App-based COVID-19 surveillance and prediction: The COVID Symptom Study Sweden (preprint)

Background The response of the Swedish authorities to the COVID-19 pandemic was less restrictive than in most countries during the first year, with infection and death rates substantially higher than in neighbouring Nordic countries. Because access to PCR testing was limited during the first wave (February to June 2020) and regional data were reported with delay, adequate monitoring of community disease spread was hampered. The app-based COVID Symptom Study was launched in Sweden to disseminate real-time estimates of disease spread and to collect prospective data for research. The aim of this study was to describe the research project, develop models for estimation of COVID-19 prevalence and to evaluate it for prediction of hospital admissions for COVID-19. Methods We enrolled 143 531 study participants ([≥]18 years) throughout Sweden, who contributed 10.6 million daily symptom reports between April 29, 2020 and February 10, 2021. Data from 19 161 self-reported PCR tests were used to create a symptom-based algorithm to estimate daily prevalence of symptomatic COVID-19. The prediction model was validated using external datasets. We further utilized the model estimates to forecast subsequent new hospital admissions. Findings A prediction model for symptomatic COVID-19 based on 17 symptoms, age, and sex yielded an area under the ROC curve of 0.78 (95% CI 0.74-0.83) in an external validation dataset of 943 PCR-tested symptomatic individuals. App-based surveillance proved particularly useful for predicting hospital trends in times of insufficient testing capacity and registration delays. During the first wave, our prediction model estimates demonstrated a lower mean error (0.38 average new daily hospitalizations per 100 000 inhabitants per week (95% CI 0.32, 0.45)) for subsequent hospitalizations in the ten most populated counties, than a model based on confirmed case data (0.72 (0.64, 0.81)). The model further correctly identified on average three out of five counties (95% CI 2.3, 3.7) with the highest rates of hospitalizations the following week during the first wave and four out of five (3.0, 4.6) during the second wave. Interpretation The experience of the COVID Symptom Study highlights the important role citizens can play in real-time monitoring of infectious diseases, and how app-based data collection may be used for data-driven rapid responses to public health challenges.

Anosmia and other SARS-CoV-2 positive test-associated symptoms, across three national, digital surveillance platforms as the COVID-19 pandemic and response unfolded: an observation study (preprint)

Background: Multiple participatory surveillance platforms were developed across the world in response to the COVID-19 pandemic, providing a real-time understanding of community-wide COVID-19 epidemiology. During this time, testing criteria broadened and healthcare policies matured. We sought to test whether there were consistent associations of symptoms with SARS-CoV-2 test status across three national surveillance platforms, during periods of testing and policy changes, and whether inconsistencies could better inform our understanding and future studies as the COVID-19 pandemic progresses. Methods: Four months (1st April 2020 to 31st July 2020) of observation through three volunteer COVID-19 digital surveillance platforms targeting communities in three countries (Israel, United Kingdom, and United States). Logistic regression of self-reported symptom on self-reported SARS-CoV-2 test status (or test access), adjusted for age and sex, in each of the study cohorts. Odds ratios over time were compared to known changes in testing policies and fluctuations in COVID-19 incidence. Findings: Anosmia/ageusia was the strongest, most consistent symptom associated with a positive COVID-19 test, based on 658325 tests (5% positive) from over 10 million respondents in three digital surveillance platforms using longitudinal and cross-sectional survey methodologies. During higher-incidence periods with broader testing criteria, core COVID-19 symptoms were more strongly associated with test status. Lower incidence periods had, overall, larger confidence intervals. Interpretation: The strong association of anosmia/ageusia with self-reported SARS-CoV-2 test positivity is omnipresent, supporting its validity as a reliable COVID-19 signal, regardless of the participatory surveillance platform or testing policy. This analysis highlights that precise effect estimates, as well as an understanding of test access patterns to interpret differences, are best done only when incidence is high. These findings strongly support the need for testing access to be as open as possible both for real-time epidemiologic investigation and public health utility.

