key: cord-300324-95fty9yi
authors: Ni Lochlainn, M.; Lee, K. A.; Sudre, C. H.; Varsavsky, T.; Cardoso, M. J.; Menni, C.; Bowyer, R. C. E.; Nguyen, L. H.; Drew, D. A.; Ganesh, S.; Lavigne du Cadet, J.; Visconti, A.; Freydin, M. B.; Modat, M.; Graham, M. S.; Capdevila Pujol, J.; Murray, B.; El-Sayed Moustafa, J. S.; Zhang, X.; Davies, R.; Falchi, M.; Spector, T. D.; Chan, A. T.; Ourselin, S.; Steves, C. J.
title: Key predictors of attending hospital with COVID19: An association study from the COVID Symptom Tracker App in 2,618,948 individuals
date: 2020-04-29
journal: nan
DOI: 10.1101/2020.04.25.20079251
sha: 
doc_id: 300324
cord_uid: 95fty9yi

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

As the coronavirus disease 2019 (COVID-19) pandemic escalates and countries struggle to contain the virus, healthcare systems are under increasing pressure as unprecedented numbers require hospitalisation and respiratory support. Emergency departments and intensive care units worldwide are under strain, and medical resources are being diverted to tackle the crisis. There is a pressing need to identify risk factors for severe disease, and particularly to identify key predictors of hospitalisation amongst patients with COVID- 19. In order to address this, we used self-reported data collected on the COVID Symptom Tracker app (1) to identify key demographic risk factors for hospitalisation and the need for respiratory support in the context of COVID-19.

As part of the COronovirus Pandemic Epidemiology (COPE) Consortium (1), the COVID Symptom Tracker smartphone application ("app") co-developed by King's College London, Massachusetts General Hospital, and Zoe Global Limited was launched in the UK on 24th March 2020 and was available in the US beginning 29 th March 2020 (1,2). Individuals without symptoms are encouraged to use the app. It captured self-reported information related to COVID-19 symptoms. On first use, the app records the user's self-reported location, age, and core health risk factors (Table 1) . At this point height and weight are self-reported, allowing calculation of body mass index (BMI). With continued use, participants provide daily updates on symptoms, information on health care visits, COVID-19 testing results, and whether they are seeking healthcare, including the level of intervention and related outcomes.

1

What sex were you assigned at birth? 2

What is your height? 3

What is your weight? 4

Do you have lung disease or asthma? 5 Do you smoke? 6 Do you have heart disease? 7

Do you have diabetes? 8 Do you have kidney disease? 9

What treatment did you receive while in the hospital / What treatment are you receiving right now? a) None b) Oxygen and fluids (breathing support administered through an oxygen mask, no pressure applied c) Non-invasive ventilation Breathing support administered through an oxygen mask, which pushes oxygen into your lungs d) Invasive ventilation; Breathing support administered through an inserted tube. People are usually asleep for this procedure Table 1 . Relevant questions asked in the COVID Symptom Tracker App

Exposure, outcome and covariates were all ascertained via the app. A subset of individuals reported being tested for COVID-19. BMI was calculated as kg/m 2 and considered as a categorical variable (3). Classical symptoms of COVID-19 are defined as "high fever and persistent cough for several days". This was either directly selfreported as an outcome of an enrolment question or if participants reported jointly fever and persistent cough for more than two days. Visit to hospital was recorded if the location was ever self-recorded as hospital or "back from hospital".

Data from the app were downloaded and only records where the self-reported characteristics fell within the following ranges were utilised for further analyses: age between 16 (18 in the US) and 100; BMI between 16 and 55 kg/m 2 .

