Skip to main content

COVID-19 outbreak at a large homeless shelter in Boston: Implications for universal testing

Our take —

This study, published as a preprint and thus not yet peer reviewed, reports on the infection of 36.0% of clients in a homeless shelter in Boston, the large majority of whom were asymptomatic. As a cross-sectional analysis, it cannot show whether people were infected within the shelter, and the authors’ assessment of risk factors for infection are likely biased. This study should primarily be used only to estimate prevalence of infection and symptoms in a homeless shelter setting.

Study design


Study population and setting

This study was conducted at a Boston homeless shelter among clients in the last week of March 2020, among 408 participants. Following a cluster outbreak linked to the shelter, the study used universal testing of all clients at the facility, regardless of symptoms, over a two-day period. The population was primarily men, with the plurality being self-reported white, and the majority being over 50 years of age.

Summary of Main Findings

The study found, of the 408 participants, 36.0% (147) tested positive for SARS-CoV-2 infection, and only a minority of these met COVID-19 clinical case definition from symptoms. 1.0% (n=4) had fever, 8.1% (n=33) had cough, and 0.7% (n=3) had shortness of breath. There was a statistically significant association between being male and having positive SARS-CoV-2 infection, but no association between infection and other demographic characteristics.

Study Strengths

This study employed a universal testing strategy among a population that is historically difficult to reach for research. It had sufficient sample size to conduct this analysis.


The study describes infection in a transient population, though due to the cross-sectional design, can make no claim as to what degree infection transmission occurred within the facility. It does not specify, of those who had symptoms, the amount of overlap between symptom groups, and does not report the number of individuals who had any of the major symptoms. Of its few inferential statistics, none of them are adjusted for major confounders, which likely biases these associations. Overall, the information they collected on participants was also limited to major demographics, and no other data were presented on factors that could likely increase the risk of infection, such as whether an individual was employed and going to work.

Value added

This study was one of the first within a homeless shelter population, and provides important knowledge about the magnitude of potential infections, including asymptomatic infections in this setting.