Detecting COVID-19 infection hotspots in England using large-scale self-reported data from a mobile application (preprint)

Background As many countries seek to slow the spread of COVID-19 without reimposing national restrictions, it has become important to track the disease at a local level to identify areas in need of targeted intervention. Methods We performed modelling on longitudinal, self-reported data from users of the COVID Symptom Study app in England between 24 March and 29 September, 2020. Combining a symptom-based predictive model for COVID-19 positivity and RT-PCR tests provided by the Department of Health we were able to estimate disease incidence, prevalence and effective reproduction number. Geographically granular estimates were used to highlight regions with rapidly increasing case numbers, or hotspots. Findings More than 2.6 million app users in England provided 115 million daily reports of their symptoms, and recorded the results of 170,000 PCR tests. On a national level our estimates of incidence and prevalence showed similar sensitivity to changes as two national community surveys: the ONS and REACT studies. On a geographically granular level, our estimates were able to highlight regions before they were subject to local government lockdowns. Between 12 May and 29 September we were able to flag between 35-80% of regions appearing in the Government's hotspot list. Interpretation Self-reported data from mobile applications can provide a cost-effective and agile resource to inform a fast-moving pandemic, serving as an independent and complementary resource to more traditional instruments for disease surveillance.

Attributes and predictors of Long-COVID: analysis of COVID cases and their symptoms collected by the Covid Symptoms Study App (preprint)

Reports of "Long-COVID", are rising but little is known about prevalence, risk factors, or whether it is possible to predict a protracted course early in the disease. We analysed data from 4182 incident cases of COVID-19 who logged their symptoms prospectively in the COVID Symptom Study app. 558 (13.3%) had symptoms lasting >28 days, 189 (4.5%) for >8 weeks and 95 (2.3%) for >12 weeks. Long-COVID was characterised by symptoms of fatigue, headache, dyspnoea and anosmia and was more likely with increasing age, BMI and female sex. Experiencing more than five symptoms during the first week of illness was associated with Long-COVID, OR=3.53 [2.76;4.50]. Our model to predict long-COVID at 7 days, which gained a ROC-AUC of 76%, was replicated in an independent sample of 2472 antibody positive individuals. This model could be used to identify individuals for clinical trials to reduce long-term symptoms and target education and rehabilitation services.

Symptom clusters in Covid19: A potential clinical prediction tool from the COVID Symptom study app (preprint)

As no one symptom can predict disease severity or the need for dedicated medical support in COVID-19, we asked if documenting symptom time series over the first few days informs outcome. Unsupervised time series clustering over symptom presentation was performed on data collected from a training dataset of completed cases enlisted early from the COVID Symptom Study Smartphone application, yielding six distinct symptom presentations. Clustering was validated on an independent replication dataset between May 1- May 28th, 2020. Using the first 5 days of symptom logging, the ROC-AUC of need for respiratory support was 78.8%, substantially outperforming personal characteristics alone (ROC-AUC 69.5%). Such an approach could be used to monitor at-risk patients and predict medical resource requirements days before they are required.

Key predictors of attending hospital with COVID19: An association study from the COVID Symptom Tracker App in 2,618,948 individuals (preprint)

Objectives: We aimed to identify key demographic risk factors for hospital attendance with COVID-19 infection. Design: Community survey Setting: The COVID Symptom Tracker mobile application co-developed by physicians and scientists at Kings College London, Massachusetts General Hospital, Boston and Zoe Global Limited was launched in the UK and US on 24th and 29th March 2020 respectively. It captured self-reported information related to COVID-19 symptoms and testing. Participants: 2,618,948 users of the COVID Symptom Tracker App. UK (95.7%) and US (4.3%) population. Data cut-off for this analysis was 21st April 2020. Main outcome measures: Visit to hospital and for those who attended hospital, the need for respiratory support in three subgroups (i) self-reported COVID-19 infection with classical symptoms (SR-COVID-19), (ii) self-reported positive COVID-19 test results (T-COVID-19), and (iii) imputed/predicted COVID-19 infection based on symptomatology (I-COVID-19). Multivariate logistic regressions for each outcome and each subgroup were adjusted for age and gender, with sensitivity analyses adjusted for comorbidities. Classical symptoms were defined as high fever and persistent cough for several days. Results: Older age and all comorbidities tested were found to be associated with increased odds of requiring hospital care for COVID-19. Obesity (BMI >30) predicted hospital care in all models, with odds ratios (OR) varying from 1.20 [1.11; 1.31] to 1.40 [1.23; 1.60] across population groups. Pre-existing lung disease and diabetes were consistently found to be associated with hospital visit with a maximum OR of 1.79 [1.64,1.95] and 1.72 [1.27; 2.31]) respectively. Findings were similar when assessing the need for respiratory support, for which age and male gender played an additional role. Conclusions: Being older, obese, diabetic or suffering from pre-existing lung, heart or renal disease placed participants at increased risk of visiting hospital with COVID-19. It is of utmost importance for governments and the scientific and medical communities to work together to find evidence-based means of protecting those deemed most vulnerable from COVID-19. Trial registration: The App Ethics have been approved by KCL ethics Committee REMAS ID 18210, review reference LRS-19/20-18210

