Characterization of Influenza-Like Illness Burden Using Commercial Wearable Sensor Data and Patient-Reported Outcomes: Mixed Methods Cohort Study

Abstract

Background: The burden of influenza-like illness (ILI) is typically estimated via hospitalizations and deaths. However, ILI-associated morbidity that does not require hospitalization remains poorly characterized.

Objective: The main objective of this study was to characterize ILI burden using commercial wearable sensor data and investigate the extent to which these data correlate with self-reported illness severity and duration. Furthermore, we aimed to determine whether ILI-associated changes in wearable sensor data differed between care-seeking and non-care-seeking populations as well as between those with confirmed influenza infection and those with ILI symptoms only.

Methods: This study comprised participants enrolled in either the FluStudy2020 or the Home Testing of Respiratory Illness (HTRI) study; both studies were similar in design and conducted between December 2019 and October 2020 in the United States. The participants self-reported ILI-related symptoms and health care-seeking behaviors via daily, biweekly, and monthly surveys. Wearable sensor data were recorded for 120 and 150 days for FluStudy2020 and HTRI, respectively. The following features were assessed: total daily steps, active time (time spent with >50 steps per minute), sleep duration, sleep efficiency, and resting heart rate. ILI-related changes in wearable sensor data were compared between the participants who sought health care and those who did not and between the participants who tested positive for influenza and those with symptoms only. Correlative analyses were performed between wearable sensor data and patient-reported outcomes.

Results: After combining the FluStudy2020 and HTRI data sets, the final ILI population comprised 2435 participants. Compared with healthy days (baseline), the participants with ILI exhibited significantly reduced total daily steps, active time, and sleep efficiency as well as increased sleep duration and resting heart rate. Deviations from baseline typically began before symptom onset and were greater in the participants who sought health care than in those who did not and greater in the participants who tested positive for influenza than in those with symptoms only. During an ILI event, changes in wearable sensor data consistently varied with those in patient-reported outcomes.

Conclusions: Our results underscore the potential of wearable sensors to discriminate not only between individuals with and without influenza infections but also between care-seeking and non-care-seeking populations, which may have future application in health care resource planning.

Trial registration: Clinicaltrials.gov NCT04245800; https://ichgcp.net/clinical-trials-registry/NCT04245800.

Keywords: influenza; influenza-like illness; person-generated health care data; wearable sensor.

Conflict of interest statement

Conflicts of Interest: VH and DC are former employees of Genentech, Inc, a member of the Roche Group. ST-B is a current employee of Genentech, Inc. AS, ER, and EP are employees of Evidation Health. FD is a former employee of F. Hoffmann-La Roche Ltd. LF is a cofounder of Evidation Health, a company that makes use of person-generated health data in clinical research, health care, and public health applications. VU is an employee of Roche Products Ltd.

©Victoria Hunter, Allison Shapiro, Devika Chawla, Faye Drawnel, Ernesto Ramirez, Elizabeth Phillips, Sara Tadesse-Bell, Luca Foschini, Vincent Ukachukwu. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 23.03.2023.

Figures

Figure 1

Study design overview and flow diagram for preparing analysis data sets. (A) Study design overview for the FluStudy2020 and Home Testing of Respiratory Illness (HTRI) studies. During the enrollment period, the participants completed a screener survey and a baseline survey and connected their commercial wearable sensor (Fitbit) to the Evidation platform. The participants in the HTRI study were issued with an influenza diagnostic kit during enrollment to be used as instructed based on symptoms reported in the daily survey. Once enrolled, participants completed daily surveys for 120 days, in addition to monthly and biweekly surveys. Wearable sensor data were collected for 150 and 120 days in the FluStudy2020 and HTRI studies, respectively. The participants completed the Daily follow-up A survey on days when they reported feeling symptomatic (yellow and red boxes). They completed the Daily follow-up B, healthy survey on days when they reported feeling healthy (white boxes), and Daily follow-up B, recovering survey on days when they reported feeling healthy but were recovering from a recent influenza-like illness (ILI; green boxes). The labels highlight selected examples of events that the participants could self-report on surveys, as well as study-related activities (such as an influenza diagnostic kit being triggered by the participant’s self-reported symptoms). (B) Participant flow diagram for the FluStudy2020 and HTRI studies.

