Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID- 19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Articles Identification and validation of clinical phenotypes with prognostic implications in patients admitted to hospital with COVID-19: a multicentre cohort study Belén Gutiérrez-Gutiérrez, María Dolores del Toro, Alberto M Borobia, Antonio Carcas, Inmaculada Jarrín, María Yllescas, Pablo Ryan, Jerónimo Pachón, Jordi Carratalà, Juan Berenguer, Jose Ramón Arribas, Jesús Rodríguez-Baño, on behalf of the REIPI-SEIMC COVID-19 group and COVID@HULP groups* Summary Background The clinical presentation of COVID-19 in patients admitted to hospital is heterogeneous. We aimed to Lancet Infect Dis 2021 determine whether clinical phenotypes of patients with COVID-19 can be derived from clinical data, to assess the Published Online reproducibility of these phenotypes and correlation with prognosis, and to derive and validate a simplified probabilistic February 23, 2021 https://doi.org/10.1016/ model for phenotype assignment. Phenotype identification was not primarily intended as a predictive tool for mortality. S1473-3099(21)00019-0 See Online/Comment Methods In this study, we used data from two cohorts: the COVID-19@Spain cohort, a retrospective cohort including https://doi.org/10.1016/ 4035 consecutive adult patients admitted to 127 hospitals in Spain with COVID-19 between Feb 2 and March 17, 2020, S1473-3099(21)00073-6 and the COVID-19@HULP cohort, including 2226 consecutive adult patients admitted to a teaching hospital in Madrid *Members of the REIPI-SEIMC between Feb 25 and April 19, 2020. The COVID-19@Spain cohort was divided into a derivation cohort, comprising COVID-19 group and COVID@ 2667 randomly selected patients, and an internal validation cohort, comprising the remaining 1368 patients. The HULP group are listed in appendix pp 23–29 COVID-19@HULP cohort was used as an external validation cohort. A probabilistic model for phenotype assignment Unidad Clínica de was derived in the derivation cohort using multinomial logistic regression and validated in the internal validation Enfermedades Infecciosas, cohort. The model was also applied to the external validation cohort. 30-day mortality and other prognostic variables Microbiología y Medicina were assessed in the derived phenotypes and in the phenotypes assigned by the probabilistic model. Preventiva, Hospital Universitario Virgen Macarena, Seville, Spain Findings Three distinct phenotypes were derived in the derivation cohort (n=2667)—phenotype A (516 [19%] patients), (B Gutiérrez-Gutiérrez PhD, phenotype B (1955 [73%]) and phenotype C (196 [7%])—and reproduced in the internal validation cohort (n=1368)— M D del Toro PhD, phenotype A (233 [17%] patients), phenotype B (1019 [74%]), and phenotype C (116 [8%]). Patients with phenotype A J Rodríguez-Baño PhD); were younger, were less frequently male, had mild viral symptoms, and had normal inflammatory parameters. Patients Departamento de Medicina, Universidad de Sevilla, Seville, with phenotype B included more patients with obesity, lymphocytopenia, and moderately elevated inflammatory Spain (B Gutiérrez-Gutiérrez, parameters. Patients with phenotype C included older patients with more comorbidities and even higher inflammatory M D del Toro, Prof J Pachón PhD, parameters than phenotype B. We developed a simplified probabilistic model (validated in the internal validation cohort) J Rodríguez-Baño); Instituto de for phenotype assignment, including 16 variables. In the derivation cohort, 30-day mortality rates were 2·5% Biomedicina de Sevilla, Seville, Spain (B Gutiérrez-Gutiérrez, (95% CI 1·4–4·3) for patients with phenotype A, 30·5% (28·5–32·6) for patients with phenotype B, and 60·7% M D del Toro, Prof J Pachón, (53·7–67·2) for patients with phenotype C (log-rank test p<0·0001). The predicted phenotypes in the internal validation J Rodríguez-Baño); cohort and external validation cohort showed similar mortality rates to the assigned phenotypes (internal validation Departamento de Farmacología Clínica, Hospital cohort: 5·3% [95% CI 3·4–8·1] for phenotype A, 31·3% [28·5–34·2] for phenotype B, and 59·5% [48·8–69·3] for Universitario La Paz, phenotype C; external validation cohort: 3·7% [2·0–6·4] for phenotype A, 23·7% [21·8–25·7] for phenotype B, and Universidad Autónoma de 51·4% [41·9–60·7] for phenotype C). Madrid, Madrid, Spain (A M Borobia PhD, A Carcas PhD); Instituto de Investigación Interpretation Patients admitted to hospital with COVID-19 can be classified into three phenotypes that correlate with La Paz, Madrid, Spain mortality. We developed and validated a simplified tool for the probabilistic assignment of patients into phenotypes. (A M Borobia, A Carcas, These results might help to better classify patients for clinical management, but the pathophysiological mechanisms J R Arribas MD); Centro Nacional of the phenotypes must be investigated. de Epidemiología, Instituto de Salud Carlos III, Madrid, Spain (I Jarrín MSc); Fundación Funding Instituto de Salud Carlos III, Spanish Ministry of Science and Innovation, and Fundación SEIMC/GeSIDA. SEIMC/GeSIDA, Madrid, Spain (M Yllescas PharmD); Copyright © 2021 Elsevier Ltd. All rights reserved. Servicio de Medicina Interna, Hospital Universitario Infanta Leonor, Madrid, Spain Introduction protection due to previous infections with other corona­ (P Ryan PhD); Unidad Clínica de Patients admitted to hospital with COVID-19 show various viruses,7 genetic determinants,8 and other non-genetic- Enfermedades Infecciosas, clinical signs and symptoms and laboratory abnor­ mediated factors such as age and underlying conditions.3,4 Microbiología y Medicina Preventiva, Hospital malities.1–5 Some of these features have been found to be We hypothesise that patients admitted to hospital with Universitario Virgen del Rocío, predictors of mortality.3,4 The reasons for this heterogeneous COVID-19 might be classified into few clinical patterns Seville, Spain (Prof J Pachón); presentation are not fully understood. However, it