Background
In Australia, the Victoria State Government uses wastewater surveillance to detect new incursions of COVID-19. Wastewater is analysed for SARS-CoV-2 shed by infected individuals. The Government is evaluating the program to ensure efficiency and appropriate resource allocation. This study estimated the accuracy of wastewater surveillance for detecting SARS-CoV-2 infected people.

Methods
We conducted a diagnostic test accuracy study from 25 August 2020 to 27 September 2021 in Victoria, Australia, using data from state surveillance databases. Accuracy was determined by cross-classifying wastewater results (index test) with the presence of known SARS-CoV-2 infected people resident in catchments (reference test). Infected people were included at -2 to +7 days around symptom onset or testing date for symptomatic and asymptomatic individuals, respectively. Analyses were restricted to long sampling methods and to catchments with more than 30 residents. Sensitivity and specificity were calculated as the proportion of positive and negative reference test results with matching index test results, respectively. Calculations were then stratified by lockdown (stay-at-home orders) and non-lockdown days.

Results
We included 25,142 wastewater results. In the crude analysis, sensitivity was 42% and specificity was 92%. During lockdown days (n=8,477), sensitivity was 49% and specificity was 83%. During non-lockdown days (n=16,665), sensitivity was 30% and specificity was 96%.

Conclusion
Given its high specificity (high true-negative and low false-positive proportions), catchments with positive wastewater detections likely contain infected residents, and should be, and have been, targeted for public health action. Its moderate sensitivity (moderate true-positive proportion) is biologically plausible, as faecal shedding is variable between individuals, and the virus may degrade in sewage. The estimated sensitivity is likely closer to the true value during lockdowns; people are shedding in their places of residence resulting in less misclassification between the index test and the reference test. These results will inform the Government’s current program evaluation.