LAUSR

Results from the recently published meta-analysis by Vai et al. [1]. indicate that psychiatric disorders may be associated with an increased risk of death after SARS-CoV-2 infection (pooled unadjusted OR= 2.00, 95%CI= 1.58–2.54; 23 studies including 43,938 participants with any psychiatric disorder and 1,425,793 control participants). These findings suggest that psychiatric disorders per se may be risk factors of death in COVID-19. However, a critical limitation for interpreting these findings is that only 9 of 23 studies included in that meta-analysis adjusted for a limited number of comorbid medical conditions. Because comorbid medical illnesses are more prevalent in people with psychiatric disorders than in the general population [2] and are associated with increased risk of COVID-19-related mortality [3, 4], this suggests that the association between psychiatric disorders and increased mortality in patients with COVID-19 may be confounded by medical comorbidities. To examine the potential influence of comorbid medical conditions in the relationship between psychiatric disorders and risk of COVID19-related mortality, we examined the association between an International Classification of Diseases, Tenth Revision (ICD-10) diagnosis of psychiatric disorder (F01-F99) and mortality in a large multicenter retrospective observational study of patients hospitalized in Greater Paris University hospitals for laboratory-confirmed COVID19 between January 24th, 2020, and May 1st, 2021. Study design and ethical approval are detailed elsewhere [5–7]. A study flowchart is provided in Supplementary Fig. 1. Multivariable logistic regression models adjusting successively for age, sex, hospital, period of hospitalization, current hospitalization status, and number of ICD-10 medical conditions were used. We also performed sensitivity analyses. First, we reproduced the analyses while adjusting for the Elixhauser Comorbidity Index (ECI) instead of the number of ICD-10 medical conditions. This index is a method of categorizing comorbidities of patients based on ICD-10 diagnosis codes in which each comorbidity category is dichotomous, i.e. either present or not present. It was initially designed to predict hospital resource use and in-hospital mortality [8]. Second, to assess the potential effect of unmeasured confounding, we computed E-values [9]. The E-value quantifies the minimum strength of association that an unmeasured confounder must have with both the predictor and the outcome, while simultaneously considering the measured covariates, to negate the observed association [9]. The lowest possible E-value is 1 (i.e., no unmeasured confounding is needed to explain away the observed association). Of 49,089 adult patients hospitalized for COVID-19, 3768 (7.7%) had an ICD-10 diagnosis of psychiatric disorder. Distributions of age, sex, hospital, period of hospitalization, current hospitalization status, and medical comorbidities according to a psychiatric disorder diagnosis, and their associations with mortality are detailed in Supplementary Tables 1–3 and Supplementary Fig. 2. During a median follow-up period of 45 days (SD= 122.1), death occurred in 1001 of 3768 (26.6%) patients with a psychiatric disorder diagnosis versus 3,780 of 45,321 (8.3%) in patients without this diagnosis (OR= 3.98; 95%CI= 3.67–4.31; p < 0.001). Similar to the results of Vai et al [1], this association remained significant but was substantially reduced when adjusting for age and sex (AOR= 1.71; 95%CI= 1.57–1.86; p < 0.001; degrees of freedom (df)= 7). Additional adjustment for hospital, period of hospitalization, and current hospitalization did not alter the significance of the association (AOR= 1.35; 95%CI= 1.24–1.48; p < 0.001; df= 10). However, when further adjusting for the number of medical conditions, this association was significant and reversed (AOR= 0.87; 95%CI= 0.79–0.96; p= 0.005; df= 12) (Fig. 1; Supplementary Table 4). The main results were not substantially modified when adjusting for the Elixhauser Comorbidity Index instead of the