could be (phenotypes) according to their demographics, underlying Servei de Malalties Infeccioses, related to factors such as viral load,6 partial immune conditions, signs, symptoms, radiological findings, Hospital Universitari de www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 1 Articles Bellvitge, Barcelona, Spain (Prof J Carratalà PhD); Instituto Research in context de Investigación Biomédica de Bellvitge, Barcelona, Spain Evidence before this study radiological features at presentation among patients (Prof J Carratalà); Universitat de We searched PubMed, Scopus, and medRxiv from Jan 9 to admitted to hospital with COVID-19. The phenotypes Barcelona, Barcelona, Sept 30, 2020, using the terms [“COVID-19” OR “SARS-CoV-2”] were shown to have clinical implications, since they Spain (Prof J Carratalà); Servicio de Microbiología Clínica y AND [“phenotypes” OR “clinical features”], with no language were associated with patient prognosis. Furthermore, Enfermedades Infecciosas, restrictions, to detect any published study identifying and we developed and validated a simplified probabilistic model Hospital General Universitario characterising phenotypes among patients with COVID-19. for phenotype assignment. This model is available as a tool Gregorio Marañón, Madrid, We found one study that identified three phenotypes in a online to facilitate the probabilistic classification of patients Spain (J Berenguer PhD); Instituto de Investigación cohort of 85 patients admitted to the intensive care unit, which with COVID-19 who are admitted to hospital into Sanitaria Gregorio Marañón, were correlated with mortality, and one preprint study in which phenotypes. Madrid, Spain (J Berenguer); phenotypes were investigated in ambulatory patients with Unidad de Enfermedades Implications of all the available evidence self-declaration of symptoms. We also found studies referring Infecciosas, Servicio de Identification of COVID-19 phenotypes allows investigation Medicina Interna, Hospital to distress syndrome-associated phenotypes or of potential differences in their underlying pathophysiological Universitario La Paz, Madrid, hyperinflammatory phenotypes. mechanisms, which could allow better pathogenesis-targeted Spain (J R Arribas) Added value of this study approaches for therapies in the design and selection of Correspondence to: Dr Jesús Rodríguez-Baño, To our knowledge, this is the first study investigating the participants in clinical trials, depending on the mechanism of Servicio de Enfermedades existence and characterisation of clinical phenotypes for action of specific drugs and their use in clinical management. Infecciosas, Hospital COVID-19 patients at hospital admission. We identified three Furthermore, phenotype assignment would be helpful in Universitario Virgen Macarena, distinct clinical phenotypes on the basis of demographics, identifying low-risk patients and patients who might need 41009 Seville, Spain jesusrb@us.es underlying conditions, clinical and laboratory data, and closer monitoring during admission. See Online for appendix For the online phenotype and laboratory data at presentation. If they exist, these Hospitals Virgen Macarena and Virgen del Rocío ethics assignment tool see http://fen- phenotypes might denote different pathophysiological committee (Seville, Spain), which waived the need covid.com/index.html routes and outcomes and be useful for better classifying to obtain written informed consent because of the patients for testing and treatment strategies. observational nature of the study. STROBE recom­ The objectives of this study were to determine whether mendations were followed (appendix pp 2–3). clinical phenotypes of patients with COVID-19 can be We discussed the objectives of the study, the study derived from clinical data, to assess their reproducibility design, and results with several health-care workers who and correlation with prognosis, and to derive and validate a had had COVID-19. simplified probabilistic model for phenotype assignment. Phenotype derivation Methods We considered 69 variables to derive the clinical Databases phenotypes. The variables were selected based on the In this study, we used data from two cohorts: available information about the features of patients the COVID-19@Spain cohort, a retrospective cohort admitted to hospital1–3 and the early clinical experience including 4035 consecutive adult patients admitted to gained at the participating sites. All data were collected at 127 hospitals in Spain with COVID-19 between Feb 2 and hospital admission and included age, sex, race or ethnicity, March 17, 2020, and the COVID-19@HULP cohort, comorbidities, drugs previously used for underlying including 2226 consecutive adult patients admitted diseases, COVID-19-related signs and symptoms at to a teaching hospital in Madrid between Feb 25 presentation, laboratory data, and chest radiographical and April 19, 2020. The cohort designs and patient data (table 1). As our objective was to explore the existence characteristics were previously reported in detail.4,5 of phenotypes, we did not preselect any variables. 41 patients in the COVID-19@HULP cohort who were The proportion of missing data per variable in the also included in the COVID-19@Spain cohort were COVID-19@Spain cohort is shown in the appendix (pp 5–6). excluded from the COVID-19@HULP cohort for the The Little MCAR test was used to verify that missing data current study (2185 remaining patients in this cohort). were at random, and imputation was done using the The COVID-19@Spain cohort was divided into a Markov chain Monte Carlo method. derivation cohort, comprising 2667 randomly selected Analyses to identify the phenotypes were first done in patients, selected using the SPSS function for selection of the derivation cohort. We assessed the distributions of random samples from a database, and an internal values and missing data, and correlation among the validation cohort, comprising the remaining 1368 patients. variables, using the χ² test and Pearson’s