Separate logistic regression models were fit to predict two different outcomes: 1. Visit to hospital as outcome were fit to test for association between i) self-reported obesity and ii) chronic lung disease and asthma, heart disease, diabetes and kidney disease in the following groups: 1) self-reported COVID-19 infection with classical symptoms (SR-COVID19); 2) self-reported positive COVID-19 test results (T-COVID19); 3) imputed/predicted COVID-19 infection based on symptomatology (I-COVID19) Imputation for testing positive for COVID was performed using the data at day of maximum sum of symptoms and applying a logistic regression using coefficients defined previously (2) . 100 samples from the coefficient distribution were sampled to create the predictions. A participant was considered as positive 50% of the time. Please see Supplementary Table 1 of the coefficients for the logistic regression. 2. The need for respiratory support (oxygen or ventilation) as the outcome to test for similar association with comorbidities in the specified groups was carried out, for participants who had provided information about the treatment received in hospital.

Sign of recovery was defined as a decrease of more than two points in the sum of reported symptoms compared to the maximum sum of symptoms.

The total studied sample comprised 2,618,948 individuals who supplied their self-assessment on the COVID Symptom Tracker app after quality control, including 27.4% males. 16 .3% of the sample reported at least one of the following comorbidities: diabetes, lung, heart or kidney disease. Table 2 summarises the demographic characteristics in the different population groups following the removal of those for whom the outcome (recovery or hospital visit) was still unknown.

SR-COVID-19 Figure 2 illustrates the differrence in age and BMI distribution across the different groups. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. .

Requirement for respiratory support (RS) Amongst those who reported a visit to hospital and then subsequently updated the app with the medical support they received, respiratory support (oxygen and/or ventilation) was reported in 43.0, 23.1 and 19.1% of the cases, for the T-COVID-19, SR-COVID-19 and I-COVID-19 groups respectively. Plotted ORs for needing respiratory support, according to the three defined groups are shown in Figure 3 .

Pre-existing lung disease and diabetes were consistently associated with a higher risk of requiring respiratory support with OR ranging from 1.52 for lung disease in T-COVID19 to 2.03 for diabetes in SR-COVID19. Individuals with obesity were notably more likely to require medical support for the T-COVID19 and I-COVID19 groups with OR of 1. 33 19 , T-COVID-19 and I-COVID-19 respectively. Age was also more predictive of treatment course than of hospital visit with OR around 1.02 in all three groups.

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. .

In this study we found that age, obesity, diabetes and pre-existing lung, renal and cardiac disease, were risk factors for a hospital visit with COVID-19 amongst a large but relatively young, community-based population of app users.

While a number of studies have now reported the increased risk of SARS-CoV-2 in those with diabetes, hypertension and cardiovascular disease (CVD) (4-9), all of which are associated with obesity, the reporting of body mass index (BMI) or obesity has been left out of most initial studies. Only 2 studies have reported BMI in the context of COVID-19 (Table 3 ). The omission of obesity from initial cohort studies is particularly lamentable when one considers that during the H1N1 outbreak ('swine flu'), obesity was found to be an independent risk factor for ventilation, as well as increased morbidity and mortality (10, 11) . Since then, Kwong et al. have published a study that explored the relationship between BMI and seasonal influenza infection using a series of Canada's cross-sectional population-based health surveys covering twelve influenza seasons (12) . Analysis of the retrospective cohort demonstrated that obese people are at greater risk for respiratory hospitalisations during the seasonal flu periods. The World Health Organisation reports that 27.8% of adults in the UK are obese, the third highest rate in Europe. Worldwide, more than 650 million adults are obese (13). Obesity is associated with an increased risk of a wide range of diseases (14) , along with being associated with greater risk of infection, and of developing serious complications of common infections (15) . Importantly, obese individuals are at higher risk of developing Acute Respiratory Distress Syndrome (ARDS) (16), a common complication of SARS-CoV-2 (5).

