Leverage Diagnosis Intensity in Medication Recommendations

In recent years, the patient’s medical records have grown rapidly. This has led to the need of medication recommendation systems that use past diagnoses and personalize the proposed treatment. The majority of those systems prioritize medication recommendations based on the current diagnoses of the patient’s electronic health record (EHR) without considering diagnosis intensity. Thus, they fail to capture the intensity of the patient’s condition, which leads to low accuracy on their recommendations. In this study, we propose a medication recommendation system that leverages the diagnosis intensity in medication recommendation (LDIM-R). The main goal of our model is to compare the severity of diagnoses at each visit to all the previous visits. The proposed study commences by accessing the patient’s historical health record from the Medical Information Mart for Intensive Care (MIMIC-III). Subsequently, the average intensity of previous diagnoses is calculated and compared to the diagnosed intensity of the current visit. Then the proposed model considers three factors: 1) there is no discernible change in the overall diagnosed intensity; 2) the overall diagnosed intensity has declined; or 3) the overall diagnosed intensity has increased. As demonstrated by the result of our experiments over the MIMIC-III dataset, our method outperformed state-of-the-art medication recommendation systems.