Frequency Tree clustering for ICU mortality analytics using graph databases

Mortality analytics is an emerging research area that discovers and communicates meaningful patterns in clinical data to reduce mortality rates. Nonetheless, intensive care unit (ICU) mortality analytics for leading causes, such as circulatory system diseases (CDS), is still complicated due to the interactions of different mortality causes. To improve analytics accuracy and quality, clustering analysis can provide clinicians with interpretable insights about clinical data. Although robust tools to support the clinicians are still emerging. Recently, Frequency Tree (F-Tree) has been utilized effectively to cluster data points using the frequencies of data categories. While F-Tree clustering would result in interpretable data patterns and effective (accurate) clustering in various domains, straightforward adoption of F-Tree clustering in healthcare domain will be hindered by memory overhead and slow analytics retrieval. In this regard, we propose an integration between F-Tree clustering and graph databases as a ICU mortality analytics tool that enriches F-Tree with a robust retrieval engine. Modeling F-Tree using such a retrieval engine improves its clustering and association retrieval quality in terms of interpretability, scalability, and efficiency (timing) assessments.