External Validation and Modification of a New Score for Predicting Mortality in Patients with COVID-19 in High Altitude Patients. A Peruvian Study

This study aims to validate two predictive mortality scores for patients with COVID-19 to support clinical decision-making in those who require hospitalization. The tomographic patterns found can be added to the original scores to increase their predictive power. Retrospective, analytical, observational, and cross-sectional studies were carried out in two phases. 489 medical records of patients with COVID-19 hospitalized at “Daniel A. Carrión” Hospital in Huancayo (located at 3,250 meters above sea level) were reviewed to perform external validation. Two predictive scores, formed by nine (score 5) and ten variables (score 6) were evaluated. In a second step, a subgroup of 258 patients with chest CT scan results was assessed to determine the association of tomographic findings with mortality. The diagnostic precision of SAWBPIL and SAWBPI scores was high and it was found between 80% and 85%, as expressed by an area under the curve (AUC) of SAWBPIL score of 0.843 and SAWBPI score (without DHL) of 0.822. This diagnostic precision was similar to those of the original study (AUC 0.838 and 0.826, respectively) and higher than that of the CALL score (AUC 0.756). The percentage of pulmonary involvement was 54.59% in the surviving patients and 66.6% in those who died (p-value = 0.000). The performance of the modified SAWBPI score (SAWBPI-CT1.0), formed when adding the percentage of pulmonary involvement and the presence or absence of septa, reached a diagnostic precision of 84.4%, while the SAWBPIL score (SAWBPIL-CT1.0) reached a precision of 90.3%. For a value of 3.75 severity points or more, the modified SAWBPIL score reached a sensitivity of 86.8% and a specificity of 70.1% to predict mortality from COVID-19. Two new mortality prediction scores were shown to maintain their predictive capacity in the population studied. Adding tomographic data, the diagnostic precision of the score that includes LDH values reached a value of 90.3%, higher than most currently used scores.