ARTICLE https://doi.org/10.1038/s41467-021-21100-y OPEN The natural history of symptomatic COVID-19 during the first wave in Catalonia Edward Burn 1,2, Cristian Tebé3,4, Sergio Fernandez-Bertolin1, Maria Aragon 1, Martina Recalde1,5, Elena Roel 1, Albert Prats-Uribe 2, Daniel Prieto-Alhambra 2 ✉ & Talita Duarte-Salles1 1234567890():,; The natural history of coronavirus disease 2019 (COVID-19) has yet to be fully described. Here, we use patient-level data from the Information System for Research in Primary Care (SIDIAP) to summarise COVID-19 outcomes in Catalonia, Spain. We included 5,586,521 individuals from the general population. Of these, 102,002 had an outpatient diagnosis of COVID-19, 16,901 were hospitalised with COVID-19, and 5273 died after either being diagnosed or hospitalised with COVID-19 between 1st March and 6th May 2020. Older age, being male, and having comorbidities were all generally associated with worse outcomes. These findings demonstrate the continued need to protect those at high risk of poor outcomes, particularly older people, from COVID-19 and provide appropriate care for those who develop symptomatic disease. While risks of hospitalisation and death were lower for younger populations, there is a need to limit their role in community transmission. 1 Fundació Institut Universitari per a la recerca a l’Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGol), Barcelona, Spain. 2 Centre for Statistics in Medicine (CSM), Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS), University of Oxford, Oxford, UK. 3 Biostatistics Unit at Bellvitge Biomedical Research Institute (IDIBELL), L’Hospitalet, Barcelona, Spain. 4 Universitat de Barcelona, Barcelona, Spain. 5 Universitat Autònoma de Barcelona, Bellaterra, Spain. ✉email: daniel.prietoalhambra@ndorms.ox.ac.uk NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y T he natural history of coronavirus disease 2019 (COVID- associations between age, sex, and comorbidities and risks of 19) has yet to be well-established. While a wide range of COVID-19 diagnosis, hospitalisation with COVID-19, and a studies have described outcomes among patients hospita- COVID-19-related death during the first wave of COVID-19 in lised with COVID-191,2, these individuals represent only a frac- Catalonia. It should be stressed that these latter research ques- tion of those who develop symptomatic disease. Similarly, while tions are inferential, assessing the existence of relationships but outcomes for tested populations have also revealed important not the underlying mechanisms or reasons for them8. insights3,4, COVID-19 testing has often been prioritised on the basis of individuals’ symptoms or perceived risk of outcomes, Results making inferences about specific risk factors for progression Study participants and observed outcomes. A total of 5,586,521 among these populations difficult5. participants were included in the study (Fig. 1). Based on the A full description of the natural history of COVID-19 from study eligibility criteria described above, 169,575 individuals were symptomatic to severe disease is needed. Such a description excluded due to a lack of a year of prior history, one for having a requires comprehensive patient-level data that captures incident, COVID-19 positive test prior, 307 for having a prior COVID-19 symptomatic cases from a representative population, with sub- diagnosis, and three for having a COVID-19 hospitalisation prior sequent longitudinal follow-up, and where outcomes such as to the 1st March 2020. A further 1207 individuals who were hospitalisations and mortality, both inside and outside of the hospitalised on the 1st March 2020 and 40,999 who were care hospital setting, can be identified. Moreover, assessing the impact home residents were also excluded (a flow chart is provided in of chronic health conditions on the course of the disease requires Supplementary Fig. 1). The characteristics of the final study comprehensive data on patients’ medical histories. Linked real population are summarised in Table 1. A histogram of the age world data from countries with universal taxpayer-funded pri- distribution of the study population can be seen in Fig. 2, with a mary care-based health systems where general practitioners are summary of comorbidities shown in Fig. 3. the first point of contact for care, and where this role has been Out of the included study population, 102,002 went on to maintained in the response to COVID-19, provide a unique have an outpatient diagnosis of COVID-19 (67-day cumulative opportunity for this purpose. incidence: 1.83%). Of those diagnosed in outpatient settings, In Spain, one of the European countries worst hit by the 2581 died with COVID-19 without being hospitalised (45-day COVID-19 pandemic, primary care has continued to play an cumulative incidence: 3.01%). In total, 16,901 individuals had a important role in the response to the disease. In Catalonia, an hospitalisation with COVID-19, with 8860 of them having autonomous region of Spain with a devolved health system, more previously had an outpatient diagnosis (45-day cumulative than 120,000 outpatient cases of COVID-19 were diagnosed incidence: 9.06%). Of those hospitalised, 2692 had died with between 15th March and 24th April 20206. Meanwhile, a COVID-19 by the end of follow-up (45-day cumulative nationwide seroprevalence study conducted between 27th April incidence: 19.16%), see Fig. 1 and Table 2. An extensive to 11th May, 2020 found the prevalence of IgG antibodies to be summary of the observed data can be seen alongside the study around 5% in Spain, and 7% in Catalonia7. Consequently, despite code at https://github.com/SIDIAP/MultiStateCovid-19. the burden of disease already experienced, there is a need to better understand the features of COVID-19 so as to inform the continued regional, national, and global responses to it. The association between age and risk of COVID-19 diagnosis, In this study, our first aim was to summarise COVID-19 hospitalisation, and death. Age profiles varied by transition, see related outcomes in Catalonia as experienced during the first Table 1. While the median age of the study population as a wave of the pandemic. Subsequently, we aimed to describe the whole was 43 (interquartile range: 25–60), those diagnosed with Fig. 1 Multi-state model of COVID-19. The entire study population began in the general population state as of the 1st March 2020, with progression through the model possible up to 6th May 2020. 