Physiological presentation and risk factors of long COVID in the UK using smartphones and wearable devices: a longitudinal, citizen science, case-control study

Abstract

Background: The emergence of long COVID as a COVID-19 sequela was largely syndromic in characterisation. Digital health technologies such as wearable devices open the possibility to study this condition with passive, objective data in addition to self-reported symptoms. We aimed to quantify the prevalence and severity of symptoms across collected mobile health metrics over 12 weeks following COVID-19 diagnosis and to identify risk factors for the development of post-COVID-19 condition (also known as long COVID).

Methods: The Covid Collab study was a longitudinal, self-enrolled, community, case-control study. We recruited participants from the UK through a smartphone app, media publications, and promotion within the Fitbit app between Aug 28, 2020, and May 31, 2021. Adults (aged ≥18 years) who reported a COVID-19 diagnosis with a positive antigen or PCR test before Feb 1, 2022, were eligible for inclusion. We compared a cohort of 1200 patients who tested positive for COVID-19 with a cohort of 3600 sex-matched and age-matched controls without a COVID-19 diagnosis. Participants could provide information on COVID-19 symptoms and mental health through self-reported questionnaires (active data) and commercial wearable fitness devices (passive data). Data were compared between cohorts at three periods following diagnosis: acute COVID-19 (0-4 weeks), ongoing COVID-19 (4-12 weeks), and post-COVID-19 (12-16 weeks). We assessed sociodemographic and mobile health risk factors for the development of long COVID (defined as either a persistent change in a physiological signal or self-reported symptoms for ≥12 weeks after COVID-19 diagnosis).

Findings: By Aug 1, 2022, 17 667 participants had enrolled into the study, of whom 1200 (6·8%) cases and 3600 (20·4%) controls were included in the analyses. Compared with baseline (65 beats per min), resting heart rate increased significantly during the acute (0·47 beats per min; odds ratio [OR] 1·06 [95% CI 1·03-1·09]; p<0·0001), ongoing (0·99 beats per min; 1·11 [1·08-1·14]; p<0·0001), and post-COVID-19 (0·52 beats per min; 1·04 [1·02-1·07]; p=0·0017) phases. An increased level of historical activity in the period from 24 months to 6 months preceding COVID-19 diagnosis was protective against long COVID (coefficient -0·017 [95% CI -0·030 to -0·003]; p=0·015). Depressive symptoms were persistently elevated following COVID-19 (OR 1·03 [95% CI 1·01-1·06]; p=0·0033) and were a potential risk factor for developing long COVID (1·14 [1·07-1·22]; p<0·0001).

Interpretation: Mobile health technologies and commercial wearable devices might prove to be a useful resource for tracking recovery from COVID-19 and the prevalence of its long-term sequelae, as well as representing an abundant source of historical data. Mental wellbeing can be impacted negatively for an extended period following COVID-19.

Funding: National Institute for Health and Care Research (NIHR), NIHR Maudsley Biomedical Research Centre, UK Research and Innovation, and Medical Research Council.

Figure 1

Physiological signals and self-reported measures of mental health over time Metrics plotted by cohort, by sex in the COVID-19 cohort, and by age group in the COVID-19 cohort. Dates range from 8 weeks before to 24 weeks after COVID-19 diagnosis. Physiological signals (resting heart rate, RMSSD, step count, and sleep duration) measured through wearable devices (passive data) and self-reported measures (PHQ-8 score, GAD-7 score, arousal, and valence) through survey responses (active data). References for changes in physiological signals are baseline values. Arousal and valence scores are reported on a visual analogue scale ranging from –1 to +1. The shaded area corresponds to 95% CIs over a 14-day window for self-reported measures, a 1-day window for physiological signals by cohort and sex, and a 7-day window for physiological signals by age group. GAD-7=seven-item assessment of Generalised Anxiety Disorder. PHQ-8=eight-item Patient Health Questionnaire. RMSSD=root mean square of successive differences between heartbeats.

Figure 2

A heatmap of prompted self-reported symptoms The heatmap represents counts of self-reported symptom severity among patients in the COVID-19 cohort 24 weeks before and after the date of diagnosis. Although it was possible to input additional symptoms, these symptoms were specifically prompted for in the survey. The colour intensity indicates the number of participants who reported a symptom (log scale). Severity could be reported on a three-point scale (mild, medium, or severe).