Predictive symptoms for COVID-19 in the community: REACT-1 study of over 1 million people.

BACKGROUND: Rapid detection, isolation, and contact tracing of community COVID-19 cases are essential measures to limit the community spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). We aimed to identify a parsimonious set of symptoms that jointly predict COVID-19 and investigated whether predictive symptoms differ between the B.1.1.7 (Alpha) lineage (predominating as of April 2021 in the US, UK, and elsewhere) and wild type. METHODS AND
FINDINGS: We obtained throat and nose swabs with valid SARS-CoV-2 PCR test results from 1,147,370 volunteers aged 5 years and above (6,450 positive cases) in the REal-time Assessment of Community Transmission-1 (REACT-1) study. This study involved repeated community-based random surveys of prevalence in England (study rounds 2 to 8, June 2020 to January 2021, response rates 22%-27%). Participants were asked about symptoms occurring in the week prior to testing. Viral genome sequencing was carried out for PCR-positive samples with N-gene cycle threshold value < 34 (N = 1,079) in round 8 (January 2021). In univariate analysis, all 26 surveyed symptoms were associated with PCR positivity compared with non-symptomatic people. Stability selection (1,000 penalized logistic regression models with 50% subsampling) among people reporting at least 1 symptom identified 7 symptoms as jointly and positively predictive of PCR positivity in rounds 2-7 (June to December 2020): loss or change of sense of smell, loss or change of sense of taste, fever, new persistent cough, chills, appetite loss, and muscle aches. The resulting model (rounds 2-7) predicted PCR positivity in round 8 with area under the curve (AUC) of 0.77. The same 7 symptoms were selected as jointly predictive of B.1.1.7 infection in round 8, although when comparing B.1.1.7 with wild type, new persistent cough and sore throat were more predictive of B.1.1.7 infection while loss or change of sense of smell was more predictive of the wild type. The main limitations of our study are (i) potential participation bias despite random sampling of named individuals from the National Health Service register and weighting designed to achieve a representative sample of the population of England and (ii) the necessary reliance on self-reported symptoms, which may be prone to recall bias and may therefore lead to biased estimates of symptom prevalence in England.
CONCLUSIONS: Where testing capacity is limited, it is important to use tests in the most efficient way possible. We identified a set of 7 symptoms that, when considered together, maximize detection of COVID-19 in the community, including infection with the B.1.1.7 lineage.

Introduction

To control the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic, rapid identification and isolation of infected individuals is essential [1,2], together with testing and isolation of their contacts [3,4]. A range of symptoms have been identified as associated with COVID-19. According to the US Centers for Disease Control and Prevention, these include fever or chills, cough, shortness of breath or difficulty breathing, fatigue, muscle or body aches, headache, new loss of taste or smell, sore throat, congestion or runny nose, nausea or vomiting, and diarrhea [5]. However, it is unclear which symptoms are the most informative of a COVID-19 diagnosis. This is important in settings where test supply is limited.

A novel SARS-CoV-2 variant of concern, VOC 20DEC-01 (lineage B.1.1.7), was first identified in England in September 2020 and became the dominant lineage in the UK within 4 months [6]. As of April 2021, it had been detected in 114 countries [7] and had become the dominant lineage in the US, Europe, and elsewhere [7,8].

Here, we used data from the REal-time Assessment of Community Transmission-1 (REACT-1) study to identify a parsimonious set of symptoms that, taken together, are the most predictive of SARS-CoV-2 PCR positivity. For data collected during January 2021 we also compare predictive symptoms for B.1.1.7 infection versus wild type, identified via viral genome sequencing.

Methods

Study population

REACT-1 is a series of community prevalence surveys of SARS-CoV-2 virus swab positivity in England, conducted at approximately monthly intervals since May 2020. Using the National Health Service patient register across the 315 lower-tier local authorities (LTLAs) in England, recruitment letters were sent to a random nationally representative sample of individuals aged 5 years and over, with a separate (non-overlapping) sample selected at each round. We therefore approached a different base sample at each round and did not have repeat samples to account for. Participant sampling aimed to achieve approximately equal numbers of participants in each LTLA. Random samples were drawn stratified by LTLA, with larger numbers selected in some LTLAs to address variable response rates at the LTLA level. Up to 160,000 valid responses and viable swabs were obtained at each round [9]. Participation involved a self-administered throat and nasal swab, and the completion of a short online or telephone questionnaire including information on demographic variables, household composition, behaviors, and recent symptoms. The questionnaires used are available on the study website (https://www.imperial.ac.uk/medicine/research-and-impact/groups/react-study/react-1-study-materials/). At the time of survey completion, participants were unaware of the result of their swab test. In the survey, participants were asked about new symptoms occurring in the week preceding the swab. These included a set of 26 clinically relevant symptoms potentially related to COVID-19: (i) loss or change of sense of smell and loss or change of sense of taste, (ii) coryzal symptoms (runny nose, sneezing, blocked nose, sore eyes, sore throat, and hoarse voice), (iii) gastrointestinal symptoms (appetite loss, nausea/vomiting, diarrhea, and abdominal pain/belly ache), (iv) fatigue-related symptoms (tiredness, severe fatigue, heavy arms/legs, and difficulty sleeping), (v) respiratory and cardiac symptoms (new persistent cough, shortness of breath, chest pain, and tight chest), and (vi) other flu-like and miscellaneous symptoms (fever, muscle aches, chills, headache, dizziness, and numbness/tingling). Data from round 1 were excluded as the symptom questions asked in that round were not consistent with those in the subsequent rounds [9].

