International Journal of Obesity www.nature.com/ijo ARTICLE OPEN Epidemiology and Population Health Characteristics and outcomes of 627 044 COVID-19 patients living with and without obesity in the United States, Spain, and the United Kingdom Martina Recalde1,2,36, Elena Roel 1,2,36, Andrea Pistillo 1, Anthony G. Sena 3,4, Albert Prats-Uribe 5, Waheed-Ul-Rahman Ahmed6,7, Heba Alghoul8, Thamir M. Alshammari 9, Osaid Alser10, Carlos Areia 11, Edward Burn1,5, Paula Casajust12, Dalia Dawoud13, Scott L. DuVall14,15, Thomas Falconer16, Sergio Fernández-Bertolín1, Asieh Golozar17,18, Mengchun Gong19, Lana Yin Hui Lai20, Jennifer C. E. Lane6, Kristine E. Lynch14,15, Michael E. Matheny21,22, Paras P. Mehta23, Daniel R. Morales24, Karthik Natarjan16,25, Fredrik Nyberg26, Jose D. Posada27, Christian G. Reich28, Peter R. Rijnbeek4, Lisa M. Schilling29, Karishma Shah6, Nigam H. Shah27, Vignesh Subbian30, Lin Zhang31,32, Hong Zhu33,34, Patrick Ryan3,16, Daniel Prieto-Alhambra4,5, ✉ Kristin Kostka28,35,37 and Talita Duarte-Salles1,37 © The Author(s) 2021 BACKGROUND: A detailed characterization of patients with COVID-19 living with obesity has not yet been undertaken. We aimed to describe and compare the demographics, medical conditions, and outcomes of COVID-19 patients living with obesity (PLWO) to those of patients living without obesity. METHODS: We conducted a cohort study based on outpatient/inpatient care and claims data from January to June 2020 from Spain, the UK, and the US. We used six databases standardized to the OMOP common data model. We defined two non-mutually exclusive cohorts of patients diagnosed and/or hospitalized with COVID-19; patients were followed from index date to 30 days or death. We report the frequency of demographics, prior medical conditions, and 30-days outcomes (hospitalization, events, and death) by obesity status. RESULTS: We included 627 044 (Spain: 122 058, UK: 2336, and US: 502 650) diagnosed and 160 013 (Spain: 18 197, US: 141 816) hospitalized patients with COVID-19. The prevalence of obesity was higher among patients hospitalized (39.9%, 95%CI: 39.8−40.0) than among those diagnosed with COVID-19 (33.1%; 95%CI: 33.0−33.2). In both cohorts, PLWO were more often female. Hospitalized PLWO were younger than patients without obesity. Overall, COVID-19 PLWO were more likely to have prior medical conditions, present with cardiovascular and respiratory events during hospitalization, or require intensive services compared to COVID-19 patients without obesity. CONCLUSION: We show that PLWO differ from patients without obesity in a wide range of medical conditions and present with more severe forms of COVID-19, with higher hospitalization rates and intensive services requirements. These findings can help guiding preventive strategies of COVID-19 infection and complications and generating hypotheses for causal inference studies. International Journal of Obesity; https://doi.org/10.1038/s41366-021-00893-4 INTRODUCTION similarities with other viral infections such as seasonal influenza or Obesity is associated with increased mortality and is a well-known H1N1, people with obesity were soon labeled as “at-risk” risk factor for chronic conditions, such as diabetes, hypertension, individuals [5]. Since obesity is a worldwide public health priority, cardiovascular disease, and cancer [1, 2]. Due to its proinflamma- granular information on patients with COVID-19 and obesity is tory state that impairs the immune response, obesity has also needed to guide preventive strategies as well as to generate been related to an increased risk of viral infections [3]. The novel hypotheses for etiological studies [6]. coronavirus disease 2019 (COVID-19), caused by the severe acute A review and meta-analysis of 75 studies reported that obesity respiratory syndrome coronavirus 2 (SARS-CoV-2), emerged in is a risk factor for testing positive for SARS-CoV-2, for severe December 2019 in Wuhan, China, and rapidly spread around the COVID-19 and for COVID-19 related mortality [7]. While undoubt- world [4]. This new virus causes a respiratory tract infection with edly relevant to the field, these studies mainly focused on clinical manifestations ranging from asymptomatic/mild symp- exploring multiple risk factors related to COVID-19 and thus did toms to severe illness requiring intensive services. Partly due to its not offer a detailed characterization of patients with COVID-19 A full list of author affiliations appears at the end of the paper. Received: 23 November 2020 Revised: 7 June 2021 Accepted: 24 June 2021 M. Recalde et al. 2 living with obesity. For instance, an exhaustive description of the ever-recorded obesity diagnosis (Supplementary Appendix 3) and/or a medical conditions and COVID-19 related outcomes, such as body mass index (BMI) measurement between 30 and 60 kg/m2 and/or a thromboembolic events, among these patients is lacking. Other bodyweight measurement between 120 and 200 kg prior or at index date. current limitations include the susceptibility to collider bias of We included upper cut-off thresholds to discard implausible observations. studies reporting “risk factors” of COVID-19 infection and Patients without obesity were those who did not fulfill the obesity definition. progression due to sampling mechanisms (e.g., subsamples of tested or hospitalized populations) [8]. A large characterization study focussing exclusively on patients with COVID-19 living with Baseline characteristics and outcomes of interest obesity using real-world data from different health settings and Demographics (sex and age) were obtained at the index date. More than countries could address the limitations of the previous evidence. 15 000 medical conditions from up to one year prior to the index date In this study, we aimed to describe and compare the were identified based on the Systematized Nomenclature of Medicine demographics, medical conditions, and outcomes of COVID-19 (SNOMED) hierarchy, with all descendant codes included [15]. Specific definitions for comorbidities of particular interest were created; the patients living with obesity (PLWO) to those of COVID-19 patients detailed definitions of these variables can be consulted in Supplementary living without obesity, in inpatient or outpatient settings. Appendix 3. We reported here a list of key comorbidities based on their prevalence in the cohorts of the participating sites, as well as on their clinical relevance to obesity and the COVID-19 research field [17]. METHODS Our 30-day outcomes of interest for the diagnosed cohort were Study design, setting, and data sources hospitalization and fatality. For the hospitalized cohort, the 30-days We conducted a multinational cohort study using routinely collected outcomes were a requirement of intensive services (IS) (identified by a healthcare data from January to June 2020 from Spain, the United recorded mechanical ventilation and/or a tracheostomy and/or extra- Kingdom (UK), and the United States (US). This study was part of the corporeal membrane oxygenation procedure), respiratory, cardiovascular, “Characterizing Health Associated Risks, and Your Baseline Disease In thromboembolic, and other events and fatality. SARS-COV-2 (CHARYBDIS)” study (protocol available for download at https://www.ohdsi.org/wp-content/uploads/2020/07/Protocol_COVID- 19-Charybdis-Characterisation_V5.docx) designed