Sepsis Mortality in Adults: Analysis using a Decision Tree-Based Algorithm

Introduction: Sepsis causes the body to damage its organs and tissues in the presence of infection. The objective was to develop an algorithm based on decision trees to analyze and classify mortality groups due to sepsis in adults. Material and methods: Analytical and cross-sectional study of a secondary database of 102,389 adults. The variables were: hospital outcome, age group, sex, septic episodes, and hospitalized time. The decision tree was used for automatic chi-square interaction detection. Results: In young adults, the decision tree included sex, days hospitalized, and episodes of sepsis. In intermediate adults: age, sex, days hospitalized, and episodes of sepsis. In older people: sex, age, and days hospitalized. In young, intermediate, and older adults, it correctly classified 98.30 %, 96.90 %, and 89.80 % of cases, respectively. In adults aged 18 to 59 years, 9.40 %, 4 %, and 0.90 % died after the third, fourth, and fifth septic episode, respectively. In adults over 60 years of age, 4.60 %, 1.80 %, and 0.80 % died in the third, fourth, and fifth episode, respectively. The percentages of patients alive since the second readmission were higher in older adults. Conclusions: Age, sex, number of sepsis episodes, and length of hospital stay predict mortality from sepsis in adults. The number of sepsis episodes influences mortality in adults up to 59 years of age, but not in older adults, while age influences mortality in intermediate and older adults. Decision trees allow generating efficient predictive and classification models, which can complement the clinical and epidemiological profile of patients hospitalized for sepsis.