3164 — Measuring Disease Severity using Latent Class Modeling of Comorbidities

Ward RC, Health Equity and Rural Outreach Innovation Center, Ralph H. Johnson Department of Veterans Affairs Medical Center, Charleston, SC, USA; Gebregziabher M, Health Equity and Rural Outreach Innovation Center, Ralph H. Johnson Department of Veterans Affairs Medical Center, Charleston, SC, USA; Payne E, Health Equity and Rural Outreach Innovation Center, Ralph H. Johnson Department of Veterans Affairs Medical Center, Charleston, SC, USA; Walker R, Health Equity and Rural Outreach Innovation Center, Ralph H. Johnson Department of Veterans Affairs Medical Center, Charleston, SC, USA; Egede LE, Health Equity and Rural Outreach Innovation Center, Ralph H. Johnson Department of Veterans Affairs Medical Center, Charleston, SC, USA;

Objectives:
In order to mitigate the potential for bias, adjustment for disease burden or comorbidity is essential in statistical models of health outcomes research. Commonly used adjustment methods include Deyo-Charlson, Elixhauser or including selected diagnostic indicators of individual comorbidities. In this study, we propose a new summary measure of disease severity developed via latent class analysis (LCA). We studied the operating characteristics of the proposed measure and compared them to existing methods to determine the most optimal risk-adjustment with applications to data from a national cohort of veterans with diabetes.

Methods:
We propose a measure of disease severity which is developed through latent class analysis of diagnostic indicators of comorbidities. We use statistical information criteria to choose the best latent class model that can serve as an alternative summary measure of disease severity. The operating characteristics of the proposed measure are examined using the receiver operating characteristics (ROC) and cross-validation techniques.

Results:
We applied this approach to retrospective data involving a large cohort of 892,223 Veterans with HbA1c to demonstrate that improved risk adjustment over existing comorbidity indices can be achieved through a summary measure derived through LCA analysis.

Implications:
The proposed summary measure of disease severity leads to higher levels of risk adjustment. Instead of categorizing the count of diagnostic indicators from ICD-9 codes, the proposed measure provides for more robust adjustment for disease severity.

Impacts:
Researchers should consider alternative methods to previously validated comorbidity scores when adjusting for risk in outcomes research. Such gains in risk prediction modeling can have a wide applicability, including better predictions of patient outcomes or hospital readmissions.