by the Observational Data analysis Health Data Sciences and Informatics (OHDSI) community. All data were We described the number of patients included and the prevalence of 1234567890();,: standardized to the Observational Medical Outcomes Partnership obesity in each database as well as the demographics, comorbidities, and (OMOP) Common Data Model (CDM) [9]. The OHDSI network maintains outcomes as proportions (calculated by the number of persons within a the OMOP-CDM, along with a wide range of tools developed by its given category, divided by the total number of persons) with their members to facilitate analyses of mapped data [10]. Data results for this respective 95% confidence intervals (CIs) for each database, by obesity study were extracted on July, 16th, 2020. status. To calculate these proportions in each database, we established a We included primary, outpatient and inpatient care data from electronic minimum count required (of five individuals), to minimize the risk of re- health records (EHRs) and health insurance claims data from six databases. identification of patients. To compare medical conditions across groups, Data from Spain included the Information System for Research in Primary we calculated standardized mean differences (SMDs) [18], which we Care (SIDIAP), which includes primary linked to inpatient care data summarized in Manhattan-style plots. The SMD can be used to compare covering approximately 80% of the population in Catalonia, Spain [11]. The the prevalence of a dichotomous variable between two groups and is UK data covered the Clinical Practice Research Datalink (CPRD), with independent of sample size [19]. A |SMD| > 0.1 indicates a meaningful patients from over 600 general practices in the UK [12]. Data from the US difference in the prevalence of a given condition; in the context of this included: Columbia University Irving Medical Center (CUIMC), covering study, a SMD > 0.1 indicates a higher prevalence in PLWO, whereas a New York-Presbyterian Hospital and its affiliated physician practices; IQVIA SMD < −0.1 indicates a higher prevalence among patients without obesity. Open Claims, which are pre-adjudicated claims collected from office-based This study was descriptive by nature and, therefore, statistical modeling physicians and specialists covering over 300 million lives (~80% of the US was out of scope. Differences across the groups compared should not be population); the Stanford Medicine Research Data Repository (STARR- interpreted as causal effects. OMOP), with data from Stanford Health Care [13], and the United States To ensure data privacy at all times, we employed a federated analysis Department of Veterans Affairs (VA-OMOP), covering the national approach [16]. Following a pre-specified analysis plan, a common Department of Veterans Affairs health care system which serves more analytical code for the whole CHARYBDIS study was developed for the than 9 million enrolled Veterans (of whom 93% are male). A more detailed OHDSI Methods library, available at https://github.com/ohdsi-studies/ description of the included data sources is available in Supplementary Covid19CharacterizationCharybdis, and was run locally in each data- Appendix 1. base. Individual-level data remained within host institutions and only aggregate results from each database were provided to the research team and publicly shared. All the results reported in this paper and Study participants additional data are available for consultation at a dynamic and We included two non-mutually exclusive cohorts of patients: (1) all patients interactive website, which changes over time as new databases are diagnosed with COVID-19 (clinical diagnosis and/or positive test for SARS- added and/or results are updated to CHARYBDIS (https://data.ohdsi.org/ CoV-2), and (2) all patients hospitalized with a COVID-19 diagnosis. We Covid19CharacterizationCharybdis/). considered clinical diagnoses in the definition of COVID-19 cases due to We used R version 3.6 for data visualization. All the data partners testing restrictions during the first months of the pandemic (e.g., in Spain) obtained Institutional Review Board (IRB) approval or exemption to [14]. The diagnostic codes used are described in Supplementary Appendix conduct this descriptive study. 2. Patients hospitalized with COVID-19 were identified as those having a hospitalization episode along with a clinical diagnosis or positive SARS- CoV-2 test within a time window from 21 days prior to admission up to the end of their hospitalization. We chose this time window to include patients RESULTS with a diagnosis prior to hospitalization and to allow for a record delay in Prevalence of obesity test results or diagnoses [15]. We included individuals with at least one We included 627 044 diagnosed and 160 013 hospitalized patients year of observation time prior to the index date to capture observed with COVID-19 (Table 1). The diagnosed cohort consisted of 122 baseline characteristics. In the diagnosed cohort, the index date was 058 patients from Spain (SIDIAP), 2336 from the UK (CPRD), and defined as the date of the COVID-19 clinical diagnosis or the earliest test 502 650 from the US (CUIMC: 8519; IQVIA-OpenClaims: 466 191; day registered within seven days of a first positive test, whichever occurred STARR-OMOP: 3328; VA-OMOP: 24 612). The hospitalized cohort first. In the hospitalized cohort, the index date was the day of hospitalization. Patients were followed from the index date to the earliest included 18 197 patients from Spain (SIDIAP) and 141 816 from of death, end of the observation period, or 30 days [16]. the US (CUIMC: 2600; IQVIA-OpenClaims: 133 091; STARR-OMOP: Both the diagnosed and hospitalized COVID-19 cohorts were stratified by 615; VA-OMOP: 5510). Among diagnosed and hospitalized patients, obesity status: PLWO vs patients living without obesity (from now on, 207 859 (33.1%; 95%CI: 33.0−33.2) and 63 866 (39.9%, 95%CI: 39.8 referred to as patients without obesity). Obesity was defined as having an −40.0) had obesity, respectively. In all databases, the prevalence International Journal of Obesity Table 1. Demographic characteristics of patients diagnosed and hospitalized with COVID-19 in each database, stratified by obesity status. SIDIAP (Spain) CPRD (UK) CUIMC (US) IQVIA-Open Claims (US) STARR-OMOP (US) VA-OMOP (US) With Without With Without With Without With Without With Without With Without obesity obesity obesity obesity obesity obesity obesity obesity obesity obesity obesity obesity Diagnosed with COVID-19 All, n 36 409 85 649 976 1360 3446 5073 154 325 311 866 1157 2171 11 546 13 066 All, %* (95%CI) 29.8 70.2 41.8 58.2 40.5 59.5 33.1 66.9 34.8 65.2 46.9 53.1 (29.5 (69.9−70.5) (39.8 (56.2−60.2) (39.5 (58.5−60.5) (33.0 (66.8-67) (33.2 (63.6−66.8) (46.3 (52.5−53.7) −30.1) −43.8) −41.5) −33.2) −36.4) −47.5) International Journal of Obesity Laboratory 39.2 27.2 51.5 50.3 66.7 65.5 – – 24.4 28.6 – – confirmed, % (38.9 (27−27.4) (49.5 (48.3−52.3) (65.7 (64.5−66.5) (22.9 (27.1−30.1) (95%CI) −39.5) −53.5) −67.7) −25.9) Female sex, % 62.5 55.9 56.2 57.2 60.9 56.4 61.0 51.7 52.9 52.8 13.2 19.0 (95%CI) (62.2 (55.6−56.2) (54.2 (55.2−59.2) (59.9 (55.3−57.5) (60.9 (51.6-51.8) (51.2 (51.1−54.5) (12.8 (18.5−19.5) −62.8) −58.2) −61.9) −61.1) −54.6) −13.6) Age, % (95%CI) <18 years 1.2 4.7 1.2 3.5 1.0 3.9 (1.1−1.3) (4.6−4.8) (1−1.4) (3.1−3.9) (0.97 (3.8−4) −1.03) 18−64 years 58.9 75.6 55.4 52.3 65.2 65.8 66.7 58.0 66.2 69.6 53.7 58.4 (58.6 (75.4−75.8) (53.4 (50.3−54.3) (64.2 (64.8−66.8) (66.6 (57.9−58.1) (64.6 (68−71.2) (53.1 (57.8−59) −59.2) −57.4) −66.2) −66.8) −67.8) −54.3) >65 years 40.0 19.6 44.7 47.7 33.6 30.6 32.3 38.2 33.8 30.4 46.3 41.6 (39.7 (19.4−19.8) (42.7 (45.7−49.7) (32.6 (29.6−31.6) (32.2 (38.1−38.3) (32.2 (28.8−32) (45.7 (41−42.2) −40.3) −46.7) −34.6) −32.4) −35.4) −46.9) Hospitalized with COVID-19 All, n 8403 9794 – – 1408 1192 50863 82228 274 341 2918 2592 All, % (95%CI) 46.2 53.8 – – 54.2 45.8 38.2 61.8 44.6 55.4 53.0 47.0 (45.5 (53.1−54.5) (52.3 (43.9−47.7) (37.9 (61.5−62.1) (40.7 (51.5−59.3) (51.7 (45.7−48.3) −46.9) −56.1) −38.5) −48.5) −54.3) Laboratory 76.0 72.9 – – 88.5 91.6 10.6 18.3 M. Recalde et al. confirmed, % (75.4 (72.3−73.5) (87.3 (90.5−92.7) (8.2−13) (15.2−21.4) (95%CI) −76.6) −89.7) Female sex, % 51.2 39.9 – – 54.7 39.8 55.0 43.5 51.5 49.0 6.6 4.4 (95%CI) (50.5 (39.2−40.6) (52.8 (37.9−41.7) (54.7 (43.2−43.8) (47.6 (45.0−53.0) (5.9−7.3) (3.9−4.9) −51.9) −56.6) −55.3) −55.4) Age, % (95%CI) <18 years – – – – – – 0.5 2.0 – – – – (0.5−0.5) (1.9−2.1) 18−64 years 37.9 50.6 – – 53.3 34.5 51.7 38.5 64.2 57.2 37.1 27.4 (37.2 (49.9−51.3) (51.4 (32.7−36.3) (51.4−52) (38.2−38.8) (60.4−68) (53.3−61.1) (35.8 (26.2-28.6) −38.6) −55.2) −38.4) >65 years 62.1 49.4 – – 46.7 65.6 47.8 59.5 35.8 42.8 62.9 72.6 (61.4 (48.7−50.1) (44.8 (63.8−67.4) (47.5 (59.2−59.8) (32.0 (38.9−46.7) (61.6 (71.4−73.8) −62.8) −48.6) −48.1) −39.6) −64.2) Notes. * Proportion of patients with and without obesity among all patients (row percentage); - data not available or below the minimum cell count required (five individuals) Abbreviations: CI confidence interval, COVID-19 coronavirus disease 2019, CPRD Clinical Practice Research Datalink, CUIMC Columbia University Irving Medical Center, SIDIAP Information System for Research in Primary Care, STARR-OMOP Stanford Medicine Research Data Repository, UK United Kingdom, US United States, VA-OMOP United States Department of Veterans Affairs. 3 M. Recalde et al. 4 Fig. 1 Distribution of age among patients living with and without obesity in each database, stratified by COVID-19 cohort type (diagnosed and hospitalized). CPRD Clinical Practice Research Datalink, COVID-19 coronavirus disease 2019, CUIMC Columbia University Irving Medical Center, SIDIAP Information System for Research in Primary Care, STARR-OMOP Stanford Medicine Research Data Repository, VA- OMOP United States Department of Veterans Affairs. of obesity was lower among diagnosed patients than among those between PLWO and patients without obesity are available in hospitalized, with differences ranging from 5 (IQVIA-OpenClaims) Supplementary Appendices 4 and 5. In the diagnosed cohorts, to 16% (SIDIAP). PLWO consistently had a higher prevalence of comorbidities compared to those without obesity; these differences were Baseline demographics meaningful (i.e., with a SMD > 0.1, which indicates a meaningfully The sex distribution (proportions and 95% CIs) of the patients are higher prevalence among PLWO) for the majority of comorbidities reported in Table 1. Aside from VA-OMOP, in the diagnosed cohort, across databases. For example, while the prevalence of hyperten- patients with and without obesity were mostly female. The sion for PLWO ranged from 30 to 32% in Europe (SIDIAP and proportion of females was higher among PLWO compared to CPRD) and from 55 to 81% in the US, in those without obesity it patients without obesity in SIDIAP (63% vs 56%), CUIMC (61% vs ranged from 12 to 16% and from 26 to 53%, respectively. The SMD 56%), and IQVIA-OpenClaims (61% vs 52%), while in VA-OMOP the for hypertension was above 0.1 in all databases. As in the opposite was observed (13% vs 19%). No differences were observed diagnosed cohort, PLWO hospitalized with COVID-19 had a higher in CPRD and STARR-OMOP. In the hospitalized cohort, patients prevalence of comorbidities than those without obesity, and these without obesity were predominantly male (female ranged from 40 differences were meaningful for the majority of comorbidities. to 49%, VA-OMOP: 4%) but PLWO still were more commonly female However, the differences between groups were less obvious. For in all databases aside from VA-OMOP (range: 51−55%, VA-OMOP: example, heart disease differed by 20% among those diagnosed in 7%). Differences in the proportion of females between PLWO and VA-OMOP (PLWO: 60%, without obesity: 40%) and by 9% among patients without obesity ranged from 3 (VA-OMOP) to 15% (CUIMC). those hospitalized (PLWO: 74%, without obesity: 65%); although The age distribution in each database is summarized in Table 1 the SMD was still above 0.1 in all databases. with proportions and their respective 95% CIs and in Fig. 1 with histograms. In the diagnosed cohort, PLWO were slightly older 30-day outcomes of interest than those without obesity (i.e., the age distribution for PLWO was The distribution of 30-days outcomes is shown in Fig. 4, the slightly skewed to the left compared to patients without obesity). proportions with their respective 95% CI and SMDs between This was particularly marked in SIDIAP, where 40% of the PLWO PLWO and patients without obesity are available in Table 2. In the were aged above 65 years and only 20% were so without obesity. diagnosed cohorts, hospitalization rates were higher among PLWO Hospitalized patients were older than those diagnosed. In the than among those without obesity in all databases. For example, hospitalized cohorts, PLWO were fairly consistently younger than in SIDIAP the proportion of patients hospitalized was 20% for those without obesity (except for SIDIAP). The proportion of PLWO and 10% for patients without obesity. However, these patients aged above 65 ranged from 36 to 63% for PLWO and differences were meaningful (SMD > 0.1) only in three databases: from 43 to 73% for those without obesity. SIDIAP, CUIMC, and STARR-OMOP. In PLWO, fatality ranged from 5 to 12% and was higher than in patients without obesity in SIDIAP Baseline medical conditions and CUIMC (7% vs 3% and 8% vs 5%, respectively), while in CPRD We compared baseline medical conditions of PLWO to those of and VA-OMOP it was similar in both groups. SIDIAP was the only patients without obesity in the diagnosed and hospitalized cohorts database with a meaningful difference in the proportion of fatality. using SMDs, which are summarized in Fig. 2. We depicted the Overall, in the hospitalized cohort, PLWO more frequently had SMDs of 485 (CPRD) to 5050 (VA-OMOP) medical conditions in the adverse events occurring in the 30 days after the index date than diagnosed cohort, and 529 (STARR-OMOP) to 5240 (IQVIA-Open- patients without obesity. For example, PLWO required IS and Claims) in the hospitalized cohort. In both cohorts, medical presented with ARDS more frequently than patients without conditions were largely more frequent among PLWO than patients obesity in the largest databases: IQVIA-OpenClaims (IS: 13% vs without obesity. 