Here, we use REACT-1 data from rounds 2 to 7 (June to December 2020) to identify a parsimonious set of symptoms (from 1 week prior to testing) that are jointly predictive of SARS-CoV-2 PCR positivity and assess their performance among holdout community population-based data from rounds 2 to 7 and, separately, round 8 (January 2021).

PCR-positive swab samples from round 8 with N-gene cycle threshold value < 34 and sufficient sample volume underwent genome sequencing. Viral RNA was amplified using the ARTIC protocol [10], and sequencing libraries were prepared with CoronaHiT [11]. Samples were sequenced using the Illumina NextSeq 500 platform. Each run included 1 positive and 1 negative control per 96 samples. Sequencing data were analyzed using the ARTIC bioinformatic pipeline [12]. Lineages were assigned using PangoLEARN [13].

Statistical analyses

Twenty-five participants (19 in rounds 2–7 and 9 in round 8) were excluded due to missing information on sex. We used univariate logistic regression to model the risk of testing positive for SARS-CoV-2 as a function of symptoms reported in the week prior to testing. For multivariable models, round 2–7 data (restricted to people reporting any of the 26 surveyed symptoms) were split into a 70% training set (76,187 observations, of which 1,078 were positive) and a 30% test set (32,651 observations, of which 461 were positive). We adopted a variable selection approach with stability selection using least absolute shrinkage and selection operator (LASSO) penalized logistic regression with all 26 surveyed symptoms as predictors and PCR positivity as the outcome [14]. LASSO models were fit on 1,000 independent random 50% subsamples of the 70% training set. The stability selection penalty parameter was calibrated to give a per family error rate of fewer than 5 falsely selected symptoms [14,15]. The proportion of penalized models where each symptom was included (its selection proportion) was used as a measure of its importance; those with selection proportion above 50% were considered stably selected and were included in a predictive model of PCR positivity, where mean penalized odds ratios across models (where selected) were used as weightings. The resultant model was then applied to the 30% holdout test set (rounds 2–7) and, separately, to round 8 data. We then estimated the proportion of symptomatic COVID-19 cases that would be detected using this predictive model across all levels of symptomatic community testing.

As a series of sensitivity analyses, we used the same univariate and multivariate approaches to (i) investigate age-specific symptoms by stratifying our study population into 3 age groups (5 to 17 years, 18 to 54 years, and 55 years old or older); (ii) account for possibly differential predictive abilities of symptoms in relation to their sequence/timing of onset, by restricting the list of symptoms to those reported first (rather than any symptom reported in the week prior to testing); and (iii) search for (sets of) symptoms potentially discriminating B.1.1.7 versus wild-type infections among test-positive cases.

All calculations were done with the R computational environment, version 4.0.2, using the logistic LASSO algorithm as implemented in the glmnet package and in-house scripts for the stability selection (available at https://github.com/mrc-ide/reactidd/tree/master/R/symptom_prediction_paper_scripts).

Ethical approval

We obtained research ethics approval from the South Central–Berkshire B Research Ethics Committee (IRAS ID: 283787).

Patient and public involvement

Participants in the REACT-1 study were not involved in the definition of our research question or in the outcome measurements. They were not involved in developing the analytical plan or the implementation of the study. There are no plans to disseminate the results of the research directly to the study participants or the relevant patient community. However, a public advisory panel provides regular review of the study processes and results, and links to published reports from the study are available on the study website (https://www.imperial.ac.uk/medicine/research-and-impact/groups/react-study/real-time-assessment-of-community-transmission-findings/).

Results

Descriptive and univariate analyses

The key characteristics of the study population in rounds 2–7 and 8 for the full sample and for those reporting symptoms are summarized in S1 Table. Of the 979,709 respondents (after 19 exclusions) with a valid swab test in REACT-1 rounds 2–7, 870,872 (88.9%) reported no symptoms in the week prior to testing while 108,837 (11.1%) reported at least 1 of the 26 surveyed symptoms. We detected 4,168 PCR-positive cases, of whom 1,538 (36.9%) reported 1 or more symptoms in the past week. In round 8, there were 167,636 participants (after 6 exclusions), of whom 146,701 (87.5%) reported no symptoms in the past week and 20,935 (12.5%) reported at least 1 of the 26 surveyed symptoms. We detected 2,282 PCR-positive cases, of whom 1,031 (45.1%) reported 1 or more symptoms in the past week (Fig 1; S2 Table). In univariate analyses, each of the 26 surveyed symptoms was associated with PCR positivity in rounds 2–7 and round 8 (Fig 2; S2 Table). In rounds 2–7, with a mean prevalence of PCR positivity of 0.46%, the positive predictive value for any symptom was 1.4%, with an odds ratio for PCR positivity of 4.7 compared with non-symptomatic individuals. In round 8, with a PCR positivity prevalence of 1.36%, the positive predictive value for any symptom was 4.9%, with an odds ratio for PCR positivity of 6.0 compared with non-symptomatic individuals (Fig 1).

