Malnutrition is one of the major geriatric syndromes and frailty factor, this joint with the fact of elderly population growing, will situate malnutrition as a front end problem in the upcoming years. Therefore, it is important that health professionals can assess and follow up nutritional status in a proper way, using all available data related to patients. Process mining can be used to extract knowledge from information in order to understand health care processes. A classic approach to assess malnutrition usually comprises anthropometric measures as static variables, with no information about patients evolution and pathways. The aim of this work was to examine anthropometric measures from a dynamic perspective thanks to process mining tools, in order to obtain dynamic behaviour models. This paper proposes a method based on the use of process mining to discover and identify weight changes behaviour. Clustering is used as part of the pre-processing of data to manage variability, and then process mining is used to identify patterns of patients' behaviour. The method is applied through different experiments to data from 96 patients. Results grouped almost all individuals in different models based on common behaviours. Main finding shows different behaviour groups seem to have different results regarding malnutrition status for same interventions. By discovering patterns of dynamic weight change and their relation with malnutrition, nursing homes and health care professional can promote more successful intervention among patients based on their behaviour, moreover they can compare interventions' results analysing changes in behaviour between before and after the intervention.
展开▼
