2021), a trend that continued during the COVID-19 pandemic (AHCA & NCAL, 2022). Recent waves of nursing home closures in rural areas have particularly raised concerns about access to postacute and long-term care for rural residents (Healy, 2019; Leys, 2023; Stulick, 2022). Residents of closing nursing homes have to relocate to other nursing homes or find other care arrangements. Nursing homes that remain in the market could potentially absorb at least some displaced residents, although direct evidence on how closures affect incumbent nursing homes is scarce. Nursing home closure, especially in rural areas, can decrease residents’ access to postacute and long-term care services due to decreased availability of care providers (Sharma, Baten, et al., 2021; Tyler & Fennell, 2017). There is suggestive evidence that rural nursing home closure slightly increases the quality of care in nearby nursing homes possibly due to increased market concentration and decreased information asymmetries (Bowblis & Vassallo, 2014). There are also increasing concerns about ownership changes in nursing homes and the impact of ownership changes on residents and nursing home performance (Welch et al., 2022). However, there is little evidence on how nursing home closures affect the occupancy and financial performance of nearby nursing homes. This study provides evidence on this question using national data on nursing homes from 2009 to 2018.
Conceptually, the exit of a nursing home can potentially benefit nearby nursing homes that have empty beds by increasing demand for their services and revenue. Postacute care patients and long-term care residents tend to select nursing homes closer to family suggesting that they may relocate to nursing homes in the same market. However, the profit margins of incumbent nursing homes can increase or decrease depending on the marginal cost and marginal revenue of caring for displaced patients (Nyman, 1988; Scanlon, 1980). Since Medicaid is the primary payer of nursing home services (62% of residents) (Eiken et al., 2018) and since the Medicaid payment rate is often inadequate to cover the cost of caring for medically complex residents, an influx of Medicaid residents may not improve profit margins (Hunter, 2018; Medicaid and CHIP Payment and Access Commission [MACPAC], 2023; Troyer, 2002). However, an increase in the proportion of privately insured patients could increase profit margins. Therefore, the effect of closure on the profitability of incumbent nursing homes would depend on the payer mix and acuity level of new patients and residents.
As many as 85% of residents from closed nursing homes could relocate to other nursing homes in the first quarter after closure (Olenski, 2022). Although lack of direct evidence, some residents may obtain home- and community-based services (HCBS) instead of relocating to other nursing homes. Consumers generally prefer HCBS over institutionalized nursing home care (Kane & Kane, 2001), and residents may increasingly seek HCBS depending on their preferences, level of care needs and availability of providers. For example, two studies suggest that in general residents with low care needs use assisted living facilities as substitutes for nursing homes (Cornell et al., 2020; Grabowski et al., 2012). Other studies showed that the rapid growth of the assisted living industry has increased nursing home concentration and decreased nursing home financial performance (Bowblis, 2014; Lord et al., 2018). Increased assisted living facility capacity was also found to have a mixed effect on nursing home quality (Bowblis, 2012).
Nursing home closures may decrease market competition, possibly increasing the market power of incumbent nursing homes. However, the empirical evidence on the relationship between market competition and nursing home financial performance is mixed. Existing research suggests that nursing homes in less competitive markets tend to charge a higher price for private-pay residents and provide lower quality of care (Yang et al., 2021), both of which can affect profitability. Yet the direct evidence on the effects of competition on profitability is mixed with a study suggesting that nursing homes in less competitive markets had higher operating and total margins (Weech-Maldonado et al., 2010), while another study finding no association between market competition and operating margins (Weech-Maldonado et al., 2019). These two studies on profitability differ in the sample period and measures of competition. Conceptually, it is possible that nursing homes in competitive markets need to spend more resources to attract additional residents (Yang et al., 2021).
In addition to an effect on market competition, a nursing home closure can spur public attention, which may motivate incumbent nursing homes to reduce their own closure risk by improving their performance and reputation. One study of rural nursing home closures between 1998 and 2004 found suggestive evidence that neighboring facilities improved their care quality, although the study did not examine occupancy or financial outcomes (Bowblis & Vassallo, 2014).