10%; ARDS: 35% vs 31%) and VA-OMOP (IS: 22% vs 15%; 46% vs The distribution of the selected key comorbidities is shown in 41%), whereas in CUIMC and STARR-OMOP percentages were Fig. 3, and the proportions with their respective 95% CIs and SMDs similar. VA-OMOP was the only database with a meaningful International Journal of Obesity M. Recalde et al. 5 Fig. 2 Standardized mean differences in conditions among patients living with obesity compared to patients living without obesity in each database, stratified by COVID-19 cohort type (diagnosed and hospitalized). SMD < 0 means the prevalence was greater in COVID-19 patients living without obesity, SMD > 0 means the prevalence was greater in COVID-19 patients living with obesity. COVID-19 coronavirus disease 2019, CPRD Clinical Practice Research Datalink, CUIMC Columbia University Irving Medical Center, SIDIAP Information System for Research in Primary Care, SMD standardized mean difference, STARR-OMOP Stanford Medicine Research Data Repository, VA-OMOP United States Department of Veterans Affairs. Fig. 3 Comorbidities at baseline among patients living with obesity compared to patients living without obesity in each database, stratified by COVID-19 cohort type (diagnosed and hospitalized). Prevalence of comorbidities for COVID-19 patients living with obesity (red) and without obesity (blue) are depicted in overlapped horizontal bars. The gray color is the overlap between groups. E.g., in CPRD, 32% of COVID-19 patients living with obesity and 16% living without obesity have hypertension. Comorbidities with a meaningful difference (|SMD| > 0.1) between patients living with and without obesity are marked with an asterisk (*). COPD chronic obstructive pulmonary disease, COVID-19 coronavirus disease 2019, CPRD Clinical Practice Research Datalink, CUIMC Columbia University Irving Medical Center, SIDIAP Information System for Research in Primary Care, SMD standardized mean difference, STARR-OMOP Stanford Medicine Research Data Repository, VA- OMOP United States Department of Veterans Affairs. difference in the proportion of IS. Similarly, heart failure was also cardiac arrhythmia, and cardiovascular disease events were more frequent among PLWO than among patients without slightly more frequent among PLWO, although SMDs were below obesity in CUIMC: 7% vs 3%, IQVIA-OpenClaims: 7% vs 5%, 0.1 in all databases. Acute kidney injury was the only outcome STARR-OMOP: 16% vs 9%, and VA-OMOP: 23% vs 17%), these that was more frequent among patients without obesity; differences were meaningful in CUIMC and STARR-OMOP. Sepsis, however, this difference was not meaningful in any database. International Journal of Obesity M. Recalde et al. 6 Fig. 4 A comparison of 30-day events among patients living with and without obesity in each database, by COVID-19 cohort type (diagnosed and hospitalized). Proportion of outcomes for COVID-19 patients living with obesity (red) and without obesity (blue) are depicted in overlapped horizontal bars. The gray color is the overlap between groups. E.g., in the diagnosed cohort, 20 and 10% of patients living with and without obesity in SIDIAP, respectively, were hospitalized. Outcomes with a meaningful difference (|SMD| > 0.1) between patients living with obesity and patients without obesity are marked with an asterisk (*). ARDS acute respiratory distress syndrome, COVID-19 coronavirus disease 2019, CPRD Clinical Practice Research Datalink, CUIMC Columbia University Irving Medical Center, SIDIAP Information System for Research in Primary Care, SMD standardized mean difference, STARR-OMOP Stanford Medicine Research Data Repository, VA-OMOP United States Department of Veterans Affairs. As for fatality, there were no consistent nor meaningful characterization goes far beyond prior studies reporting few differences between PLWO and patients without obesity in the comorbidities and supports the generation of new hypotheses hospitalized cohort: while it was higher for PLWO in SIDIAP (14% that can be tested in future studies. In addition, for the sake of vs 11%), there were no differences in CUIMC (20% vs 21%) nor in transparency and reproducibility, we have made methods, tools, VA-OMOP (16% vs 18%). and all results publicly available. As CHARYBDIS is an ongoing study, results (included longer follow-up time) will be updated and new studies focussing on obesity could be conducted. All of the DISCUSSION above has been accomplished through the coordinated efforts of In this large cohort study including 627 044 COVID-19 patients the OHDSI community to provide a rapid response to the COVID- from Spain, the UK, and the US, we found that the prevalence of 19 pandemic. obesity was higher among COVID-19 patients hospitalized (40%) Our study also has limitations. First, we cannot exclude a compared to those diagnosed (31%). PLWO diagnosed and selection bias of COVID-19 cases due to underreporting in the hospitalized with COVID-19 were more commonly female, and context of testing restrictions and asymptomatic or paucisympto- those hospitalized were younger than patients without obesity. matic cases that usually do not seek medical care. Additionally, The extraction of more than 15 000 medical conditions revealed testing policies have varied across countries and time depending PLWO were not only more prone to have obesity-related on the course of the pandemic. Nevertheless, the inclusion of comorbidities, such as hypertension, heart disease, and type 2 patients clinically diagnosed (not tested) in different settings likely diabetes but also to more than a thousand different health provided consistency to our data, although it might have incurred conditions. After 30-days of follow-up, PLWO presented with in false positives. Second, we did not have information on BMI as a higher hospitalization rates and intensive services requirements, continuous variable, which prevented us from investigating the although these differences were only meaningful in some impact of different categories of obesity in COVID-19 outcomes. databases. This might explain the higher proportion of comorbidities and Our study has several strengths, such as its large amount of outcomes observed in the US databases, as PLWO from the US data. By bringing together harmonized data using a federated might have higher BMIs than those from Europe [20]. In addition, approach, we have conducted a large-scale study while respecting our definition of obesity included diagnoses and measurements the confidentiality of patient records. The international approach recorded at any time prior to or at the index date, and therefore of this study is a strong asset given that we are investigating the some individuals might have been misclassified due to changes in intersection of two major global threats, namely the obesity BMI since the most recently recorded status. However, previous epidemic and the COVID-19 pandemic. The former, together with evidence shows that BMI trajectories in adults are relatively stable, the diverse healthcare settings and populations described in this with a tendency to increase with age [21]. Therefore adults with study, increase the generalizability of our findings. Further, we obesity are likely to still have obesity over time. Finally, this study provide a wide overview of the characteristics and outcomes of was underpinned by routinely collected data which can raise patients with and without obesity, using data visualization tools to concerns about the quality of the data. Some databases are prone summarize large amounts of medical data. This exhaustive to oversampling certain groups of people as a result of how these International Journal of Obesity Table 