Drug recommendation system for geriatric patients based on bayesian networks and evolutionary computation

Geriatric people face health problems, mainly with chronic diseases such as hypertension, diabetes, osteoarthritis, among others, which require continuous treatment. The prescription of multiple medications is a common practice in that population, which increase the risk of unwanted or dangerous drug interactions. The quantity of drugs is constantly growing, as are they interactions. It is therefore desirable to have support systems for medical that digest all available data and warn for possible drug interactions. In this paper we proposed a drug recommendation system that takes into account pre-existing diseases of the geriatric patient, current symptoms and verification of drug interactions. A Bayesian network model of the patient was built to allow reasoning in situations of limited evidence of the patient. The system uses also a genetic algorithm, which seeks the best drug combination based on the available patient information. The system showed consistency in simulated settings, which were validated by a specialist.