JMIR Publications

ABSTRACT

Background:

Temperature-taking behaviors vary with levels of circulating infectious illness; however, little is known about how these behaviors differ by demographic characteristics.

Objective:

We investigated differences in temperature-taking frequency and the proportion of readings that were feverish among demographic groups (age, gender, urban/rural status) over influenza offseason, influenza season, and Waves 1, 2, and 3 of the COVID-19 pandemic.

Methods:

Using data from smart thermometers collected from May 1, 2019, to February 28, 2021, across the United States, we calculated the frequency of temperature-taking and the proportion of temperature readings that were feverish. Mixed effects negative binomial and mixed effects logistic regression analyses were performed to identify demographic characteristics associated with temperature-taking frequency and the proportion of feverish readings, respectively. Separate models were fit over five study periods: influenza offseason (n=122,480), influenza season (n=174,191), Wave 1 of COVID-19 (n=350,385), Wave 2 (n=366,489), and Wave 3 (n=391,578).

Results:

Both temperature-taking frequency and the proportion of feverish readings differed by study period (ANOVA: P<.001) and were highest during influenza season. During all periods, children aged 2-5 years and 6-11 years had significantly higher frequencies of temperature-taking than users aged 19-30 years, and children had the highest proportion of feverish readings of all age groups, after adjusting for covariates. During Wave 1 of COVID-19, users over age 60 years had 1.79 times (95% CI: 1.76, 1.83) the rate of temperature-taking as users aged 19-30 years, and 74% lower odds (95% CI: 72-75%) of a reading being feverish. Across all periods, males had significantly lower temperature-taking frequency and significantly higher odds of having a feverish reading compared to females. Users living in urban areas had significantly higher frequencies of temperature-taking than rural users during all periods, except Wave 2 of COVID-19, and urban users had higher odds of a reading being feverish in all study periods except Wave 1 of COVID-19.

Conclusions:

Temperature-taking behavior and the proportion of readings that were feverish are associated with both population disease levels and individual demographic characteristics. Differences in the health behavior of temperature-taking may reflect changes in both perceived and actual illness risk. Specifically, older adults may have experienced an increase in perceived risk during the first three waves of COVID-19, leading to increased rates of temperature monitoring, even when their odds of fever were lower than younger adults.