2. Occurrence of 30-day events, in % (95%CI), among patients living with and without obesity in each database, by COVID-19 cohort type (diagnosed and hospitalized). SIDIAP (Spain) CPRD (UK) CUIMC (US) IQVIA-Open Claims (US) STARR-OMOP (US) VA-OMOP (US) With Without SMD With Without SMD With Without SMD With Without SMD With Without SMD With Without SMD obesity obesity obesity obesity obesity obesity obesity obesity obesity obesity obesity obesity Patients 36 409 854 649 NA 976 1360 NA 3446 5073 NA 154 325 311 866 NA 1157 2171 NA 11 546 13 066 NA diagnosed, n Hospitalization 20.3 10.3 0.20 – – – 36.4 20.2 0.26 32.1 25.8 0.10 22.1 15.8 0.11 16.6 12.8 0.08 (19.9 (10.2 (34.8 (19.1 (31.9 (25.6 (19.7 (14.3 (15.9 (12.2 −20.7) −10.4) −38.0) −21.3) −32.3) −26.0) −24.5) −17.3) −17.3) −13.4) International Journal of Obesity Death 7.1 (6.8 3.3 (3.3 0.12 12.3 11.3 (9.6 0.02 8.0 (7.1 5.0 (4.4 0.09 – – – – – – 5.1 (4.7 4.8 (4.4 0.01 −7.4) −3.3) (10.2 −13.0) −8.9) −5.6) −5.5) −5.2) −14.4) Patients 8403 9794 NA – – NA 1408 1192 NA 50 863 8228 NA 274 341 NA 2918 2592 NA hospitalized, n Intensive service – – – – – – 2.3 (1.5 2.0 (1.2 0.01 13.0 9.7 (9.1 0.06 9.1 (5.7 6.2 (3.6 0.08 21.9 14.5 0.13 requirement −3.1) −2.8) (12.7 −10.3) −12.5) −8.8) (20.4 (13.1 −13.3) −23.4) −15.9) Death 13.9 10.9 0.06 – – – 19.5 20.6 −0.02 – – – – – – 15.9 18.4 −0.05 (13.2 (10.3 (17.4 (18.3 (14.6 (16.9 −14.6) −11.5) −21.6) −22.9) −17.2) −19.9) Cardiovascular events Acute – – – – – – 1.3 (0.7 1.1 (0.5 0.01 2.1 (2.0 2.2 (1.9 −0.01 – – – 5.5 (4.7 6.9 (5.9 −0.03 myocardial −1.9) −1.7) −2.2) −2.5) −6.3) −7.9) infarction Cardiac – – – – – – 9.8 (8.2 9.3 (7.7 0.01 13.1 11.7 0.03 16.8 14.1 0.05 33.4 31.4 0.03 arrhythmia* −11.4) −10.9) (12.8 (11.0 (12.4 (10.4 (31.7 (29.6 −13.4) −12.4) −21.2) −17.8) −35.1) −33.2) Heart failure* – – – – – – 6.5 (5.2 3.1 (2.1 0.11 7.0 (6.8 5.1 (4.6 0.06 15.7 9.1 (6.0− 0.14 22.9 17.4 0.10 −7.8) −4.1) −7.2) −5.6) (11.4 (21.4 (15.9 −20.0) −24.4) −18.9) Stroke – – – – – – 2.6 (1.8 2.5 (1.6 0.00 1.7 (1.6 1.9 (1.6 −0.01 – – – 2.4 (1.8 4.0 (3.2 −0.06 −3.4) −3.4) −1.8) −2.2) −3.0) −4.8) Total – – – – – – 7.3 (5.9 5.9 (4.6 0.04 8.2(8.0 7.6 (7.0 0.02 11.7 8.8 (5.8 0.07 20.6 19.0 0.03 M. Recalde et al. cardiovascular −8.7) −7.2) −8.4) −8.2) (7.9 −11.8) (19.1 (17.5 disease events −15.5) −22.1) −20.5) Thromboembolic events Deep vein – – – – – – 2.1 (1.4 1.8 (1.0 0.02 2.2 (2.1 1.7 (1.4 0.03 – – – 4.9 (4.1 3.8 (3.1 0.04 thrombosis −2.8) −2.6) −2.3) −2.0) −5.7) −4.5) Pulmonary – – – – – – 2.2 (1.4 1.8 (1.0 0.02 1.8 (1.7 1.5 (1.2 0.03 – – – 3.8 (3.1 3.3 (2.6 0.02 embolism −3.0) −2.6) −1.9) −1.8) −4.5) −4.0) Other events Acute kidney – – – – – – 18.1 25.4 −0.01 11.1 10.4 (9.7 0.04 8.4 (5.1 9.1 (6.0 0.08 17.1 18.6 0.08 injury* (16.1 (22.9 (10.8 −11.1) −11.7) −12.2) (15.7 (17.1 −20.1) −27.9) −11.4) −18.5) −20.1) Acute – – – – – – – – – 0.2 (0.2 0.2 (0.1 0.00 – – – 0.4 (0.2 0.7 (0.4 −0.02 pancreatitis −0.2) −0.3) −0.6) −1.0) ARDS* – – – – 17.9 15.2 0.05 34.9 30.5 0.07 14.6 10.0 (6.8 0.10 46.2 41.0 0.07 (15.9 (13.2 (34.5 (29.5 (10.4 −13.2) (44.4 (39.1 −19.9) −17.2) −35.3) −31.5) −18.8) −48.0) −42.9) – – – – – – 0.02 0.00 – – – −0.06 7 M. Recalde et al. 8 data are captured (e.g., the Veterans Affairs system historically Abbreviations: ARDS acute respiratory distress syndrome, CI confidence interval, COVID-19 coronavirus disease 2019, CPRD Clinical Practice Research Datalink, CUIMC Columbia University Irving Medical Center, NA Notes. - data not available or below the minimum cell count required (five individuals); events marked with an * were recorded only during hospitalization. SMD < 0 means the prevalence was greater in COVID- not applicable, SIDIAP Information System for Research in Primary Care, SMD standardized mean difference, STARR-OMOP Stanford Medicine Research Data Repository, UK United Kingdom, US United States, VA- 19 patients living without obesity, SMD > 0 means the prevalence was greater in COVID-19 patients living with obesity. A SMD > |0.1| indicates a meaningful difference in the prevalence of a given condition SMD 0.06 0.04 serves more men than women, routine claims data may only Without reflect health outcomes in commercially insured populations, etc.). obesity Obesity, comorbidities, and outcomes were assessed based on 4.7 (3.9 0.8 (0.5 −23.6) −5.5) −1.1) having a record of a condition/measurement, therefore they may (20.4 22.0 VA-OMOP (US) be underestimated. In addition, outcomes such as hospitalization or intensive services requirements are also influenced by factors obesity 3.1 (2.5 1.6 (1.1 external to the patient’s condition (i.e., bed availability, criteria for −25.9) −3.7) −2.1) With (22.7 24.3 admission), which might differ across databases. Even still, the consistency of our findings across databases that differ by setting and country lends credence to the generalizability of our findings. SMD 0.04 Given the prevalence of obesity in Spain (24%), the UK (27%), – and the US (37%), a high proportion of PLWO among COVID-19 cases was expected [20]. However, the prevalence of obesity Without obesity STARR-OMOP (US) 7.9 (5.0 −10.8) among diagnosed COVID-19 patients was higher than the general population in four databases: SIDIAP (Spain): 30%; CPRD (UK): 42%; – CUIMC and VA-OMOP (US): 41 and 47%, respectively, which is obesity suggestive of an increased risk of diagnosis in PLWO. In addition, 9.5 (6.0 −13.0) With the prevalence of obesity was higher in hospitalized COVID-19 – patients, with an overall prevalence of obesity of 40%, which is in line with three cohort studies from the US that reported that 40, SMD 42, and 48% of inpatients were living with obesity [17, 22, 23]. A 0.00 0.02 large meta-analysis of observational studies reported that obesity IQVIA-Open Claims (US) is associated with a higher risk of testing positive for SARS-CoV-2 Without obesity or being diagnosed with COVID-19 as well as of being hospitalized 1.3 (1.1 0.2 (0.1 −17.0) −1.5) −0.3) (15.4 with COVID-19 [7]. While this could be due to an increased 16.2 vulnerability to SARS-CoV-2 in PLWO, other hypotheses should be considered in future studies. On the one hand, individuals with 1.3 (1.2- obesity 0.2 (0.2 −17.7) obesity could be more likely to seek care and be tested for SARS- −1.4) −0.2) With (17.1 17.4 CoV-2 since they are (presumably) a high-risk population, have multiple comorbidities, and are more prone to respiratory symptoms due to their compromised pulmonary function [2, 7]. −0.02 SMD On the other hand, given the fact that obesity disproportionately – affects disadvantaged populations, potential differential expo- sures across subpopulation groups should also be explored Without obesity 1.0 (0.4 6.0 (4.7 (e.g., differential occupational risks) [2]. −1.6) −7.3) Women predominated among hospitalized patients with – CUIMC (US) obesity, even though obesity rates are similar in both sexes in the three countries [20]. Although male sex is a well-established obesity 1.7 (1.0 0.8 (0.3 5.6 (4.4 −2.4) −1.3) −6.8) risk factor for COVID-19-related hospitalization and death, little is With known about the role of obesity on COVID-19 outcomes stratified by sex [14, 23–25]. Recent studies addressing this issue in SMD secondary analyses have reported inconsistent results. A study – – conducted among UK Biobank participants found that the impact of BMI in COVID-19-related death was higher among females Without obesity