Prediction of hospitalization time probability for COVID-19 patients with statistical and machine learning methods using blood parameters.

OBJECTIVES

Coronavirus disease 2019 (COVID-19) may induce life-threatening complications and lead to death in the patients.

METHOD

The aim of this study was to describe a predictive model for the disease outcome (length of hospitalization and mortality) by using blood parameters results at the admission time of 201 patients with positive RT-PCR test for the infection. Variables including; age, sex, comorbidity risk factors, the length of hospitalization, and 25 blood parameters results at the time of admission were considered.

RESULTS

After analyzing the data, it was observed that several factors, such as hypocalcemia, hyponatremia, red blood cell microcytosis, monocytopenia, thrombocytosis, comorbidity risk factors (diabetes, dialysis, cardiovascular diseases, and hypertension), and age over 50 years had a significant impact on the length of hospitalization and mortality of the patients (<0.05).

CONCLUSION

Based on the data analysis, the authors found that the proportional hazard assumption was not established. Therefore, the authors opted to use the accelerated failure time model for our analysis. Among the various models considered, the log-normal model provided the best fit. Considering the analysis of laboratory results at the time of admission, the authors propose that thrombocytosis, red blood cell microcytosis, monocytopenia, hypocalcemia, hyponatremia, comorbidity factors, and age over 50 years can serve as predictive markers for estimating hospitalization length and mortality. These findings suggest that these factors may play a significant role in predicting patient outcomes.