RESULTS
During the study period, 380,089 people tested positive. A flowchart describing patient evolution is shown in Figure 1. The descriptive data of the entire sample is available in online supplementary Table 1.
The variables identified in the multivariable model related to death were greater age; being male; baseline diseases such as peripheral vascular disease, heart failure, heart, cerebrovascular, liver, and kidney disease, dementia, diabetes, hemiplegia, specific lung diseases such as interstitial pulmonary disease and cystic fibrosis; and history of malignant tumors. Among the basal treatments, use of diuretics and chronic systemic steroids were also related to death. We created a score from 0 to 68, with four categories and cut-off points at 23, 33 and 41 points. The AUCs for the categorized score were 0.9381, 0.9383 and 0.9384, in train, test and Omicron samples, respectively (Table 1).
The variables related to adverse evolution identified in the multivariable model were older age; being male; baseline diseases such as, heart failure, heart and cerebrovascular disease, dementia, kidney disease, diabetes, specific lung diseases such as interstitial pulmonary disease; and history of malignant tumors. Among the basal treatments, the use of diuretics and chronic systemic steroids were also related to adverse evolution. We created a score from 0 to 59, with four categories and cut-off points at 14, 21 and 29 points. The AUCs for this model were 0.8789, 0.8717 and 0.8990, in the train, test and Omicron samples, respectively. (Table 2).
Finally, the variables related to hospital admission identified in the multivariable model were older age; being male; baseline diseases such as heart failure, heart, cerebrovascular, liver and kidney disease, arterial hypertension, dyslipidemia, diabetes, specific lung diseases such as interstitial pulmonary disease and cystic fibrosis;HIV; and history of malignant tumors. Among the basal treatments, the use of NSAIDs, heparin, bronchodilators, immunosuppressants, diuretics and chronic systemic steroids were also related to hospital admission. We created a score from 0 to 54, with four categories and cut-off points at 13, 19, and 26 points. The AUCs of this model were 0.7879, 0.7852 and 0.7968, in the train, test and Omicron samples, respectively. (Table 3).
For all different models and cut points, we estimated the sensitivity, specificity and Net Benefit percentages (Table 4) while the risk/probability of event was represented for each outcome and risk category (Figure 2).