Epidemiological evaluation of sewage surveillance as a tool to detect the presence of COVID-19 cases in a low case load setting

ABSTRACT

In low prevalence settings the development of sensitive and specific quantitative Reverse Transcription Polymerase Chain Reaction tests to detect SARS-CoV − 2 (the virus causing COVID-19) in sewage presents the possibility of using sewage sampling as a diagnostic test for the presence of infected people in the catchment of the sampled sewer. However, the usefulness of such surveillance has not been quantified. In this study in the Australian state of Victoria between August and October 2020 the location of each known SARS-CoV − 2-infected person was determined on each day from two days before onset to 55 days after, in 46 metropolitan and rural sewer catchments sampled weekly – a total of 71 positive and 275 negative samples, and 354,155 person-days of location data. These were categorised by time since onset and distance from the sampling site. The odds of detection in sewage were between 5 and 20 times higher where known cases were present, with less effect of distance than time since onset. Using positive qRT-PCR in a sewage sample as a diagnostic test not just for viral RNA in the sample, but for the presence of known infected people in the catchment on the same day, the sensitivity was moderate (31% to 76%) and the specificity high (87% to 94%). The odds of detection were increased with increased numbers of known infected people but decreased with increased distance and time since onset. The probability of detection of the viral subgenome in sewage samples was about 10% when one known infected person was present, and this increased with higher numbers of known infected people and greater proximity to the sampling site. Sewage surveillance can be used to detect people infected with SARS-CoV − 2 in the catchment, directing a search for infectious clinical cases and other public health actions. However, detection at least eight weeks after onset may be due to existing cases rather than new ones, and, although not zero, the probability of detecting a single case is low.