Fig 1

Flow chart showing numbers of participants by symptom status and PCR result.

(A) Rounds 2–7 and (B) round 8 of the REACT-1 study.

Fig 2

Results from univariate logistic regression models of PCR positivity for 26 surveyed symptoms.

Effect size estimates are expressed as odds ratios (95% confidence intervals) in rounds 2–7 (left) and round 8 (right).

Multivariable analyses and community case detection

Seven symptoms were selected as jointly positively predictive of PCR positivity in a LASSO stability selection model trained on round 2–7 data: loss or change of sense of smell, loss or change of sense of taste, fever, new persistent cough, chills, appetite loss, and muscle aches. In addition, numbness/tingling was selected as jointly but negatively predictive of PCR positivity. The resultant model gave an area under the curve (AUC) of 0.75 for holdout round 2–7 test data and 0.77 for round 8 data, with no improvement in AUC upon inclusion of other symptoms (Fig 3). Testing people in the community with at least 1 of the 7 selected positively predictive symptoms gave sensitivity, specificity, and positive predictive value of 71%, 64%, and 1.9% in holdout round 2–7 test data and 74%, 64%, and 9.7% in round 8 data, respectively. Performance of the stability selection model at varying levels of community testing is shown in S1 Fig; for example, testing 10% of symptomatic individuals would detect around half of all symptomatic cases.

Fig 3

Selected symptoms predictive of COVID-19.

Results of LASSO stability selection using 1,000 models (with 50% subsamples of training data from rounds 2–7). Mean (penalized) log odds ratios (log ORs) across all models are shown in the top panel. Positive regression coefficients are presented in teal, and negative in red. Only symptoms selected at least once are displayed. The selection proportions (selection prop.; proportion of 1,000 models that included each symptom) are shown in the middle panel; the horizontal dashed line shows the selection threshold of 50%. Symptoms are ordered according to their selection proportions, and selected symptoms are in black. The bottom panel shows the area under the curve (AUC) of models adding each variable in order of selection proportion (from left to right) in both holdout data from rounds 2–7 (grey) and data from round 8 (red).

B.1.1.7 (Alpha) lineage versus wild-type symptoms

Of the 2,282 positive cases in round 8, 1,088 had cycle threshold value < 34 and underwent viral genome sequencing; 898 (82.5%) were B.1.1.7, 181 (16.6%) were wild type, and 8 were other lineages. Characteristics of the positive cases with B.1.1.7 and wild type in round 8 were similar except for the region of the case, with a marked excess of B.1.1.7 cases in East of England, South East, and London (S3 Table). LASSO stability selection models fit on round 8 data with an outcome of B.1.1.7 versus PCR negative selected the same 7 positively predictive symptoms as in rounds 2–7 (Fig 4). In univariate analysis within round 8, there was slightly higher reporting of any symptom for B.1.1.7 infection versus wild type. Specifically, we found a higher prevalence of sore throat, new persistent cough, fever, difficulty sleeping, dizziness, and nausea/vomiting at a nominal 0.05 significance level (Fig 5A). Stability selection models identified sore throat and new persistent cough as jointly and positively associated with B.1.1.7 infection compared with wild-type infection, while loss or change of sense of smell was selected as negatively associated, i.e., more predictive of wild type infection (Fig 5B).

Fig 4

Symptoms predictive of B.1.1.7 infection.

LASSO stability selection for symptoms predictive of B.1.1.7 (Alpha) lineage infection versus symptomatic people (aged 5+ years) testing PCR negative in round 8. Mean log odds ratio (Log OR) and selection proportion (selection prop.) are represented for each symptom in the top and bottom panels, respectively. Positive regression coefficients are presented in teal, and negative in red.

Fig 5

B.1.1.7 versus wild-type symptoms.

Comparison of symptom profile in B.1.1.7 (Alpha) lineage versus wild-type infection among 1,124 people testing positive in round 8 (other lineages excluded, N = 8). (A) Proportion of people reporting each symptom by lineage (left panel), and the differences in proportions with 95% confidence intervals (right panel). (B) Results of LASSO stability selection (using 1,000 models with 50% subsampling) with B.1.1.7 infection, versus wild-type infection as the outcome, summarized by the mean log odds ratio (Log OR) and selection proportion (selection prop.) for each of the symptoms selected at least once. Positive regression coefficients are presented in teal, and negative in red. The horizontal dashed line represents the selection threshold of 50%.