Geo-social gradients in predicted COVID-19 prevalence and severity in Great Britain: results from 2,266,235 users of the COVID-19 Symptoms Tracker app (preprint)

Understanding the geographical distribution of COVID-19 through the general population is key to the provision of adequate healthcare services. Using self-reported data from 2,266,235 unique GB users of the COVID Symptom Tracker app, we find that COVID-19 prevalence and severity became rapidly distributed across the UK within a month of the WHO declaration of the pandemic, with significant evidence of "urban hot-spots". We found a geo-social gradient associated with disease severity and prevalence suggesting resources should focus on urban areas and areas of higher deprivation. Our results demonstrate use of self-reported data to inform public health policy and resource allocation.

Loss of smell and taste in combination with other symptoms is a strong predictor of COVID-19 infection (preprint)

ImportanceA strategy for preventing further spread of the ongoing COVID-19 epidemic is to detect infections and isolate infected individuals without the need of extensive bio-specimen testing. ObjectivesHere we investigate the prevalence of loss of smell and taste among COVID-19 diagnosed individuals and we identify the combination of symptoms, besides loss of smell and taste, most likely to correspond to a positive COVID-19 diagnosis in non-severe cases. DesignCommunity survey. Setting and ParticipantsSubscribers of RADAR COVID-19, an app that was launched for use among the UK general population asking about COVID-19 symptoms. Main ExposureLoss of smell and taste. Main Outcome MeasuresCOVID-19. ResultsBetween 24 and 29 March 2020, 1,573,103 individuals reported their symptoms via the app; 26% reported suffering from one or more symptoms of COVID-19. Of those, n=1702 reported having had a RT-PCR COVID-19 test and gave full report on symptoms including loss of smell and taste; 579 were positive and 1123 negative. In this subset, we find that loss of smell and taste were present in 59% of COVID-19 positive individuals compared to 18% of those negative to the test, yielding an odds ratio (OR) of COVID-19 diagnosis of OR[95%CI]=6.59[5.25; 8.27], P= 1.90x10-59. We also find that a combination of loss of smell and taste, fever, persistent cough, fatigue, diarrhoea, abdominal pain and loss of appetite is predictive of COVID-19 positive test with sensitivity 0.54[0.44; 0.63], specificity 0.86[0.80; 0.90], ROC-AUC 0.77[0.72; 0.82] in the test set, and cross-validation ROC-AUC 0.75[0.72; 0.77]. When applied to the 410,598 individuals reporting symptoms but not formally tested, our model predicted that 13.06%[12.97%;13.15] of these might have been already infected by the virus. Conclusions and RelevanceOur study suggests that loss of taste and smell is a strong predictor of having been infected by the COVID-19 virus. Also, the combination of symptoms that could be used to identify and isolate individuals includes anosmia, fever, persistent cough, diarrhoea, fatigue, abdominal pain and loss of appetite. This is particularly relevant to healthcare and other key workers in constant contact with the public who have not yet been tested for COVID-19. Key pointsO_ST_ABSWhat is already known on this topicC_ST_ABSO_LIThe spread of COVID-19 can be reduced by identifying and isolating infected individuals but it is not possible to test everyone and priority has been given in most countries to individuals presenting symptoms of the disease. C_LIO_LICOVID-19 symptoms, such as fever, cough, aches, fatigue are common in many other viral infections C_LIO_LIThere is therefore a need to identify symptom combinations that can rightly pinpoint to infected individuals C_LI What this study addsO_LIAmong individuals showing symptoms severe enough to be given a COVID-19 RT-PCR test in the UK the prevalence of loss of smell (anosmia) was 3-fold higher (59%) in those positive to the test than among those negative to the test (18%). C_LIO_LIWe developed a mathematical model combining symptoms to predict individuals likely to be COVID-19 positive and applied this to over 400,000 individuals in the general population presenting some of the COVID-19 symptoms. C_LIO_LIWe find that [~]13% of those presenting symptoms are likely to have or have had a COVID-19 infection. The proportion was slightly higher in women than in men but is comparable in all age groups, and corresponds to 3.4% of those who filled the app report. C_LI