Spatio-Temporal Data Mining for Public Health by Using Large-Scale Datasets

Our research focuses on analyzing large-scale health datasets from the New York State SPARCS system to understand spatio-temporal trends and patterns in healthcare. By aggregating data based on clinical classification systems, diagnostic-related groups, and spatial indicators like county and zip code, our study aims to identify significant trends that could inform healthcare resource allocation and policy decisions. Over six years, comprising 13.5 million patient records, the research identified temporal trends and spatial distributions of various health conditions. Notably, cardiovascular disease (CVD), chronic obstructive pulmonary disease, pneumonia, schizophrenia, and bipolar disorders were among the top conditions in terms of occurrence. In terms of costs, heart failure, schizophrenia, chronic obstructive pulmonary disease, and nervous system disorders ranked the highest. We generated spatial prevalence maps and calculated spatial distribution statistics including Moran's Index, to pinpoint hotspots of these diseases across different counties. The findings suggest the need for improved messaging and awareness campaigns to address the steady heart failure rates and the increasing trend in recent years, indicating a necessity for preventative measures to reduce CVD incidence. Additionally, the prevalence of mental illnesses like schizophrenia and bipolar disorder underscores the importance of exploring preventative measures that could benefit both cardiovascular health and mental health, particularly through cost-effective lifestyle interventions. Overall, the research provides valuable insights into the distribution and trends of various health conditions, highlighting areas where targeted interventions could lead to improved healthcare outcomes and resource efficiencies.