BACKGROUND
Current public health surveillance infrastructure primarily relies on laboratory reporting of COVID-19 to assess the burden of disease. A combination of limited diagnostic testing capacity, reporting delays, and the increased use of non-reportable at-home tests have hindered the ability to track the spread of COVID-19. Novel epidemiological approaches have been implemented to supplement gaps in the current public health surveillance system.
OBJECTIVE
To investigate if there is a lead time and determine the correlation between population-level fever incidence recorded by smartphone-connected thermometers and reported COVID-19 case incidence in five US states
METHODS
To compare the incidence curves of fever-based surveillance and reported COVID-19 cases from November 2, 2021 - March 30, 2022 in five states, we compared their respective maxima (“peak-to-peak” analysis) and conducted a cross-correlation analysis.
RESULTS
A total of 276,981 temperatures were recorded in the study population, of which 48,091 were febrile and the plurality of which were from California. Using both the peak-to-peak analysis and cross correlation analysis, fever incidence preceded COVID-19 cases in all five states. For the peak-to-peak method, lead times ranged from 2 (Texas) to 15 days (Ohio). In the cross-correlation analysis, lead time ranged from 6 to 8 days, with Ohio having the shortest lead time and New York, Pennsylvania, and Texas having the longest. Fever incidence and reported COVID-19 cases were highly correlated in all five states (r ≥ 0.9).
CONCLUSIONS
Thermometer-based fever incidence consistently produced a lead time with a high correlation of COVID-19 cases in the five states that were studied. In conjunction with early warning and laboratory diagnostic data, these results can inform health policies and resource allocation of vaccines and testing. Future studies can analyze lead time in other states and scales of geography, compare fever surveillance to other surveillance sources, and stratify fever data by age, gender, and race and/or ethnicity, if available, to identify any sentinel groups.