Type 2 diabetes mellitus (T2DM) is strongly associated with future cardiovascular (CV) events, including heart failure (HF) and its sequelae such as hospitalization (HFH) and death. Identifying T2DM patients at highest risk of HFH may be clinically valuable so therapeutic regimens can be optimized.

Accordingly, this study aimed to develop and validate a risk prediction model for HFH among patients with T2DM. This electronic medical record- (EMR-) based, retrospective cohort study included patients with new or preexisting diagnoses of T2DM receiving health care services longitudinally through the Geisinger Health System. Study patients were assigned an index date at an office visit at least two years following their first EMR-documented encounter between 1/1/03 and 11/10/15. Candidate predictors of HFH were defined according to EMR documentation in the 2-year period prior to the index date from demographics, vital signs, medical history including diagnoses and procedures, CV-related symptoms, and laboratory tests. Cox proportional hazards regression with stepwise variable selection identified the strongest predictors of HFH. Internal validation was performed by randomly splitting the sample 1:1. Among 77,385 T2DM patients meeting study criteria, median (IQR) age of patients at the index date was 62 (52, 72) years and 49% were male. Over a mean (SD) follow-up of 5.8 (4.2) years, 2903 (3.8%) HFH occurred, and estimated HFH incidence rates at 1, 3, and 5 years after the index date were 0.5%, 1.8%, and 3.0%. The development set model containing the 15 strongest HFH predictors had a c-statistic of 0.833, and the validation set c-statistic was 0.824. The five strongest predictors were age, a history of HF, coronary artery disease, blood urea nitrogen, and HbA1c. The patients with 10% highest risk based on the model had ≥2.0% annual risk of HFH. This proposed model for HFH among T2DM demonstrated strong predictive capability and may be valuable in guiding therapeutic decision making.


B. Williams: None. P. Gandhi: Employee; Self; Boehringer Ingelheim Pharmaceuticals, Inc..

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.