Although to our knowledge no published studies have evaluated nursing home closure effects on occupancy and profitability of other nursing homes, several studies have evaluated the effect of hospital closures on other hospitals. Studies suggest that hospital closures are associated with increased patient volumes (e.g., inpatient admissions and emergency room visits) in neighboring hospitals (Lindrooth et al., 2003; McKay & Dorner, 1996; Ramedani et al., 2022; Song & Saghafian, 2020). However, there is mixed evidence on hospital closure effects on finances of neighboring hospitals. One study found that rural hospital closure did not lower average costs or improve the profitability of neighboring hospitals (McKay & Dorner, 1996). Another study found that urban hospital closures led to increased inpatient and emergency admissions at competitor hospitals and improved their operating efficiency (Lindrooth et al., 2003). Yet another study on hospital closure in California reported a decline in the operating margins of hospitals around the closing hospital possibly due to increased admissions of high-risk, high-cost patients from the closed hospital (Hodgson et al., 2015). Finally, a recent study reported more annual patient discharges per bed at neighboring hospitals following a hospital closure (Song & Saghafian, 2020). Given the differences in operations and finances between nursing homes and hospitals, the extent to which the findings on hospital closures apply to nursing home closures is not clear.
New Contribution
How nursing home closure affects incumbent nursing homes’ performance is a crucial consideration for policymakers and local official addressing the nursing home closure and access issues in the long-term care system. However, to our knowledge, no studies have examined the spillover effects of nursing home closure on occupancy rates and profit margins of incumbent nursing homes within the same market. This study aims to address this knowledge gap and makes three key contributions. First, the nursing home market structure is different in rural and urban areas. Using the universe of nursing homes in the United States from 2009 to 2018, we define nursing home markets using both county boundaries and various radii, and examine the potential heterogeneous effects by rurality. Second, except for our primary outcomes of interest, we evaluate additional outcomes related to revenues, expenses, and resident mix to investigate potential mechanisms underlying any changes in profits. Finally, to account for the treatment effects heterogeneity over time in the presence of different treatment timings, we employ the Callaway and Sant’Anna (C&S) difference-in-differences model to estimate overall average effects and treatment effects separately by year. Evidence from this study contributes to the previous work on the multifaceted impact of nursing home closure on the local long-term care system.
Method
Data
We use the Medicare Provider of Service (POS) 2019 file to identify nursing home closures and entrants between 2009 and 2018. POS contains information for all active or inactive nursing homes participating in Medicare and/or Medicaid. A facility is considered closed if its termination status indicates voluntary or involuntary closure. Following previous research, we identify all closures (or entrants), including hospital-based nursing home closures (or entrants), within our study period based on a facility’s termination date from (or participation date in) Medicare and/or Medicaid (Xu et al., 2023). We also verify closure status by ensuring that these facilities are not observed in the Nursing Home Compare or LTCFocus data sets following the year of termination in the POS data (Sharma, Baten, et al., 2021; Xu et al., 2023). It is possible that termination status in the POS data may not be a sufficient indicator for closure because a facility terminated from Medicare and/or Medicaid program may have another nursing home in the same address under a different provider ID within a short period after closure (Castle, 2005). To address this concern, in a sensitivity analysis, we define a facility as closed if it is terminated from Medicare and/or Medicaid in the POS data and does not have another nursing facility in POS data at the same address within 3 years of termination (Bowblis & Vassallo, 2014; Castle, 2005; Sojourner et al., 2015).
We obtain information on financial performance from the Medicare cost reports. Similar to previous studies, we combine any reports with partial-year data and use full-year cost report data (Bowblis & Brunt, 2014; Sharma, Hefele, et al., 2021). We obtain information on facility and resident characteristics from the LTCfocus data. We link the LTCfocus and Medicare Cost Report data using federal identifiers for nursing homes.
We use the 2013 Urban Influence Code (UIC) information from the Economic Research Service, US Department of Agriculture, to classify counties into metropolitan (UICs 1,2), micropolitan (UICs 3,5,8), and noncore counties (UICs 4,6,7,9-12; hereinafter referred to as rural counties) (USDA ERS—Urban Influence Codes, n.d.). We obtain county-level data on household income, unemployment rate, percent of the population 65 or older, and the number of home health agencies from the Area Health Resource File.
