Association of Receipt of Paycheck Protection Program Loans With Staffing Patterns Among US Nursing Homes

This economic evaluation assesses the association between Paycheck Protection Program loan receipt and nursing home staffing hours.


Introduction
The COVID-19 pandemic magnified several long-standing problems within nursing homes (NHs), including critical staffing shortages. 1 In March 2020, at the start of the pandemic, more than 1 in 5 NHs in the US reported severe staffing shortages. 1 The most commonly reported staff shortage was in nurse aides (18.4%), licensed nurses (15.9%), clinical staff (2.5%), and other staff (9.8%). 2 Staffing shortages in NHs have threatened the quality of resident care significantly, especially when these shortages have occurred among frontline care staff. 1,2 Nursing homes with a greater proportion of Medicare residents were less likely to experience staffing shortages. 2 Compared with Medicaid, which often pays less than reported operating costs, Medicare provides higher reimbursement to NHs, allowing for more investment in staff and quality. 2,3 This reimbursement plan suggests that the funding NHs receive is directly associated with staffing efforts. 2 As such, stakeholders have advocated for increased reimbursement to NHs; however, NH advocacy groups, such as LeadingAge, have warned against simply raising funds for NHs without ensuring that they will go toward areas important to resident care, such as staffing. 4 To provide a solution that ensures that staffing receives adequate investment, a number of states, federal entities, and most recently, the National Academies of Sciences, Engineering, and Medicine have called for specified proportions of NH funding (eg, Medicare and Medicaid payments) to be appropriated toward staffing. 5 Information is still needed on what percentage of funding would be sufficient to appropriate toward staffing levels as well as whether and how appropriation is associated with NH staffing levels. In a recent study on NH spending and staffing levels, Bowblis et al 6 identified that the median proportion of revenues spent on nursing staff in 2019 was 33.9%. When examining staffing levels across a threshold of 25% to 45% of revenues, the authors found that facilities with higher shares of Medicaid residents spent a larger share of revenues on nursing staff but had lower staffing levels. That study, however, was cross-sectional and only reported NHs as having thresholds of appropriation between 25% to 45%, whereas levels of 60% to 75% have been proposed as being ideal.
One opportunity for examining how appropriated funding at a higher level may affect staffing is through the Paycheck Protection Program (PPP). The PPP was a loan program to support small businesses, including NHs, during the COVID-19 pandemic in the form of providing a direct incentive for small businesses to keep their workers on payroll. The loan was forgivable if recipients appropriated at least 60% to 75% of the loan they received toward payroll and staffing. 7 Assessing the PPP may inform how appropriating a specified proportion of funding toward staffing efforts might influence staffing levels.
We sought to understand the characteristics of NHs that received PPP loans and the changes in staffing levels (ie, hours) after they received the loans compared with NHs that did not receive PPP loans. These findings may help policy makers better understand whether the PPP may be a model for future programs seeking to appropriate specific funds to NH staffing.

Design
This economic evaluation used a difference-in-differences event-study design to estimate the association of PPP loan receipt with staffing hours at NHs that received a PPP loan compared with NHs that did not receive a PPP loan. With this approach, we were able to account for other local, state, and federal efforts that may have been occurring at the time when the PPP loans were available, providing a natural experiment. This study met predetermined criteria for institutional review board exemption according to the Common Rule due to the use of deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline and relevant parts of the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline. 8

Data Sources
Eight data sources were used for this study: the Small Business Administration (SBA), 9 Nursing Home Compare, 10 LTCFocus, 11  Rural-Urban Continuum Codes. 16 All data sources are publicly available except the MDS.