2 NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y ARTICLE Table 1 Patient characteristics of the study population and by multistate transition. From general population From diagnosed with COVID-19 From hospitalised with COVID-19 General To diagnosed To hospitalised To death To hospitalised To death To death population with COVID-19 with COVID-19 with COVID-19 N 5,586,521 102,002 8041 11,327 8860 2581 2692 Days since index 24 [19.0, 32.0] 33 26 [21.0, 35.0] date (median [IQR]) [27.0, 40.0] Age (median [IQR]) 43 [25.0, 60.0] 46 [35.0, 58.0] 71 [58.0, 80.0] 85 60 [49.0, 73.0] 86 80 [73.0, 86.0] [76.0, 90.0] [80.0, 91.0] Age gr (%) Under 18 966,680 (17.3) 4536 (4.4) 34 (0.4) 9 (0.1) 40 (0.5) <5 <5 18–39 1,437,297 (25.7) 30,530 (29.9) 426 (5.3) 80 (0.7) 850 (9.6) 5 (0.2) 11 (0.4) 40–59 1,785,852 (32.0) 44,493 (43.6) 1776 (22.1) 657 (5.8) 3441 (38.8) 63 (2.4) 142 (5.3) 60–69 615,538 (11.0) 10,305 (10.1) 1558 (19.4) 1026 (9.1) 1673 (18.9) 131 (5.1) 292 (10.8) 70–79 468,662 (8.4) 6120 (6.0) 2162 (26.9) 1967 (17.4) 1602 (18.1) 406 (15.7) 822 (30.5) 80 or older 312,492 (5.6) 6018 (5.9) 2085 (25.9) 7588 (67.0) 1254 (14.2) 1976 (76.6) 1424 (52.9) Sex = female (%) 2,831,062 (50.7) 59,006 (57.8) 3439 (42.8) 5672 (50.1) 3989 (45.0) 1467 (56.8) 1075 (39.9) Region (%) Barcelona 4,169,520 (74.6) 82,702 (81.1) 7201 (89.6) 8694 (76.8) 7604 (85.8) 2164 (83.8) 2393 (88.9) Girona 557,006 (10.0) 9929 (9.7) 432 (5.4) 942 (8.3) 657 (7.4) 217 (8.4) 158 (5.9) Lleida 370,570 (6.6) 5621 (5.5) 211 (2.6) 728 (6.4) 391 (4.4) 108 (4.2) 94 (3.5) Tarragona 489,378 (8.8) 3748 (3.7) 196 (2.4) 958 (8.5) 207 (2.3) 92 (3.6) 45 (1.7) Missing 47 (0.0) <5 <5 5 (0.0) <5 <5 <5 Charlson comorbidity index (%) 0 4,558,137 (81.6) 80,214 (78.6) 3627 (45.1) 1613 (14.2) 5385 (60.8) 365 (14.1) 583 (21.7) 1 400,005 (7.2) 8996 (8.8) 1036 (12.9) 1607 (14.2) 1031 (11.6) 519 (20.1) 427 (15.9) 2 357,892 (6.4) 6846 (6.7) 1409 (17.5) 2304 (20.3) 1164 (13.1) 460 (17.8) 559 (20.8) 3+ 270,487 (4.8) 5946 (5.8) 1969 (24.5) 5803 (51.2) 1280 (14.4) 1237 (47.9) 1123 (41.7) Autoimmune 273,496 (4.9) 6251 (6.1) 806 (10.0) 1285 (11.3) 727 (8.2) 275 (10.7) 326 (12.1) condition (%) Chronic kidney 202,102 (3.6) 3934 (3.9) 1408 (17.5) 3710 (32.8) 892 (10.1) 887 (34.4) 819 (30.4) disease (%) COPD (%) 119,950 (2.1) 2359 (2.3) 740 (9.2) 1457 (12.9) 463 (5.2) 261 (10.1) 363 (13.5) Dementia (%) 42,579 (0.8) 1933 (1.9) 429 (5.3) 2701 (23.8) 294 (3.3) 990 (38.4) 376 (14.0) Heart disease (%) 533,261 (9.5) 10,732 (10.5) 2611 (32.5) 5686 (50.2) 1903 (21.5) 1175 (45.5) 1324 (49.2) Hyperlipidemia (%) 515,045 (9.2) 10,820 (10.6) 1498 (18.6) 1580 (13.9) 1507 (17.0) 381 (14.8) 508 (18.9) Hypertension (%) 689,436 (12.3) 13,747 (13.5) 2405 (29.9) 3834 (33.8) 2206 (24.9) 896 (34.7) 932 (34.6) Malignant 291,535 (5.2) 5595 (5.5) 1327 (16.5) 3751 (33.1) 991 (11.2) 588 (22.8) 670 (24.9) neoplasm (%) Obesity (%) 928,163 (16.6) 20,761 (20.4) 3216 (40.0) 3184 (28.1) 3103 (35.0) 695 (26.9) 1088 (40.4) Type 2 diabetes (%) 317,505 (5.7) 6000 (5.9) 1644 (20.4) 2612 (23.1) 1284 (14.5) 590 (22.9) 710 (26.4) Counts of less than 5 have been obscured to protect patient privacy. COPD chronic obstructive pulmonary disease, IQR interquartile range. COVID-19 were 46 (35–58), and those hospitalised without an to a 65-year-old), but in April risks were highest for oldest ages outpatient diagnosis were 71 (58–80). Individuals hospitalised (with a hazard ratio of 1.68 [1.61–1.76] for a 90-year-old relative after a diagnosis of COVID-19 were on average 60 (49–73) years to a 65-year-old). old, and those who died after being diagnosed with COVID-19 Risk of hospitalisation with COVID-19 after an outpatient (but who were not admitted to hospital beforehand) had an diagnosis of COVID-19 peaked from around age 75, with median age of 85 (76–90). Individuals who died after being estimated hazard ratios of 0.24 (0.23–0.26) for a 45-year-old, hospitalised had a median age of 80 (73–86). 1.20 (1.17–1.23) for a 70-year-old, and 1.22 (1.14–1.31) for a 90- Estimated hazard ratios for age are shown in Fig. 4 and year-old, all relative to a reference age of 65 years old. summarised in full detail in Supplementary Tables 1 to 19. A Older age was associated with an increased risk of hospitalisa- non-linear relationship can be seen for outpatient diagnosis with tion with COVID-19 without a prior diagnosis, death after being COVID-19, with a peak in risks among those aged around 45. hospitalised with COVID-19, and, in particular, death after being Relative to a reference age of 65, estimated hazard ratios were 0.70 diagnosed but without being hospitalised. An age of 90 years old (95% confidence interval: 0.69–0.72) for a 20-year-old, 1.57 was associated with hazard ratios, relative to 65 years old, of 2.91 (1.55–1.60) for a 45-year-old in the overall model, with age as the (2.75–3.08), 8.28 (7.34–9.34), and 26.00 (23.62–28.62) for each of sole explanatory factor. For those at oldest ages, relative hazard these, respectively. For an age of 20, these relative hazard ratios ratios differed in March as compared to April (ie there was non- were all estimated to be less than 0.1. Models estimated separately proportionality in hazards, which can also be seen in the log-log by region gave broadly comparable results, see Fig. 4. Associations plot shown in Supplementary Fig. 2). In March, oldest age was between age and diagnosis with COVID-19 varied by sex, with associated with a lower risk of diagnosis relative to a 65-year-old the peak in risk at middle age more pronounced among women, (with a hazard ratio of 0.65 [0.62–0.68] for a 90-year-old relative while increased risks at oldest ages were seemingly driven by NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y Fig. 2 Histogram of age, by sex, for the study population and by transition in the multistate model. The ages, split by sex, of the study population as a whole (general population) and of those individuals making transition from general population to diagnosed with COVID-19 and hospitalised with COVID- 19, from diagnosed with COVID-19 to hospitalised with COVID-19 and death, and from hospitalised with COVID-19 to death. males. Adjustment for comorbidities generally attenuated the comorbidities made relatively little difference to these estimates (see associations for older ages to some degree, but not in their Supplementary Tables 16 to 18). entirety (see Supplementary Tables). The association between comorbidities and risk of outpatient The association between sex and risk of COVID-19 diagnosis, COVID-19 diagnosis, hospitalisation, and death. Comorbidity hospitalisation, and death. Compared to the study population as profiles varied across different stages of the natural history a whole, which was 51% female, a greater proportion of females of COVID-19, see Table 1 and Fig. 2. The highest proportion were diagnosed with COVID-19 (58%), but more males (57%) of otherwise healthy participants (Charlson index score of 0) transitioned directly to being hospitalised with COVID-19 with- was seen among those diagnosed with COVID-19 in outpatient out a previous outpatient diagnosis. Of those diagnosed, more settings (79%), and lowest amongst those who died after an males (55%) were subsequently hospitalised, but more females outpatient diagnosis without hospital admission (14%). While, (57%) died without having been hospitalised. Of those hospita- for example, the prevalence of chronic kidney disease, COPD, lised, 60% of those who died were male, see Table 1. obesity, and type 2 diabetes were 4%, 2%, 17%, and 6%, respec- After adjustment for age, male sex was associated with a reduced tively, among the study population as a whole, the prevalence of risk of outpatient COVID-19 diagnosis (hazard ratio: 0.75 [95% CI: these conditions was 