In the developed world age is now the leading risk factor for many of our most common diseases, including CVD, cancer and neurodegenerative disease (17) . Older adults are most affected by annual influenza outbreaks with 75% of influenza-associated deaths in the US in the 2018-2019 season being adults aged >65 years (18) . Initial data emerging from the countries first affected by the COVID-19 pandemic reveals that older adults are again disproportionately affected, with multiple studies reporting higher mortality, a more severe disease course and longer length of hospital stay amongst the older population (5, (7) (8) (9) 19, 20) . The increased vulnerability of older adults to infection is often attributed to 'immunosenescence', a term which refers to decline in immunity with advancing age (21). Immunosenescence is associated with increased susceptibility to infection (21), re-. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . emergence of latent infections (22) , reduced immune surveillance, contributing to diseases such as cancer (23) and the reduced efficacy of vaccines (24, 25) . Related to this is the concept of 'inflammageing', referring to the increase in systemic inflammation noted in older adults (26) . Thus, the ageing immune system of an average older adult has higher numbers of circulating inflammatory cells but each with reduced ability to fight infection. However, ageing is a heterogeneous process, and some older adults, especially the physically active have immune systems more comparable to younger adults (27) . We show older age as a risk factor for hospitalisation with COVID-19 but also urge more definition and careful consideration by policymakers and healthcare researchers alike. Advice from governments recommending them to stay indoors and "shield" or "cocoon" (28, 29) , due to their increased risk of morbidity and mortality from COVID-19 may be appropriate on average, but more sophisticated approaches (including immunity testing) should be conducted to avoid blanket decision making on the basis of age alone. The number of people over 60 has doubled since 1980, now exceeding 962 million. Careful consideration must be given to older people in the development of diagnostic, therapeutic, rehabilitation and vaccination programs, in order for such programs to maximise autonomy and minimise the burden on older people both from COVID-19 and the measures introduced to manage it.

Research in other respiratory viruses clearly suggests that patients with chronic respiratory diseases, particularly chronic obstructive pulmonary disease (COPD) and asthma, would be at increased risk of SARS-CoV-2 infection, and of more severe presentations of COVID-19 (30, 31) . However, chronic lung disease appears to be under-represented in the literature to date for COVID-19 (32); a similar pattern was seen with SARS. With the exception of one single-centre retrospective SARS study (33) , clinical studies to date that report on comorbidities in either COVID-19 or SARS, reveal little evidence that chronic lung disease is a key predictor for hospitalisation in these groups (Table 4 ). To place this table in context, the World Health Organisation (WHO) estimates COPD prevalence to range between 4-20%, depending on the country, and rising considerably with age (34) .The lower reported prevalence of chronic lung disease in the literature amongst patients diagnosed with COVID-19 is unexplained. It is possible that, in contrast to the diagnosis of diabetes, there was substantial underdiagnosis or poor recognition of pulmonary disease in patients with COVID-19 patients described in early Chinese cohort studies. Notably, the COVID Symptom Tracker App asked participants: "Do you have lung disease or asthma?", and we believe that the group answering yes to this question likely includes a large proportion of individuals who self-reported mild asthma (or a previous history of mild asthma). Such diagnoses would possibly not be deemed sufficiently significant for hospital doctors to document as a relevant comorbidity when reporting patients in cohort studies, and so we caution against overinterpretation of the lung disease data herein.

As previously mentioned, many of the initial studies on COVID-19 reported an increased risk of SARS-CoV-2 in those with diabetes, hypertension and CVD (4-9), In the case of diabetes, COVID-19 has also been associated with developing diabetic ketoacidosis, even amongst those who usually have good glycaemic control (35, 36) . Less well-reported in the context of the pandemic, however, is renal disease. A number of the initial studies from China found that pre-existing renal disease was a risk factor for more severe COVID-19 infection, as summarised in a meta-analysis by Henry and Lippi (37) ; an overall OR of 3.03 (95% CI 1.09 -8.47) was found for the association of chronic kidney disease to severe COVID-19 across 4 early studies. In addition, it is becoming apparent that COVID-19 can cause acute kidney injury (AKI) in COVID-19, and this confers a worse prognosis: a Chinese prospective cohort study (n=701) reported that 44% and 27% of COVID-19 patients had proteinuria and haematuria on admission, respectively, with 5.1% of this cohort developed AKI during their hospital stay (38) . Importantly, this study also reported higher mortality from COVID-19 for those who had preexisting kidney disease. As we learn more about AKI in COVID-19, it is becoming paramount to disentangle whether pre-existing renal disease predisposes patients to significant AKI in COVID-19, as one might expect. While we did find a significantly increased risk of attending hospital for those with renal disease, relatively few reported this condition in our cohort (Table 2) , which may represent a problem with self-reporting bias. More research is needed to examine the effect of pre-existing renal disease on COVID-19 infection.