Sensitivity analyses

In age-stratified analyses of round 2–7 data, 6 (5–17 years) or 7 (18–54 and 55+ years) symptoms were jointly selected. New persistent cough was not selected for those aged 5–17 years but was for adults, while chills, fever, loss or change of sense of smell, and loss or change of sense of taste were selected in all age groups. Additional age-specific selected symptoms (positive coefficients) were headache (5–17 years), appetite loss (18–54 and 55+ years), and muscle aches (18–54 years), and age-specific negatively associated symptoms (negative coefficients) were runny nose (5–17 years) and numbness/tingling (55+ years) (S2 Fig).

Considering first reported symptom, instead of any symptom, as the predictor, in univariate analysis, all first reported symptoms among the 26 surveyed were associated with PCR positivity in both round 2–7 and round 8 data (S3 Fig). In multivariable analysis, LASSO stability selection on round 2–7 data included headache as positively predictive of PCR positivity instead of appetite loss (S4 Fig); otherwise, the selected (positively associated) symptom set was the same as in Fig 3, although with lower predictive performance (AUC of 0.68 and 0.66 for holdout round 2–7 test data and round 8 data, respectively).

Discussion

In this study of over 1 million people in England, we found that 7 symptoms stably and jointly predicted SARS-CoV-2 PCR positivity; these were loss or change of sense of smell, loss or change of sense of taste, fever, new persistent cough, chills, appetite loss, and muscle aches. The first 4 of these symptoms are currently used in the UK to determine eligibility for community PCR testing, selected at a time (May 2020) when testing capacity was limited. Based on our findings, this symptom set is too restrictive, and in order to improve PCR positivity detection rates and consequently improve control of viral transmission via isolation measures, we would propose to extend the list of symptoms used for triage to all 7 symptoms we identified. This approach would have the advantage of increasing the yield of detected cases, leading to greater numbers of infected people being required to self-isolate, thereby reducing the pool of infection in the community.

The use of the 7 symptoms we identified for PCR test allocation would result in 30% to 40% of symptomatic individuals in England being eligible for a test (versus 10% currently) and, if all those eligible were tested, would result in the detection of 70% to 75% of the positive cases. Further extending the list of symptoms for test allocation would increase the number of tests performed (up to 3-fold) and potentially result in a 100% positive detection rate if all symptomatic individuals were tested, but with increasing prevalence rates would exceed testing capacity. We believe that our approach, relying on these 7 symptoms, provides a reasonable balance between number of tests performed and detection rates in England, but alternative approaches could be envisaged depending on national population size, composition, and resources. In particular, testing individuals reporting any COVID-19-related symptoms has been implemented in some countries that have adopted a zero COVID-19 policy (e.g., Australia and New Zealand). Such investments have been justified by arguing that, if successful, this approach leads to an earlier and more efficient control of transmission of SARS-CoV-2, in turn reducing the prevalence of the infection and hence of symptoms, thus reducing the testing capacity needs.

Irrespective of the test allocation strategy, prevention efficacy strongly relies on the accurate reporting of symptoms by individuals in the general population. To ensure adherence to self-isolation measures and lessen the burden of such measures in socially disadvantaged individuals, efficient financial and other support policies should be encouraged. Failure to implement such targeted policies could introduce a social gradient in the willingness to report symptoms [16] and attend for testing, which in turn could lead to increased transmission of SARS-CoV-2 in more deprived populations, as has been observed in England [17].

Other studies investigating sets of symptoms predictive of PCR positivity have been published [18]. As in our study, these identified loss or change of sense of smell, loss or change of sense of taste, fever, and new persistent cough as consistent predictors of infection. Of the 3 additional symptoms we identify (chills, appetite loss, and muscle aches), appetite loss was found previously to be associated with PCR positivity in the UK [18]. Unlike other studies, REACT-1 involves community-based random samples of individuals in the population, and with over 1 million participants, it provides reliable, reproducible, and representative estimates of prevalence and prediction of PCR positivity in England from a combination of informative symptoms.

The second wave of COVID-19 in England coincided with the emergence of VOC 20DEC-01 (lineage B.1.1.7), which became dominant in the UK by January 2021 [6] and subsequently in many countries around the world as of April 2021 [7]. B.1.1.7 is defined by 17 mutations; 8 of these affect the viral spike protein, the means by which SARS-CoV-2 binds to angiotensin-converting enzyme 2 (ACE2) and enters host cells. These spike mutations might confer an evolutionary benefit. For example, the spike deletion ΔH69/ΔV70 enhances viral infectivity in vitro [19], and N501Y may enhance spike binding affinity to ACE2 [20]. From modeling, it has been estimated that B.1.1.7 is 40% to 90% more transmissible than earlier lineages [21] and might be associated with an increased risk of hospitalization and death [22,23].

Here we show that the same symptoms were selected as jointly predictive of B.1.1.7 infection as for earlier lineages. However, when we compared B.1.1.7 with wild-type infections, new persistent cough and sore throat were jointly selected as predictive of B.1.1.7 infections, while loss or change of sense of smell was predictive of wild-type infections. This is consistent with findings from the UK Office for National Statistics Coronavirus (COVID-19) Infection Survey, where people testing PCR positive for lineages compatible with B.1.1.7 were less likely to report loss or change of sense of taste or smell (both symptoms combined), and more likely to report cough, compared with non-B.1.1.7-compatible lineages [24].

