Mobile outreach testing for COVID-19 in twenty homeless shelters in Toronto, Canada

BackgroundIt is unclear what the best strategy is for detecting COVID-19 among homeless shelter residents and what individual factors are associated with positivity. MethodsWe conducted a retrospective chart audit obtaining repeated cross-sectional data from outreach testing done at homeless shelters between April 1st and July 31st, 2020 in Toronto, Canada. We compared the positivity rate for shelters tested because of an outbreak (at least one known case) versus surveillance (no known cases). A patient-level analysis examined differences in demographic, health, and behavioural characteristics of residents who did and did not test positive for COVID-19. FindingsOne thousand nasopharyngeal swabs were done on 872 unique residents at 20 shelter locations. Among the 504 tests done in outbreak settings, 69 (14%) were positive and 1 (0.2%) was indeterminate. Among the 496 tests done for surveillance, 11 (2%) were positive and none were indeterminate. Shelter residents who tested positive were significantly less likely to have a health insurance card (54% vs 72%, p=0.03) or have visited another shelter in the last 14 days (0% vs 18%, p<0.01) compared to those who tested negative; There was no association between COVID-19 positivity and medical history (p=0.40) or symptoms (p=0.43). InterpretationOur findings support testing of asymptomatic shelter residents for COVID-19 when a positive case is identified at the same shelter but suggest limited utility of testing all shelter residents in the absence of a known case. Visiting another shelter in the last 14 days is associated with a decreased risk of COVID-19 positivity.