key: cord-0874509-an66eqjc
authors: Iyanda, Ayodeji Emmanuel; Adu Boakye, Kwadwo
title: A 2-year pandemic period analysis of facility and county-level characteristics of nursing home coronavirus deaths in the United States, January 1, 2020 – December 18, 2021
date: 2022-02-21
journal: Geriatr Nurs
DOI: 10.1016/j.gerinurse.2022.02.013
sha: d9466989202688339dde1b0e5f0da0535e55d00b
doc_id: 874509
cord_uid: an66eqjc

Nursing home residents are highly susceptible to COVID-19 infection and complications. We used a generalized linear mixed Poisson model and spatial statistics to examine the determinants of COVID-19 deaths in 13,350 nursing homes in the first 2-year pandemic period using the Centers for Medicare and Medicaid Services and county-level related data. The average prevalence of COVID-19 mortality among residents was 9.02 per 100 nursing home beds in the first 2-year of the pandemic. Fully-adjusted mixed model shows that nursing homes COVID-19 deaths reduced by 5% (Q2 versus: Q1 IRR = 0.949, 95% CI 0.901 – 0.999), 14.4% (Q3 versus Q1: IRR = 0.815, 95% CI 0.718 – 0.926), and 25% (Q2 versus: Q1 IRR = 0.751, 95% CI 0.701 – 0.805) of facility ratings. Spatial analysis showed a significant hotspot of nursing home COVID-19 deaths in the Northeast US. This study contributes tonursing home quality assessment for improving residents' health.

Coronavirus disease 2019 , caused by the Severe Acute Respiratory Coronavirus 2 (SARS-CoV-2), has a pronounced incidence rate and higher mortality in people aged 60 and above and in people with underlying medical conditions. 1, 2 Notably, people in long-term care facilities (LTCFs) such as nursing homes (NHs) are highly susceptible to SARS-CoV-2 infection. Communal settings in NHs and LTCFs increase the risk of exposure for residents mainly because of the adverse health outcomes, the advanced age of residents, proximity to other highly vulnerable facilities such as meat and livestock plants, interacting with exposed staff and family members, and the in-and-out-movement of health care workers among facilities and communities. [3] [4] [5] [6] [7] [8] [9] Contact network and quality health outcome models help us to understand the medical geography of nursing home COVID-19 mortality in the United States (US).

Quality health outcome models 10 assume that multiple factors, including general facility conditions (staffing, number of beds, presentable environment), affect the quality of care relative to the health outcomes in LTCFs. 11 Systematic and meta-analysis studies have found a strong relationship between NHs facility ownership status and health outcomes. 12, 13 For example, a meta-analysis examining quality care and health outcomes concluded that for-profit NHs provided poor quality care. 13 Research on community spread of SARS-CoV-2 at the beginning of the pandemic links health care facility characteristics such as the lack of emergency preparedness and training to handle the pandemic, shortage of personal protective equipment, staff shortages, and general physical conditions. [14] [15] [16] [17] However, these studies overlooked the role of facility ratings as a determinant of nursing home COVID-19 outbreaks. Nevertheless, the importance of facility quality rating has been confirmed in several studies investigating heart failure, patients with dementia, and consumer demand and choices about nursing homes in different locations. [18] [19] [20] [21] Studies based on epidemiological models emphasize how the contact networks lead to SARS-CoV-2 spread in society. [22] [23] [24] [25] The micro-and macro-contextual conditions (MM2C) are essential in transmitting SARS-COV-2 in LTHCs, and they determine control and prevention strategies.

The MM2Cs can be modeled spatially andsystematically to examine the complex situations of disease spread in nursing homes. Chen et al.'s study using device-level geolocation data from 50 million smartphones in the US provided empirical evidence of visitor infection in NHs. 25 Sugg et al. used GIS techniques to model community-level determinants of COVID-19 transmission in NHs compared to the general population. 26 The study showed regional similarity in COVID-19 transmission in Northeast US.