correlation The COVID-19@HULP cohort was used as an external coefficient for categorical and continuous variables, validation cohort. An overview of the analyses done in respectively. We excluded highly correlated variables. We the derivation and validation cohort is shown in the did a two-step cluster analysis using both continuous and appendix (p 19). The study was approved by the University categorical variables, which provided the optimal number 2 www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 Articles of clusters. We used silhouette analysis to assess the Features of the patients in the phenotypes obtained were quality of the cluster derivation. We did a sensitivity compared using χ² test and Kruskal-Wallis test for analysis excluding variables with more than 50% missing categorical and continuous variables, respectively. We data. visualised the patterns of distribution of the variables in Phenotype A vs phenotype C Phenotype B vs phenotype C OR (95% CI) p value OR (95% CI) p value Demographics Age (per year) 0·92 (0·90–0·93) <0·0001 0·96 (0·95–0·97) <0·0001 Female sex 1·79 (1·27–2·54) 0·0014 1·32 (0·97–1·82) 0·089 Race or ethnicity White 0·48 (0·06–4·16) 0·50 0·48 (0·06–3·62) 0·48 Black 0·80 (0·04–17·20) 0·89 0·10 (0·01–2·29) 0·15 Hispanic 2·08 (0·20–21·48) 0·54 1·08 (0·12–9·74) 0·94 Asian 0·20 (0·01–6·66) 0·37 0·55 (0·03–9·68) 0·68 Arab 0·50 (0·03–7·45) 0·61 0·40 (0·03–4·82) 0·47 Other 1 (ref) ·· 1 (ref) ·· Comorbidities Chronic heart disease 0·12 (0·08–0·17) <0·0001 0·23 (0·17–0·31) <0·0001 Hypertension 0·08 (0·05–0·12) <0·0001 0·19 (0·12–0·28) <0·0001 Chronic lung disease 0·19 (0·12–0·29) <0·0001 0·54 (0·39–0·74) <0·0001 Asthma 1·75 (0·83–3·66) 0·14 1·73 (0·87–3·44) 0·12 Chronic kidney disease (stage 4) 0·05 (0·03–0·10) <0·0001 0·06 (0·04–0·09) <0·0001 Liver cirrhosis 0·76 (0·23–2·56) 0·65 0·75 (0·26–2·13) 0·59 Chronic neurological disease 0·45 (0·27–0·76) 0·0031 0·65 (0·43–1·00) 0·057 Active solid malignancy 0·63 (0·34–1·17) 0·14 0·73 (0·43–1·24) 0·21 Active haematological malignancy 0·83 (0·29–2·43) 0·74 0·90 (0·35–2·29) 0·82 HIV/AIDS 1·52 (0·17–14·29) 0·71 1·92 (0·26–14·29) 0·53 Obesity (body-mass index >30 kg/m²) 0·28 (0·18–0·45) <0·0001 0·52 (0·37–0·74) 0·0043 Diabetes 0·12 (0·08–0·18) <0·0001 0·31 (0·23–0·42) <0·0001 Chronic inflammatory disease 1·33 (0·62–2·84) 0·47 1·20 (0·60–2·42) 0·60 Dementia 0·19 (0·10–0·35) <0·0001 0·57 (0·38–0·87) 0·0086 Malnutrition 0·40 (0·21–0·76) 0·012 0·51 (0·31–0·86) 0·016 Smoking status Never 2·67 (1·88–3·81) <0·0001 1·26 (0·93–1·71) 0·14 Current smoker 2·56 (1·37–4·79) 0·0037 1·11 (0·63–1·97) 0·71 Former smoker 1 (ref) ·· 1 (ref) ·· Treatments for underlying conditions Angiotensin converting enzyme inhibitors 0·39 (0·25–0·59) <0·0001 0·76 (0·54–1·08) 0·12 Angiotensin receptor blockers 0·41 (0·27–0·62) <0·0001 0·57 (0·40–0·80) 0·0010 Inhaled corticosteroids 0·40 (0·24–0·67) <0·0001 0·79 (0·53–1·19) 0·26 Systemic corticosteroids 0·61 (0·31–1·20) 0·15 0·69 (0·39–1·23) 0·22 Cancer chemotherapy 1·15 (0·41–3·23) 0·80 1·12 (0·45–2·86) 0·80 Biological drugs 1·08 (0·42–2·78) 0·87 0·65 (0·27–1·54) 0·32 Infection data at admission Non-focal symptoms Reported fever 1·85 (1·27–2·63) 0·0014 2·17 (1·59–3·03) <0·0001 Temperature (per 1°C) 0·93 (0·78–1·11) 0·41 1·25 (1·06–1·46) 0·0063 Myalgia or arthralgia 2·70 (1·69–4·17) <0·0001 2·27 (1·49–3·45) <0·0001 Headache 3·03 (1·69–5·56) <0·0001 1·33 (0·76–2·33) 0·32 Skin rash 0·95 (0·18–5·00) 0·95 1·10 (0·26–4·76) 0·89 Anosmia 3·51 (0·81–15·15) 0·098 1·77 (0·42–7·41) 0·44 Altered mental status 0·25 (0·15–0·43) <0·0001 0·67 (0·45–0·99) 0·042 (Table 1 continues on next page) www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 3 Articles Phenotype A vs phenotype C Phenotype B vs phenotype C OR (95% CI) p value OR (95% CI) p value (Continued from previous page) Inflammation White blood cells (per 10³ cells/µL) 0·79 (0·76–0·83) <0·0001 0·83 (0·80–0·86) <0·0001 Lymphocytes (per 10³ cells/µL) 1·11 (0·97–1·28) 0·14 0·99 (0·86–1·14) 0·87 Neutrophils (per 10³ cells/µL) 0·74 (0·70–0·78) <0·0001 0·82 (0·79–0·85) <0·0001 D-dimer (per 10³ µg/L) 0·79 (0·66–0·93) 0·0050 0·98 (0·95–1·01) 0·18 Procalcitonin (per 1 ng/mL) 0·09 (0·04–0·17) <0·0001 0·52 (0·44–0·61) <0·0001 C-reactive protein (per 102 mg/L) 0·92 (0·84–1·02) 0·11 0·97 (0·95–0·99) 0·0090 IL‐6 (per 10² µg/mL) 0·17 (0·11–0·27) <0·0001 1·00 (0·92–1·09) 0·97 Ferritin (per 10³ ng/mL) 0·19 (0·11–0·31) <0·0001 1·30 (0·89–1·89) 0·17 Cardiovascular Heart rate per minute (per unit) 1·00 (0·99–1·01) 0·69 1·01 (1·00–1·02) 0·15 Systolic blood pressure (per 1 mmHg) 1·00 (0·99–1·00) 0·56 1·00 (0·99–1·00) 0·62 Diastolic blood pressure (per 1 mmHg) 1·02 (1·01–1·04) <0·0001 1·02 (1·01–1·03) <0·0001 Respiratory tract Chest pain 1·45 (0·86–2·38) 0·16 0·98 (0·61–1·59) 0·94 Dyspnoea 0·19 (0·13–0·27) <0·0001 0·65 (0·48–0·88) 0·0061 Cough 1·22 (0·87–1·72) 0·25 1·61 (1·19–2·22) 0·0021 Expectoration 0·46 (0·32–0·68) <0·0001 0·74 (0·53–1·01) 0·055 Haemoptysis 0·51 (0·20–1·30) 0·16 0·48 (0·22–1·04) 0·062 Respiratory rate per min (per unit) 0·80 (0·77–0·83) <0·001 0·94 (0·92–0·97) <0·0001 Oxygen saturation, room air, pulse oximetry (per 1·62 (1·55–1·70) <0·0001 1·09 (1·07–1·11) <0·0001 1%) Oxygen saturation after oxygen supplementation 1·35 (1·26–1·45) <0·0001 1·07 (1·03–1·11) <0·0001 (per 1%) Oxygen saturation, room air, venous blood 1·07 (1·05–1·09) <0·0001 1·03 (1·02–1·03) <0·0001 (per 1%) PCO2, venous blood (per 1 mmHg) 1·01 (0·99–1·02) 0·61 0·98 (0·96–0·99) 0·022 Lung infiltrates on chest radiography No infiltrate 3·43 (2·32–5·08) <0·0001 0·77 (0·54–1·10) 0·15 Unilateral 2·25 (1·44–3·51) <0·0001 1·16 (0·78–1·72) 0·45 Bilateral 1 (ref) ·· 1 (ref) ·· Interstitial lung infiltrate 0·48 (0·34–0·68) <0·0001 1·18 (0·88–1·59) 0·27 Ground-glass opacity infiltrate 0·74 (0·43–1·25) 0·26 1·19 (0·75–1·85) 0·46 Liver Albumin, mean (SD; per 1 g/dL) 9·81 (6·46–14·87) <0·0001 3·37 (2·37–4·79) <0·0001 Lactic acid dehydrogenase (per 10² U/L) 0·61 (0·55–0·68) <0·0001 1·00 (0·96–1·04) 1·00 Bilirubin (per 1 mg/dL) 0·93 (0·77–1·13) 0·49 1·00 (0·96–1·04) 0·87 Renal Creatinine (per 1 × mg/dL) 0·10 (0·07–0·15) <0·0001 0·13 (0·10–0·17) <0·0001 Sodium (per 1 × mEq/L) 1·07 (1·03–1·11) 0·0011 1·00 (0·97–1·04) 0·78 Potassium (per 1 × mEq/L) 0·25 (0·18–0·34) <0·0001 0·24 (0·18–0·31) <0·0001 Haematological Haemoglobin (per 1 × g/dL) 1·66 (1·53–1·81) <0·0001 1·60 (1·48–1·72) <0·0001 Haematocrit (per 1%) 1·19 (1·15–1·23) <0·0001 1·16 (1·13–1·20) <0·0001 Platelets (per 10⁵/µL) 0·86 (0·76–0·99) 0·031 0·80 (0·71–0·90) <0·0001 Activated partial thromboplastin time (per 1 × s) 0·99 (0·98–1·00) 0·038 0·99 (0·99–1·00) 0·075 International normalised ratio (per unit) 0·18 (0·12–0·28) <0·0001 0·32 (0·26–0·40) <0·0001 Other Creatine phosphokinase (per 10² × U/L) 1·01 (0·95–1·08) 0·71 1·02 (0·96–1·08) 0·50 Blood glucose (per 1 × mg/dL) 0·98 (0·98–0·98) <0·0001 0·99 (0·99–0·99) <0·0001 OR=odds ratio. Table 1: Bivariate analysis of variables associated with phenotypes in the derivation cohort 4 www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 Articles the different phenotypes using chord diagrams and phenotypes, such as age and oxygen saturation,4 we did a heatmaps after grouping variables into comorbidities and stratified analysis to check if any mortality association system-related or organ-related data (appendix p 4). A was maintained in all strata of these variables. two-step cluster analysis was also done in the internal All analyses were done with IBM SPSS Statistics 26, validation cohort to check the reproducibility of phenotype SPM 8.2, and R version 3.6.0. identification. Role of the funding source Derivation and validation of a parsimonious The funders of the study had no role in study design, probabilistic model for phenotypes data collection, data analysis, data interpretation, or As the number of variables used to derive the phenotypes writing of the report. was very high, assigning patients to phenotypes was neither intuitive nor applicable for clinical practice. Results Therefore, we developed a simplified probabilistic model The features of the patients in the cohorts used for this to assign patients into the pheno­types. As we identified study were previously reported in detail.4,5 A two-step three phenotypes, we did a multinomial logistic regression cluster analysis of variables collected at hospital admission analysis in the derivation cohort. First, we analysed the identified three clinical phenotypes in the derivation bivariate association of each variable of the phenotypes cohort: phenotype A (516 [19%] of 2667 patients), using the χ² and Kruskal-Wallis tests for categorical and phenotype B (1955 [73%] of 2667 patients), and phenotype C continuous variables, respectively. Those with p<0·20 (196 [7%] of 2667 patients). The silhouette score was 0·6, were included in a multinomial logistic regression model; indicating good quality of clustering. Exclusion of variables the variance inflation factor value was used to detect the with a high proportion of missing data did not cause any potential occurrence of collinearity and interactions evident changes (data not shown). were tested. The variables were selected using a manual The baseline characteristics of the derivation and internal backward selection process. The ability of the final model validation cohorts are present in the appendix (pp 7–12). to predict the phenotypes as identified by the derivation Overall, patients with phenotype A were younger (mean process was checked by calculating the area under the age 55·2 years [SD 18·4] vs 68·7 years [15·9] and 77·2 years receiver operating characteristic curves (AUROC) with [10·9] in phenotypes B and C, respectively), were less 95% CIs for the three phenotypes. We also tested the frequently male (55% vs 63% and 69%), presented more predictive ability of the model in 60 randomly chosen frequently with headache (19% vs 9% and 7%), myalgia subcohorts (using a tool in the SPM software) with 80%, (29% vs 26% and 13%), and chest pain (15% vs 11% and 60%, or 40% of the sample size of the derivation cohort. 11%), had higher lymphocyte count (mean 1439 cells/µL The probabilistic model for phenotype assignment was [SD 1761] vs 1094 cells/µL [1424] and 1096 cells/µL [1170]), used in two ways. First, we applied the model to the and had lower levels of inflammatory parameters internal validation cohort to check its ability to predict the such as C-reactive protein, IL-6, ferritin, or lactic phenotypes obtained from this cohort. Second, we applied acid dehydrogenase (appendix pp 7–9). Patients with the model to both the internal and external validation phenotype B more frequently reported fever (83% vs 80% cohorts to obtain a probabilistic assignment of patients to and 69% in phenotypes A and C, respectively) and cough the phenotypes (model-derived formulae used for (74% vs 68% and 63%), more frequently lacked pulmonary probability calculations are in the appendix p 4). Patients infiltrates in chest radiography (20% vs 46% and 25%), were assigned to the phenotype with the highest belonging more frequently had interstitial infiltrates (45% vs 25% and probability according to the model-derived formula. We 41%), and had higher levels of ferritin (mean 809·5 ng/mL checked the distribution of variables among the assigned [SD 588·4] vs 616·4 ng/mL [219·7] and 752·8 ng/mL phenotypes. [320·4]) and creatine phosphokinase (mean 164·3 U/L [SD 464·0] vs 150·8 U/L [368·2] and 141·4 U/L [199·0]; appendix Prognostic assessment of the phenotypes pp 7–9). Patients with phenotype C more frequently had We compared the 30-day mortality of patients in the chronic heart disease (56% vs 13% and 23% in phenotypes different phenotypes in the derivation cohort with A and B, respectively), hypertension (86% vs 31% and Kaplan-Meier curves and log-rank tests, and calculated 53%), chronic lung disease (31% vs 8% and 19%), stage 4 hazard ratios (HRs) with 95% CIs. We also collected data chronic kidney disease (34% vs 3% and 3%), obesity (body- on complications that occurred during treatment in mass index >30 kg/m²; 23% vs 8% and 14%), diabetes (48% hospital (listed in