Persistent COVID-19 symptoms in a community study of 606,434 people in England

Abstract

Long COVID remains a broadly defined syndrome, with estimates of prevalence and duration varying widely. We use data from rounds 3-5 of the REACT-2 study (n = 508,707; September 2020 - February 2021), a representative community survey of adults in England, and replication data from round 6 (n = 97,717; May 2021) to estimate the prevalence and identify predictors of persistent symptoms lasting 12 weeks or more; and unsupervised learning to cluster individuals by reported symptoms. At 12 weeks in rounds 3-5, 37.7% experienced at least one symptom, falling to 21.6% in round 6. Female sex, increasing age, obesity, smoking, vaping, hospitalisation with COVID-19, deprivation, and being a healthcare worker are associated with higher probability of persistent symptoms in rounds 3-5, and Asian ethnicity with lower probability. Clustering analysis identifies a subset of participants with predominantly respiratory symptoms. Managing the long-term sequelae of COVID-19 will remain a major challenge for affected individuals and their families and for health services.

Conflict of interest statement

The authors declare no competing interests.

© 2022. The Author(s).

Figures

Fig. 1. Study population flow chart.

An overview of primary and replication study population size, with exclusions and proportion of participants experiencing symptoms 12 weeks after symptom onset.

Fig. 2. Persistence of symptoms over time.

Plots showing persistence of symptoms as a proportion of those who reported symptoms at any time, among n = 71,642 respondents for whom we had 150 days’ observation time. Women have higher rates of persistent symptoms; a slower decline in symptom prevalence is observed after 12 weeks in both sexes. The vertical dashed lines show 4 and 12 weeks post symptom onset, respectively.

Fig. 3. Symptom prevalence in September 2020–February 2021, and in May 2021.

Prevalence of 37 symptoms surveyed across rounds 3–6 of REACT-2. Top panel shows symptoms that were surveyed in all rounds (n = 606,434 observations); middle panel shows symptoms surveyed in rounds 3–5 only (n = 508,707 observations); bottom panel shows symptoms surveyed in round 6 only (n = 97,727 observations). Right panel compares symptom prevalence in the main study cohort (REACT-2 rounds 3–5, surveyed between October 2020 and February 2021) with the replication cohort (REACT-2 round 6, surveyed in May 2021). Asterisks indicate symptoms that were grouped in the round 6 survey. Green bars in the right panel indicate a decrease in symptom prevalence in round 6 compared with rounds 3–5. Error bars indicate 95% binomial confidence intervals of the prevalence.

Fig. 4. Modelling of persistent symptoms as a function of biological and demographic variables.

a Logistic regression models with one or more symptoms at 12 weeks (y/n) as the binary outcome variable, both adjusted for age-sex and mutually adjusted*; b mean contribution to area under the curve (AUC) that each variable makes to a multivariable boosted tree model, derived by permuting each variable in turn (1000× to obtain a distribution) and quantifying the change in model performance; c modelled probability of persistent symptoms at 12 weeks as a function of age and sex, using generalised additive models with splines on age and interactions between age and sex. All models were fit on n = 71,642 respondents for whom we had 150 days’ observation time. Age, sex, adiposity household income, healthcare/care home worker, deprivation, current smoker status and prior hospitalisation with COVID-19 are the strongest predictors of persistent symptoms in multivariable modelling, while Asian ethnicity is associated with a lower risk of persistent symptoms at 12 weeks. Box plots in panel b show median, first and third quartiles; whiskers indicate 1.5 × the interquartile range; data beyond this range are plotted as points. Note: Owing to missing data in some variables, the total n for the mutually adjusted model in panel a is 55,730.

Fig. 5. Results of clustering on symptom profile at 12 weeks.

Clustering was conducted using CLARA (partitioning around medoids) algorithm. Two stable clusters were identified at 12 weeks. Cluster L1 (“tiredness cluster”) had high prevalence of tiredness. Cluster L2 (“respiratory cluster”) was a smaller subset of 4,441 participants who had high prevalence of shortness of breath and tight chest as well as chest pain. Panel a shows symptom prevalence by cluster. Panel b shows the distribution of symptom counts by cluster (median 2 symptoms for cluster L1 [n = 15,799] and 3 symptoms for cluster L2 [n = 4441]). Box plots in panel b show median, first and third quartiles; whiskers indicate 1.5*the interquartile range; data beyond this range are plotted as points. Panel c shows the self-reported symptom severity and medical treatment sought by cluster (with those who were no longer symptomatic at 12 weeks for comparison).

Fig. 6. Persistence of individual symptoms, by symptom cluster.

Persistence of symptoms for all symptomatic participants (top panel) and in 12-week symptoms clusters L1 and L2 (bottom panels). Dashed lines show 4 and 12 weeks post symptom onset, respectively.