30, 14, 40, and 26% among those who died 0.74–0.75]). Conversely, male sex was associated with an increased after being hospitalised with COVID-19. Of those individuals that risk of all other transitions (Fig. 5). Age-adjusted hazard ratios for died after diagnosis of COVID-19 without having being hospi- males were 1.64 (1.57–1.72) for hospitalisation with COVID-19 talised, 38% had dementia (see Table 2). without outpatient diagnosis, 1.77 (1.70–1.85) for hospitalisation After adjustment for age and sex, dementia had the strongest after outpatient diagnosis of COVID-19, 1.33 (1.23–1.44) for death association with risk of outpatient diagnosis with COVID-19 after outpatient diagnosis of COVID-19, and 1.31 (1.21–1.42) for (hazard ratio: 3.09 [95% CI: 2.94–3.24]), see Fig. 6. Aside from death after hospitalisation with COVID-19. The overall increased malignant neoplasm, all other comorbidities were associated with risk of diagnosis with COVID-19 for females appeared to be driven increased risks of outpatient diagnosis, with estimated hazard by younger women, with an age-adjusted hazard ratio of 0.72 ratios between 1.08 (1.05–1.11) for type 2 diabetes to 1.23 (0.71–0.73) for males, relative to females, among those 70 years old (1.21–1.25) for obesity. Similarly, all conditions were associated or younger, while the hazard ratio for those over 70 years old was with an increased risk of hospital admission for COVID-19 1.11 (1.07–1.15). The increased risk of death following diagnosis without a previous outpatient diagnosis, with obesity (hazard with COVID-19 appeared more pronounced among men younger ratio: 1.78 [1.70–1.87]) and dementia (1.86 [1.68–2.06]) asso- than 70, while the increased risk of death following hospitalisation ciated with the greatest excess risks. with COVID-19 for men appeared to be more pronounced in Dementia was associated with a reduced risk of hospitalisation March (see Fig. 6 and Supplementary Table 17). Estimates were after diagnosis, but an increased risk of death without broadly consistent across by region, and further adjustment for hospitalisation, with age and sex adjusted hazard ratios of 0.66 4 NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y ARTICLE Fig. 3 Prevalence of comorbidities among the general population and by transition in the multistate model. The proportion of the study population as a whole (general population) with a comorbidity of interest, and of those individuals making transition from general population to diagnosed with COVID-19 and hospitalised with COVID-19, from diagnosed with COVID-19 to hospitalised with COVID-19 and death, and from hospitalised with COVID-19 to death. COPD chronic obstructive pulmonary disease. (0.58–0.75) and 2.84 (2.60–3.09), respectively. Aside from COPD with an even greater increase in risk for younger, and particularly where no association was seen, all other conditions were female, study participants for a number of transitions. Obesity, associated with an increased risk of hospitalisation after out- for example, was associated with hazard ratios of 2.69 (2.43–2.97) patient diagnosis, ranging from a hazard ratio of 1.05 (0.98–1.13) and 2.11 (1.95–2.28) for hospitalisation from general population for malignant neoplasm to 1.57 (1.50–1.64) for obesity. and hospitalisation after a diagnosis with COVID-19 among While fewer conditions appeared associated with an increased females, 70 years old or younger. While most associations risk of death following outpatient diagnosis (without hospitalisa- between comorbidities and transitions were consistent across tion), malignant neoplasm had a hazard ratio of 1.13 (1.03–1.24), calendar time, dementia was associated with a greater increase in chronic kidney disease had one of 1.24 (1.14–1.35), and type 2 risk during April than in March. Dementia was associated with diabetes one of 1.42 (1.29–1.56). Among those hospitalised, while age and sex adjusted hazard ratios of 2.07 (1.91–2.25) and 1.33 little difference in outcomes was seen for those with hyperlipi- (1.15–1.54) for transitions from general population to diagnosis daemia and hypertension after adjusting for age and sex, other with COVID-19 and hospitalisation with COVID-19 in March. conditions were associated with increased risks with hazard ratios These increased to 4.26 (4.00–4.54) and 2.93 (2.52–3.40) in April, ranging from 1.09 (1.01–1.18) for obesity to 1.98 (1.77–2.22) for respectively (see Supplementary Table 20 and Supplementary dementia. Fig. 3). Associations between comorbidities and transitions were The association between the Charlson comorbidity index and seen to be broadly consistent across regions (see Supplementary transitions was most clearly seen for the risks of hospitalisation Table 21 and Supplementary Fig. 4). without a previous diagnosis and of death following an outpatient diagnosis with COVID-19 without hospitalisation. A score of three or more was, relative to a score of zero, associated with age Discussion and sex adjusted hazard ratios of 2.33 (2.18–2.48) and 2.59 In this large cohort study, primary care data from 5,586,521 (2.28–2.94), respectively. individuals with linked COVID-19 testing, hospitalisation, and Results for analyses stratified by age and sex are also shown in mortality data has allowed for a detailed characterisation of the Fig. 6. It can be seen that various comorbidities were associated natural history of symptomatic COVID-19 as seen during the NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications 5 6 Table 2 Time at risk and observed outcomes by state populations. From general population From diagnosed with COVID-19 From hospitalised with COVID-19 Follow-up in days To diagnosis To hospitalised To death Follow-up in days To hospitalised To death Follow-up in days To death ARTICLE with COVID-19 with COVID-19 with COVID-19 Median (min, Events Events Events (cumulative Median (min, Events Events Median (min, Events IQR, max) (cumulative (cumulative incidence at IQR, max) (cumulative (cumulative IQR, max) (cumulative incidence at incidence at 67 days incidence at incidence at incidence at 67 days) 67 days) 45 days) 45 days) 45 days) All 67 (1, 67–67, 67) 102,002 (1.83%) 8041 (0.14%) 11,327 (0.20%) 35 (0, 19–44, 66) 8860 (9.06%) 2581 (3.01%) 36 (0, 25–43, 65) 2692 (19.16%) Age Under 18 67 (1, 67–67, 67) 4536 (0.47%) 34 (0.0035%) 9 (0.00093%) 29 (0, 16–40, 65) 40 (0.96%) <5 28 (0, 19–41.75, 55) <5 Age 18–39 67 (1, 67–67, 67) 30,530 (2.13%) 426 (0.03%) 80 (0.0056%) 37 (0, 22 to 44, 66) 850 (2.91%) 5 (0.031%) 36 (0, 26.75–43, 65) 11 (1.15%) Age 40–59 67 (1, 67–67, 67) 44,493 (2.49%) 1776 (0.10%) 657 (0.037%) 37 (0, 22–44, 66) 3441 (8.05%) 63 (0.18%) 37 (0, 29–43, 65) 142 (3.54%) Age 60–69 67 (1, 67–67, 67) 10,305 (1.67%) 1558 (0.25%) 1026 (0.17%) 34 (0, 14–44, 65) 1673 (16.82%) 131 (1.51%) 38 (0, 30–44, 64) 292 (10.80%) Age 70–79 67 (1, 67–67, 67) 6120 (1.31%) 2162 (0.46%) 1967 (0.42%) 24 (0, 7–40, 65) 1602 (27.21%) 406 (8.07%) 37 (0, 26–43, 65) 822 (24.97%) Age 80 67 (1, 67–67, 67) 6018 (1.93%) 2085 (0.67%) 7588 (2.43%) 16 (0, 7–31, 62) 1254 (21.72%) 1976 (40.84%) 27 (0, 14–39.5, 65) 1424 (51.31%) or older