the appendix p 9). These variables were vs 10% and 22%), and acute altered mental status (18% vs also analysed in the validation cohorts, in which patients 5% and 12%); had higher levels of neutrophils (mean were assigned to the phenotype with the highest 8539 cells/µL [SD 6656] vs 4112 cells/µL [2511] and probability according to the probabilistic model-derived 4892 cells/µL [2844]), D-dimer (mean 1343·1 µg/L [SD formula. Since any association of phenotypes with 2419·7] vs 715·8 µg/L [986·5] and 986·3 µg/L [3290·5]), mortality might be caused by a different distribution of a procalcitonin (mean 0·70 ng/mL [SD 0·96] vs 0·17 ng/mL few strong independent prognostic variables in the [0·26] and 0·27 ng/mL [0·51]), C-reactive protein (mean www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 5 Articles Respiratory ascular Cardiov 10 0 10 0 0 Live n r tio 10 ma 10 am Infl 0 0 Re na l al 0 foc 1 n− No 10 0 Ha 0 em ato 10 ies log 10 idit ica l orb Com 0 0 Othe 10 phics r Demogra 10 0 0 pe A 100 Phenoty 10 90 0 10 80 20 70 B pe oty 60 en 30 Ph Phe not 50 0 ype C 04 40 10 30 20 Figure 1: Chord diagram of the distribution of groups of variables in the phenotypes in the derivation cohort Variables are grouped into categories. The phenotypes are shown in different colours: phenotype A is green, phenotype B is blue, and phenotype C is red. For each phenotype, if a variable mean (for continuous variables) or proportion (for categorical variables) is significantly different to the mean or proportion in the full derivation cohort, a ribbon connects the phenotype and the variable group. The width of the ribbons correlates with the number of variables that are significantly different from those in the derivation cohort for that phenotype. 127·1 mg/L [SD 119·8] vs 47·4 mg/L [68·3] and 88·8 mg/L 0·6, and the distribution of variables in the phenotypes [84·2]), creatinine (mean 2·76 mg/dL [SD 2·11] vs was as in the derivation cohort, except for the proportion of 0·96 mg/dL [0·56] and 0·99 mg/dL [0·36]), and potassium patients with liver cirrhosis and active solid malig­nancies (mean 4·5 mEq/L [SD 0·7] vs 4·0 mEq/L [0·5] and (which were not significantly different in the derivation 4·0 mEq/L [0·5]); and had poorer oxygenation parameters cohort but were more frequent in phenotype C than in (appendix pp 7–9, figures 1, 2). phenotype A or phenotype B in the internal validation We repeated the two-step cluster analysis in the internal cohort), haematological malignancy (no difference in the validation cohort. This analysis also selected three clusters derivation cohort but less frequent in phenotype A than in with a very similar distribution of patients to the derivation phenotypes B and C in the internal validation cohort), and cohort: phenotype A (233 [17%] of 1368 patients), phe­no­ ferritin and creatine phosphokinase concentrations (which type B (1019 [74%] of 1368 patients), and phenotype C were higher in phenotype B than in phenotypes A and C in (116 [8%] of 1368 patients). The silhouette score was also the derivation cohort and in phenotype C than in 6 www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 Articles phenotypes A and B in the internal validation cohort; appendix pp 10–12). Age To develop a simple way to assign patients to a phenotype, Demographics and non-focal symptoms we developed and validated a parsimonious probabilistic Temperature model for belonging to phenotypes. We first did a bivariate White blood cells analysis of the association of the different variables with Lymphocytes phenotype A versus phenotype C and phenotype B versus Neutrophils phenotype C in the derivation cohort. We found a Inflammation D-dimer Procalcitonin significant crude association with phenotype for many C-reactive protein variables (table 1). After a variable selection process, we Interleukin-6 developed a final multinomial logistic regression model Ferritin with 16 variables, including age, sex, chronic lung disease, Heart rate Cardiovascular Systolic blood pressure obesity, diastolic blood pressure, oxygen saturation Diastolic blood pressure (room air), white blood cell count, neutrophils, haematocrit, Respiratory rate coagulation international normalised ratio, C-reactive Oxygen saturation, room air pulse oximetry protein, glucose, creatinine, sodium, potassium, and type Respiratory tract Oxygen saturation,after oxygen of lung infiltrate on chest radiograph (table 2). Therefore, supplementation we derived a simplified probabilistic model for patient PCO2 assignment to phenotypes. The AUROC of the model for Oxygen saturation room air, venous blood the observed data in the derivation cohort showed Albumin very good predictive ability for the three phenotypes Liver Lactic acid dehydrogenase Bilirubin (0·86, 95% CI 0·85–0·88 for phenotype A, 0·88, Creatinine 0·86–0·89 for phenotype B, and 0·99, 0·99–0·99 for Renal Sodium phenotype C). The predictive ability was similar in smaller, Potassium randomly selected subcohorts (appendix p 13). Haemoglobin Haematocrit The capacity of the model to correctly assign patients to Haematological Platelets phenotypes was validated in the internal validation Activated partial thromboplastin time cohort for the phenotypes directly derived from that International normalised ratio Creatine phosphokinase cohort. The ability of the model to predict the observed Other phenotypes in the internal validation cohort was also Blood glucose high (AUROC 0·86, 95% CI 0·84–0·89 for phenotype A; A B C 0·86, 0·84–0·88 for phenotype B; and 0·95, 0·93–0·98 Phenotype for phenotype C; appendix p 22). –1 –0·8 –0·6 –0·4 –0·2 Mean 0·4 0·8 1·2 1·6 2 The probabilistic model was then applied to the internal and external validation cohorts to obtain the Figure 2: Heatmap of the distribution of continuous variables in the individual probability of being assigned a specific phenotypes in the derivation cohort phenotype. The number of patients in the internal A colour gradient is used to show differences in mean values in relation to the validation cohort assigned to phenotypes A, B, and C full derivation cohort, towards red for higher values and blue for lower values. The colour gradient indicates the number of SDs that the mean value in the according to their highest probability were 263 (19%), subcohort of interest is below or above the mean value in the full cohort. 