Sample
We identify 1,335 nursing home closures from 807 counties in 50 states and DC between 2009 and 2018. The initial sample includes 129,395 facility-year observations on 13,737 incumbent nursing homes that were in continuous operation during the study period. We exclude data of closed nursing homes from the analytical sample. We also exclude 1,323 facility-year observations (<1%) with missing dependent variables or erroneous data such as negative revenues or expenses (Park et al., 2011; Weech-Maldonado et al., 2010). Consistent with the literature on Medicare cost reports, we exclude 4,710 observations (~4%) with outlier profit margins above the 99th percentile or below the 1st percentile of the sample (Sharma et al., 2017; Sharma & Xu, 2021). We also exclude 2,418 incumbent nursing homes that were exposed to a closure in 2009 as those would have no pretreatment period to compare to. In regression analyses, the data for each nursing home in the treated group was restricted to 5 years before and 4 years after a closure. The main analytical sample includes 90,587 facility-year data on 11,146 unique nursing homes, with 5,233 nursing homes located in counties that had a closure (treated group) and 5,913 nursing homes located in counties that did not have closures over the study period (control group).
Variables
The key independent variable in our analysis is the closure of a nursing home in a given nursing home market. In the main model, we define a nursing home market at the county level. Several studies use counties as the market level for nursing homes due to funding and regulatory policies, resident preferences, and the availability of county-level data on indicators of nursing home access, utilization, and performance (Grabowski, 2008; Lin, 2015; Park & Werner, 2011). For markets that experienced multiple closures, we use the first closure during the study period as the primary event for incumbent nursing homes. Incumbent nursing homes located in the same market where the closed facility was located are included in the treatment group. Nursing homes located in markets that did not experience any closure during the study period are the control group.
The primary outcome measures of interest include annual average occupancy rate, net profit margin, and operating margin. Occupancy rate is the number of residents divided by the total number of beds from the LTCfocus data. Financial measures from the Medicare cost report are net profit margin measuring overall financial performance and operating margin measuring profitability from providing services to nursing home residents. Following a previous study (Weech-Maldonado et al., 2010), we calculate net profit margin as (total revenues – total expenses) divided by total revenues, and operating margin as (net patient revenues – operating expenses) divided by net patient revenues. Information on revenues and expenses are obtained from the Medicare Cost Report, Worksheet G3 (see details in Supplemental Table S1 online).
To investigate potential mechanisms, we evaluate additional outcomes related to revenues, expenses, and resident mix. We focus on three measures of resident mix: percent of Medicaid-paid residents, percent of Medicare-paid residents, and the Resource Utilization Group (RUG) case mix index. We use the natural logs of revenue and expense measures in the regression analysis described below as these financial data are heavily skewed. All financial measures are inflation-adjusted to 2018 U.S. dollars.
Statistical Analysis
There were nursing home closures each year and new nursing homes were exposed to closure (treated) in each year between 2010 and 2018. Hence, our analytical sample includes nine different cohorts of “treated” nursing homes based on the first year of treatment. With varying treatment timing, two-way fixed effects (TWFE) difference estimates are biased when treatment effects vary over time (Goodman-Bacon, 2021). To address staggered treatment and this potential bias, we employ the Callaway and Sant’Anna (C&S) difference-in-differences model (Callaway & Sant’Anna, 2020). In this approach, we use never treated as controls. We compare changes in outcomes between incumbent nursing homes in markets with any closure and nursing homes in markets without a closure over the same period. The C&S model compares all possible two-group/two-period combinations and then aggregates them into an overall average treatment effect (Callaway & Sant’Anna, 2020).