Sex Male 67 (1, 67–67, 67) 42,996 (1.56%) 4602 (0.17%) 5655 (0.21%) 35 (0, 16–44, 66) 4871 (11.80%) 1114 (3.13%) 37 (0, 26–43, 65) 1617 (20.46%) Sex Female 67 (1, 67–67, 67) 59,006 (2.09%) 3439 (0.12%) 5672 (0.20%) 35 (0, 20–44, 66) 3989 (7.06%) 1467 (2.93%) 36 (0, 24 to 43, 65) 1075 (17.43%) Region Barcelona 67 (1, 67–67, 67) 82,702 (1.98%) 7201 (0.17%) 8694 (0.21%) 36 (0, 19–44, 66) 7604 (9.56%) 2164 (3.10%) 37 (0, 26–43, 65) 2393 (19.20%) Region Girona 67 (1, 67–67, 67) 9929 (1.78%) 432 (0.078%) 942 (0.17%) 35 (0, 18–44, 66) 657 (6.99%) 217 (2.55%) 32 (0, 17–41, 65) 158 (17.48%) Region Lleida 67 (2, 67–67, 67) 5621 (1.52%) 211 (0.057%) 728 (0.20%) 33 (0, 16–43, 65) 391 (7.35%) 108 (2.45%) 35 (0, 22–41, 64) 94 (22.27%) Region Tarragona 67 (1, 67–67, 67) 3748 (0.77%) 196 (0.04%) 958 (0.20%) 30 (0, 16–42, 66) 207 (5.82%) 92 (3.00%) 33 (0, 22–42, 64) 45 (15.92%) Charlson 0 67 (1, 67–67, 67) 80,214 (1.76%) 3627 (0.08%) 1613 (0.035%) 36 (0, 21–44, 66) 5385 (7.01%) 365 (0.55%) 37 (0, 29–43, 65) 583 (8.20%) Charlson 1 67 (1, 67–67, 67) 8996 (2.25%) 1036 (0.26%) 1607 (0.40%) 33 (0, 15–43, 65) 1031 (11.93%) 519 (6.82%) 36 (0, 23–43, 65) 427 (24.22%) Charlson 2 67 (2, 67–67, 67) 6846 (1.91%) 1409 (0.39%) 2304 (0.64%) 30 (0, 11–43, 65) 1164 (17.70%) 460 (8.16%) 36 (0, 23–43, 65) 559 (25.21%) Charlson 3+ 67 (1, 67–67, 67) 5946 (2.20%) 1969 (0.73%) 5803 (2.15%) 19 (0, 7–36, 64) 1280 (22.53%) 1237 (25.66%) 31 (0, 17–41, 65) 1123 (41.51%) Autoimmune 67 (2, 67–67, 67) 6251 (2.29%) 806 (0.29%) 1285 (0.47%) 34 (0, 16–44, 64) 727 (12.10%) 275 (5.09%) 35 (0, 22–42, 65) 326 (25.11%) condition Chronic kidney 67 (1, 67–67, 67) 3934 (1.95%) 1408 (0.70%) 3710 (1.84%) 19 (0, 7–35, 62) 892 (23.67%) 887 (27.36%) 31 (0, 17–41, 64) 819 (42.57%) disease COPD 67 (2, 67–67, 67) 2359 (1.97%) 740 (0.62%) 1457 (1.22%) 22 (0, 8–40, 65) 463 (20.64%) 261 (13.68%) 33 (0, 19–42, 65) 363 (36.26%) Dementia 67 (1, 67–67, 67) 1933 (4.54%) 429 (1.01%) 2701 (6.35%) 15 (0, 8–25, 59) 294 (15.92%) 990 (64.07%) 21 (0.5, 12 to 34, 65) 376 (64.59%) Heart disease 67 (1, 67–67, 67) 10,732 (2.01%) 2611 (0.49%) 5686 (1.07%) 28 (0, 10–41, 65) 1903 (18.46%) 1175 (13.20%) 34 (0, 19–42, 65) 1324 (34.22%) Hyperlipidemia 67 (1, 67–67, 67) 10,820 (2.10%) 1498 (0.29%) 1580 (0.31%) 34 (0, 15–43, 65) 1507 (14.48%) 381 (4.18%) 37 (0, 27–43, 65) 508 (20.40%) Malignant 67 (1, 67–67, 67) 5595 (1.92%) 1327 (0.46%) 3751 (1.29%) 29 (0, 10–42, 65) 991 (18.37%) 588 (12.70%) 34 (0, 20–43, 65) 670 (34.19%) neoplasm Obesity 67 (1, 67–67, 67) 20,761 (2.24%) 3216 (0.35%) 3184 (0.34%) 33 (0, 14–43, 66) 3103 (15.52%) 695 (3.94%) 37 (0, 25–43, 65) 1088 (20.03%) Type 2 67 (2, 67–67, 67) 6000 (1.89%) 1644 (0.52%) 2612 (0.82%) 25 (0, 8–41, 64) 1284 (22.30%) 590 (11.77%) 35 (0, 22–42, 65) 710 (28.50%) diabetes Counts of less than 5 have been obscured to protect patient privacy. COPD chronic obstructive pulmonary disease, IQR interquartile range. NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y ARTICLE Fig. 4 Age and COVID-19 transitions. Estimated hazard ratios for age (relative to a reference age of 65) from cause-specific Cox models for each transition. Dotted lines represent 95% confidence intervals. The models shown included age as the only explanatory variable, with the overall models and models stratified by calendar month, sex, and region presented. NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y Fig. 5 Sex and COVID-19 transitions. Estimated hazard ratios for male sex (relative to female) from cause-specific Cox models for each transition. Points give estimated hazard ratios, with lines representing 95% confidence intervals. The models shown included sex and age as explanatory variables, with the overall models and models stratified by calendar month, sex, and region presented. 8 NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y ARTICLE Fig. 6 Comorbidities and COVID-19 transitions. Estimated hazard ratios for comorbidities of interest from cause-specific Cox models for each transition. Points give estimated hazard ratios, with lines representing 95% confidence intervals. Models were estimated separately for each comorbidity of interest, with adjustment for age and sex. COPD chronic obstructive pulmonary disease. first-wave of the epidemic in Spain. Over 100,000 outpatient during the height of the pandemic in Spain. The increase in risk COVID-19 diagnoses, close to 17,000 hospitalisations with for age was most strongly seen for the transition from diagnosis COVID-19, and almost 5300 COVID-19 related deaths were to death with no admission to hospital seen between the two, observed between 1st March and 6th May. Of these deaths, half while the risk of hospitalisation itself was also seen to fall at were among individuals who were diagnosed with COVID-19 in oldest ages. This may reflect, in part, rationing of health care outpatient settings but were not seen to have been admitted to resources, with younger patients likely prioritised for the hospital. receipt of hospital care, and intensive services if admitted. With Older age was consistently associated with increased risks of half of the deaths in this study observed among those not hospitalisation and mortality, most dramatically for risk of death hospitalised after diagnosis, who had an median age of 86, it is following an outpatient diagnosis of COVID-19 without a sub- of utmost importance that similar individuals should be given sequent hospital admission. There was though a notable peak in appropriate access to care in the future. risk of outpatient diagnosis for people in middle-age during Our findings on outpatient diagnoses of COVID-19 may March, whereas in April risks were highest for oldest ages. A provide further insights on community transmission of differential association was seen with sex; while women were at COVID-19. A peak in diagnosis of COVID-19 was seen among increased risk of outpatient diagnosis, men were consistently at individuals around the age of 45 years old during March, but by higher risk of hospitalisation and death. Finally, the specific April the risks of diagnosis were highest for those at oldest ages. comorbidities studied (autoimmune condition, chronic kidney While this could to some degree reflect differences in health disease, chronic obstructive pulmonary disease, dementia, heart seeking behaviour across age groups and changes in the way in disease, hyperlipidemia, hypertension, malignant neoplasm, which diagnoses were made and recorded, it likely also reflects obesity, and type 2 diabetes), and a composite comorbidity index differences in infection rates across age groups and transmis- (Charlson Comorbidity Index) were associated with worse out- sion dynamics. A large study in Spain found that ser- comes. In a number of cases the increase in risk associated with oprevalence rose until plateauing at around age of 45 based on comorbidities appeared to be particularly pronounced for women point-care-tests, but that seroprevalence was lower for those aged under 70. older than 85 compared to younger adults given immunoassay Older age has consistently been found to be a risk factor for results7. Meanwhile, a