Our study has limitations. First, it is uncertain to what extent our findings are generalizable to other settings. However, our sampling procedure was designed to ensure good representation across the whole population in England, including capturing sociodemographic and ethnic diversity. Second, our data relied on a time-resolved series of reported symptoms from self-administered questionnaires. These may be subject to recall bias and may not precisely represent the individual dynamics of symptom onset. However, our self-reported data are based on representative community-based samples and may not share the same limitations as the routine reporting of symptoms on which test allocation and isolation measures are based [25]. Furthermore, participants were unaware of their test results at the time of symptom report, which would limit reporting and information bias. Third, as sampling in each round was cross-sectional, some individuals may have been infected (and had symptoms) more than 1 week before the swab was obtained but were no longer symptomatic at the time of the study or might have gone on to develop symptoms after testing. Fourth, despite our careful sample weighting to account for possible variations in response rates by sex, age, region, deprivation, and ethnicity, we cannot rule out residual selection bias. Finally, a number of B.1.1.7 infections may have been included in PCR positive samples from rounds 5 to 7 (estimated proportions of B.1.1.7 infection among all SARS-CoV-2 infections in England are 0.7% [95% CI 0.6%, 0.9%] and 10.6% [95% CI 10.1%, 11.2%] for November [round 6] and December [round 7], respectively; S5 Fig) [26]. Overall, B.1.1.7 would represent less than 4% of the PCR-positive cases in rounds 2–7, and because our results were suggestive of a consistent symptomatology for both B.1.1.7 and wild-type infections, possible confounding of our results by the inclusion of early B.1.1.7 incident cases is limited.

Our main analyses fit models to the population at all ages. In age-stratified analyses, however, we found a different symptom profile among children and adolescents (ages 5–17 years), where headache (positive association) and runny nose (negative association) were selected, but new persistent cough was not. This may have implications for symptomatic testing in school-aged children. Headache was also selected (instead of appetite loss) for first reported symptom at all ages (5+ years), but a model derived from first reported symptom was less predictive of PCR positivity than the model based on all symptoms within the week prior to testing.

In summary, we show that using a combination of 7 symptoms to determine test eligibility would maximize the case detection rate in the community under testing capacity constraints such as those faced in England between June 2020 and January 2021. This has policy relevance for countries where there is limited testing capacity. We identified the same symptom set for predicting B.1.1.7, which by April 2021 had become the predominant lineage in the UK, US, and many other countries worldwide.

Univariate effect of each symptom reported in the week prior to testing on risk of PCR positivity in rounds 2–7 and round 8.

(DOCX)

Click here for additional data file.

Age-stratified LASSO stability selection on testing PCR positive.

(DOCX)

Click here for additional data file.

Univariate effect of first reported symptom on risk of PCR positivity in rounds 2–7 and round 8.

(DOCX)

Click here for additional data file.

LASSO stability selection for first reported symptom.

(DOCX)

Click here for additional data file.

Proportion of B.1.1.7 variant cases among all cases in England, October 2020–January 2021.

(DOCX)

Click here for additional data file.

Supplementary methods.

(DOCX)

Click here for additional data file.

Key characteristics of the REACT-1 study population for rounds 2–7 and round 8.

Results are presented for the full study population.

(DOCX)

Click here for additional data file.

Descriptive statistics for symptoms and SARS-CoV-2 PCR test results among REACT-1 participants, rounds 2–7 and round 8.

(DOCX)

Click here for additional data file.

Characteristics of the round 8 PCR-positive participants in whom lineage data were available.

Numbers are presented for cases with wild-type (N = 181) and B.1.1.7 (N = 898) infections. The statistical significance of differences between these 2 groups is evaluated using a chi-squared test for categorical variables. We report the p-value for the null hypothesis of no difference.

(DOCX)

Click here for additional data file.

12 May 2021

Dear Dr Chadeau-Hyam,

Thank you for submitting your manuscript entitled "Predictive symptoms for COVID-19 in the community: REACT-1 study of over one million people" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by May 14 2021 11:59PM.

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Callam Davidson

Associate Editor

PLOS Medicine

1 Jul 2021

Dear Dr. Chadeau-Hyam,

Thank you very much for submitting your manuscript "Predictive symptoms for COVID-19 in the community: REACT-1 study of over one million people" (PMEDICINE-D-21-02112R1) for consideration at PLOS Medicine.

Your paper was discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments fully. You will appreciate that we cannot make a decision about publication until we have seen the revised manuscript and your response, and we expect to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We hope to receive your revised manuscript by Jul 22 2021 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

Please let me know if you have any questions, and we look forward to receiving your revised manuscript.

Sincerely,

Callam Davidson,

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Please update the final sentence of the ‘Methods and Findings’ section of your abstract to list 2-3 of the study’s main limitations (the sentence should begin ‘Limitations of this study include…’ or similar).