We begin with a C&S model without any covariates. Under the (unconditional) parallel trends assumption, the C&S model should give us unbiased estimates of treatment effects. However, it is more plausible that the trends affecting occupancy and financial performance of nursing homes with different characteristics (e.g., for-profit status) differ. Therefore, we estimate the model controlling for a set of facility, resident and market characteristics to identify the treatment effects under the conditional parallel trend assumption which states that parallel trends hold conditional on a set of time-invariant observed characteristics. The selection of covariates that potentially affect nursing home occupancy and financial performance is informed by previous literature (Castle et al., 2009; Pradhan et al., 2013; Zinn et al., 2009). Specifically, we include facility characteristics (number of beds, for-profit status, chain affiliation), staffing level (nursing staff hours per president day for certified nursing assistants [CNA], licensed practical nurse [LPN], and registered nurse [RN]), patient characteristics (percent of Medicaid-paid residents and percent Medicare-paid residents, average age, percent female, percent White, and average RUG case mix index), and market factors (market concentration, number of home health agencies, percent population 65 or older, household income, and unemployment rate). C&S model only uses time-invariant baseline covariates for the estimation of the propensity score and outcome regressions; in doing so, it overcomes the “bad control” problem when time-varying covariates are affected by the treatment.
We also estimate an event study to evaluate pretrends as well as treatment effects over time. In addition to visually examining the statistical significance of each coefficient in preclosure periods from event study plots, we report the p-value for the pretrend test that evaluates whether all coefficients of the preclosure periods are jointly equal to zero.
Because the C&S model is essentially a series of difference-in-differences models for each treatment cohort, it practically accounts for year and cohort effects. We use the doubly robust difference-in-difference estimator based on stabilized inverse probability weighting and ordinary least squares. We cluster standard errors at the facility and the county level to account for correlation for the same facility over time and correlation for facilities located in the same county. We determine statistical significance at p < .05. All analyses were performed using STATA Version 17 (StataCorp, 2021). As noted above, we estimate closure effects separately in metropolitan, micropolitan, and rural counties.
We conduct several robustness checks to assess whether estimates are sensitive to our sample selection criteria or model choices. First, we exclude nursing homes located in counties with any nursing home entrants and/or multiple closures to estimate the effects of a single nursing home closure. Second, data from the year before closure can be noisy, hence we re-estimate the model excluding data from the year before closure and using two years before closure as the reference period. Third, to account for the change from RUG-III to RUG-IV in late 2010, we limit the sample to 2011 and onward. Fourth, instead of counties, we use the distance from the closed nursing home to define a nursing home market and identify nursing homes that were exposed to any closure in that market as noted above. Consistent with previous studies, we use different radii (15, 10, or 5 miles) to define a nursing home market (Bowblis & Vassallo, 2014; Cornell et al., 2020; Tyler & Fennell, 2017). In those models, we identify treated nursing homes within 15, 10, or 5 miles around a closed nursing home. Nursing homes that are not within a market experiencing closure based on the chosen radius are assigned to the control group. We use nursing home latitude and longitude coordinates from LTCFocus data to calculate distances between two nursing homes using STATA’s geonear command. Fifth, our identification of nursing home closure using the termination status in the POS data may capture temporary closures and new facilities with different provider ID and/or names can reopen in the same address soon after closure. Following previous literature (Bowblis & Vassallo, 2014; Castle, 2005; Sojourner et al., 2015), we verified each closure identified in the POS data by ensuring that there was no other facility at the same address for at least 3 years following the closure. Sixth, we aggregate nursing home data at the county level and re-estimate the C&S model using county-year panel data. Finally, we estimate a conventional TWFE model.
Results
Sample Characteristics
Supplemental Table S2 summarizes the characteristics of nursing homes in the analytical sample in 2009 (first study year). Compared to nursing homes located in never-closure counties, nursing homes in ever-closure counties were similar in occupancy rates (85% vs. 84%), net profit margins (2.69% vs. 3.05%), and operating margins (−0.54% vs. 0.33%). Nursing homes in ever-closure counties had more beds (121 vs. 105) and were less likely to be chain-affiliated (57% vs. 61%). Nurse staffing levels are similar at baseline between nursing homes located in ever-closure counties and never-closure counties. In terms of county characteristics, ever-closure counties were more likely to be metropolitan (52% vs. 37%), unconcentrated (30% vs. 6%) and have more home health agencies (7 vs. 2) on average and higher household income (US$54,016 vs. US$49,475) in 2009.
