COVID-19 among people experiencing homelessness in England: a modelling study

Abstract

Background: People experiencing homelessness are vulnerable to COVID-19 due to the risk of transmission in shared accommodation and the high prevalence of comorbidities. In England, as in some other countries, preventive policies have been implemented to protect this population. We aimed to estimate the avoided deaths and health-care use among people experiencing homelessness during the so-called first wave of COVID-19 in England-ie, the peak of infections occurring between February and May, 2020-and the potential impact of COVID-19 on this population in the future.

Methods: We used a discrete-time Markov chain model of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection that included compartments for susceptible, exposed, infectious, and removed individuals, to explore the impact of the pandemic on 46 565 individuals experiencing homelessness: 35 817 living in 1065 hostels for homeless people, 3616 sleeping in 143 night shelters, and 7132 sleeping outside. We ran the model under scenarios varying the incidence of infection in the general population and the availability of prevention measures: specialist hotel accommodation, infection control in homeless settings, and mixing with the general population. We divided our scenarios into first wave scenarios (covering Feb 1-May 31, 2020) and future scenarios (covering June 1, 2020-Jan 31, 2021). For each scenario, we ran the model 200 times and reported the median and 95% prediction interval (2·5% and 97·5% quantiles) of the total number of cases, the number of deaths, the number hospital admissions, and the number of intensive care unit (ICU) admissions.

Findings: Up to May 31, 2020, we calibrated the model to 4% of the homeless population acquiring SARS-CoV-2, and estimated that 24 deaths (95% prediction interval 16-34) occurred. In this first wave of SARS-CoV-2 infections in England, we estimated that the preventive measures imposed might have avoided 21 092 infections (19 777-22 147), 266 deaths (226-301), 1164 hospital admissions (1079-1254), and 338 ICU admissions (305-374) among the homeless population. If preventive measures are continued, we projected a small number of additional cases between June 1, 2020, and Jan 31, 2021, with 1754 infections (1543-1960), 31 deaths (21-45), 122 hospital admissions (100-148), and 35 ICU admissions (23-47) with a second wave in the general population. However, if preventive measures are lifted, outbreaks in homeless settings might lead to larger numbers of infections and deaths, even with low incidence in the general population. In a scenario with no second wave and relaxed measures in homeless settings in England, we projected 12 151 infections (10 718-13 349), 184 deaths (151-217), 733 hospital admissions (635-822), and 213 ICU admissions (178-251) between June 1, 2020, and Jan 31, 2021.

Interpretation: Outbreaks of SARS-CoV-2 in homeless settings can lead to a high attack rate among people experiencing homelessness, even if incidence remains low in the general population. Avoidance of deaths depends on prevention of transmission within settings such as hostels and night shelters.

Funding: National Institute for Health Research, Wellcome, and Medical Research Council.

Figure 1

State transitions in the model Individuals are classified into susceptible, exposed, infectious (asymptomatic or symptomatic), and removed states (ie, no longer susceptible). Light and dark blue boxes represent individuals in community settings: living in a hostel, sleeping in a night shelter, sleeping outside, or staying in a COVID-PROTECT hotel. Pink boxes represent individuals in health-care settings (COVID-CARE or hospital). Each of these locations is modelled as a number of separate closed subgroups, based on data about homeless accommodation in England. ICU=intensive care unit.

Figure 2

New infections of SARS-CoV-2 among people experiencing homelessness in England, scenarios C, D, E, F, G, and H Scenario A is an estimate of the historical impact of COVID-19 on people experiencing homelessness, and it leads into the future scenarios C–H (table 2). Scenario B is a counterfactual historical scenario that does not lead into the future scenarios, and is therefore not included in the figure. Results from 200 model runs are presented. The dark blue line shows the model run producing the median number of cumulative new cases. SARS-CoV-2=severe acute respiratory syndrome coronavirus 2.

Figure 3

Illustrative timeline of outbreaks in hostels 50 hostels were selected at random for this figure, from a single model run. Each line represents a single hostel, with the total of all 50 hostels shown in red.