compared to men, while others have found a higher effect among males, opposite effects of sex in different age strata or null – – differences [26–29]. Thus, the intersection between sex/gender CPRD (UK) and obesity in relation to COVID-19-outcomes warrants further obesity investigation. Because sex-stratification was beyond the pre- With specified analysis plan of our study, we were unable to report OMOP United States Department of Veterans Affairs. – – our results by sex, which could have provided valuable insights on the matter. We intend, however, to address this issue in upcoming SMD studies from the CHARYBDIS project. – – We also found that hospitalized PLWO were younger than those Without without obesity. Although younger individuals have less risk of obesity infections and complications than older people due to having SIDIAP (Spain) fewer comorbidities and a stronger immune system, this is not the – – case for those with obesity [2, 7, 30–32]. Some authors have obesity postulated that PLWO younger than 60 years could have a greater With risk of severe COVID-19 outcomes [33]. PLWO also had many more – – Table 2 continued comorbidities than patients without obesity. Unsurprisingly, the highest differences were observed in obesity-related conditions, Gastrointestinal Hepatic failure such as hypertension, diabetes, and heart disease, which have been identified as risk factors for severe COVID-19 outcomes bleeding [14, 17, 25, 34, 35]. However, as our findings revealed, PLWO with Sepsis* COVID-19 differ from patients without obesity in a wider range of medical conditions than previously described. Future etiological International Journal of Obesity M. Recalde et al. 9 studies aiming to disentangle the effect of obesity in COVID-19 [Preprint]. 2020 [cited 2020 Aug 20]. Available from: https://arxiv.org/abs/ outcomes should have this information present and consider data- 2003.10534v1 driven techniques to account for confounding, such as propensity 14. Burn E, Tebé C, Fernandez-Bertolin S, Aragon M, Recalde M, Roel E, et al. The score estimation and its adjustment methods [18]. natural history of symptomatic COVID-19 during the first wave in Catalonia. Nat Commun. 2021;12:1–12. Finally, PLWO experienced adverse events more frequently than 15. Burn E, You SC, Sena AG, Kostka K, Abedtash H, Abrahão MTF, et al. Deep phe- those without obesity, particularly hospitalization and the require- notyping of 34,128 adult patients hospitalised with COVID-19 in an international ment of intensive services. Certainly, our results must be interpreted network study. Nat Commun. 2020;11:5009. carefully considering the differences in demographics and comor- 16. Observational Health Data Sciences and Informatics. The Book of OHDSI (Inde- bidities between these groups. Interestingly, in patients hospita- pendently published). 2019. https://ohdsi.github.io/TheBookOfOhdsi/. lized, we did not observe clear differences in fatality between 17. Richardson S, Hirsch JS, Narasimhan M, Crawford JM, McGinn T, Davidson KW, patients with and without obesity. While two meta-analyses et al. Presenting characteristics, comorbidities, and outcomes among 5700 reported that obesity is associated with a higher risk of COVID-19 patients hospitalized with COVID-19 in the New York City area. JAMA. related mortality; other large observational studies from the US and 2020;323:2052–9. 18. Austin PC. An introduction to propensity score methods for reducing the effects the UK using finer categories of BMI only found an association of confounding in observational studies. Multivariate Behav Res. 2011;46: with mortality for morbid obesity (BMIs ≥ 35 kg/m2 or ≥40 kg/m2) 399–424. [7, 25, 28, 29, 34]. Given the scarcity of evidence regarding the 19. Austin PC. Using the standardized difference to compare the prevalence of a frequency of specific adverse events during hospitalization among binary variable between two groups in observational research. Commun Stat PLWO, our findings are of special interest to the field and should be Simul Comput. 2009;38:1228–34. addressed in upcoming etiological studies. 20. World Health Organization. WHO Overweight and obesity. World Health Orga- In this large international cohort, we showed that among nization. 2020. https://www.who.int/data/gho/data/themes/topics/topic-details/ COVID-19 cases, PLWO were more likely to be female, have more GHO/ncd-risk-factors. comorbidities, and worse outcomes than patients without obesity. 21. Kelly SP, Lennon H, Sperrin M, Matthews C, Freedman ND, Albanes D, et al. Body mass index trajectories across adulthood and smoking in relation to prostate The prevalence of obesity was higher among hospitalized patients cancer risks: the NIH-AARP diet and health study. Int J Epidemiol. 2019;48:464–73. with COVID-19 compared to patients diagnosed with COVID-19. 22. Garg S, Kim L, Whitaker M, O’Halloran A, Cummings C, Holstein R, et al. Hospi- Our results may be useful in guiding clinical practice and aid talization rates and characteristics of patients hospitalized with laboratory- future preventative strategies for patients living with obesity, as confirmed coronavirus disease 2019—COVID-NET, 14 states, March 1–30, 2020. well as providing useful data to support subsequent etiological MMWR Morb Mortal Wkly Rep. 2020;69:458–64. studies focussed on obesity and COVID-19. 23. Petrilli CM, Jones SA, Yang J, Rajagopalan H, O’Donnell L, Chernyak Y, et al. Factors associated with hospital admission and critical illness among 5279 people with coronavirus disease 2019 in New York City: prospective cohort study. BMJ. REFERENCES 2020;369:m1966. 1. Di Angelantonio E, Bhupathiraju SN, Wormser D, Gao P, Kaptoge S, de Gonzalez 24. Docherty AB, Harrison EM, Green CA, Hardwick HE, Pius R, Norman L, et al. AB, et al. Body-mass index and all-cause mortality: individual-participant-data Features of 20 133 UK patients in hospital with covid-19 using the ISARIC WHO meta-analysis of 239 prospective studies in four continents. Lancet. Clinical Characterisation Protocol: prospective observational cohort study. BMJ. 2016;388:776–86. 2020;369:m1985. 2. Hu FB. Obesity Epidemiology. New York: Oxford University Press, 2009. 25. Gupta S, Hayek SS, Wang W, Chan L, Mathews KS, Melamed ML, et al. Factors 3. Moser JAS, Galindo-Fraga A, Ortiz-Hernández AA, Gu W, Hunsberger S, Galán- associated with death in critically Ill patients with coronavirus disease 2019 in the Herrera JF. et al.Underweight, overweight, and obesity as independent risk fac- US. JAMA Intern Med. 2020;180:1436–47. tors for hospitalization in adults and children from influenza and other respiratory 26. Anderson MR, Geleris J, Anderson DR, Zucker J, Nobel YR, Freedberg D, et al. Body viruses.Influenza Other Respir Viruses.2019;13:3–9. mass index and risk for intubation or death in SARS-CoV-2 infection: a retro- 4. WHO. WHO Director-General’s opening remarks at the media briefing on COVID- spective cohort study. Ann Intern Med. 2020;173:782–90. 19—11 March 2020. WHO. 2020. https://www.who.int/directorgeneral/speeches/ 27. Sattar N, Ho FK, Gill JM, Ghouri N, Gray SR, Celis-Morales CA, et al. BMI and future detail/who-director-general-s-opening-remarks-at-the-media-briefing-oncovid- risk for COVID-19 infection and death across sex, age and ethnicity: preliminary 19---11-march-2020. findings from UK biobank. Diabetes Metab Syndr Clin Res Rev. 2020;14:1149–51. 