Main Sample Difference-in-Differences Estimates
Table 1 reports the average treatment effect estimates from the C&S difference-in-differences analysis without any covariates (Model 1) or with a full set of covariates listed in Supplemental Table S2 (Model 2). Overall, the estimates are similar between the two models. The estimates of nursing home closure effects on the occupancy rates in the overall sample and metropolitan subsample are smaller and no longer significant at the 5% level after controlling for covariates. Based on the p-value for the pretrend test, the conditional parallel trends assumption holds in all regressions in Model 2, whereas there is a violation in the unconditional parallel trends assumption when evaluating the closure effect on occupancy rates in the rural subsample. Hence, Model 2 in which the CSDID model controls for a full set baseline facility and county characteristics is preferred. Model 2 suggests that in the overall, metropolitan and micropolitan samples, estimates of nursing home closure effects on the occupancy rates and profit margins of incumbent nursing homes in the same county are small and statistically nonsignificant for all outcomes. However, in rural counties, the occupancy rates of incumbent nursing homes increase by 2.68 percentage points (p < .001) following a closure. There are no significant effects on net profit margin and operating margin.
Effects of Nursing Home Closure on Occupancy Rates and Profit Margins of Nursing Homes in the Same County.
Note. All regression models were estimated using Callaway and Sant’Anna (C&S) difference-in-differences model. Aggregated group-time average treatment effects (ATT) were estimated using -dripw- method and never treated as the control group. Year and facility fixed effects are accounted for in the model. The year before closure is used as the universal base period. Model 1 does not include any covariates; Model 2 control for covariates listed in Supplemental Table S2. The null hypothesis of the pretrend test is that all aggregated group-time ATTs in pretreatment periods jointly equal zero. Robust standard errors (in parentheses) were two-way clustered at the nursing home and county level.
p < .05. **p < .01. ***p < .001.
Figure 1 shows the event study plots of the nursing home closure effects on occupancy rates of nursing homes in the same county for the overall sample and subsamples by rurality. Overall, we find no significant differences in nursing home occupancy rate trends between ever- and never-closure counties in the years before closure, although the point estimate 3 years before closure in rural counties rises to be statistically significant at the 5% level. The year-by-year estimates postclosure suggest some differences in effects over time in the micropolitan and rural samples. In micropolitan counties, there is a trend toward a decline in occupancy after 3 years. In rural counties, incumbent nursing homes have the largest increase in occupancy rates throughout the postclosure period.
Event Study Plots of the Effect of Nursing Home Closure on Occupancy Rates of Nursing Homes in the Same County.
Figures 2 and 3 present the event study plots for the effects of nursing home closure on the net profit margins and operating margins of nursing homes in the same county for the overall sample and subsamples by rurality. Overall, there are no significant preclosure trend differences in the net profit margins or operating margins between ever- and never-closure counties before closure. However, in the rural sample, we note that the point estimate for the net profit margin three years before closure is statistically significant. Overall, we see no consistent pattern of closure effects on profit margins of incumbent nursing homes. In rural counties, there is a positive effect on the operating margins of incumbent nursing homes in the fourth year after closure, which is statistically significant.
Event Study Plots of the Effect of Nursing Home Closure on Net Profit Margins of Nursing Homes in the Same County.
Event Study Plots of the Effect of Nursing Home Closure on Operating Margins of Nursing Homes in the Same County.
Supplemental Table S3 online reports effect estimates for the additional financial and resident mix outcomes. In rural counties, there are significant increases in total revenue, net patient revenue, total expenses, and total operating expenses with the increase in revenue outweighing the increase in expenses. In the overall sample and metropolitan sample, there are significant decreases in the resident mix.
Estimates From Alternate Models
Table 2 shows the estimated effects of nursing home closure using a sample excluding nursing homes located in counties that experienced any nursing home entrants or multiple closures. The results are largely consistent with estimates from the main model. There is a small increase in the occupancy rates by 0.63 percentage points in the overall sample, which is statistically significant (p < .05). The magnitude of the increases in the occupancy rates in the rural sample is slightly smaller than that obtained from the main model. Excluding data from the year before closure also gives largely consistent estimates (Table 3).
Effects of Nursing Home Closure on Occupancy Rates and Profit Margins of Nursing Homes in the Same County: Excluding Nursing Homes Located in Counties With Nursing Home Entrants and/or Multiple Nursing Home Closures.