Despite the literature consistently showing males to be at higher risk for more severe COVID-19 (7, 39, 40) , we did not find gender assigned at birth to be a predictor of attending a hospital in this study. However, male gender was associated with a substantially greater risk of requiring respiratory support, in all groups. Men have been shown to be less likely to utilise medical services than women (41) and large studies have consistently shown women to report greater numbers of physical symptoms than men (41, 42) . Such health behaviours may lead to men having a higher threshold for visiting the hospital with COVID-19 than women, along perhaps with less awareness of early symptoms. This may result in males being more unwell when they do seek medical help and contribute somewhat to the stark difference in mortality between genders (43) (44) (45) .

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . Our sample size is the largest reported to date, adding significant strength to many predictors of hospitalisation or severe COVID-19 that have been identified in smaller cohort studies to date. We hope that our data will guide clinicians to protect their at-risk patient cohorts and policymakers to consider these vulnerable groups when planning and allocating resource, both during this pandemic and for future similar eventualities. Clearly many questions remain, and the mechanisms by which a number of these predictors of hospitalisation are unclear; much research is needed to define these questions, and quickly.

Our study has a number of limitations. First, all the data collected is self-reported, and questions on comorbidities were somewhat simplified to ease reporting at large scale on an app. Both symptoms and test results may be subject to reporting bias. Secondly, the sampling using an app will under-represent individuals without smartphone devices, including older participants, and is likely to under-represent those severely affected by the disease. Additionally, we are reporting visits, rather than admissions, to hospital; we do not know how many visits resulted in an inpatient stay. While we believe that our sampling provides useful information about the risk of most symptomatic infection, it will not provide insight into very severe disease as the most unwell patients may not record hospitalisation due to incapacitation or even death. Additionally, COVID-19 diagnoses, where confirmed by testing, were likely to be based on RT-PCR which is thought to be between 66-80% sensitive for a single test (46, 47) . Another important caveat of note is that the individuals on which the model was trained are highly selected because COVID-19 tests are not performed at random (48) . The participants were tested because they either displayed severe symptoms, were in contact with COVID-19 positive individuals, were healthcare workers or had travelled to an area of particular risk. Additionally, the app captured whether participants had been diagnosed with COVID-19 but did not specifically ask when. Given that symptoms are recorded at the time of data entry, it is possible that some individuals may no longer have been symptomatic from the virus.

The key predictors of hospitalisation in the context of COVID-19 affect many in our society. Lung disease, diabetes, heart disease, advanced age and increased BMI were associated with risk of hospitalisation with COVID. Careful planning of the use of immune testing and contact tracing could be particularly relevant for people in these higher-risk groups. This may help minimise the risk that patients living with comorbidities and older people are disproportionately isolated in the months to come. Finally, the presence of these factors should be regarded as an important factor in future risk stratification models for 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 most vulnerable from COVID-19 in a way which minimises their risk of economic, mental and social implications of isolation.

In the UK, app Ethics has been approved by KCL ethics Committee REMAS ID 18210, review reference LRS-19/20-18210. In the US, the informed consent process was approved by the Partners Human Research Committee (Protocol 2020P000909). All subscribers provided informed consent before submitting responses. The study was registered with clinicaltrials.gov as NCT04331509. . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 29, 2020. . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 29, 2020. Table 4 . Studies which reported on pre-existing lung disease in the context of COVID-19 infection . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 29, 2020. .