seroprevalence study in Geneva found worse outcomes in COVID-19. It has been associated with an that those aged between 20 and 49 had the highest likelihood of increased risk of hospitalisation after testing positive for being seropositive11. Consequently, while they face lower risks COVID-193,4, worse outcomes among those hospitalised1,2,9, of poor outcomes if infected, younger age groups were infected and an increased risk of COVID-19-related mortality among and appear to have contributed to the spread of the virus in the the general population10. Similarly, in our study we find older community. age to be associated with poorer outcomes. Our findings suggest Being male has also been associated with worse prognosis in that not only is older age associated with increased disease COVID-19, with increased risks for both hospitalisation among severity, but that there was also likely inequitable access to care those tested3, and worse outcomes among those hospitalised2,3,9. NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y Our data are compatible with previous literature, with risks of classifications are broad, and we have not attempted to split hospital admission and mortality all increased for males. How- conditions by severity or concurrent medication use. ever, we found the opposite effect for outpatient diagnosis, with The purpose of this study was descriptive in nature. The aim women being at increased risk of diagnosis with COVID-19 in the was not prediction, nor was it causal inference. In particular, it community. While this finding contrasts with two UK studies, should be noted that associations between specific comorbidities which reported a higher risk of testing positive for SARS-CoV-2 and outcomes do not necessarily reflect a causal relationship. among men12,13, it is congruent with the findings from a study Assessing whether a particular chronic condition is the cause from China that reported a higher attack rate among women14. In worse outcomes in COVID-19 will require further consideration Spain, seroprevalence was seen to be similar for males and of, and accounting for, relevant confounding factors. The females7. Further research is therefore needed to understand this associations which have been described here are also not finding, which could be explained by differences in exposure and/ immutable, but rather sit in the context of the first wave of or vulnerability to severe acute respiratory syndrome coronavirus COVID-19 in Spain. While there is substantial evidence that 2 (SARS-CoV-2), and differences in health seeking behaviour individuals at older age are at highest risk of severe disease if between men and women. infected, effective shielding strategies would affect the risks of Comorbidities have been associated with worse outcomes in COVID-19 diagnoses and hospitalisations among this group. COVID-19 in this study. In line with our findings, those with a Similarly, ensuring that there is sufficient capacity to provide higher Charlson score testing positive for COVID-19 have pre- appropriate care would likely have a particular impact on the viously been seen to have an increased risk of hospitalisation and proportion of patients that die after a diagnosis with COVID-19 death4. Dementia was seen in our study to be associated with a without a hospital admission. much-increased risk of diagnosis with COVID-19. For those In conclusion, in this descriptive study we provide a compre- diagnosed, dementia was associated with a reduced risk of hos- hensive overview of the natural history of symptomatic COVID- pitalisation, but an increased risk of death without hospitalisation. 19 in Catalonia, Spain, during the first wave of the pandemic. The As with age, the association between dementia and risk of findings from this study can help inform the continued response diagnosis and hospitalisation with COVID-19 in the general to COVID-19 both in Spain and elsewhere. Our research has population were seen to change over calendar time, with helped to reveal a clear need to protect at risk populations, par- increased risks particularly pronounced in April. Other comor- ticularly older people, whilst also considering middle-age popu- bidities were seen to be associated with worse outcomes, which lations at a particularly high risk of milder infections and is in accordance with previous findings on outcomes for indivi- therefore likely key in the community transmission of the disease. duals with chronic kidney disease, COPD, heart disease, hyper- lipidemia, hypertension, cancer, obesity, and type 2 diabetes Methods mellitus2–4,12,13,15. Our findings have also generally shown the Study design. This study was informed by primary care records from Catalonia, associations between type 2 diabetes, obesity, and chronic kidney Spain, which were linked to COVID-19 test results, hospital records, and mortality disease and COVID-19 outcomes to be most pronounced among data. The resulting database was comprehensive, allowing for the identification of various key events in the progression of COVID-19 over the study period, which women under 70. Fewer studies have assessed the relationship began on the 1st March 2020 and ended on the 6th May 2020. A multistate cohort between autoimmune conditions and COVID-19 outcomes. As model provided the framework for analysis, allowing for a systematic consideration with other comorbidities, we found autoimmune conditions to be of transitions for the general population to diagnosis of COVID-19, hospitalisation associated with an increased risk of poorer outcomes in COVID- with COVID-19, and COVID-19 mortality. 19. Further research is though certainly merited to consider each of these conditions in turn and consider the impact of other Study participants, setting, and data source. Individual-level routinely-collected primary care data were extracted from the Information System for Research in factors such as these individuals’ medication use. Primary Care (SIDIAP; www.sidiap.org) database, which captures patient records Our study was informed by routinely-collected health care data from approximately 80% of the Catalan population and is representative in geo- with various interactions between individuals and the health graphy, age, and sex16. Linkage was made at an individual-level to COVID-19 RT- system identified, covering outpatient diagnoses, COVID-19 PCR testing data, hospital data, and regional mortality data. testing, and hospitalisations, and with linked mortality data. The entire database has been mapped to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM). This provided a means of This though unavoidably misses health outcomes experienced by structuring the data to a standardised format, and allowed for the application of individuals that do not lead to any interaction with the health analytical tools developed by the open-science Observational Health Data Sciences care system, with both asymptomatic individuals and a sizeable and Informatics (OHDSI) network17. proportion of mild symptomatic cases unlikely to be seen. Some All individuals in SIDIAP as of the 1st March 2020 