Note. All regression models were estimated using Callaway and Sant’Anna (C&S) difference-in-differences model. Aggregated group-time average treatment effects (ATT) were estimated using -dripw- method and never treated as the control group. Year and facility fixed effects are accounted for in the model. The year before closure is used as the universal base period. All models control covariates listed in Supplemental Table S2. The null hypothesis of the pretrend test is that all aggregated group-time ATTs in pretreatment periods jointly equal zero. Robust standard errors (in parentheses) were two-way clustered at the nursing home and county level.
p < .05. **p < .01. ***p < .001.
Effects of Nursing Home Closure on Occupancy Rates and Profit Margins of Nursing Homes in the Same County: Excluding Data From the Year Before Closure.
Note. All regression models were estimated using Callaway and Sant’Anna (C&S) difference-in-differences model. Aggregated group-time average treatment effects (ATT) were estimated using -dripw- method and never treated as the control group. Year and facility fixed effects are accounted for in the model. Data from the year before closure is excluded and two years prior to closure is used as the universal base period. All models control covariates listed in Supplemental Table S2. The null hypothesis of the pretrend test is that all aggregated group-time ATTs in pretreatment periods jointly equal zero. Robust standard errors (in parentheses) were two-way clustered at the nursing home and county level.
p < .05. **p < .01. ***p < .001.
Using data from 2011 and onward also yields similar estimates (Table 4) to the main model except that the magnitude of the increases in the occupancy rates and profit margins in the rural sample are greater and all statistically significant. Specifically, the net profit margin and operating margin increase by 1.41 percentage points (p < .05) and 1.67 percentage points (p < .05), respectively, for incumbent nursing homes in the rural counties following a closure. Supplemental Table S4 shows effect estimates for the additional financial and resident mix outcomes using data from 2011 and onward. The estimated effects on revenues and expenses are consistent with the main model but the estimated effects on resident mix outcomes are quite different from the model. There are no significant changes in resident mix outcomes in the overall, metropolitan, and micropolitan sample; however, there is a significant increase (2.32 percentage points) in the percent of Medicare-paid residents in the rural sample.
Effects of Nursing Home Closure on Occupancy Rates and Profit Margins of Nursing Homes in the Same County: Using Data From 2011 and Onward.
Note. All regression models were estimated using Callaway and Sant’Anna (C&S) difference-in-differences model. Aggregated group-time average treatment effects (ATT) were estimated using -dripw- method and never treated as the control group. Year and facility fixed effects are accounted for in the model. The year before closure is used as the universal base period. All models control covariates listed in Supplemental Table S2. The null hypothesis of the pretrend test is that all aggregated group-time ATTs in pretreatment periods jointly equal zero. Robust standard errors (in parentheses) were two-way clustered at the nursing home and county level.
p < .05. **p < .01. ***p < .001.
Supplemental Table S5 shows the estimated effects of nursing home closure using distance-based market definitions. The results are largely consistent with our main findings using a county-based market definition. There are significant increases in occupancy rates, net profit margins, and/or operating margins for incumbent nursing homes located within 10 miles or 15 miles of a closed facility in rural areas. Applying a 5-mile radius market definition, increases in profit margins for incumbent nursing homes in rural counties become statistically nonsignificant. There is also a small increase in occupancy rates in the overall sample using a 5-mile radius market definition. However, when using a radius-based market definition, pretrend test is rejected in a few models, suggesting potential violations of conditional parallel trends.
The results are consistent when we use a more stringent definition of nursing home closure which required a closure of a nursing home followed by no other nursing homes opening at the same address for the subsequent 3 years (Table S6). The estimates based on aggregated county-year panel data (Table S7) are consistent with the results based on nursing homes as the unit of analysis and indicate significant increases in occupancy and net profit margins in rural counties. The TWFE estimates (Table S8) are generally consistent with the main C&S estimates.
Discussion
In one of the first studies that examine the effects of nursing home closures on occupancy rates and finances of other nursing homes in the same market, we find that, in the overall sample, there are no discernable effects from closures on the occupancy rates and profit margins of incumbent nursing homes. However, in rural areas, nursing home closure increases the occupancy rates of incumbent nursing homes by about 2.7 percentage points. This finding is robust to alternative sample definitions including using a more stringent definition of nursing home closure. There is some evidence suggesting that closure also has positive effects on the profit margins of incumbent nursing homes in rural areas. For example, limiting the sample from 2011 (instead of 2009) and onward, we find rural nursing home closure increases the net profit margin and operating margin of incumbent nursing homes by 1.4 and 1.7 percentage points, respectively.