Please ensure you have provided all the necessary information that would allow a reader to access the data used in this study (for example, contact information for the project steering committee) in your response to the relevant question in the submission form. See https://journals.plos.org/plosmedicine/s/data-availability for more information.

If the study has an associated protocol and/or analysis plan, please can these be provided as supplementary materials.

Please add a completed STROBE checklist as a supplementary file, labelled "S1_STROBE_Checklist" or similar and referred to as such in the Methods section.

In the checklist, please refer to individual items by section (e.g., "Methods") and paragraph number, not by line or page numbers as these generally change in the event of publication.

Please update your in-text citations to appear before punctuation.

The final paragraph of the ‘Introduction’ can be shortened to state the study’s aims only; elements of discussion can be relocated to the ‘Discussion’ section.

Please remove the ‘Transparency’ and ‘Conflict of Interest’ statements from the end of the main text. The relevant information should be captured in your responses to the submission forms and, in the event of publication, will be presented as metadata.

Comments from the reviewers:

Reviewer #1: Thanks for the opportunity to review your manuscript. My role is as a statistical reviewer, so my comments and questions focus on the study design, data, and analysis presented in the manuscript. I have put general questions first, and then followed these with queries specific to a section of the manuscript (with a line/page reference).

This manuscript presents a study designed to identify key symptoms of COVID-19. This data is from community prevalence surveys in the UK where participants are recruited, questioned about symptom history and have a swab tested for COVID-19. The symptoms were identified by splitting surveys 2-7 into training and test data, and variable selection was achieved by using LASSO logistic regression with all the recorded symptoms. Within the test data sets, a resampling approach was used to estimate a stability of selection parameter, with variables selected 50% of the time were selected. This model was validated against the test set from surveys 2-7 (same temporality) and in survey 8. This is a complex but robust approach to variable selection with both LASSO and the resampling within training data. There was a fairly good level of performance in the round 8 data (AUC=0.77) given that the symptoms are all self-report.

There are some new approaches used here that I am not familiar with (i.e. particularly the approach to stability selection penalty parameter), is the code used for the analysis able to be shared? This would help me understand what you did and be able to fully recommend it.

A pre-registered study protocol is available online for the overall REACT study. Was an analysis plan for this study created, and if so, is this available for the review?

I have checked but I couldn't find what the overall recruitment rate for the REACT-1 study was. It looks as though the study uses quota sampling to meet the sample size objectives, how many potential participants at each wave were approached but didn't respond, relative to those that did participant in the study? (see question below about survey design and generalisability)

P7, L131. What software (R?) and packages were used to complete the analysis?

P7, L132. To clarify, the main analysis was repeated by COVID strain, age groups, and using first reported symptoms with each sub-group repeated separately from the others?

P11, L235. Do you mean literally mean a 'predictive' model be applied in other contexts with the 7 symptoms or that policy around access to testing, isolation etc. should expand to include the extra symptoms?

P11, L259. This is a good point - is there any data available from the overall survey that can show that this was achieved? I tried looking but couldn't find any.

P11, L263. 'Inaccurate' is very broad - what is a likely form of recall bias, and would induce spurious associations, or conceal real associations with this methodology? I have seen 'negative control' type of questions added to surveys e.g. symptoms that could not be associated with the disease included to see if there is an association with the outcome (although this might be hard given the range of symptoms COVID seems to bring about).

Reviewer #2: This is an excellent study, methodologically well done, well written, and of public health importance, which leaves me little to say.

I have only a few suggestions:

(1) B117 was spreading in England from Sep 2020, and so would have been present in Round 5-7 samples. This confounds the comparison of Round 8 with earlier rounds, so please provide an estimate of what percentage of infections in England were likely B117 in earlier months, based on existing public domain estimates. It would have been interesting to show an additional analysis comparing Rounds 2-4 vs Round 8 for similar reasons, but this is not essential.

(2) P11/L234-236

This is a very important point, which perhaps could be stated more clearly for the reader. The important message here is that in countries where testing is being rationed by being based only on the first four symptoms, then almost certainly increasing the symptom list by the other 3 symptoms will be almost as efficient, whilst identifying a significantly larger number of infections.

(3) P12/L238-239

It's not clear to me what the intention was of making the reference to economically disadvantaging individuals. The issue of economic disadvantage at the individual level applies equally however restrictive the testing regime, since almost everyone who tests positive will be infected. I think the authors are trying to convey an economic point at the level of the population. But from an economic perspective, the best way to minimize the economic disadvantage of isolation must be to reduce the number of infected persons in the population, which is best achieved by reducing transmission (Reff) as much as possible. To achieve this, isolating a greater fraction of infected cases is always better, so increasing the number of individuals so identified in the short-term will always reduce the net number of people needing isolation in the longer-term.

(4) I refer to the following papers:

Hellewell J, Abbott S, Gimma A, Bosse NI, Jarvis CI, Russell TW, et al. Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. Lancet Glob Health. 2020;8(4):e488-e96.

