Prediction of the risk of an unfavorable outcome in COVID‑19 using the classification tree method, taking into account the age and number of comorbid pathology according to the infectious hospital.
Published: 2024-05-02
-
Objective: to determine the probability of the risk of an unfavorable outcome using the classification tree method in patients with COVID‑19 who were treated in an infectious hospital based on the analysis of the contribution of such predictors as age and the amount of comorbid pathology.
Materials and methods. The data of outpatient records of 5304 patients who were treated in an infectious diseases hospital with a diagnosis of COVID‑19 from January 1, 2021 to January 1, 2022 were analyzed. The age of the examined patients was 62 [56–66] years. Among 5,304 patients, there were 2,891 males (54,5%) and 2,413 females (45,5%). The patients were divided into age groups according to the WHO classification. The frequency of comorbid pathology was analyzed taking into account the nosological unit of the disease registered in at least 1% of the patients included in the study.
Results. In the studied sample, the elderly prevailed in a larger percentage – 46.8%. The presence of one or more comorbid pathology was revealed in the prevailing number of hospitalized – in 5,244 people (98,9%). The most common comorbid pathology in the examined patients was arterial hypertension in 2038 people (38,4%), coronary heart disease in 1997 people (37,7%) and type 2 diabetes mellitus in 1629 people (30,7%). A classification tree was constructed to predict the risk of the probability of an unfavorable outcome in patients with COVID‑19.The minimum number of observations in the parent node in the classification tree was 400 people, in the child node – 200 people. In the resulting classification tree, 8 terminal nodes were observed.
Conclusion. The probability of the risk of an unfavorable outcome in the analyzed sample of patients increases with an increase in the number of comorbid pathology and the age of patients. According to the forecast using the classification tree method, the greatest probability of risk (3,2 times) of an unfavorable outcome in relation to the general sample was among elderly persons+centenarians with more than three comorbid pathology. The developed classification tree showed a high probability of correct predictions (80%). The sensitivity of the resulting model was 77,9%, the specificity was 64,2%.