From an urban structure perspective, researchers acknowledge the role of compact space and urbanization in the diffusion of diseases. However, studies investigating the association between facility characteristics and health outcomes underemphasized the importance of geography and facility proximity to other highly susceptible facilities such as meat and livestock plants (e.g., Bui et al. 14 , Li et al. 27 He et al. 28 ) in COVID-19 infection, hospitalization, and mortality rates.

Facility location is an important spatial determinant of health, and addressing disparities in health outcomes is a place-based issue. 29, 30 Study Objectives and Hypotheses

The purpose of this study is to further the World Health Organization's 1 (WHO) objectives toward infection and prevention control (IPC) in LTCFs. Our study examines the consequence of failed IPC strategies leading to COVID-19 deaths by assessing facility and community-level characteristics. In addition, we examined the geographic distribution of nursing home COVID-19 deaths across the contiguous US. We hypothesize that (1) nursing home characteristics (e.g., facility ratings, nursing home facility ownership, and facility locations) determine the prevalence of nursing home COVID-19 deaths, and (2) nursing home COVID-19 deaths are randomly distributed across the US.Findings from this study contribute to disease prevention that would improve the health and well-being of older populations in nursing homes.

This cross-sectional study examined the prevalence of nursing home COVID-19 deaths since the beginning of the pandemic in the US. In addition, we examined the association between NHs characteristics (e.g., facility ratings, nursing home facility ownership, and facility locations) the prevalence of nursing home COVID-19 deaths. We used the nursing home facility dataset and county-level dataset.

Nursing home data sources: We linked facility profile data and nursing homes COVID-19 health data from the Centers for Medicare and Medicaid Services (CMS). 31 We used the most recently County-level data sources: Non-nursing home data included the county-level demographic data from the American Community Survey (ACS) provided by the US Census Bureau. 33 We selected the 5-year estimates from the 2015-2019 ACS (DPO5) with a specific interest in the estimates of the older population and race/ethnicity variables. 34 Also, we included a county-level air quality index from the National Environmental Public Health Tracking website. 35 County-level COVID-19 vaccination rates and social vulnerability index (SVI) were retrieved from the CDC Wonder, 36 while urban-rural continuum data came from the U.S. Department of Agriculture. 37 For our spatial analysis, we included butcher locations data provided by ESRI Business Analytics to examine connections between the nursing home and community transmission that may contribute to COVID-19 infection and deaths.

Out of the 15,450 nursing homes in the United States, the full facility-level dataset used in this study contained only records of 15,250 nursing homes. Thus, we successfully joined 13, 335 nursing homes addresses to other county-level datasets using the county names as the primary key identifier. This current study was exempted from ethical approval because the data is publicly available online and has no direct human contact.

Outcome variable: The primary outcome variable in this study was the cumulative COVID-19 deaths among residents in nursing homes across the US from January 1, 2020, and December 18, 2021 which represents the 2-year pandemic period in this study. COVID-19 attributed deaths among nursing home residents are self-reported by nursing homes administrators. We treated COVID-19 deaths as a count variable in our analysis because there are several facilities without any reported COVID-19 related deaths. We determined the prevalence rates by dividing the cumulative reported resident COVID-19 related deaths by facility size, calculated as the number of nursing home beds per 100 residents.

The main facility-level variable of interest was the facility ratings based on several characteristics, including quality measure rating, health inspection rating, staff rating, short-term rating, longtime rating, and registered nurse staff rating. The composite facility rating 31 was primarily used as an exposure variable at Level 1. Other facility-level control variables used were coronavirus infection rates among staff and residents, contributing to COVID-19 mortality. We also included data on resident and staff coronavirus vaccination because it is expected that an increase in vaccination among staff and residents should reduce the prevalence and incidence (new death reports) in nursing homes and society. Additionally, we used the 2013 National Center for Health Statistics of urban-rural continuum codes to form a classification scheme that distinguishes metropolitan counties by the population size of their metro area and nonmetropolitan counties by the degree of urbanization and adjacency to a metro area. 37 Hence, the urbanization status for each nursing home facility was based on the corresponding rural-urban category (see Appendix A).