SBA
The SBA maintains data on all small businesses that received PPP loans, including NHs. 9 The PPP originated under the 2020 CARES Act. 17 Almost any small business that existed before February 15, 2020, was able to apply for a forgivable loan of up to $10 million to retain their workforce and assist with operational expenses. Nursing homes included in this data set meet the SBA definition of a small business as having fewer than 500 employees or, in the case of nursing care facilities, less than $30 000 000 in annual revenues. At least 60% to 75% of the PPP loan needed to be used for payroll costs for the loan to be forgiven. To apply for loan forgiveness, the small business had to provide documentation for all payroll periods along with nonpayroll expenses that overlapped with the covered period. 18 Loan amounts were determined by calculating the average monthly cost of employee salaries and the loan amounts paid up to 8 weeks of payroll costs and benefits. After June 5, 2020, the default spend-down period became the shorter of 24 weeks or time remaining until December 31, 2020. Entities receiving loans before June 5 could opt to extend from 8 weeks to 24 weeks. A substantial number of skilled nursing facilities successfully applied for these funds. 19 We focused on funds that were disbursed during the first round of the loans, which occurred from April through August 2020.

Nursing Home Compare
Data in Nursing Home Compare are collected quarterly and allow for identification of and comparisons between NHs that are certified by the CMS. 10 The data set includes information about the quality of care provided by these NHs, including how they have performed on health inspections and how they are staffed, enforcement actions the government has taken against them, and how well their residents are treated in specific areas of care. Variables of interest from this data source were facility characteristics, staffing quality score, overall star rating, and ownership.

LTCFocus
LTCFocus is a data set organized by Brown University that provides the facility and aggregated resident characteristics of all Medicare-and Medicaid-certified NHs in the US compiled from multiple data sources. 11 We used 2018 data, which were the most up-to-date data available at the time of our analysis.

CMS Payroll Based Journal
The CMS Payroll Based Journal provides reports on direct care staffing information. 12 We used staffing hours for quarters 1 through 4 of 2020.

MDS
The MDS, owned by CMS, is a standardized assessment tool that measures the health status of NH residents. 13 These assessments are performed and recorded by NH staff and include information on resident characteristics, such as race and ethnicity, functional and cognitive status, psychosocial functioning, geriatric syndromes (eg, dementia, depression), and end-of-life care wishes. Variables of interest from this data source included race and ethnicity aggregated to the facility level.

Area Deprivation Index
The Area Deprivation Index is a measure created as a composite index with greater validity, robustness, and explanatory power than single-area measures in documenting the extent of social disparities in health and mortality 20 and is based on a measure created by the Health Resources & Services Administration. 21 This index is a nationally relative measure with scores from 1 to 100 at the census-block level. Higher scores indicate greater socioeconomic disadvantage in the neighborhood.
A national ranking of 85th percentile or higher was considered severely disadvantaged based on prior research. The 2018 version of the index was used in this study. 14

USDA Rural-Urban Continuum Codes
USDA Rural-Urban Continuum Codes distinguish metropolitan counties by the population size of their metropolitan area and nonmetropolitan counties by degree of urbanization and adjacency to a metropolitan area. 16 We present information on the rurality and urbanicity of the NHs.

Healthcare Cost Report Information System Skilled Nursing Facility Cost Data Set
Medicare-certified facilities are required to submit an annual cost report to a Medicare administrative contractor. The cost report contains institutional information, such as facility characteristics, utilization data, cost and charges by cost center (in total and for Medicare), Medicare settlement data, and financial statement data. 15 We used these data to calculate total revenue for the NH.

Measures Demographic Variables
Demographic characteristics of the NHs that were used for analyses included the percentage of Black residents in the NH (we focused on the percentage of Black residents because other studies have found poorer staffing and resident outcomes in nursing homes with greater percentages of Black residents 22,23 ), region, urbanicity (urban vs rural), Area Deprivation Index (dichotomized, Ն85th percentile [severely disadvantaged] vs <85 percentile [less disadvantaged]), staffing quality rating (range, 1 to 5; a higher rating suggests that the facility has sufficient staffing levels and a higher likelihood of providing quality care and attention to its residents), overall star rating (range, 1-5), bed size, total residents per day, percentage occupancy, ownership (for profit, nonprofit, or government), chain ownership, and percentage of Medicaid residents in the NH.