were identified. So that study participants had sufficient prior observation time for comorbidities to be of the clinical diagnoses observed in the study may also represent identified, any individuals with less than one year of prior history available were false positives, given the uncertainty surrounding the presentation excluded. As the index date for all individuals in the multistate model, described of the disease during the study period. We did not require clinical below, began on the 1st March, any individual who had a clinical diagnosis or diagnoses to be confirmed by the presence of a positive RT-PCR positive test result for COVID-19 between the 1st January and 29th February 2020 was excluded. In addition, because the starting state in the model was test, as such tests were not being routinely performed in out- representative of individuals living in the community in Catalonia, individuals who patient settings in Spain during the study period. Deaths from were hospitalised or a care home resident as of the 1st March 2020 were also COVID-19 where individuals were not tested or diagnosed excluded. Study follow-up began on the 1st March 2020 and ended on the 6th May beforehand will also not have been identified. 2020 (the last date of available data). While providing a broad picture of clinical trajectories, each of the studied states and transitions can, and should, be considered Variables. Individuals’ age and sex were extracted. Health conditions, using in further detail. Analyses of the provision of intensive care individual’s observed medical history were identified. The Charlson comorbidity index was calculated and scores were categorised as 0, 1, 2, or 3+. Ten specific admissions during hospitalisation is one such example, but would health conditions were also extracted, all of interest as potential risk factors for the require further data with sufficient granularity on inpatient progression of COVID-19. Autoimmune condition (which included type 1 dia- treatment. In addition, emergency room presentations that did betes, rheumatoid arthritis, psoriasis, psoriatic arthritis, multiple sclerosis, systemic not lead to a hospital admission were also not assessed in this lupus erythematosus, Addison’s disease, Grave’s disease, Sjorgen’s syndrome, Hashimoto thyroiditis, Myasthenia gravis, vasculitis, pernicious anaemia, celiac study. Similarly, the set of comorbidities considered here are only disease, scleroderma, sarcoidosis, ulcerative colitis, and Crohn’s disease), chronic a subset of the myriad set of conditions that are of interest kidney disease, chronic obstructive pulmonary disease (COPD), dementia, heart when considering potential risk factors for COVID-19. Our disease, hyperlipidemia, hypertension, malignant neoplasm excluding non- 10 NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y ARTICLE melanoma skin cancer, type 2 diabetes mellitus were each identified on the basis of which observed associations could be explained by differences in comorbidity diagnosis codes. Obesity was identified either by a diagnosis code, a record of a profiles across age groups, models were also estimated with additional adjustment body mass index measurement between 30 and 60 kg/m2, or a recorded weight subsequently made for comorbidities, with the Charlson score and the ten specific between 120 and 200 kg within 5 years of the index date. The region in which an conditions of interest described above also included as explanatory factors in individual lived; Barcelona (inclusive of the Barcelona, Metro area north, Metro the model. area south, and Catalunya Central health regions), Girona, Lleida (inclusive of both Lleida and Alt Pirineu i Aran health regions), or Tarragona (inclusive of the Describing the association between sex and risk of COVID-19 diagnosis, hospitali- Tarragona and Terres de l’Ebre health regions) as of their index date was identified, sation, and death. Our objective here was to describe the overall association as was whether they were a care home resident. between sex and outcomes, accounting for differences in age, during the first wave Study outcomes included an outpatient clinical diagnosis of COVID-19, a of COVID-19 in Catalonia. Therefore, the primary models included sex and age, hospitalisation with COVID-19, and death. Outpatient COVID-19 diagnoses were with the sex the explanatory factor of interest. identified on the basis of the first observation of a compatible clinical code being As above, the relationship between sex and the risk of each transition in the recorded for COVID-19 disease (such as ICD-10-CM codes B34.2 “Coronavirus multistate model was assessed by estimating cause-specific Cox models. Non- infection, unspecified” and B97.29 “Other coronavirus as the cause of diseases linearity in age was incorporated in the same way as for the previous research classified elsewhere”). Hospitalisation with COVID-19 was identified as a hospital objective. To consider whether associations between sex and outcomes varied by admission, a hospital stay of at least one night, where the individual had a positive age, models were also estimated separately for those aged 70 or younger and those RT-PCR test result or a clinical diagnosis of COVID-19 over the 21 days prior to aged over 70 (with age assumed to have a linear relationship within each of the their admission up to the end of their hospital stay18. Mortality was identified using strata). Stratified models were also estimated by region and calendar time. An region-wide mortality data, and so included both deaths during hospitalisations additional set of models were also estimated with adjustment made for and in the community. comorbidities. Describing the association between comorbidities and risk of transitions. Our Statistical methods objective here was to describe the overall association between comorbidities and Multistate model. A multistate model provided the framework for the analysis. outcomes, accounting for differences in age and sex, during the first wave of Multistate models allow for a consideration of individuals progression to multiple COVID-19 in Catalonia. The primary models included the specific comorbidity of events of interest, extending on competing risk models by also describing transi- interest (the explanatory factor of interest), age, and sex. tions to intermediate events19. In the context of COVID-19, clinical diagnoses of The relationship between comorbidities and each transition was assessed by the disease and hospitalisations with the disease can be considered as key inter- estimating cause-specific Cox models. Models were estimated for each comorbidity mediate events between not being infected (or at least, not having been identified as of interest separately, controlling for age and sex and with non-linearity in age being so) on one end to death on the other. incorporated in the same way as described above. To consider whether associations The structure of the multi-state model used in this study is shown in Fig. 1. between comorbidities and outcomes varied by age and sex, models were There were four states: general