Elevated Ambient Temperature Associated With Reduced Infectious Disease Test Positivity Rates: Retrospective Observational Analysis of Statewide COVID-19 Testing and Weather Across California Counties

Background: From medication usage to the time of day, a number of external factors are known to alter human body temperature (BT), even in the absence of underlying pathology. In select cases, clinical guidance already suggests the consideration of clinical and demographic factors when interpreting BT, such as a decreased threshold for fever as age increases. Recent work has indicated factors impacting BT extend to environmental conditions including ambient temperature. However, the effect sizes of these relationships are often small, and it remains unclear if such relationships result in a meaningful impact on real-world health care practices.

Objective: Temperature remains a common element in public health screening efforts. Leveraging the unique testing and reporting infrastructure developed around the COVID-19 pandemic, this paper uses a unique resource of daily-level statewide testing data to assess the relationship between ambient temperatures and positivity rates. As fever was a primary symptom that triggered diagnostic testing for COVID-19, this work hypothesizes that environmentally mediated BT increases would not reflect pathology, leading to decreased COVID-19 test positivity rates as temperature rises.

Methods: Statewide COVID-19 polymerase chain reaction testing data curated by the California Department of Public Health were used to obtain the daily number of total tests and positivity rates for all counties across the state. These data were combined with ambient temperature data provided by the National Centers for Environmental Information for a period of 133 days between widespread testing availability and vaccine approval. A mixed-effects beta-regression model was used to estimate daily COVID-19 test positivity rate as a function of ambient temperature, population, and estimates of COVID prevalence, with nested random effects for a day of the week within unique counties across the state.

Results: Considering over 19 million tests performed over 4 months and across 45 distinct counties, adjusted model results highlighted a significant negative association between daily ambient temperature and testing positivity rate (P<.001). Results of the model are strengthened as, using the same testing data, this relationship was not present in a sensitivity analysis using random daily temperatures drawn from the range of observed values (P=.52).

Conclusions: These results support the underlying hypothesis and demonstrate the relationship between environmental factors and BT can impact an essential public health activity. As health care continues to operate using thresholds of BT as anchor points (ie, ≥100.4 as fever) it is increasingly important to develop approaches to integrate the array of factors known to influence BT measurement. Moreover, as weather data are not often readily available in the same systems as patient data, these findings present a compelling case for future research into when and how environmental context can best be used to improve the interpretation of patient data.