We used geospatial and nonspatial techniques to achieve the current research objectives presented under the 'Study Objectives and Hypoteses' subsection.

Nonspatial analysis: We examined descriptive statistics (mean, median, standard deviation, frequency, and percentages) for continuous and categorical variables used in this present study.

Kolmogorov-Smirnov test and visualization of the histogram of the outcome variable (COVID-19 deaths) indicate a rightly skewed distribution with many zeros. Following a similar approach, 14 we used nonparametric analysis of variance (ANOVA) when k > 2 to determine whether the average count of nursing home COVID-19 death varies among three categorical explanatory variables (i.e., quartiles of facility ratings, ownership status, and urbanization status).

We assessed all continuous variables for multicollinearity using the variance impact factor (VIF) and retained variables with a VIF value of less than three.

In the preliminary analyses, we included several county-level variables including a measure of the urbanization status (i.e., urbanity), proportion of the population ages 65 and above who had received at least two doses of COVID-19 vaccines, SVI, and air quality index. We selected these variables based on a priori knowledge of potential facility and community conditions that increase potential exposure of older people living in LTCFs 26, 38, 39 and existing epidemiological findings indicating the impact of location on health outcomes from different facilities. 40, 41 Consequently, we included the six classifications of rural-urban and the proportion of the urban population ages 65 and above who were completely vaccinated at the county level. The urbanization status assesses the role of place as a determinant factor of COVID-19 facility deaths in the US and serves as the primary county-level variable in our multilevel and spatial analyses.

In the multivariate analysis, we fit Poisson regression using the generalized linear mixed model (GLMM) in SPSS version 28 42 to examine the association between facility ratings and prevalence of nursing home COVID-19 deaths across the US. Because NHs are nested within each county, facility ratings are expected to cluster at the county level based on the idea of MM2Cs. Hence, we used a Poisson generalized mixed-effect model with a random intercept for facility ratings to account for the clustered data structure at the county level.

Next, we estimate Poisson regression for facility attributed COVID-19 deaths from nursing homes using pseudo-likelihood estimation (PL) rather than ordinary least squares estimation used in OLS regression because of the zero-inflation in COVID-19 mortality 43 . We adjusted for multiple comparisons using the least significant difference (LSD). This study estimated five multilevel Poisson regression models to evaluate facility-based and county-based measures' contribution to COVID-19 deaths in US nursing homes. We added variables in sequence to the five different multilevel Poisson regression models. In model 1, we fit a null model which only and vaccination mandates) using the unique State FIP codes. All models assumed that the random effects structure follows a variance components structure and used robust estimation to account for model misspecification and Poisson assumptions (i.e., robust covariances). The exponential coefficients were interpreted using the incidence rate ratio (IRR) with 95% confidence intervals. Detailed statements on sensitivity analysis are presented in Appendix B. (Table 1 ). Table 1 indicates that the average COVID-19 deaths was highest in the poorest rated facilities (Quartile, Q1) but decreased as facility ratings increased in the fourth quartile (Test Statistics (df=3) = 226.41, p < 0.000, 2-tailed).

The prevalence of nursing home COVID-19 deaths also varies by urbanization status (Test Statistics (df=5) = 39.127, p < 0.000, 2-tailed). As shown in Table 1 , resident deaths attributed to COVID-19 were higher in 'for-profit' nursing home facilities compared to 'Government-owned' nursing homes or 'not-for-profit' facilities (Test Statistics (df=2) = 102.58, p < 0.000, 2-tailed). To examine the relationship between the dependent and categorical explanatory variables, four variables were overlaid: Log of COVID-19 deaths, facility ratings (Quartile 1-Quartile 4), facility ownership (3 levels), and urbanization status (6 categories). Fig. 2 shows that resident deaths attributed to COVID-19 vary by facility ownership across sublevels of urbanization.