population, outpatient diagnosis with COVID-19, subsequently estimated stratifying by both age (70 or younger or above 70) and sex, hospitalised with COVID-19, and death. The model is progressive with all with age also included as an explanatory factor within each of the strata as a linear individuals beginning in the general population state (as described above, term. For dementia, models were only estimated for the older age strata. Stratified individuals included in the study had no prior history of COVID-19 and were not models were also estimated by region and calendar time. hospitalised on their index date). Six transitions were possible: from general population to outpatient diagnosis with COVID-19, from general population to hospitalised with COVID-19 (i.e., individuals who did not get a clinical diagnosis Reporting summary. Further information on research design is available in the Nature prior to hospital admission), from general population to death, from outpatient Research Reporting Summary linked to this article. diagnosis with COVID-19 to hospitalised with COVID-19, from outpatient diagnosis with COVID-19 to death, and from hospitalised with COVID-19 to death. Given the research objectives, the analyses focused on the five transitions Data availability In accordance with current European and national law, the data used in this study is only directly related to COVID-19, with the general population to death used primarily as a means of accounting for the competing risk of death for the study population available for the researchers participating in this study. Thus, we are not allowed to (although it should be noted that individuals who died with COVID-19, but had distribute or make publicly available the data to other parties. However, researchers from not been diagnosed or received a positive test will also be included in this public institutions can request data from SIDIAP if they comply with certain transition). requirements. Further information is available online (https://www.sidiap.org/index.php/ The index date for all individuals, from which follow-up began, was 1st March menu-solicitudes-en/application-proccedure) or by contacting Anna Moleras 2020. For any given transition in the model, an individual´s end of follow-up was (amoleras@idiapjgol.org). whichever came first: their exit from the database (administrative censoring), the occurrence of the event of interest, the occurrence of a competing event, or the end Code availability of the study period (6th May 2020). The mapping of source SIDIAP data to the OMOP CDM was facilitated by various open- source OHDSI software, which included usagi (https://github.com/OHDSI/Usagi), to Describing the association between age and risk of COVID-19 diagnosis, hospitali- help define mappings from source codes to the standard concepts used in the CDM, and sation, and death. Our objective here was to describe the overall association Achilles (https://github.com/OHDSI/Achilles), to help assess data quality after mapping. between age and outcomes seen during the first wave of COVID-19 in Catalonia. Data analysis was performed in R version 4.0.0. The R packages used in the analysis Consequently, the primary models of interest included age as the sole explanatory included OHDSI CohortDiagnostics (https://github.com/OHDSI/CohortDiagnostics), factor. numerous tidyverse packages21, mstate22, and rms23. The analytic code used in this study The relationship between age and the risk of each transition in the multistate is freely available at https://github.com/SIDIAP/MultiStateCovid-19. Cohort definitions model was assessed by estimating cause-specific Cox models, with hazard ratios were adapted from those used in the OHDSI Seek COVER study and the OHDSI and 95% confidence intervals (95% CIs) calculated. A non-linear relationship CHARYBDIS project24. A web application which summarises the cohort definitions used between age and the risk of transitions was considered by fitting age with a is available at https://livedataoxford.shinyapps.io/MultiStateCovidCohorts/. polynomial of degree 2 (i.e., quadratic) and with restricted cubic splines (with 3, 4, or 5 knots)20. Comparisons between these, and a model where age was assumed to have a linear relationship, were made using Akaike information criterion (AIC) Received: 11 August 2020; Accepted: 7 January 2021; and Bayesian Information Criterion (BIC). Where a more parsimonious model was seen to well-approximate a more complex one, the simpler model was favoured. Non-proportionality in hazards was considered through visual inspection of log–log plots. To consider whether associations between age and outcomes varied by sex, models were also estimated separately for males and females. Similarly, models were also estimated separately for each region. To consider the impact of calendar References time on the association between age and the risk of transitions from general 1. Zhou, F. et al. Clinical course and risk factors for mortality of adult inpatients population to outpatient diagnosis with COVID-19 and hospitalised with COVID- with COVID-19 in Wuhan, China: a retrospective cohort study. Lancet 395, 19, models were estimated separately for March and April (i.e., with follow-up split 1054–1062 (2020). into these distinct periods). To consider the impact of calendar time on the 2. Docherty, A. B. et al. Features of 20 133 UK patients in hospital with covid-19 association between age and subsequent transitions (i.e., from outpatient diagnosis using the ISARIC WHO Clinical Characterisation Protocol: prospective with COVID-19 to hospitalised with COVID-19, outpatient diagnosis with observational cohort study. BMJ 369, m1985 (2020). COVID-19 to death, and from hospitalised with COVID-19 to death), models were 3. Petrilli, C. M. et al. Factors associated with hospital admission and critical estimated separately for those who arrived into these starting states in March and illness among 5279 people with coronavirus disease 2019 in New York City: those who transitioned into them in April. Additionally, to consider the degree to prospective cohort study. BMJ 369, m1966 (2020). NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21100-y 4. Reilev, M. et al. Characteristics and predictors of hospitalization and death in performed during the OHDSI COVID-19 studyathon. The mapping of data to the the first 11 122 cases with a positive RT-PCR test for SARS-CoV-2 in OMOP CDM has been supported by a taskforce from the European Health Data and Denmark: a nationwide cohort. Int. J. Epidemiol. https://doi.org/10.1093/ije/ Evidence Network (EHDEN). This project is funded by the Health Department from dyaa140 (2020). the Generalitat de Catalunya with a grant for research projects on SARS-CoV-2 and 5. Griffith, G. J. et al. Collider bias undermines our understanding of COVID-19 COVID-19 disease organised by the Direcció General de Recerca i Innovació en Salut. disease risk and severity. Nat. Commun. 