GIScience 2016 Short Paper Proceedings Deriving Hospital Catchment Areas from Mobile Phone Data Bernd Resch 1,2 , Azmat Arif 1 , Gautier Krings 3 , Guillaume Vankeerberghen 3 , Marc Buekenhout 4 University of Salzburg, Department of Geoinformatics – Z_GIS, Schillerstrasse 30, 5020 Salzburg, Austria Email: bernd.resch@sbg.ac.at, azmat.arif@stud.sbg.ac.at Harvard University, Center for Geographical Analysis, 1737 Cambridge Street, Cambridge, MA 02138, USA Real Impact Analytics, 5 Place du Champ de Mars, 1050 Brussels, Belgium Email: {gautier.krings; guillaume.vankeerberghen}@realimpactanalytics.com Email: mbuekenhout@msn.com Abstract Delineating catchment areas of medical facilities is essential for estimating the quality of a health-care system and to maximise the efficiency of health service provision. One critical shortcoming of previous approaches are manifested in their comprehensive assumptions about a hospital’s patients by using census data or gravity models. In contrast, our approach uses anonymised mobile and landline phone data to derive hospital catchment areas. Our goal is not to assess the quality of the health care system, but to identify the geographic areas, in which people actually use a hospital. Thus, our results reveal new insights into the catchment areas of hospitals by minimising assumptions about demographic factors. 1. Introduction and Related Work Adequate provision of health services is a central priority of health professionals and policy makers worldwide. More, the efficiency of health care systems, i.e., the provision of best possible service with minimum resources, is critical to public providers (Fransen et al. 2015). These requirements have led to a number of studies to analyse catchment areas and service quality of medical facilities. Even though geospatial analysis methods have existed for decades, there is still a general lack of studies that have mapped and examined health service catchments in practice, not only from a theoretical viewpoint (Schuurman et al. 2006). Previous approaches for delineating medical catchment areas comprise statistical population-to-provider ratios, gravitational models, travel cost estimation, analysis of the physical distance between hospitals, census-based patient-origin analysis, commuter-based approaches of modelling spatial accessibility (Fransen et al. 2015; Wang and Wheeler 2015), or the two-step floating catchment area method based on the physician-to-population ratio (Luo and Wang 2003). The major drawback of these approaches is that they make far- reaching assumptions about a hospital’s patients by applying census data, travel times or gravity models. Moreover, they do not take heterogeneous activity and mobility patterns into account or only derive them from static census data. Thus, the approach proposed in this paper uses anonymised mobile and landline phone data to delineate hospital catchment areas. Like this, we aim to identify the geographic areas, in which people use a hospital instead of assessing the quality of the health care system per se. Therefore, we analyse calls to and from hospitals in Trinidad and Tobago. This goes beyond just using patient records in that we are able to draw conclusions from a wider range of communication with a hospital (enquiries, arrangement of appointments, follow-up care, visitors, etc.), beyond patients’ hospital stays.