Our findings of an increase in occupancy rates for incumbent nursing homes in rural areas echo findings from studies on hospital closures reporting an increase in patient volumes for nearby hospitals (Lindrooth et al., 2003; McKay & Dorner, 1996; Ramedani et al., 2022; Song & Saghafian, 2020). Using administrative data on the universe of nursing home residents between 2001 and 2014, a recent working paper found that about 85% of displaced residents of closed nursing homes transferred to other nursing homes (Olenski, 2022). Researchers have shown that there are fewer HCBS providers in rural market areas around closed nursing homes than in urban market areas (Tyler & Fennell, 2017). Since geographic proximity is a dominant factor for residents when selecting long-term care services (Kane & Kane, 2001), it is possible that displaced residents of closed nursing homes in rural areas are more likely to transfer to incumbent nursing homes in the same market.
Nursing homes in rural areas, on average, are smaller (79 vs. 97 vs. 117) and have lower occupancy rates (75% vs. 76% vs. 80%) than nursing homes in micropolitan and metropolitan counties (Sharma et al., 2022). Reports have shown that rural nursing homes struggle with low occupancy rates (Healy, 2019; Kacik, 2019). A 2.7 percentage-point increase in occupancy rate can be meaningful to some rural nursing homes. Prior studies have shown that higher occupancy leads to better production efficiency and correlates with higher profit margins (Nyman & Bricker, 1989; Weech-Maldonado et al., 2010).
Although our main model shows no evidence of spillover effects of nursing home closure on profit margins, some alternative sample specifications suggest that nursing home closure has positive effects on the profit margins of incumbent nursing homes in rural areas. In the hospital literature, the evidence on the spillover effects of hospital closure on the finances of nearby hospitals is mixed as noted above (Hodgson et al., 2015; McKay & Dorner, 1996). One study using California hospital data showed that hospital closure decreased profit margins for nearby non-profit hospitals but found no changes in the payer mix, suggesting increased costs of treating patients (Hodgson et al., 2015). In the main sample using data from 2009 to 2018, we find a slight decrease in the payer mix or resident acuity level of incumbent nursing homes in the overall sample (mainly metropolitan areas) following a closure. This finding of changes in the case mix may be complicated by the transition from RUG-III to RUG-IV in late 2010 and greater HCBS supply and utilization in urban areas (Coburn et al., 2019; Siconolfi et al., 2019). Limiting the sample to 2011 and onward, we did not observe such effects. However, in rural areas, incumbent nursing homes saw an increased share of residents paid by Medicare following a closure. Increasing Medicare revenues is one of the main strategies for nursing homes to ensure profitability (Harrington et al., 2011). This is also consistent with observed greater increases in revenues than the expenses of incumbent nursing homes in rural areas. Improved finances potentially help the remaining nursing homes in their struggle for survival and allow nursing homes to invest in the quality of care (Park & Werner, 2011; Sharma & Xu, 2021; Weech-Maldonado et al., 2019).
This study provides empirical evidence to policymakers debating nursing home closure and access. Rural nursing home closure is particularly concerning due to its impact on access to nursing home care considering the longer travel distances to nursing homes and the limited availability of HCBS providers (Tyler & Fennell, 2017). However, the exits of less efficient firms can improve the overall market efficiency. It is questionable whether the exits of some nursing homes help stabilize markets and ensure continued access to nursing home care. Our findings suggest that closure of nursing homes leads to increases in occupancy at nearby nursing homes in rural areas and possibly helps improve their financial viability. Since rural areas have more nursing home beds per capita on average (Sharma et al., 2022), some areas may be able to withstand nursing home closure without compromising on access to care. However, this may not be the case in all areas, as closures could largely increase distances to available nursing homes and create nursing home deserts. Therefore, it is important to assess the availability of other nursing homes or nursing home alternatives nearby before a nursing home closes and evaluate the possibility of converting some of the closing nursing homes to adult day care centers, assisted living facilities, or home health agencies to ensure continued access to postacute and long-term care services. Understanding how local, state, and national policymakers can facilitate this process is important.
