key: cord-1039909-uoecn9if
authors: Alacevich, Caterina; Cavalli, Nicolò; Giuntella, Osea; Lagravinese, Raffaele; Moscone, Francesco; Nicodemo, Catia
title: The presence of care homes and excess deaths during the COVID‐19 pandemic: Evidence from Italy
date: 2021-04-21
journal: Health Econ
DOI: 10.1002/hec.4277
sha: 869f063aab7672de4604bd1ca82842b9f02385bd
doc_id: 1039909
cord_uid: uoecn9if

This paper explores the relationship between the spatial distribution of excess deaths and the presence of care home facilities during the first wave of the COVID‐19 outbreak in Italy. Using registry‐based mortality data for Lombardy, one of the areas most severely hit by the pandemic we show that the presence of a care home in a municipality is associated with significantly higher excess death rates in the population. This effect appears to be driven by excess mortality in the elderly population of 70 years old and older. Our results are robust to controlling for the number of residents in each care home, suggesting that the presence of such facilities may have acted as one of factors contributing to the diffusion of COVID‐19 at the local level.

sub-national level (Arpino et al., 2020) . Questions thus remain over which other proximate mechanisms may articulate the role of the age structure in shaping COVID-19 outcomes at the population level (Nepomuceno, et al., 2020) . In this study, we estimate the concentration of excess mortality in municipalities with care homes. We use registry-based daily mortality records by municipality (comuni) from Lombardy (Italy), among the areas hardest-hit by the pandemic worldwide. Our identification strategy relies on estimating differences in excess mortality between municipalities with and without care homes, with respect to the previous 5 years' average. Italy was the first western country to experience a severe COVID-19 outbreak. Excess deaths in Lombardy's care homes are at the center of media and policy attention as well as judicial inquiries for a lack of clear rules on containment practices such as the reception of patients with COVID-19-related symptoms, testing, availability of protective gear, and contact with outside visitors. 1 Besides the share of elderly persons, attention on their living conditions is rapidly rising amidst dramatic reports of death rates in longterm residential and nursing facilities, henceforth care homes, described as`besieged castles' during the pandemic (Trabucchi & De Leo, 2020; Vandoros, 2020) . In several of these facilities, lack of adequate rules on containment practices have in fact determined a rapid and lethal spread of SARS-CoV-2. Furthermore, the fact that care homes do not have an infrastructure of acute medicine may have exacerbated the challenge of the pandemic. Although there is consensus in the epidemiological literature on the role played by the environment, including in long-term care, in determining the spread of infectious diseases, up to date rigorous, non-anecdotal evidence on how care homes might have added on the burden of COVID-19 is currently lacking. Our analyses show that the excess mortality of adults aged 70+ was significantly higher in municipalities with at least one care home, reaching up to 18 additional daily excess deaths per 100,000 inhabitants by the end of March. On the contrary, we do not find a significantly robust linear relationship between the number of beds in care homes and excess mortality. Our findings show that the result does not simply map the number of vulnerable people resident in care homes, who are prone to higher mortality risk due to preconditions, but are driven by the presence of care homes. To consider the locally correlated nature of the pandemic, we show a spatial autoregressive model (SAR), confirming our results.

We use death registry data for Lombardy (ISTAT, 2020), which provide daily deaths by age between January 1st and March 31st, for the years 2015 to 2020, by municipality of residence. The sample includes 131,040 observations from 1440 municipalities (95% of Lombardy). We measure municipality-level excess mortality rates as the difference between (daily or total) deaths per 100,000 individuals by age group between January 1st and March 31st 2020 and the corresponding 2015-2019 average. Differencing total deaths with respect to previous years allows us to account for the differential baseline mortality in municipalities with care homes 2 (Vandoros, 2020) . Another advantage over official COVID-19 deaths is that our measure does not depend on how deaths are classified. We link this dataset with the location of residential care homes (RSAs) by municipality. 3,4,5 Figure 1 illustrates the spatial distribution of care homes and excess mortality rates. Care homes are present in 34% of the municipalities and distributed widely across the region. Excess death rates present some geographical clustering.