Large metropolitan and large fringe metropolitan areas: COVID-19 deaths in the governmentowned facility were higher in the first, third, and fourth quartile than nonprofit facilities in 'Large metropolitan areas. We observe that the 'for-profit' type of facility consistently had higher COVID-19 deaths. Similarly, COVID-19 deaths were higher in government nursing home facilities in 'large fringe metropolitan areas' across the subdivision of facility ratings except in the third quartile. Also, we noticed that COVID-19 prevalence in 'for-profit' nursing home facilities was higher in the upper quartile but lowest in the facility in the third quartile.

Medium metro and micropolitan areas: Compared to the second and third quartile, the prevalence of COVID-19 deaths in the government-owned facilities in the medium metro areas was lower in the first and upper quartiles. As seen in Fig. 2 , COVID-19 mortality was higher in 'for-profit' facilities in the first quartile (Q1) in medium metropolitan and higher in the micropolitan areas in the second and third quartile of facility ratings. Compared to government and not-for-profit facilities, nursing home COVID-19 death was higher in poorest rated 'forprofit' facilities than high-rated facilities in micropolitan areas.

Nonurban and small metro areas: The prevalence of COVID-19 deaths in nursing homes in nonurban areas decreased with increased facility ratings for the three categories of facility ownership. In small metro areas, COVID-19 deaths decreased in the first and second quartiles in government-owned facilities and increased in 'for-profit' owned facilities in the third and fourth quartiles. Adjusting for other variables in the full model, nursing home COVID-19 deaths significantly reduced consistently from poorly rated to highly rated nursing home facilities (Table 2) .

Linear and interaction analyses show a similar pattern of association between the prevalence of COVID-19 facility deaths and facility rating (Appendix C). COVID-19 facility deaths decreased with a 16.4% increase in nursing home facility rating (IRR = 0.843, 95% CI 0.802-0.887), controlling for facility ownership and geographic factors. In addition, we observed a 16.7% decrease in COVID-19 facility death in 'for-profit' nursing homes compared to not-for-profit facilities. We also saw a 17% statistically significant decrease in COVID-19 facility deaths in micropolitan areas. However, there was no significant interaction between facility rating and urbanization status. Next, the GLM results for count data which ignores the potential impact of clustering, were consistent with the main analyses. There were statistically significant negative associations for Q4, Q3, and COVID-19 facility deaths (see Appendix D). However, the statistically significant association observed in GLMM for Q3 disappeared in the GLM due to failure to account for clustering or random effects. In addition, the model with removed zero records of COVID-19 deaths shows a significant negative association with Q3 and Q4 of overall facility rating (Appendix E). Summarily, the association between COVID-19 deaths in nursing homes and facility rating remained consistent in all the models.

Forest-based Classification and Regression showed that proximity to butchery sites was the most important explanatory variable explaining the geographic pattern of COVID-19 deaths in nursing homes followed by the numbers of staff and residents who had received 2-dose COVID-19 vaccination (Appendix F). Fig 3 shows the spatial pattern of predicted nursing home COVID-19 deaths and the hotspot of predicted nursing home COVID-19 deaths in Northeastern US. 

We investigated facility-and community-level determinants of nursing home COVID-19 deaths for the first 2-year pandemic period using the Centers for Medicare and Medicaid Services and other county-level data. This assessment is important for public health intervention of COVID-19 related death among the adult population in NHs. We found that COVID-19 deaths were higher in the facilities in the lowest quartile of facility ratings and least in facilities with high facility ratings. Our results also indicate that the prevalence of COVID-19 deaths varies by facility ownership, notably increasing resident death in 'for-profit' facilities. Additionally, we found a high COVID-19 facility deaths prevalence in NHs in large fringe metropolitan areas compared to non-core or less-urbanized areas, and spatial analysis indicated a significant hotspot of predicted COVID-19 deaths in the Northeastern US.