Our study has some limitations. First, a nursing home terminated from the Medicare and/or Medicaid program may continue to operate under a different provider number. We used the address information to confirm whether a terminated nursing home closed and the results from the sensitivity analysis are consistent with the main findings. Nonetheless, we do not have information to confirm whether each nursing home closure was full or partial (e.g., converting to other types of long-term care facilities). If some closures are over identified in our study, we would be underestimating the spillover effects of full closures in this study.
Second, federal regulations require a nursing home to give residents a 60-day notice before it closes and make a plan for resident relocation (Notification of Facility Closure—CMS, n.d.). As such, the termination date from Medicare and/or Medicaid program can be different from the actual timing of facility closure. We use the year of termination as a proxy for treatment year (i.e., exposure to a closure) for incumbent nursing homes, which may capture some anticipation effects but may also introduce random measurement error that attenuates estimates toward the null especially for the first year of the event study. Excluding the data from the year before closure generates similar estimated effects of nursing home closure.
Third, we use county boundaries to define nursing home markets and consider incumbent nursing homes located in the same county as the closed facility to be treated in our main analyses. There are no standard measures to define markets for nursing homes or other long-term care services, and each method presents advantages and limitations (Grabowski, 2008). There is little consensus on whether to differentiate market definitions between urban and rural areas. We address this issue by using an alternative radius-based definition to assign treatment status to incumbent nursing homes, and results are generally consistent across definitions (Bowblis & Vassallo, 2014; Cornell et al., 2020; Tyler & Fennell, 2017).
Fourth, although we generally observe no systematic pretrends that would bias our estimates, there can be contemporaneous events such as changes in payment rates and the supply of HCBS that confound the relationship between exposure to closure and occupancy rates and profit margins.
Fifth, although we identified closures of all Medicare- and/or Medicaid-nursing homes, our analytical sample of active nursing homes excludes hospital-based nursing homes and Medicaid-only certified facilities as they do not report financial information in this data. However, hospital-based nursing homes and Medicaid-only certified facilities represent a small proportion of the nursing home industry and therefore their exclusion should not meaningfully limit the generalizability of our findings.
Finally, this study does not examine the effects of the closure on the quality of care of incumbent nursing homes or resident access to different types of care due to the limitation of scope and data availability. However, occupancy rates are one of the first-order outcomes in the causal pathway of how a nursing home closure affects the nearby nursing homes and financial resources are essential for nursing homes to invest in quality of care. Future research should investigate other effects of nursing home closure on nearby incumbent nursing homes.
Our findings suggest that nursing home closure improves the occupancy rates of remaining nursing homes in the same market in rural areas but there is little evidence of closure effects on occupancy rates and financial performance in metropolitan and micropolitan areas. Higher occupancy rates and potentially absorbing more residents paid by Medicare are beneficial for the financial performance of rural nursing homes, which often struggle to survive. Compared to urban areas, rural areas have a higher supply of nursing homes but a lower supply of HCBS providers. Nursing home closure has multifaceted impacts on service availability, care access and the local long-term care system. While rural nursing home closure has attracted substantial policy attention, nursing home regulators and local officials should consider the long-term care market heterogeneity. This includes evaluating the availability and accessibility of alternative nursing homes and/or HCBS providers when considering interventions targeted at nursing home closure. Future research should evaluate the spillover effects of nursing home closure on the quality of care in nearby nursing homes using resident-level data and other intermediary outcomes such as staffing. It would also be worth investigating the spillover effects of nursing home closure on alternative care providers such as home health agencies and assisted living facilities.
Supplemental Material
sj-docx-1-mcr-10.1177_10775587241296182 – Supplemental material for Effects of Nursing Home Closures on Occupancy and Finances of Nearby Nursing Homes
Supplemental material, sj-docx-1-mcr-10.1177_10775587241296182 for Effects of Nursing Home Closures on Occupancy and Finances of Nearby Nursing Homes by Lili Xu, Hari Sharma and George L. Wehby in Medical Care Research and Review
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
