Network Analysis: an approach to the study of drug-drug relations. Co-medication in Norwegian elderly and severe drug-drug interactions as examples

Abstract

Network analysis (NA) has been used for studying many social aspects. Employing of network analysis as an approach in the field of public health, to study the relations between patients or health workers and their potential effects in many medical perspectives, took a good share of researchers’ efforts as well. Few attempts have been conducted to use network analysis to study drug-drug relations using medicines as the main actors in the network instead of persons. We aimed at two primary objectives; a methodological objective and a clinical one. The methodological is to define the co-medication in a more reliable way and to use NA as an approach to map and extract the co-medication patterns in the elderly. Afterwards, to comment on the relevant clinical information represented in these networks. We represented two examples of drug-drug relations in the form of networks; i) The elderly co-medication in Norway in three years (2012-2014). The data was extracted from the Norwegian Prescription Dataset (NorPD) and included 61,930,313 prescriptions of 342,451 men (45%) and 419,455 women (55%), in total 761,906 patients. The mean age of the study population is 75 years. ii) The severe Drug-drug Interactions (DDI) based on the drug-drug interactions from the Prescribing and Dispensing Support dataset (FEST) with a total of 57,151 sever interactions. In our thesis, co-medication was defined as treatment episodes. Determining these episodes depends on the time of prescriptions’ dispensing, the Defined Daily Dose (DDD) of each drug, assuming 80% of patients’ adherence. We used the Proportion of Days Covered (PDC) to measure the adherence. We allowed a gap of 14 days as an accepted medical-free period between the treatment episodes. After defining the treatment episodes for each patient, a prevalence point was chosen to study the co-medication pattern in it. This approach in defining co-medication allows flexibility in choosing the studied prevalence points. Six different elderly co-medications patterns were extracted from our primary network. Comparing co-medication patterns in two prevalence time points, with a one-year difference, revealed changes in use, number of users, and prescribed patterns. We used “betweenness centrality”, a specific NA measure, to obtain the drugs with the most contribution in the severe interactions. The network showed 662 severe DDI in the studied treatment episode with a range of 1 to 2320 patients who were exposed to these severe interactions. We concluded that network analysis, as an approach, can be effectively used in visualizing and studying drug-drug relations with some unique descriptive measures.