JMIR Publications

ABSTRACT

Background:

Clinical data on multiple chronic conditions (MCC) can be represented using graphical models to study their interaction and identify the path toward MCC development. However, the graphical models representing MCC are often complex and difficult to analyze. Therefore, it is necessary to provide a concise representation of MCC graphical models.

Objective:

The objective of this work is to summarize the complex graphical models of multiple chronic conditions interactions based on Eigen analysis of graph Laplacian to streamline the comprehension and analysis.

Methods:

We develop three algorithms which utilize the 2nd eigenvalue analysis of the graph (EAGL) Laplacian to summarize complex graphical models of multiple chronic conditions by reducing less significant edges. The first algorithm learns a sparse probabilistic graphical model of MCC interactions directly from data. The second algorithm summarizes an existing probabilistic graphical model of MCC interactions when a supporting dataset is available. Finally, the third algorithm, which is a variation of the second algorithm, summarizes the existing graphical model of MCC interactions with no supporting data.

Results:

We use two datasets of: (1) 257,633 veteran patients who have been monitored for the emergence of five multiple chronic conditions including brain injury (TBI), post-traumatic stress disorder (PTSD), depression (Depr), substance abuse (SuAb), and back pain (BaPa) over five years, and (2) co-appearance of 100 most common terms in the literature of multiple chronic conditions to validate the performance of the proposed model. The proposed summarization algorithms demonstrate considerable performance in extracting major connections between MCC without reducing the predictive accuracy of the resulting graphical models.

Conclusions:

Using graph summarization can improve the interpretability and predictive power of complex graphical models of MCC.