During the first wave of the pandemic in the Spring of 2020, 85.6% of COVID-19 deaths occurred among the over 70 (ISS, 2020). To account for differential mortality risks between age groups, we consider separately the 70+, 50-69, and 0-49. To investigate the relationship between excess mortality and care homes, we use municipalities as units of analysis. In Figure 2 , we plot excess daily death rates by age group in 2020 for municipalities with and without a care home. To account for the fact that 2020 is a leap year and there is no corresponding data for February 29 in the previous years, we have excluded the 29th of February from the sample. Including February 29 does not significantly affect our estimates. Figure 4A in the Appendix shows no evidence of significant trends before 2020.

Before the first announced COVID-19 case (February 21st), there was no significant difference. Since the second week of March, municipalities with a care home exhibited significantly higher excess daily deaths for individuals aged 70 + (panel 3). By the end of March, they reached up to 18 additional excess deaths per 100,000 inhabitants.

To consider the differential timing of the outbreak, in the fourth panel we normalize time by the first day in which a municipality recorded a positive excess death rate after the epidemic onset. This is done separately for each age group. Next, we regress daily excess death rates on the presence of care homes by age group (age) as follows: F I G U R E 2 Excess daily death rates in 2020 in municipalities with and without care homes. Own elaborations based on ISTAT, Lombardy, and Ministry of Health data (2020). Panel 4 normalizes the number of days before and after the first excess death recorded in a municipality within a positive 7-days moving average (day 0), after the regional pandemic onset [Colour figure can be viewed at wileyonlinelibrary.com]?

τ d are days after January 1st-fixed effects, CareHome (0-1) indicates the presence of at least one care home in municipality m, RedArea (0-1) marks the first complete lockdown for ten municipalities 6 (February 23rd). χ m is a vector of municipality characteristics: presence of a hospital, population (%) above age 70, women (%) in the age group, and province fixed effects. ΔExcess Death Rate age d;m is the difference between the daily death rate in m by age group in 2020 and its 2015-2019 average. Table A1 (Appendix) describes the variables. Table 1 reports the Results. For the 70+ (column 1), the coefficient of care homes is positive and significant, with 4.91 excess daily deaths per 100,000 individuals (42% of the mean, 11.6). For the other age groups (50-69 and 0-49), coefficients are positive but not significant. In the Appendix, we report the results of an analysis of the relationship between excess deaths in the group 70+, per capita care home beds, and the presence of large care homes. The results suggest that the extensive margin (the presence of a care home) is more relevant than the intensive one (the number of beds in care homes), and there may be a non-linear relationship between care homes size and excess deaths (see also  Table A2 ). These results are not affected by intra-week variation in reporting as shown in Table A3 which re-estimates the main results using a 7-day moving average. 7,8 F I G U R E 2 (Continued) ALACEVICH ET AL.

We confirm the results for people 70+ using total excess death rates as a dependent variable (column 4). Next, we perform the Moran's I test for spatial correlation in residuals. The result of Moran is positive and significant, indicating spatial dependence: contiguous municipalities tend to show similar total excess mortality rates. 9 To account for this, as robustness, we estimate a SAR model for total excess mortality rates using contiguity matrix assigned weights 1 when municipality i and j are contiguous according to the queen contiguity criterion and 0 otherwise (Baltagi et al., 2018) . Column 5 reports the results. The spatial autocorrelation adjustment (spatial parameter) exhibits a statistically significant coefficient of 0.10. When comparing the coefficient of Care Homes with the ordinary least squares model (column 4 vs. column 5), we find that the size declines only slightly (4.36 vs. 4.19) . For robustness analysis, we report in Appendix also a difference-in-differences estimation that accounts for fixed municipality characteristics that may correlate with excess deaths. This specification confirms our results.

Our paper provides evidence of significantly higher excess death rates in municipalities where a care home is present, compared to municipalities without any care home. These findings suggest that care homes drive higher death rates, and this relationship goes beyond the fact that their presence implies a larger population of at-risk individuals. However, as official data on COVID-19 deaths in care homes for Italy are not available, our analysis cannot precisely differentiate the proportion of fatalities that happened inside a care home from our death counts. Because of a lack of information on pre-existing co-morbidities at the municipality level, our identification strategy also assumes that there were no significant changes in the prevalence of underlying co-morbidities in 2020 compared to the 2015-2019 average. Finally, we do not have information on the safety practices adopted by each home care in response to the pandemic. In this respect, policy choices may also be relevant, with reports emerging that the Lombardy region promoted practices of discharge of COVID-19 patients into care homes -a factor that is likely to have further the spread of the disease. Existing data show in fact that care homes in Lombardy registered a mortality rate of 6.5%, against a national average of 3.1% (ISS, 2020) . Although this weighs on the generalizability of our results, it should be noted that similar practices of discharge into care homes have been reported in other countries, such as the United Kingdom, see NHS report 2020. As Abbreviation: OLS, ordinary least squares.