1021 (75%), and 84 (6%), respectively (appendix pp 14–15). The corresponding figures for the external validation cohort were 323 (15%), 1757 (80%), and (95% CI 1·0–5·6) for phenotype A, 31·0% (28·2–33·9) for 105 (5%; appendix p 16). In the internal validation phenotype B, and 53·4% (44·4–62·2) for phenotype C cohort, the distribution of all variables in the three (appendix p 17). Regarding the phenotypes assigned on the predicted phenotypes was similar to that in the basis of the probabilistic model, the mortality rates in the derivation cohort (appendix pp 14–15). For the external internal validation cohort were 5·3% (95% CI 3·4–8·1) validation cohort, not all variables collected in the for phenotype A, 31·3% (28·5–34·2) for phenotype B, derivation cohort were available. Therefore, we checked and 59·5% (48·8–69·3) for phenotype C (figure 3; the distribution of the variables included in the model, appendix p 17) and in the external validation cohort were which was similar to that in the derivation cohort 3·7% (2·0–6·4) for phenotype A, 23·7% (21·8–25·7) for (appendix p 16). phenotype B, and 51·4% (41·9–60·7) for phenotype C (the In the derivation cohort, 30-day mortality rates were external validation cohort only had in-hospital mortality 2·5% (95% CI 1·4–4·3) for patients with phenotype A, and not 30-day mortality data; figure 3; appendix p 17). All 30·5% (28·5–32·6) for patients with phenotype B, and mortality data are summarised in the appendix (p 17). 60·7% (53·7–67·2) for patients with phenotype C (figure 3; The proportion of patients in the derivation cohort who appendix p 17). In the internal validation cohort, the needed intensive care unit care or had transfusion- mortality in the reproduced phenotypes was 2·6% requiring anaemia, pleural effusion, acute kidney failure, www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 7 Articles Phenotype A vs phenotype C Phenotype B vs phenotype C OR (95% CI) p value OR (95% CI) p value Age (per year) 0·93 (0·90–0·96) <0·0001 0·96 (0·93–0·99) 0·0051 Female sex 0·68 (0·33–1·41) 0·30 0·44 (0·22–0·89) 0·021 Chronic lung disease 0·55 (0·26–1·16) 0·10 0·79 (0·42–1·54) 0·48 Obesity (body-mass index >30 kg/m²) 0·49 (0·20–1·23) 0·12 0·71 (0·31–1·64) 0·42 White blood cells (per 10³ cells/µL) 0·80 (0·73–0·87) <0·0001 0·73 (0·68–0·79) <0·0011 Neutrophils (per 10³ cells/µL) 0·89 (0·80–0·99) 0·032 0·99 (0·90–1·08) 0·86 C-reactive protein (per 10² mg/L) 0·95 (0·91–1·00) 0·055 0·94 (0·90–0·99) 0·011 Diastolic blood pressure (per 1 mmHg) 1·03 (1·01–1·05) 0·011 1·02 (1·01–1·04) 0·013 Oxygen saturation, room air, pulse oximetry (per 1%) 1·56 (1·46–1·66) <0·0001 1·11 (1·07–1·16) <0·0001 Lung infiltrate on chest radiography No infiltrate 4·07 (1·83–9·02) 0·00055 1·17 (0·55–2·49) 0·69 Unilateral 3·50 (1·51–8·06) 0·0032 2·05 (0·93–4·51) 0·071 Bilateral 1 (ref) ·· 1 (ref) ·· Creatinine (per 1 mg/dL) 0·09 (0·05–0·15) <0·0001 0·06 (0·04–0·10) <0·0001 Sodium (per 1 mEq/L) 1·09 (1·02–1·17) 0·010 1·04 (0·98–1·11) 0·14 Potassium (per 1 mEq/L) 0·37 (0·21–0·67) 0·00093 0·26 (0·15–0·45) <0·0001 Haematocrit (per 1%) 1·29 (1·21–1·38) <0·0001 1·27 (1·19–1·35) <0·0001 International normalised ratio (per unit) 0·12 (0·07–0·22) <0·0001 0·12 (0·08–0·18) <0·0001 Blood glucose (per 1 mg/dL) 0·99 (0·98–0·99) <0·0001 0·99 (0·98–0·99) <0·0001 The variance inflation factor value was less than 2 in all cases. OR=odds ratio. Table 2: Multinomial logistic regression model for the prediction of phenotypes in the derivation cohort heart failure, bacterial pneumonia, acute respiratory Clinical phenotypes have been described for patients distress syndrome, or cardior­ espiratory arrest during with severe acute respiratory distress with potential admission was significantly increased in phenotype C implications for respiratory support therapy.9 Phenotypes compared with phenotypes A and B and significantly based only on self-declaration of symptoms by non- decreased in phenotype A compared with phenotypes B hospitalised patients with COVID-19 using an app have and C; differences were not significant for stroke, been reported.10 Clinical phenotypes have been identified ischaemic coronary event, liver failure, or disseminated in patients with sepsis,11 and a so-called hyperinflammatory intravascular coagulation (appendix pp 7–9). Results were phenotype has been proposed in patients with similar in the internal validation cohort, with the exception COVID-19.12,13 However, to our knowledge, only one other that liver failure was more frequent in phenotype B study14 has specifically investigated the existence of (appendix p 15). diverse clinical phenotypes for patients with COVID-19 at To check whether the association of the phenotypes hospital admission; three phenotypes were also identified with mortality was maintained after considering different in that study14 on the basis of clinical and laboratory distributions of strong mortality predictors across the features, using hierarchical clustering in 85 patients phenotypes, such as age and oxygen saturation, we did a admitted to the intensive care unit, with a small number stratified analysis per strata of these variables in the of variables. derivation cohort. In all strata, phenotypes were In our study, the phenotypes we identified were significantly associated with mortality (appendix p 18). associated with patient prognosis. By contrast with studies that generate outcome prediction scores or Discussion identify outcome predictors, in which the independent We identified three phenotypes based on demographics, predictive association of each variable with the outcome underlying conditions, clinical and laboratory data, and is assessed, phenotypes provide information about how radiological features at presentation