RESULTS
Between March 9th and May 6th, 2020, a total of 1370 patients presented at the Emergency Department (ED) of Saint-Louis Hospital. Out of these, 1364 underwent SARS-CoV-2 testing via PCR, and 370 were found to have a positive nasopharyngeal sample for SARS-CoV-2. Leftover samples were available for 295 patients, comprising 92 individuals who were discharged to their homes, 160 who were admitted to a medical ward, and 43 who were admitted to the medical intensive care unit ()ICU.
Patients differed across the different admission groups with regard to age, comorbidities, duration since the onset of initial symptoms, presence of dyspnea, and requirement for oxygen (Table 1). The TTV load in nasopharyngeal samples obtained upon ED admission exhibited a notably higher median value among patients admitted to the ICU (Median: 3.02 log10 copies/mL) compared to those who were discharged home (2.22) or were admitted to a medical ward (2.24) (p=0.006) (Table 1).
In the univariate analysis, factors such as diabetes, obesity, hepatitis, fever, dyspnea, oxygen requirement, and TTV load were identified as predictors of ICU admission. A subsequent multivariate analysis utilizing model selection based on Akaike’s information criterion (AIC) and incorporating multiple imputation by chained equations (MICE) yielded the same contributing factors (Table 2).
Analysis of the probability of ICU admission, using splines to alleviate the constraint of log-linearity, exhibited a plateau characterized by a similar threshold (Figure 1). Employing ROC curve analysis, we established the optimal cutoff value for TTV load that predicts ICU admission to be 2.91 log10 copies/mL, yielding a sensitivity of 0.605 and a specificity of 0.69. This cutoff value was employed to apply the previously selected multivariate model in predicting ICU admission based on the dichotomization of TTV load. Consequently, factors such as obesity, hepatitis, time since first symptoms, body temperature, dyspnea, and a TTV load exceeding 2.91 log10 copies/mL emerged as predictors of ICU admission (Figure 2).