The observed association between facility rating and COVID-19 deaths is consistent with several previous State-level studies that found significant associations between CMS facility rating and COVID-19 cases and fatalities in nursing homes 14, 27, 28 . For example, He et al. 28 deaths. associated with a lower probability of a larger outbreak and fewer deaths." 38(p2466) Our findings on the effect of urbanization status on COVID-19 facility death were consistent with Gorge and Konetzka's study, which found the largest magnitude effects for counties with a higher percentage of metropolitan status. 38 However, the study did not examine the nuanced variation in COVID-19 deaths by urbanization status. Complemented by the spatial analysis, our nonspatial analysis showed that the average count of COVID-19 deaths significantly varies by urbanization status along Bos-Wash megapolis, which extends from Boston through New York and Philadelphia down to Washington, D.C. It is important to note that although this present study found that COVID-19 mortality significantly reduced in facilities in micropolitan and noncore/rural areas, this association disappeared after adjusting for staff and residents COVID-19 cases and vaccination status and remained statistically insignificant in the fully-adjusted model. This effect seems attenuated by community-level total COVID-19 vaccination suggesting the role of vaccination in reducing deaths in society and NHs. Health disparity is a place-based issue; hence, improving the locational conditions of these facilities could make a difference in improving health outcomes. 29 Additionally, we found a statistically significant increase in COVID-19 death in 'for-profit' and government-owned nursing homes. However, the incidence rate ratio of COVID-19 deaths in government-owned nursing homes was 22.3 percent and 6.4 percent in 'for-profit' nursing homes.

This means that COVID-19 deaths prevalence was (predicted) three times higher in governmentowned nursing homes facilities than 'for-profit'-owned nursing homes facilities in the multivariate analysis. Similar to a study in Real Madrid, Spain, 46 this current study found increased COVID-19 death in government and 'for-profit' nursing homes may suggest differences in nursing home management and perhaps funding status. Damian et al.'s study found that the differences in nursing home facility mortality were related to facility subsidy, ownership status, and facility size. 46 Cohen and Spector studied the effect of different reimbursement types on NHs mortality and found lower death rates in public and not-for-profit compared to for-profit in a sample of 2663 residents from 658 nursing homes. 46 

This study has broad research implications in caring for older population. In addition, findings from this work can further inform research and policy actions at the facility and regional levels. First, investigating the facility level disparity of COVID-19 deaths in government and for-profitowned warrants in-depth study, especially in poorly rated nursing home facilities. Second, research investigating medical, social, and environmental factors contributing to the pattern of COVID-19 deaths in the Northeast US is required. Thirdly, research and policy considering the travel patterns of unvaccinated nurses (travel nurses) may help understand and curb the excess deaths in nursing homes in the US. Therefore, further investigation of facility conditions at the micro-meso-and macro-level may provide informative insights that can drive local and regional policies to address nursing home excess deaths from infectious disease, including COVID-19.

This investigation of COVID-19 related deaths in nursing homes in the US had a number of limitations. We did not include all individual facility quality ratings but instead used the overall facility rating due to the evidence of mediation of the individual rating scores. Furthermore, this study failed to control for residents' demographic characteristics and underlying health conditions, which may modify the main effect of facility ratings in our models. We used countylevel demographic variables in our models, which may not truly account for the demographic of residents at the county level. We also acknowledge possible misdiagnosis of non-SARS-CoV-2 mortality that may have contributed to incorrectly reporting to CMS, which may have caused over-or under-estimation in our analyses. Despite these limitations, this study has some strengths. This is the work to present a holistic report of facility overall quality measure on the COVID-19 deaths in nursing homes for the first 2-year pandemic period in the US.

The findings on nursing homes COVID-19 deaths presented in this study echo the importance of facility-based intervention toward eradicating SARS-CoV-2 infection. Nursing home COVID-19 deaths reduced in high-ranking facilities, and NHs in the Northeast US had a significant hotspot of nursing home COVID-19 deaths between January 2020 and December 2021. We suggest increasing COVID-19 vaccination in most nursing homes in most at-risk locations close to meat and livestock plants. Factors that contribute to increased death in 'for-profit' and governmentowned facilities warrant further research. Finally, findings from this 2-year pandemic period analysis can provide valuable guidance for managing long-term care facilities in the US as we advance in the pandemic and other regions of the world.