among patients the population can be classified according to clustering admitted to hospital with COVID-19. The phenotypes, of variables and how such clusters are associated with the despite not intended to be used for predicting mortality, outcome. As age and oxygen saturation are strong had clinical implications, as we observed associations independent predictors of mortality,4 we did a stratified with patient prognosis. We also developed a simplified analysis of these variables. The results of this analysis probabilistic model that is potentially applicable to other suggest that the association of phenotypes with mortality cohorts. is not only due to the different distribution of these Clinical presentation of COVID-19 is polymorphic. variables in the phenotypes, but that the phenotypes are 8 www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 Articles consistently associated with different mortality risks. A However, the phenotypes are not expected to provide 100 Phenotype accurate prediction of prognosis, as done by predictive A modelling, as the outcome rates in the phenotypes B depend on the exact distribution of the strongest outcome 80 C Cumulative incidence of death (%) predictors in each population to which the phenotypes HR phenotype C (vs phenotype A)=37·91 (95% CI 21·38–67·23), are applied. In this sense, phenotypes are complementary log-rank p<0·0001 60 to predictive scores. Beyond that, the phenotypes might reflect different profiles of pathogen and host interactions, as a consequence of different infecting viral 40 HR phenotype B (vs phenotype A)=14·10 (95% CI 8·14–24·42), load, natural or acquired humoral and cellular immune log-rank p<0·0001 response against SARS-CoV-2, or cell–receptor features 20 and expression, alongside host genetic background.6–8 Since the databases used in this study only included phenotypic profiles and manifestations, we cannot 0 provide information about underlying immunological or Number at risk virological mechanisms. Future studies could reproduce A 516 513 508 507 506 505 503 the phenotypes and investigate their correlations with B 1955 1822 1612 1481 1420 1388 1364 C 196 147 108 92 79 77 77 virological, immunological, and genetic data. We did not analyse the duration of disease at hospital B admission because the start of symptoms can be difficult 100 to assess in many patients and can be confused with manifestations related to chronic conditions; in our 80 experience, this is particularly frequent in older patients Cumulative incidence of death (%) with comorbidities. The duration of symptoms could be HR phenotype C (vs phenotype A)=17·71 (95% CI 9·79–32·06), log-rank p<0·0001 relevant to differentiate between the viral and infla­ 60 mmatory phases of the disease,13 but a clear cutoff in the number of days to differentiate between the phases 40 HR phenotype B (vs phenotype A)=6·81 (95%CI 3·98–11·64), cannot currently be defined. log-rank p<0·0001 Classification of patients into phenotypes might be useful to design treatment strategies. Very low-risk patients 20 (eg, those with phenotype A who are younger than 60 years or with oxygen saturation >95%), who would need lower 0 degrees of watchfulness and care, might be identified and Number at risk discharged for ambulatory follow-up. Patients without A 263 261 256 252 251 250 249 initial criteria for being admitted to the intensive care unit B 1021 951 831 760 733 717 702 C 84 62 45 40 37 35 34 but with phenotype B or phenotype C could be closely monitored during admission. As some aspects of the C pathophysiology of the infection in patients with different 100 phenotypes might be different, the therapeutic approach might need to be tailored on a patient-by-patient basis.15 Since phenotype C comprises patients with laboratory 80 Cumulative incidence of death (%) parameters suggestive of a hyperinflammatory state, such patients might be selected to investigate the efficacy of 60 anti-inflammatory drugs. This strategy would allow more HR phenotype C (vs phenotype A)=17·91 (95% CI 9·58–33·49), log-rank p<0·0001 specific and efficient design of randomised trials. However, whether these phenotypes are useful for clinical purposes 40 requires further investigation of the underlying mechanisms and more specific studies. HR phenotype B (vs phenotype A)=6·75 (95% CI 3·81–11·98), 20 log-rank p<0·0001 Since the phenotypes were identified using a high number of variables, it would be difficult to apply 0 0 5 10 15 20 25 30 Follow-up (days) Number at risk Figure 3: Probability of death up to day 30 according to phenotypes in the A 323 303 301 297 294 292 292 derivation cohort (A), internal validation cohort (B) and external validation B 1757 1668 1517 1440 1396 1363 1344 cohort (C) C 105 90 70 65 58 56 52 HR=hazard ratio. www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0 9 Articles them clinically in the absence of automated big data approval of a proposal by the REIPI-SEIMC COVID-19 and COVID@ management. Therefore, we developed and validated a HULP groups boards, with a signed data access agreement. simplified probabilistic prediction model for phenotype Acknowledgments assignment. A publicly available calculator and app have The study was funded by Instituto de Salud Carlos III, Spanish Ministry of Science and Innovation (COV20/01031), and Fundación SEIMC/GeSIDA. been developed to facilitate the classification of patients Additionally, JR-B, BG-G, JB, IJ, JC, JP, and JRA received funding for admitted to hospital with COVID-19 into phenotypes, research from Plan Nacional de I+D+i 2013–2016 and Instituto de Salud using the probabilistic model for phenotype assignment. Carlos III, Subdirección General de Redes y Centros de Investigación Limitations of our study are the high proportion of Cooperativa, Ministerio de Ciencia, Innovación y Universidades (cofinanced by European Development Regional Fund “A way to achieve patients classified into phenotype B, reflecting the Europe”), and operative programme Intelligent Growth 2014–2020 through profile of the patients