Rapid implementation of mobile technology for real-time epidemiology of COVID-19. medRxiv

Loss of smell and taste in combination with other symptoms is a strong predictor of COVID-19 infection. medRxiv

Prevalence of comorbidities in the novel Wuhan coronavirus (COVID-19) infection: a systematic review and meta-analysis

Clinical Characteristics of Coronavirus Disease 2019 in China

Prevalence of Underlying Diseases in Hospitalized Patients with COVID-19: a Systematic Review and Meta-Analysis. Arch Acad Emerg Med

Clinical course and outcomes of critically ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective, observational study

Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

The impact of obesity on the immune response to infection

Obesity and its Implications for COVID-19 Mortality

Obesity and respiratory hospitalizations during influenza seasons in Ontario, Canada: a cohort study

Health impacts of obesity

ARDS in Obese Patients: Specificities and Management

Critical Care

Ageing as a risk factor for disease

Estimated Influenza Illnesses, Medical visits, Hospitalizations, and Deaths in the United States -2018-2019 influenza season

Clinical features of COVID-19 in elderly patients: A comparison with young and middle-aged patients

Epidemiological and Clinical Characteristics of 17 Hospitalized Patients with 2019 Novel Coronavirus Infections Outside Wuhan, China. medRxiv

Immunosenescence: Emerging challenges for an ageing population

Systematic review of incidence and complications of herpes zoster: Towards a global perspective

Features of immunosenescence in women newly diagnosed with breast cancer

Personalized vaccinology: A review. Vaccine

Vaccines for the elderly: Current use and future challenges

Inflammageing: chronic inflammation in ageing, cardiovascular disease, and frailty HHS Public Access

An investigation into the relationship between age and physiological function in highly active older adults

Vulnerable people could need to "cocoon

Social distancing, shielding, and self-isolation

Role of viral respiratory infections in asthma and asthma exacerbations

Do chronic respiratory diseases or their treatment affect the risk of SARS-CoV-2 infection?

Severe acute respiratory syndrome: prognostic implications of chest radiographic findings in 52 patients

COVID 19 infection may cause ketosis and ketoacidosis. Diabetes, Obes Metab

COVID-19 and Diabetes: Knowledge in Progress

Chronic kidney disease is associated with severe coronavirus disease 2019 (COVID-19) infection. International Urology and Nephrology

Kidney disease is associated with in-hospital death of patients with COVID-19

Clinical Characteristics of Coronavirus Disease 2019 in China

Clinical characteristics of 140 patients infected with SARS CoV 2 in Wuhan

Gender differences of symptom reporting and medical health care utilization in the German population

Gender differences in the reporting of physical and somatoform symptoms

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

COVID-19 Surveillance Group. Characteristics of COVID-19 patients dying in Italy: report based on available data on

Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet. 2020/01/28

Correlation of Chest CT and RT-PCR Testing in Coronavirus Disease 2019 (COVID-19) in China: A Report of 1014 Cases

Essentials for Radiologists on COVID-19: An Update-Radiology Scientific Expert Panel

COVID-19: investigation and initial clinical management of possible cases -GOV.UK [Internet

Clinical characteristics and outcomes of 112 cardiovascular disease patients infected by 2019-nCoV. Zhonghua Xin Xue Guan Bing Za Zhi

Preliminary Estimates of the Prevalence of Selected Underlying Health Conditions Among Patients with Coronavirus Disease 2019 -United States

Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study

Clinical and biochemical indexes from 2019-nCoV infected patients linked to viral loads and lung injury. Sci China Life Sci

Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan

Clinical Characteristics of Imported Cases of Coronavirus Disease 2019 (COVID-19) in Jiangsu Province: A Multicenter Descriptive Study

Clinical characteristics of novel coronavirus cases in tertiary hospitals in Hubei Province

Clinical findings in a group of patients infected with the 2019 novel coronavirus (SARS-Cov-2) outside of Wuhan, China: retrospective case series

Clinical features and outcomes of severe acute respiratory syndrome and predictive factors for acute respiratory distress syndrome

Clinical features and short-term outcomes of 144 patients with SARS in the greater Toronto area

A major outbreak of severe acute respiratory syndrome in Hong Kong