5. Louie JK, Acosta M, Samuel MC, Schechter R, Vugia DJ, Harriman K, et al. A novel 28. Tartof SY, Qian L, Hong V, Wei R, Nadjafi RF, Fischer H, et al. Obesity and mortality risk factor for a novel virus: obesity and 2009 pandemic influenza a (H1N1). Clin among patients diagnosed with COVID-19: results from an integrated health care Infect Dis. 2011;52:301–12. organization. Ann Intern Med. 2020;173:773–81. 6. GOV.UK. New obesity strategy unveiled as country urged to lose weight to beat 29. Du Y, Lv Y, Zha W, Zhou N, Hong X. Association of body mass index (BMI) with coronavirus (COVID-19) and protect the NHS. GOV.UK. 2020. https://www.gov.uk/ critical COVID-19 and in-hospital mortality: a dose-response meta-analysis. government/news/new-obesity-strategy-unveiled-as-country-urged-to-lose- Metabolism. 2021;117:154373. weight-to-beat-coronavirus-covid-19-and-protect-the-nhs. 30. Divo MJ, Martinez CH, Mannino DM. Ageing and the epidemiology of multi- 7. Popkin BM, Du S, Green WD, Beck MA, Algaith T, Herbst CH, et al. Individuals with morbidity. Eur Respir J. 2014;44:1055–68. obesity and COVID-19: a global perspective on the epidemiology and biological 31. Simon AK, Hollander GA, McMichael A. Evolution of the immune system in relationships. Obes Rev. 2020;21:e13128. humans from infancy to old age. Proc R Soc B Biol Sci. 2015;282:20143085. 8. Griffith GJ, Morris TT, Tudball MJ, Herbert A, Mancano G, Pike L, et al. Collider bias 32. Muscogiuri G, Pugliese G, Laudisio D, Castellucci B, Barrea L, Savastano S, et al. undermines our understanding of COVID-19 disease risk and severity. Nat The impact of obesity on immune response to infection: plausible mechanisms Commun. 2020;11:1–12. and outcomes. Obes Rev. 2021;22:e13216. 9. Voss EA, Makadia R, Matcho A, Ma Q, Knoll C, Schuemie M, et al. Feasibility and 33. Lighter J, Phillips M, Hochman S, Sterling S, Johnson D, Francois F, et al. Obesity in utility of applications of the common data model to multiple, disparate obser- patients younger than 60 years is a risk factor for COVID-19 hospital admission. vational health databases. J Am Med Informatics Assoc. 2015;22:553–64. Clin Infect Dis. 2020;71:896–7. 10. Hripcsak G, Duke JD, Shah NH, Reich CG, Huser V, Schuemie MJ, et al. Observa- 34. Williamson EJ, Walker AJ, Bhaskaran K, Bacon S, Bates C, Morton CE, et al. Factors tional Health Data Sciences and Informatics (OHDSI): opportunities for observa- associated with COVID-19-related death using OpenSAFELY. Nature. tional researchers. Stud Health Technol Inform. 2015;216:574–8. 2020;584:430–6. 11. Bolíbar B, Fina Avilés F, Morros R, Del Mar Garcia-Gil M, Hermosilla E, Ramos R, 35. Reilev M, Kristensen KB, Pottegård A, Lund LC, Hallas J, Ernst MT, et al. Char- et al. Base de datos SIDIAP: La historia clínica informatizada de Atención Primaria acteristics and predictors of hospitalization and death in the first 11 122 cases como fuente de información para la investigación epidemiológica. Med Clin. with a positive RT-PCR test for SARS-CoV-2 in Denmark: a nationwide cohort. Int J 2012;138:617–21. Epidemiol. 2020;49:1468–81. 12. Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, Staa Tvan, et al. Data resource profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol. 2015;44:827–36. ACKNOWLEDGEMENTS 13. Datta S, Posada J, Olson G, Li W, O’Reilly C, Balraj D, et al. A new paradigm for We would like to acknowledge the patients who suffered from or died of this accelerating clinical data science at Stanford Medicine. arXiv:2003.10534v1 devastating disease, and their families and carers. We would also like to thank the International Journal of Obesity M. Recalde et al. 10 healthcare professionals involved in the management of COVID-19 during these of Regents, grants from Agency for Healthcare Research and Quality, outside the challenging times, from primary care to intensive care units. submitted work; Dr Prieto-Alhambra reports grants and other from AMGEN; grants, non-financial support and other from UCB Biopharma; grants from Les Laboratoires Servier, outside the submitted work; and Janssen, on behalf of IMI-funded EHDEN and AUTHOR CONTRIBUTIONS EMIF consortiums, and Synapse Management Partners have supported training MR, ER, APU, PR, DPA, KK, and TDS conceived and designed the study. SLD, TF, KEL, programs organized by DPA’s department and open for external participants. Ms. MEM, KN, JDP, CGR, NHS, PR, KK, and TDS coordinated data contributions at their Kostka and Dr. Reich report being employees of IQVIA Inc at the time the analysis was respective sites. AP, AGS, TF, SFB, JDP, KK, and TDS analyzed the data; MR, ER, and AP conducted. Dr. Rijnbeek reports grants from Innovative Medicines Initiative, grants produced the figures and tables. MR, ER, EB, DRM, FN, PR, LMS, DPA, KK, and TDS from Janssen Research and Development, during the conduct of the study. Dr. Ryan interpreted the data. MR, ER, and TDS searched the literature and wrote the first draft is an employee of Janssen Research and Development and shareholder of Johnson & with insightful contributions from EB, LYHL, JCEL, DRM, FN, PR, LMS, DPA, and KK. All Johnson. The views expressed are those of the authors and do not necessarily authors contributed to the revision of the first draft, reviewed and approved the final represent the views or policy of the Department of Veterans Affairs or the United version of the paper. States Government. No other relationships or activities that could appear to have influenced the submitted work. FUNDING The European Health Data & Evidence Network has received funding from the ETHICAL APPROVAL Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement No All the data partners received Institutional Review Board (IRB) approval or exemption. 806968. The JU receives support from the European Union’s Horizon 2020 research STARR-OMOP had approval from IRB Panel #8 (RB-53248) registered to Leland and innovation program and EFPIA. This research received partial support from the Stanford Junior University under the Stanford Human Research Protection Program National Institute for Health Research (NIHR) Oxford Biomedical Research Center (HRPP). The use of VA data was reviewed by the Department of Veterans Affairs (BRC), US National Institutes of Health, US Department of Veterans Affairs, Janssen Central IRB, was determined to meet the criteria for exemption under Exemption Research & Development, and IQVIA. The University of Oxford received funding Category 4 (3), and approved for Waiver of HIPAA Authorization. The research was related to this work from the Bill & Melinda Gates Foundation (Investment ID INV- approved by the Columbia University Institutional Review Board as an OHDSI 016201 and INV-019257). APU has received funding from the Medical Research network study. The use of SIDIAP was approved by the Clinical Research Ethics Council (MRC) [MR/K501256/1, MR/N013468/1] and Fundación Alfonso Martín Committee of the IDIAPJGol (project code: 20/070-PCV). The use of CPRD was Escudero (FAME) (APU). VINCI [VA HSR RES 13-457] (SLD, MEM, KEL). JCEL has approved by the Independent Scientific Advisory Committee (ISAC) (protocol number received funding from the Medical Research Council (MR/K501256/1) and Versus 20_059RA2). The use of IQVIA-OpenClaims was exempted from IRB approval. Arthritis (21605). MR is funded by Wereld