PRM71 - Prediction Of Costs Associated With The Hospital Management Of Hiv Patients In France Using An Advanced Data Mining Approach On The French Medical Information System Database

Abstract

The PMSI (medical information system database) is the French national administrative database containing medical information and patients’ features for millions of hospital stays. Most medical studies only use descriptive methods to analyze patient cohorts. The objective of this study is to assess the capability of an innovative Data Mining algorithm to discover hidden values in health data and to bring a better understanding of diseases. We applied the approach to predict the costs associated to HIV patients’ hospital management. We only selected hospital stays with an HIV code, HIV being the principal cause of hospitalization or not. Patients hospitalized with an HIV code in 2013 were extracted and followed up for one year. 10 groups of comorbidities and 5 types of opportunistic infections (OIs) linked to HIV were also identified, and their presence was tracked among these patients. Data were analyzed with an Enhanced Decision Tree technique, in order to explain HIV hospitalization costs depending on non-linear combinations of age, gender and the presence of comorbidities or OIs. 30,294 patients with 70,180 hospital stays were included, for a total cost of 180 million euros. Our Enhanced Decision Tree was able to determine 165 different patient profiles, created automatically to maximize the gathering of patients with similar features. The most discriminative variables for the cost of hospitalization were infections not associated to HIV, bacterial OI, cancer, fungal infections and endocrino-metabolic complications, whereas age, psychiatric or hepatic comorbidities were not discriminative. The annual average cost per patient profiles ranged from €1,680 to €42,650. This exploratory study shows that Enhanced Decision Trees are relevant to identify patient profiles from big databases and may have predictive capabilities. It could help identifying leverages to prevent hospitalizations costs. Further research should be done, adding therapeutic and biological parameters.