Kerr et al, 2021, Nature Communications. Controlling COVID-19 via test-trace-quarantine. https://www.nature.com/articles/s41467-021-23276-9

I think efficiency is important in a context where testing capacity is limited, so this paper helps inform the testing policy in those contexts. However, to successfully control COVID-19, efficiency is not the only criterion. Effectiveness is also important, ie maximization of the fraction of all infections detected is critical to any strategy that aims to reduce R below 1, especially one that seeks to avoid damaging social mobility restrictions. I would suggest the authors discuss the implications for the less constrained settings, and scenarios where the primary goal is to stop transmission.

Specifically, the authors should discuss further their finding that the 7 symptoms would still only identify ~70% of symptomatic individuals, or 30-40% of all infected cases (after accounting for asymptomatic infections. Given that a large fraction of those with the relevant symptoms will not voluntarily test in the most ideal circumstances, what this means is that in practice the actual percentage of symptomatic cases that will be tested and detected will be far less. What the findings demonstrate is that the kind of strategy that Australia and New Zealand pursued makes a lot of sense. They used the full list of 26 symptoms to trigger testing (eg https://www.nsw.gov.au/covid-19/health-and-wellbeing/symptoms-and-testing). This does result in a much lower positivity rate than would be achieved by restricting to the more efficient subset of 7 symptoms (<0.1% vs 1%), but it achieves much better control of transmission (sufficient to achieve short-term elimination), and in the long run required less overall testing than in the UK (2 per 1000/d vs 20 per 1000/day). So for balance, the issue of (cost) efficiency needs to be considered in both the short term and long-term perspectives.

Reviewer #3: This study seeks to answer a policy-important question in a large representative population. This is a prevalence study using repeated cross-sectional surveys in a representative sample of the community in England aiming to identify a set of symptoms to predict SARS-CoV-2 PCR positivity and distinguish between wild type and B.1.1.7 strains. The surveys consisted of self-administered nose / throat swabs and questionnaires to a randomly selected sample. Significant symptoms were identified over 6 rounds of testing between June and December 2020 and were validated in a sub-population from these first 6 rounds and in an additional 7th round of testing in January 2021 when the B.1.1.7 strain was predominantly circulating. A large population of over 1 million people over the age of 5 were included. Seven symptoms were identified, and the authors have concluded that these may be more efficient than the current four symptoms recommended for testing eligibility within the UK national testing programme at present. This is one of the largest studies of this type that has been conducted and has monitored changes over the course of the pandemic.

Major issues

1. It would be useful to have a table or description of the included population (eg age, sex, region, occupations, socio-economic status, number of households) and number of participants per round of testing.

Minor Issues

2.1 At what time point were surveys administered - before or after the swab? Did they relate to just the day that the swab was administered or to a period of time before sampling?

2.2 Were baseline questionnaires administered to allow chronic symptoms to be distinguished from acute?

2.3 How were the symptoms that were included in the questionnaire selected?

2.4 Were all swabs and viral isolates processed in the same laboratory? If not, is there a difference in assay performance that could affect the results?

2.5 There is a potential for selection bias due to survey response. Were there any differences between survey responders and non-responders?

2.6 There is no mention of whether sampling was carried out with or without replacement, although the study protocol states that it was with replacement. How were repeat samples in the same individuals accounted for?

2.7 Please describe any exclusion criteria if these were present.

2.8 Please describe the sample size calculation.

2.9 When comparing the wild type and the B.1.1.7 groups - were there any systematic differences (eg age) that could have introduced bias?

2.10 In discussion section, it would be useful to mention whether there have been other studies that have identified symptoms that can predict PCR positivity / developed risk scores and how this study compares to them.

Any attachments provided with reviews can be seen via the following link:

[LINK]

16 Jul 2021

Submitted filename: Response_to_reviewers_final.docx

Click here for additional data file.

6 Aug 2021

Dear Dr. Chadeau-Hyam,

Thank you very much for re-submitting your manuscript "Predictive symptoms for COVID-19 in the community: REACT-1 study of over one million people" (PMEDICINE-D-21-02112R2) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by three reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We hope to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Aug 13 2021 11:59PM.

Sincerely,

Callam Davidson,

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

Please add further detail to your data availability statement regarding how a reader could apply for access to the REACT study data via the project steering committee (a website or email address ought to be sufficient).

Please reference the study questionnaires and the address at which they are located in your Methods section.

Please remove the ‘Contributors’, ‘Funding’, and ‘Data sharing statement’ from the end of the main text. In the event of publication, this information will be published as metadata based on your responses to the submission form.

Please relocate the ‘Patient and public involvement’ and ‘Dissemination to participants and related patient and public communities’ sections to the Methods, rather than placing them at the end of the main text.

Citations should be in square brackets, not in superscript, and preceding punctuation.

Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

After updating your references to use Vancouver style, please ensure that references do not have periods after initials, use ‘et al.’ only after listing the first six authors, and do not contain any formatted text (e.g., bold, italics).

Please define any abbreviations used in figures (e.g., OR in Figures 3 and 4).

Comments from Reviewers:

Reviewer #1: Thank you for the revised manuscript and responses to my original queries.

