Real-time outlier detection in digital PCR data for wastewater-based pathogen surveillance

Wastewater-based epidemiology provides insights into the spread of infectious diseases by sampling and analyzing wastewater from wastewater treatment plant catchments. Longitudinal measurements of pathogen concentrations in wastewater have been shown to align with disease incidence and can be used to estimate indicators of epidemic growth such as effective reproduction numbers (R_t). This informs public health decision-making by providing feedback on the transmission of disease and effectiveness of interventions. However, methods for analysing growth trends from wastewater are sensitive to multiple sources of variation during shedding, in-sewer transport, sample collection, and processing. A particular challenge is posed by occasional but high outlier measurements, which can bias estimated trends and epidemiological parameters during seasonal and epidemic waves. The identification and handling of such data points in real-time wastewater monitoring is important to provide a reliable basis for public health decision-making. In this study we propose a fast and simple outlier detection method for digital PCR measurements from wastewater that allows identification of outliers in real time. We evaluated our method by assessing its potential to improve real-time trend estimates. We show that, in the majority of situations, the removal of an outlier identified by our method makes the real-time trend estimate for a particular time point more representative of the trend which is retrospectively obtained for the same time point. Our method provides a simple way for laboratories to flag potential outliers in time series data, improving data quality for further analysis.