admitted during the first weeks the following networks: Spanish Network for Research in Infectious of the epidemic in saturated hospitals, the exclusive Diseases (RD16/0016/0001 [JR-B, BG-G, MDdT], RD16/0016/0005 [JC], participation of Spanish hospitals, and the high proportion and RD16/0016/0009 [JP]) and Spanish AIDS Research Network (RD16/0025/0017 [JB], RD16/0025/0018 [JRA], RD16/0025/00XX [IJ]). of missing data for several variables. Hospital admission We thank Alejandro González-Herrero for programming of the web tool criteria might be different in other countries or at different and app. This study was presented at the ESCMID Conference on times during the pandemic; however, the cohorts we used Coronavirus Disease, Sept 23–25, 2020. included patients with varying severity of disease. Some References symptoms might not have been reported by the most 1 Guan WJ, Ni ZY, Hu Y, et al. Clinical characteristics of coronavirus disease 2019 in China. N Engl J Med 2020; 382: 1708–20. severely ill patients. Finally, the phenotypes were derived 2 Richardson S, Hirsch JS, Narasimhan M, Crawford JM, McGinn T, and validated at hospital admission and would be useful Davidson KW. Presenting characteristics, comorbidities, and for decisions at that time; whether changes in evolution outcomes among 5700 patients hospitalized with COVID-19 in the New York city area. JAMA 2020; 323: 2052–59. due to the natural history of the disease or the influence of 3 Docherty AB, Harrison EM, Green CA, et al. Features of 20 133 UK treatments modify the phenotype assignment needs patients in hospital with covid-19 using the ISARIC WHO Clinical further study. Strengths of our study include the use of Characterisation Protocol: prospective observational cohort study. BMJ 2020; 369: m1985. well characterised cohorts, the inclusion of a high number 4 Berenguer J, Ryan P, Rodríguez-Baño J, et al. Characteristics and of variables from different domains, and the validation. predictors of death among 4035 consecutively hospitalized patients In conclusion, patients admitted to hospital with with COVID-19 in Spain. Clin Microbiol Infect 2020; 26: 1525–36. COVID-19 can be classified into phenotypes that have 5 Borobia AM, Carcas AJ, Arnalich F, et al. A cohort of patients with COVID-19 in a major teaching hospital in Europe. J Clin Med 2020; prognostic implications. We developed a simplified tool 9: 1733. for the probabilistic classification of patients into pheno­ 6 Magleby R, Westblade LF, Trzebucki A, et al. Impact of SARS-CoV-2 types. Further studies are needed to elucidate the viral load on risk of intubation and mortality among hospitalized patients with coronavirus disease 2019. Clin Infect Dis 2020; underlying pathophysiological mechanisms leading to a published online June 30. https://doi.org/10.1093/cid/ciaa851. particular phenotype. 7 Mateus J, Grifoni A, Tarke A, et al. Selective and cross-reactive SARS-CoV-2 T cell epitopes in unexposed humans. Science 2020; Contributors 370: 89–94. JR-B, BG-G, MDdT, JP, JC, PR, IJ, MY, JRA, and JB conceived and 8 Ellinghaus D, Degenhardt F, Bujanda L, et al. Genomewide designed the study. All members of the REIPI-SEIMC COVID-19 association study of severe Covid-19 with respiratory failure. and COVID@HULP groups acquired the study data. JR-B, BG-G, N Engl J Med 2020; 383: 1522–34. MDdT, AMB, AC, JP, JC, PR, IJ, MY, JRA, and JB analysed and 9 Gattinoni L, Chiumello D, Caironi P, et al. COVID-19 pneumonia: interpreted the data. JRB, BGG, and MDdT drafted the manuscript. different respiratory treatments for different phenotypes? Critical revisions to the manuscript were made by all members of the Intensive Care Med 2020; 46: 1099–102. REIPI-SEIMC COVID-19 Group. JR-B, PR, JB, and JRA obtained the 10 Sudre CH, Lee KA, Lochlainn MN, et al. Symptom clusters in funds for the study. BG-G and JR-B verified all data. All authors had full Covid19: a potential clinical prediction tool from the COVID access to all the data in the study and had final responsibility for the Symptom study app. medRxiv 2020; published online June 16. decision to submit for publication. https://doi.org/10.1101/2020.06.12.20129056 (preprint). Declaration of interests 11 Seymour CW, Kennedy JN, Wang S, et al. Derivation, validation, and IJ received honoraria for participating in an advisory board for Gilead potential treatment implications of novel clinical phenotypes for sepsis. JAMA 2019; 321: 2003–17. Sciences and educational activities for ViiV. JB received research grants from AbbVie, Gilead Sciences, Merck, and ViiV, and honoraria for speaker 12 Sinha P, Calfee CS, Cherian S, et al. Prevalence of phenotypes of acute respiratory distress syndrome in critically ill patients with or advisory board participation from AbbVie, Gilead Sciences, Janssen, COVID-19: a prospective observational study. Lancet Respir Med 2020; Merck, and ViiV. JRA received fees for advisory board and speaker 8: 1209–18. participation and research grant support from Viiv, Janssen, Gilead 13 Manson JJ, Crooks C, Naja M, et al. COVID-19-associated Sciences, Merck, Teva, Alexa, and Serono. PR has received research grants hyperinflammation and escalation of patient care: a retrospective from Gilead Sciences and Merck, and honoraria for speaker or advisory longitudinal cohort study. Lancet Rheumatol 2020; 2: e594–602. board participation from Gilead Sciences, AbbVie, and ViiV. All other 14 Azoulay E, Zafrani L, Mirouse A, Lengliné E, Darmon M, Chevret S. authors declare no competing interests. Clinical phenotypes of critically ill COVID-19 patients. Data sharing Intensive Care Med 2020; 46: 1651–52. Data collected for the study, including deidentified participant data and a 15 Cao W, Li T. COVID-19: towards understanding of pathogenesis. data dictionary defining each field in the set, will be made available to Cell Res 2020; 30: 367–69. other investigators upon request to the corresponding author, after 10 www.thelancet.com/infection Published online February 23, 2021 https://doi.org/10.1016/S1473-3099(21)00019-0