Nursing home data used for this study are freely available online from https://data.cms.gov/covid-19/covid-19-nursing-home-data. Please, see 'Method' section for sources of other data used in this study.

Funding

Infection Prevention and Control Guidance for Long-Term Care Facilities in the Context of COVID-19: Interim Guidance

Estimation of case-fatality rate in COVID-19 patients with hypertension and diabetes mellitus in the New York state: a preliminary report

Prevention and control of COVID-19 in nursing homes, orphanages, and prisons

Risk Factors Associated With Mortality Among Residents With Coronavirus Disease 2019 (COVID-19) in Longterm Care Facilities in Ontario, Canada

Recommendations for protecting against and mitigating the COVID-19 pandemic in long-term care facilities

Epidemiology of Covid-19 in a long-term care facility in King County, Washington

Update: COVID-19 among workers in meat and poultry processing facilities-United States

COVID-19 among workers in meat and poultry processing facilities-19 states

Livestock plants and COVID-19 transmission

Care AA of NEP on QH. Quality Health Outcomes Model

A Hierarchical Model of Health Service Quality: Scale Development and Investigation of an Integrated Model

Nursing home profit status and quality of care: is there any evidence of an association?

Quality of care in for-profit and not-forprofit nursing homes: systematic review and meta-analysis

Association Between CMS Quality Ratings and COVID-19 Outbreaks in Nursing Homes -West Virginia

World Health Organization. Rational Use of Personal Protective Equipment for Coronavirus Disease ( COVID-19) and Considerations during Severe Shortages: Interim Guidance

Emergency preparedness during the COVID-19 pandemic: Perceptions of oncology professionals and implications for nursing management from a qualitative study

COVID-19 preparedness in US home health care agencies

Associations Between Published Quality Ratings of Skilled Nursing Facilities and Outcomes of Medicare Beneficiaries with Heart Failure

Measuring Physical and Social Environments in Nursing Homes for People with Middle-to Late-Stage Dementia

The Nursing Home Compare Report Card: Perceptions of Residents and Caregivers Regarding Quality Ratings and Nursing Home Choice

Changes in Consumer Demand Following Public Reporting of Summary Quality Ratings: An Evaluation in Nursing Homes

A network-based explanation of why most COVID-19 infection curves are linear

Analysis of mobility data to build contact networks for COVID-19

Using real-world contact networks to quantify the effectiveness of digital contact tracing and isolation strategies for Covid-19 pandemic. medRxiv

Nursing home staff networks and COVID-19

Mapping community-level determinants of COVID-19 transmission in nursing homes: A multi-scale approach

COVID-19 Infections and Deaths among Connecticut Nursing Home Residents: Facility Correlates

Is There a Link between Nursing Home Reported Quality and COVID-19 Cases? Evidence from California Skilled Nursing Facilities

Addressing Health Disparities Is a Place-Based Issue

From place-based to people-based exposure measures

COVID-19 Nursing Home Data. Centers for Medicare & Medicaid Services Data

Penalties | Provider Data Catalog

American Community Survey Office: Data Profiles. The United States Census Bureau

&g=0100000US%240500000&d=ACS%205-Year%20Estimates%20Data%20Profiles

COVID-19 Vaccinations in the United States,County

National Center for Health Statistics of Urban-Rural Classification Scheme for Counties in the US

Factors Associated With Racial Differences in Deaths Among Nursing Home Residents With COVID-19 Infection in the US

County-Level Social Determinants of Health and COVID-19 in Nursing Homes

National Center for Health Statistics: Urban Rural Classification Scheme for Counties

Association Between Childbirth Modes and Neonatal and Maternal Deaths Using a Negative Binomial Model

SPSS Statistics for Windows. IBM Corp; 2021. Accessed

Analyzing count variables in individuals and groups: Single level and multilevel models

How Forest-based Classification and Regression works. ArcGIS Pro

Facility ownership and mortality among older adults residing in care homes

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

The authors declare no conflict of interest Acknowledgment The authors appreciate the constructive comments of the editor and the two reviewers in improving this paper.