The additional explanation and available code for the Meinhausen and Buhlman stability selection approach were helpful thank you, I have also looked at the original reference and agree with you this is a an appropriate approach.

The rest of the clarifications to the manuscript address my original queries - the part of the discussion about generalisability and potential for bias looks better now. I agree with the comment about timing of self-report and availability of test results, and given the short time period between recall of symptoms and test I think the potential for bias here is limited.

My only query was about the analysis plan - on p2 of the sup information document there is the extra explanation of the stability selection approach and the link to the code, but I could not see an analysis plan or link to one. Have I inadvertently missed this? This could added to the supplementary appendix easily.

Great manuscript - I look forward to seeing whether symptoms change with delta when that data becomes available to you.

Reviewer #2: Thank you for the revisions which address the comments by me and other reviewers.

The only comment I would make is that in the last paragraph of the discussion, the use of the term "most efficient" is incorrect. Technically speaking, using the first ranked symptom would be the most efficient. What I think you mean is that the combination of seven symptoms identified would be the most optimal given the need to maximise case detection in the community under the constraints to testing capacity that existed in conditions similar to that in England in the relevant time period.

Ravi P. Rannan-Eliya

Reviewer #3: My comments have been addressed in this revised manuscript, thank you to the authors.

Any attachments provided with reviews can be seen via the following link:

[LINK]

10 Aug 2021

Submitted filename: Response_to_comments.docx

Click here for additional data file.

11 Aug 2021

Dear Dr. Chadeau-Hyam,

Thank you very much for re-submitting your manuscript "Predictive symptoms for COVID-19 in the community: REACT-1 study of over one million people" (PMEDICINE-D-21-02112R3) for review by PLOS Medicine.

There are still a small number of minor revisions that are required before we can proceed. The remaining issues that need to be addressed are listed at the end of this email. Please take these into account before resubmitting your manuscript.

In revising the manuscript for further consideration here, please ensure you address the specific points made by the editors. In your rebuttal letter you should indicate your response to the editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

We hope to receive your revised manuscript within 1 week. Please email me (cdavidson@plos.org) if you have any questions or concerns.

Sincerely,

Callam Davidson,

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

The data availability statement has changed since the last revision but not as described in your rebuttal letter – the previous revision read ‘Summary statistics and descriptive tables are available in Supplementary information; R scripts used to perform the analyses can be downloaded at https://github.com/barbarabodinier/Community_detection_COVID-19. REACT data can be accessed upon application and acceptance to the project steering committee.’ In the latest revision, the statement reads ‘Summary statistics and descriptive tables from REACT-1 study are available on the study website https://www.imperial.ac.uk/medicine/research-and-impact/groups/reactstudy/react-1-study-materials/ and, together with the R scripts used to perform the analyses, at https://github.com/barbarabodinier/Community_detection_COVID’.

My issues with this are as follows:

* The summary statistics and descriptive tables are in the supplementary materials; the web address you have provided contains the questionnaires used. Please correct the statement to include both.

* It looks as though the statement has been truncated or text is missing, as ‘and, together’ does not scan properly and the github address is now missing ‘-19’ and leads to an error page. Please check carefully.

* The email addresses you mention having added in your rebuttal later are not available (in fact reference to the project steering committee has been removed altogether). Please check and see the comment below.

* Finally (and importantly), a study author cannot be the contact person for the data – if there is a contact within the project steering committee that is not a named author, or the option to use a data access or ethics committee, please provide their details. If you foresee this being a problem, please contact me (cdavidson@plos.org) to discuss.

Line 80: Please update this bullet point to ‘Data were collected from over 1 million participants in the REACT-1 study (June 2020 to January 2021), from whom 26 symptoms were assayed and the results of a PCR test were available.

Line 83: Please update this bullet point to ‘Adopting a variable selection approach, we sought to determine the best combination of symptoms jointly and complementarily predictive of PCR positivity and compared whether these symptoms were the same between individuals infected by the wild-type virus and those infected by the B.1.1.7 variant’. The following bullet point (line 86) can then be removed.

Line 89: Please update this bullet point to ‘We identify seven symptoms that were jointly predictive of PCR positivity and appeared to vary only marginally across age groups: loss or change of sense of smell, loss or change of sense of taste, fever, new persistent cough, chills, appetite loss and muscle aches’. The following bullet point (line 92) can then be removed.

Line 99: Please update ‘but will enable’ to ‘but would enable’

Line 455: Please remove the periods after the author initials in reference 15.

12 Aug 2021

Submitted filename: Response_to_comments.docx

Click here for additional data file.

20 Aug 2021

Dear Dr Chadeau-Hyam,

On behalf of my colleagues and the Academic Editor, Dr Kretzschmar, I am pleased to inform you that we have agreed to publish your manuscript "Predictive symptoms for COVID-19 in the community: REACT-1 study of over one million people" (PMEDICINE-D-21-02112R4) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Richard Turner PhD, for Callam Davidson

Senior Editor, PLOS Medicine

rturner@plos.org