DCN

DCN identifies temporal correlations between disease states in longitudinal electronic medical records (EMR) to support epidemiological analysis and hypothesis generation.

Key Features:

• Implementation: Implemented as an R package for analysis of longitudinal EMR data.
• Cohort construction: Constructs retrospective matched cohorts from EMR records for time-to-event analysis.
• Statistical methods: Implements Cox proportional hazards regression and random forest survival analysis to assess temporal disease correlations.
• Confounder adjustment: Controls for confounding covariates including age, gender, and other demographic or medical characteristics.
• Network visualization: Generates Disease Correlation Network visualizations using graph-theory network topology rendered via JavaScript.

Scientific Applications:

• EMR screening: Screening longitudinal EMR to identify temporal associations between chronic diseases.
• Hypothesis generation: Generating novel research hypotheses about disease interrelations for follow-up studies.
• Retrospective epidemiology: Retrospective epidemiological analysis of large patient cohorts, exemplified by application to Loyola University Chicago Medical Center data (175,539 patients, 654,084 initial diagnoses across 51 conditions).

Methodology:

Implemented in R; constructs retrospective matched cohorts from EMR; applies Cox proportional hazards regression and random forest survival analysis to assess temporal correlations while adjusting for age, gender, and other demographic or medical covariates; generates graph-theory–based network representations rendered with JavaScript.