A visual analytic approach for the identification of ICU patient subpopulations using ICD diagnostic codes

A large number of clinical concepts are categorized under standardized formats that ease the manipulation, understanding, analysis, and exchange information. One most extended codifications is International Classification Diseases (ICD) used for characterizing diagnoses procedures. With formatted ICD concepts, a patient profile can be described through set sorted attributes according to relevance or chronology events. This structured data fundamental quantify similarity between patients detect relevant characteristics. Data visualization tools allow representation comprehension patterns, usually high dimensional nature, where only partial picture projected. In this paper, we provide visual analytics approach identification homogeneous cohorts by combining custom distance metrics with flexible dimensionality reduction technique. First define new metric measure diagnosis profiles concordance Second describe variation Simplified Topological Abstraction (STAD) technique enhance projection signals preserving global structure data. The MIMIC-III database implementing analysis into an interactive dashboard, providing highly expressive environment exploration comparison groups at least one identical diagnostic code. combination STAD not allows patterns but also provides layer information establish additional relationships cohorts. method tool presented here add valuable exploring heterogeneous populations. addition, applied in other domains employ ordered lists categorical