Tracking patient clusters over time enables to extract all the information available in the medico-administrative databases

Context Identifying clusters (i.e., subgroups) of patients from the analysis of medico-administrative databases is particularly important to better understand disease heterogeneity. However, the complexity of these databases, in particular due to the presence of truncated longitudinal data, requires adaptation of clustering approaches. Objective We propose here cluster-tracking approaches to identify clusters of patients from longitudinal data contained in medico-administrative databases. Material and Methods We first cluster patients at each age using either the Markov Cluster algorithm (MCL) from patient networks or Kmeans from raw data. We then track the identified clusters over ages to construct cluster-trajectories. We compared our novel approaches with three longitudinal clustering approaches by calculating the silhouette score. As a use-case, we analyzed antithrombotic drugs prescribed from 2008 to 2018 contained in the Échantillon Généraliste des Bénéficiaires (EGB), a French national cohort. Results Our cluster-tracking approaches allowed us to identify several cluster-trajectories having clinical significance. Silhouette score comparison between the different approaches reveals that the best score is obtained for the cluster-tracking approaches. Conclusion The cluster-tracking approaches are a novel and efficient alternative to identify patient clusters from medico-administrative databases by taking into account their specificities.