Comorbidity models from General Practitioners prescription data. A network analysis approach

Public health management and decision making require solid information on the population health demand, current state and history. Unfortunately, in Italy the vast amount of health related data is scattered in different repositories with various formats, and it is not readily available, even to the decision of maker institutions. Conversely, the access to prescription data from general practitioners (GP) is relatively simple, as these data are used for administrative purposes by the national health system. From these data, it has associated a morbidity state to the patients and such states are considered both over time for the same subject and within different categorized subject. Comorbidity data and then networks are defined in terms of a set of relationships between morbidity of distinct prescriptions. Networks, representing the pathologies as nodes in a graph and the relationships between them as edges (when two pathologies appear in the same prescription), are obtained and analyzed on the basis of different subsets of patients defined by age and gender. In the present contribution, we exploit GP administrative databases for healthcare systems analysis. We extract a population of 30000 subjects by considering databases of 30 GPs in a time span of 10 years. The aim is to entangle the complexity of these information by using exploratory factorial methods recently proposed in social network analysis framework (D’Esposito et al., 2014; Giordano & Vitale, 2011; Ragozini et al., 2015) for one-mode and two-mode networks, in static and time-varying ways. In addition we use community detection algorithms to extract strong comorbidity structures present in the network. The proposed strategy of analysis is able to show different patterns of derived networks. The connection of such patterns to population specific diseases as well as the time evolution of the network structure on specific subsets of patients are also analysed.