***p < 0.01, **p < 0.05, *p < 0.

1708such, our study documents the association between the presence of care homes and excess deaths during the first-wave of COVID-19 in Northern Italy. With a population estimated by the World Health Organization to have surpassed 2 million elderly worldwide in 2014, our results hold central policy significance for the establishment of future preparedness and containment policies. Our results call for a better understanding of the role of home care facilities and home care quality in the pandemic. We document a significant relationship between the presence of nursing homes and excess deaths, more work is needed to shed light on mechanisms and factors behind this result. Having access to more granular data released by governments and nursing home facilities may enable future researches to better assess the role of different factors (i.e., quality of care, externalities, protocols etc.) and whether the positive association between excess mortality and care home facilities we have detected is the result of pre-COVID existing differences in quality across care homes or the system's integration. This is crucial to shape interventions targeted at improving not only hospital quality but also care home quality and their interconnection (see also Cronin & Evans, 2020) .

No clear association emerges between intergenerational relationships and COVID-19 fatality rates from macro-level analyses

Health econometrics, contributions to economic analysis

COVID-19 in Italy: An analysis of death registry data

Nursing home quality, COVID-19 deaths, and excess mortality (No. w28012)

Demographic science aids in understanding the spread and fatality rates of COVID-19

National age and coresidence patterns shape COVID-19 vulnerability

Survey nazionale sul contagio COVID-19 nelle strutture residenziali e sociosanitarie, Istituto Superiore di Sanità. Epidemia COVID-19, Aggiornamento nazionale

Investigating the deaths of healthcare and social care workers who died after contracting coronavirus

Besides population age structure, health and other demographic factors can contribute to understanding the COVID-19 burden

Nursing homes or besieged castles: COVID-19 in northern Italy

Excess mortality during the Covid-19 pandemic: Early evidence from England and Wales

The authors declare no conflict of interest.

The data that support the findings of this study will be made openly available if it is accepted for publication.

https://orcid.org/0000-0002-8690-980X Raffaele Lagravinese https://orcid.org/0000-0002-2394-7470 ENDNOTES 1 https://www.ilpost.it/2020/04/17/regione-lombardia-rsa/ 2 "Care homes" refer to all non-acute residential facilities that host patients for long-term care needs. 3 https://www.regione.lombardia.it/wps/portal/istituzionale/HP/DettaglioServizio/servizi-e-informazioni/cittadini/salute-e-prevenzione/ strutture-sanitarie-e-sociosanitarie/residenze-anziani 4 Unfortunately, data on the specific cause of death are not available at the municipality level. We also do not have information on the type of services provided by each nursing home. 5 Death records are based on the municipality of residence. If patients died in another municipality but did not change residence, this occurrence would still be counted as a death in the place of residence 6 Bertonico, Casalpusterlengo, Castelgerundo, Castiglione d'Adda, Codogno, Fombio, Maleo, Somaglia, San Fiorano, and Terranova dei Passerini. 7 While our data do not include bed occupancy rates, there is evidence that care homes were at full capacity at the onset of the pandemic, with long waiting lists. For example, Ciminelli and Garcia-Mantico' (2020) calculated that, in Lombardy, for each care home bed available there are on average 157 people in the waiting list. 8 There are only 81 municipalities with more than 1 care home (approximately 5% of the 1440 municipalities). Furthermore, only in 7% of the municipalities there was a change in the number of nursing homes between 2015 and 2020. Excluding these municipalities does not affect the results. 9 The Moran I test is based on the contiguity matrix. We also used inverse distance matrix and used a Spatial Error Model. Results are confirmed. For sake of brevity we showed only SAR. 10 See Equation 1.

Additional supporting information may be found online in the Supporting Information section at the end of this article.