Machine Learning Algorithms to Study Features Affecting the Length of Stay in Patients with Lower Limb Fractures: A Bicentric Study

The Classification Algorithms Applied to the Inpatient Stay for Lower Limb Surgery

“…Using hospital discharge records, through the procedure code reported in the literature [19] and the year of discharge, medical records of interest were filtered. From the extracted variables, after the preprocessing stage, the independent variables used in the study (Age, Gender (M/F), Comorbidities (1/0), Complication during surgery (1/0), Pre-Op LOS) were obtained, in agreement with previous studies [22,23]. Figure 1 shows the distribution of the variables.…”

“…Through the use of predictive models, it is possible to know the LOS a priori and optimize bed management by aiding planning. This study continues a line of research begun in 2021 that involved two hospitals in Southern Italy [22,23].…”

“…Looking at the accuracy obtained in other publications, the best results were recorded for the A.O.R.N. "A. Cardarelli" in Naples [23] on a dataset of more than 200 patients. Compared with the University Hospital "San Giovanni di Dio e Ruggi d'Aragona" of Salerno [22] similar as sample size, the results are slightly better.…”

“…In this work, following when already in the literature [22,23], data from 162 patients who underwent lower limb surgery in 2019-2020 at the Evangelical Hospital "Betania" in Naples (Italy) were analyzed.…”

“…To proceed with the use of the classification algorithms, it is necessary to partition the LOS, a continuous variable, into two homogeneous groups. The Decision Tree, Random Forest, Gradient Boosted Tree, Support Vector Machine, Logistic Regression and Naïve Bayes algorithms were selected based on other studies [22,23]. DT, RF and GBT are all algorithms based on decision trees, which are optimized in RF and GBT through multiple implementation.…”