Outcome and Independent Variables
Our outcome variables were continuous and included staffing (registered nurse [RN], licensed practical nurse [LPN], and certified nursing assistant [CNA]) total weekly hours. Our main independent variable was the PPP loan, dichotomized as loan recipient vs nonrecipient.

Statistical Analysis Study Design
We merged the 8 data sources to conduct our analyses. We used business addresses as the linking identifier between PPP and Nursing Home Compare data. We then linked the additional data sets by the institution identification numbers, which were contained in the NH Compare data set and subsequent data sets. We first examined descriptive characteristics of the study sample using a Kruskal-Wallis test for continuous outcomes and a χ 2 test for categorical outcomes to compare characteristics of facilities that received a PPP loan with NHs that did not. We then estimated multiple event-study models based on weekly staffing hours for the 3 staff types (RN, LPN, and CNA) using generalized linear model linear regression models that included fixed effects for each nursing home and calendar week. Using an event-study approach allowed us to estimate the association of PPP loan receipt with staffing total weekly hours compared with changes in staffing total weekly hours experienced in non-PPP recipient NHs over the same period.

Study Sample
Only NHs that were eligible for the PPP loan were included in our analyses (NHs with <$30 million in revenue). Additionally, staffing data were not available in the CMS Payroll Based Journal data set for the first quarter of 2020. Moreover, the median loan date was April 15, 2020. Therefore, to ensure that we were able to extend the event time to more than 2 weeks before loan receipt to assess for parallel trends, we conducted a complete case analysis for NHs that had staffing data for all quarters.
For each PPP loan recipient, we identified all NHs in the same county that did not receive PPP loans during the study period. Matched neighboring facilities served as controls. We dropped any facility that was not a treatment facility or matched control. This approach allowed us to account for events that were occurring in the county that might have additionally affected staffing, such as changes in numbers of COVID-19 cases.

Event Time
The study period was from January 1 to December 23, 2020. Event time was defined on a weekly basis relative to the date of initial PPP loan receipt for the treatment facility. Pre-event time included all weeks that preceded the loan recipient date (week 0) during the study period. Postevent time included all weeks that followed the loan recipient date during the study period. Event time was truncated on a monthly basis between −12 weeks and 24 weeks. The reference week was week −1, immediately preceding loan receipt, to assess the difference between the control and PPP recipient group. Control facilities had the same event time as their matched treatment facility.
All analyses were conducted in SAS, version 9.4 (SAS Institute Inc). We report coefficients, 95% CIs, and P values. The coefficients represent the adjusted mean difference attributed to that variable.
The significance level was set at 2-sided P < .05. NHs. While total staffing hours continued to be greater among NHs that received a PPP loan compared with NHs that did not receive a PPP loan during the post-loan receipt period, 12 weeks after receipt, the magnitude in the differences in staffing hours between the PPP recipient NHs and the nonrecipient NHs generally plateaued ( Table 2 and eFigures 1-5 in Supplement 1).

