East Coast Ballroom
It’s getting hot in here: A novel application of heatmaps to health outcomes research (307824)Peter Bach, Memorial Sloan Kettering Cancer Center
*Jessica A. Lavery, Memorial Sloan Kettering Cancer Center
Katherine Panageas, Memorial Sloan Kettering Cancer Center
Keywords: data visualization, heatmap, health outcomes, genomics, hospital profiling
Heatmaps are a commonly employed data visualization tool used to identify clusters in gene expression data. We present a novel application of heatmaps and cluster analyses to health outcomes research in the setting of assessing hospital performance of cancer-directed surgery. Multiple types of cancer-directed surgeries at over 3,000 hospitals over a three-year period were examined. Hospital-level outcome measures of 30-day mortality (from date of surgery) and 30-day unplanned readmissions (from date of discharge) were analyzed using random effects models and described using risk-standardized mortality rates. Traditionally, features associated with better or worse outcomes are assessed within the framework of regression and tests of association. In this application, we used heatmaps to identify clusters of high and low performing hospitals annotated along the characteristics of surgical volume, ownership (government, non-profit, private), teaching status, location (urban, rural). This data visualization and cluster analysis offers an alternative way to explore trends in hospital profiling data.