Figure 2

Deviations from baseline levels across wearable measures during an influenza-like illness (ILI) event. Day-by-day group mean (SE) changes from baseline values for 5 wearable sensor features derived from day-level sensor data are shown. “Steps total” was the total number of steps taken per day. “Total steps count” was the total number of steps taken per day. “Proportion of steps taken with >50 spm” was the ratio of minutes in the day that the participant took over 50 steps per minute (spm) relative to the total number of minutes with at least 1 step. “Sleep duration” was the total number of minutes in the day that the participant spent asleep. “Sleep efficiency” was the ratio of minutes the participant was asleep relative to the total minutes spent in bed until the end of the main sleep event (obtained from the Fitbit application programming interface). “RHR” was heart rate (beats per minute [bpm]) during periods of inactivity (obtained from the Fitbit application programming interface). Gray lines illustrate values for the entire wearable analysis population available for that wearable sensor feature; colored lines represent values for the subset of the analysis population who tested positive for influenza. The shaded region covers days −4 to +9 relative to ILI onset, which was the time window used in statistical analyses of day-by-day ILI burden. Influenza+: influenza-positive; RHR: resting heart rate.

Figure 3

Correlations between patient-reported outcomes (y-axis) and wearable sensor data (x-axis). The color shading and the value printed in each cell indicate Spearman ρ correlation coefficient for the association between the variables in the corresponding row and column. P values of statistical associations were false discovery rate corrected for 3105 total tests (not all depicted here), following the Benjamini-Hochberg procedure. Associations that reached statistical significance are illustrated with more intense color and are annotated with an asterisk (*P<.05; **P<.01; ***P<.001). QoL: quality of life; RHR: resting heart rate; RHR δ: the change in RHR from the previous day; spm: steps per minute; TTRB: time to return to baseline.

Figure 4

Daily influenza-like illness (ILI) burden and time to return to baseline (TTRB) in health care–seeking versus non–health care–seeking populations. Colored data (plot points, lines, and boxes) represent health care–seeking participants; gray lines represent non–health care–seeking participants. (A) Results of regression analyses of day-by-day ILI burdens of health care–seeking and non–health care–seeking cohorts are shown; separate regression analyses were conducted for each wearable sensor feature. Plotted points are model-fitted estimates (95% CI error bars) for each cohort, on each day of the ILI window and for the baseline period (computed as the mean across all healthy days; represented by b on the x-axis). Diamonds above the x-axis indicate contrasts reaching statistical significance after Bonferroni correction for 14 tests; contrasts were the difference in ILI burden between day n and baseline for the health care–seeking cohort compared with the non–health care–seeking cohort. (B) TTRB and cumulative ILI burden for each sensor feature are shown for individual participants (cloud plots) and overall (density distributions) for health care–seeking and non–health care–seeking cohorts. For TTRB (first 2 columns), box plots (center line, median; box limits, upper and lower quartiles; points, outliers) are overlaid and annotated with the median value for the corresponding population. For cumulative ILI burden (third column), mean and 95% CI (point with error bars) are overlaid and annotated with the mean value. bpm: beats per minute; RHR: resting heart rate; RHR δ: the change in RHR from the previous day; spm: steps per minute.

Figure 5

Daily influenza-like illness (ILI) burden and time to return to baseline (TTRB) in the participants with confirmed influenza infection versus those with ILI symptoms only. Colored data (plot points, lines, and boxes) represent the participants who were influenza positive (influenza+); gray lines indicate those with ILI symptoms only. (A) Results of regression analyses of day-by-day ILI burdens of influenza+ and ILI symptoms–only cohorts are shown; separate regression analyses were conducted for each wearable sensor feature. Plotted points are model-fitted estimates (95% CI error bars) for each cohort, on each day of the ILI window and for the baseline period (computed as the mean across all healthy days; represented by b on the x-axis). Diamonds above the x-axis indicate contrasts reaching statistical significance after Bonferroni correction for 14 tests; contrasts were the difference in ILI burden between day n and baseline for the influenza+ cohort compared with the ILI symptoms–only cohort. (B) TTRB and cumulative ILI burden for each sensor feature are shown for individual participants (cloud plots) and overall (density distributions) for influenza+ and ILI symptoms–only cohorts. For TTRB (first 2 columns), box plots (center line, median; box limits, upper and lower quartiles; points, outliers) are overlaid and annotated with the median value for the corresponding population. For cumulative ILI burden (third column), the mean and 95% CI (point with error bars) are overlaid and annotated with the mean value. bpm: beats per minute; influenza+: influenza-positive; RHR: resting heart rate; RHR δ: the change in RHR from the previous day; spm: steps per minute.