Kanker Onderzoek Fonds (WKOF), as part of the World Cancer Research Fund International grant program [grant number: 2017/ 1630]. No funders had a direct role in this study. The views and opinions expressed TRANSPARENCY DECLARATION are those of the authors and do not necessarily reflect those of the Clinician Scientist Lead authors affirm that the paper is an honest, accurate, and transparent account of Award program, NIHR, Department of Veterans Affairs or the United States the study being reported; that no important aspects of the study have been omitted; Government, NHS, or the Department of Health, England. and that any discrepancies from the study as planned have been explained. COMPETING INTERESTS All authors have completed the ICMJE uniform disclosure form at www.icmje.org/ DATA SHARING STATEMENT Analyses were performed locally in compliance with all applicable data privacy laws. coi_disclosure.pdf and declare: Mr. Sena reports personal fees from Janssen Research Although the underlying data is not readily available to be shared, authors & Development, outside the submitted work; Dr. DuVall reports grants from Anolinx, contributing to this paper have direct access to the data sources used in this study. LLC, grants from Astellas Pharma, Inc, grants from AstraZeneca Pharmaceuticals LP, All results (e.g., aggregate statistics, not presented at a patient-level with redactions grants from Boehringer Ingelheim International GmbH, grants from Celgene for minimum cell count) are available for public inquiry. These results are inclusive of Corporation, grants from Eli Lilly and Company, grants from Genentech Inc., grants site-identifiers by contributing data sources to enable interrogation of each from Genomic Health, Inc., grants from Gilead Sciences Inc., grants from contributing site. All analytic code and result sets are made available at: https:// GlaxoSmithKline PLC, grants from Innocrin Pharmaceuticals Inc., grants from Janssen github.com/ohdsi-studies/Covid19CharacterizationCharybdis. Pharmaceuticals, Inc., grants from Kantar Health, grants from Myriad Genetic Laboratories, Inc., grants from Novartis International AG, grants from Parexel International Corporation through the University of Utah or Western Institute for Biomedical Research, outside the submitted work; Mr Ahmed reports funding from ADDITIONAL INFORMATION the NIHR Oxford Biomedical Research Center (BRC), Aziz Foundation, Wolfson Supplementary information The online version contains supplementary material Foundation, and the Royal College Surgeons of England; Dr. Golozar reports personal available at https://doi.org/10.1038/s41366-021-00893-4. fees from Regeneron Pharmaceuticals, outside the submitted work. She is a full-time employee at Regeneron Pharmaceuticals. This work was not conducted at Regeneron Correspondence and requests for materials should be addressed to T.D-S. Pharmaceuticals. Miss Lane is supported by a Medical Research Council Doctoral Research Fellowship (MR/K501256/1) and a Versus Arthritis Clinical Research Reprints and permission information is available at http://www.nature.com/ Fellowship (21605). Dr. Morales is supported by a Wellcome Trust Clinical Research reprints Development Fellowship (Grant 214588/Z/18/Z) and reports grants from Chief Scientist Office (CSO), grants from Health Data Research UK (HDR-UK), grants from Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims National Institute of Health Research (NIHR), outside the submitted work; Dr. Nyberg in published maps and institutional affiliations. reports other from AstraZeneca, outside the submitted work; Dr. Subbian reports grants from National Science Foundation, grants from State of Arizona; Arizona Board 1 Fundació Institut Universitari per a la recerca a l’Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGol), Barcelona, Spain. 2Universitat Autònoma de Barcelona, Bellaterra, Spain. 3Janssen Research & Development, Titusville, NJ, USA. 4Department of Medical Informatics, Erasmus University Medical Center, Rotterdam, The Netherlands. 5Centre for Statistics in Medicine, NDORMS, University of Oxford, Oxford, UK. 6Nuffield Department of Orthopaedics, Rheumatology, and Musculoskeletal Sciences, University of Oxford, Botnar Research Centre, Oxford, UK. 7College of Medicine and Health, University of Exeter, St Luke’s Campus, Exeter, UK. 8Faculty of Medicine, Islamic University of Gaza, Gaza, Palestine. 9College of Pharmacy, Riyadh Elm University, Riyadh, Saudi Arabia. 10Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA. 11 Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UK. 12Real-World Evidence, Trial Form Support, Barcelona, Spain. 13Cairo University, Faculty of Pharmacy, Cairo, Egypt. 14VA Informatics and Computing Infrastructure, VA Salt Lake City Health Care System, Salt Lake City, UT, USA. 15Department of Internal Medicine, University of Utah School of Medicine, Salt Lake City, UT, USA. 16Department of Biomedical Informatics, Columbia University, New York, NY, USA. 17Department of Epidemiology, Johns Hopkins School of Public, Baltimore, MD, USA. 18Pharmacoepidemiology, Regeneron Pharmaceuticals, Tarrytown, NY, USA. 19DHC Technologies co, Ltd, Beijing, China. 20Division of Cancer Sciences, School of Medical Sciences, University of Manchester, Manchester, UK. 21Tennessee Valley Healthcare System, Veterans Affairs Medical Center, Nashville, TN, USA. 22Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN, USA. 23College of Medicine, The University International Journal of Obesity M. Recalde et al. 11 of Arizona, Tucson, AZ, USA. 24Division of Population Health and Genomics, University of Dundee, Dundee, UK. 25New York-Presbyterian Hospital, New York, NY, USA. 26 School of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden. 27Department of Medicine, Stanford University, Palo Alto, CA, USA. 28Real World Solutions, IQVIA, Cambridge, MA, USA. 29Data Science to Patient Value Program, Department of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, USA. 30College of Engineering, The University of Arizona, Tucson, AZ, USA. 31School of Population Medicine and Public Health, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China. 32Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, VIC, Australia. 33Institute of Health Management, Southern Medical University, Guangzhou, China. 34Nanfang Hospital, Southern Medical University, Guangzhou, China. 35The OHDSI Center at the Roux Institute, Northeastern University, Portland, ME, USA. 37These authors contributed equally: Martina Recalde, Elena Roel. 37These authors jointly supervised this work: Kristin Kostka, Talita Duarte-Salles. ✉email: tduarte@idiapjgol.org Open Access This article is licensed under a Creative Commons article’s Creative Commons license and your intended use is not permitted by statutory Attribution 4.0 International License, which permits use, sharing, regulation or exceeds the permitted use, you will need to obtain permission directly adaptation, distribution and reproduction in any medium or format, as long as you give from the copyright holder. To view a copy of this license, visit http://creativecommons. appropriate credit to the original author(s) and the source, provide a link to the Creative org/licenses/by/4.0/. Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the © The Author(s) 2021 International Journal of Obesity