Discussion
We sought to understand whether a program that stipulated the appropriation of a specified proportion of funding toward staffing was associated with changes in staffing hours within NHs.
After linking multiple data sources that include information on NH populations and quality, small businesses, and communities, we found that receipt of PPP loans was associated with increases in staffing hours for CNAs and LPNs compared with no receipt of PPP loans among PPP-eligible NHs.
Nursing homes that received PPP loans had lower staffing hours before the COVID-19 pandemic and were more likely to have facility characteristics reflective of poorer quality of care compared with NHs that did not receive the loans.
It is important to put into context the median loan amount for NHs that received the PPP loan.
The CARES Act stimulus package 24 provided a baseline payment of $50 000 plus an additional $2500 per bed, which for an NH of 100 beds, would equate to $300 000. Through the PPP loan, NHs received a median loan amount of $664 349, which is about twice as much as that from the stimulus package. Regarding the cost of staffing, in 2021, the median cost of an RN was $77 600 annually and $37.31 hourly. 25 The median cost of a CNA was $30 290 annually and $14.56 hourly. 26 Based on these wages, the NH would have been able to pay 17 153 additional RN hours or 43 956 additional CNA hours with the loan that they received. 27 As the NH could have paid for other payroll and nonpayroll costs, such as benefits, it is unclear how much of the PPP loan was dedicated to staffing hours.
Regarding the differences in the change in staffing hours between PPP recipients and NHs that did not receive a loan, NHs that received the PPP loan were able to provide a significantly greater number of CNA hours; these equated to 2 additional CNA shifts per week as soon as 4 weeks after the Thus, when seeking to increase the hours of staff available to provide care to residents, it might have been most practical for NHs to focus on employing the greatest number of staff possible with limited funding. Lastly, it is likely that RNs were not as readily available to fill staffing shortages in NHs, as other settings, such as acute care, were experiencing shortages also and incentives to work in those settings were greater.
A similar approach to bolstering staffing and reducing vacancies and turnover through required funds toward staffing is wage pass-through laws. A wage pass-through is an additional allocation of funds provided through Medicaid reimbursement for the purpose of increasing compensation, benefits, or the number of staff for direct care workers. Twenty-one states and Washington, DC, have wage pass-through laws, in which a study found that the introduction of wage pass-throughs over the study period was associated with between 3.0% and 4.0% net increases in CNA hours per residents per day in the years following adoption. 29 Currently, the allocation of funds for wage passthroughs varies across states, and states can designate either a set dollar amount per staff hour (or resident day) or a certain percentage of a reimbursement increase to be used for wages and benefits.
Our findings provide more information to states for understanding how allocating a certain percentage of funds may change staffing levels.
Our findings demonstrated that the NHs that were most in need of funding, as evidenced by lower quality and staffing scores, were recipients of the PPP loans. Other characteristics that have been associated with lower quality were also more prevalent among PPP loan recipients, which were more likely to be for-profit NHs. These findings reflect the importance of ensuring that NHs with facility characteristics associated with poorer quality care have access to and are aware of funds for operational support.

Implications
Our findings have important policy and research implications, the first of which is supporting advocacy for the need to appropriate funds. 30 While our research was unable to provide information on variation in staffing hours based on the level of appropriation, proportions that have been proposed in the past include 60% and 75%. 31

Limitations
There are several limitations. First, because of the absence of institution identification numbers in the PPP database and having to merge the business names and addresses of NHs therein with corresponding information in the NH Compare database, there could have been discrepancies in this approach. It is possible that we did not capture all NHs that received a loan, and/or we may have erroneously identified an NH that received a PPP loan. However, this would bias findings toward the null. Second, because NHs were not required to appropriate 60% to 75% of the PPP loan to staffing if they did not care for loan forgiveness, we do not know which NHs in our sample did so. However, nearly 100% of the loans were forgiven; therefore, it is likely that the majority of the NHs in our PPP sample appropriated requisite funds to staffing. 36 Next, the PPP loans were intended to help with external stressors of NH staffing (ie, the pandemic), not intrinsic stressors; therefore, staffing changes potentially attributable to PPP loans may not be generalizable to expected changes for funding to support nonpandemic staffing shortages. Lastly, it would have been ideal to examine whether the receipt of PPP loans made NHs less likely to report a staffing shortage; however, these data only became available beginning May 2020, which did not allow for us to look at these data before or immediately after the PPP.
JAMA Network Open | Health Policy eFigure 2. Change in total hours per week for all nursing staff pre-and post-PPP loan receipt compared to non-PPP NH loan recipients eFigure 3. Change in total hours per week for CNAs pre-and post-PPP loan compared to non-PPP NH loan recipients eFigure 4. Change in total hours per week for LPNs pre-and post-PPP loan compared to non-PPP NH loan recipients eFigure 5. Change in total hours per week for RNs pre-and post-PPP loan compared to non-PPP NH loan recipients SUPPLEMENT 2. Data Sharing Statement