11, 5749 (2020). This project has received support from the European Health Data and Evidence Network 6. Prieto-Alhambra, D. et al. Filling the gaps in the characterization of the (EHDEN) project. EHDEN received funding from the Innovative Medicines Initiative 2 clinical management of COVID-19: 30-day hospital admission and fatality Joint Undertaking (JU) under grant agreement No 806968. The JU receives support from rates in a cohort of 118,150 cases diagnosed in outpatient settings in Spain. Int. the European Union’s Horizon 2020 research and innovation programme and EFPIA. J. Epidemiol. https://doi.org/10.1093/ije/dyaa190 (2020). The University of Oxford received a grant related to this work from the Bill & Melinda 7. Pollán, M. et al. Prevalence of SARS-CoV-2 in Spain (ENE-COVID): a Gates Foundation (Investment ID INV-016201), and partial support from the UK nationwide, population-based seroepidemiological study. Lancet https://doi. National Institute for Health Research (NIHR) Oxford Biomedical Research Centre. org/10.1016/S0140-6736(20)31483-5 (2020). D.P.A. is funded through a National Institute for Health Research (NIHR) Senior 8. Leek, J. T. & Peng, R. D. What is the question? Science 347, 1314–1315 (2015). Research Fellowship (Grant number SRF-2018-11-ST2-004). The views expressed in this 9. Gupta, S. et al. Factors associated with death in critically Ill patients with publication are those of the authors and not necessarily those of the NHS, the National coronavirus disease 2019 in the US. JAMA Intern. Med. https://doi.org/ Institute for Health Research or the Department of Health. A.P.-U. is supported by 10.1001/jamainternmed.2020.3596 (2020). Fundacion Alfonso Martin Escudero and the Medical Research Council (grant numbers 10. Williamson, E. J. et al. OpenSAFELY: factors associated with COVID-19 death MR/K501256/1, MR/N013468/1). in 17 million patients. Nature https://doi.org/10.1038/s41586-020-2521-4 (2020). 11. Stringhini, S. et al. Seroprevalence of anti-SARS-CoV-2 IgG antibodies in Author contributions S.F.B., M.A., E.B., and T.D.S. mapped source data to the OMOP CDM and had accessed Geneva, Switzerland (SEROCoV-POP): a population-based study. Lancet the data for the analysis. E.B., C.T., D.P.A., and T.D.S. led the data analysis. M.R. and E.R. https://doi.org/10.1016/S0140-6736(20)31304-0 (2020). performed a literature review. All authors were involved in the study conception and 12. de Lusignan, S. et al. Risk factors for SARS-CoV-2 among patients in the design, interpretation of the results, and the preparation of the manuscript. Oxford Royal College of general practitioners research and surveillance centre primary care network: a cross-sectional study. Lancet Infect. Dis. https://doi. org/10.1016/S1473-3099(20)30371-6 (2020). Competing interests 13. Ho, F. K. et al. Modifiable and non-modifiable risk factors for COVID-19: D.P.A. reports grants and other from AMGEN; grants, non-financial support and other results from UK Biobank. medRxiv https://doi.org/10.1101/ from UCB Biopharma; grants from Les Laboratoires Servier, outside the submitted work; 2020.04.28.20083295 (2020). and Janssen, on behalf of IMI-funded EHDEN and EMIF consortiums, and Synapse 14. Qian, J., Zhao, L., Ye, R.-Z., Li, X.-J. & Liu, Y.-L. Age-dependent gender Management Partners have supported training programmes organised by DPA’s differences of COVID-19 in mainland China: comparative study. Clin. Infect. department and open for external participants. C.T. reports personal fees from Amgen, Dis. https://doi.org/10.1093/cid/ciaa683 (2020). Boehringer ingelheim outside the submitted work. No other relationships or activities 15. Tartof, S. Y. et al. Obesity and mortality among patients diagnosed with that could appear to have influenced the submitted work. All other authors declare no COVID-19: results from an Integrated Health Care Organization. Ann. Intern. competing interests. Med. https://doi.org/10.7326/M20-3742 (2020). 16. García-Gil, M. D. M. et al. Construction and validation of a scoring system for the selection of high-quality data in a Spanish population primary care Additional information database (SIDIAP). Inform. Prim. Care 19, 135–145 (2011). Supplementary information The online version contains supplementary material 17. Hripcsak, G. et al. Observational health data sciences and informatics available at https://doi.org/10.1038/s41467-021-21100-y. (OHDSI): opportunities for observational researchers. Stud. Health Technol. Inform. 216, 574–578 (2015). Correspondence and requests for materials should be addressed to D.P.-A. 18. Burn, E. et al. Deep phenotyping of 34,128 adult patients hospitalised with COVID-19 in an international network study. Nat. Commun. 11, 5009 (2020). Peer review information Nature Communications thanks Leo Anthony Celi and Shemra 19. Putter, H., Fiocco, M. & Geskus, R. B. Tutorial in biostatistics: competing risks Rizzo for their contribution to the peer review of this work. Peer review reports are and multi-state models. Stat. Med. https://doi.org/10.1002/sim (2007). available. 20. Harrell, F. E. Regression Modeling Strategies. (Springer, New York, 2015). https://doi.org/10.1007/978-3-319-19425-7 Reprints and permission information is available at http://www.nature.com/reprints 21. Wickham, H. et al. Welcome to the tidyverse. J. Open Source Softw. 4, 1–6 (2019). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 22. de Wreede, L., Fiocco, M. & Putter, H. mstate: An R package for the analysis published maps and institutional affiliations. of competing risks and multi-state models. J. Stat. Softw. Artic. 38, 1–30 (2011). 23. Harrell, F. E. rms: Regression modeling strategies. R package version 5.1-0. Open Access This article is licensed under a Creative Commons https://doi.org/https://CRAN.R-project.org/package=rms (2017). Attribution 4.0 International License, which permits use, sharing, 24. Williams, R. D. et al. Seek COVER: Development and validation of a adaptation, distribution and reproduction in any medium or format, as long as you give personalized risk calculator for COVID-19 outcomes in an international appropriate credit to the original author(s) and the source, provide a link to the Creative network. medRxiv https://doi.org/10.1101/2020.05.26.20112649 (2020). 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 Acknowledgements article’s Creative Commons license and your intended use is not permitted by statutory We would like to acknowledge the patients who suffered from or died of this devastating regulation or exceeds the permitted use, you will need to obtain permission directly from disease, and their families and carers. We would also like to thank the healthcare pro- the copyright holder. To view a copy of this license, visit http://creativecommons.org/ fessionals involved in the management of COVID-19 during these challenging times, licenses/by/4.0/. from primary care to intensive care units in the Catalan healthcare system. The analysis has made use of a range of open-source, free available tools provided by the OHDSI community. The phenotypes used in this study were developed, or adapted from, work © The Author(s) 2021 12 NATURE COMMUNICATIONS | (2021)12:777 | https://doi.org/10.1038/s41467-021-21100-y | www.nature.com/naturecommunications