Predicting Early Readmission Risk among Hospitalized Patients with Diabetes

Program: Abstracts - Orals, Featured Poster Presentations, and Posters
Session: SUN 839-872-Diabetes & Obesity Management
Clinical
Sunday, June 16, 2013: 1:45 PM-3:45 PM
Expo Halls ABC (Moscone Center)

Poster Board SUN-868
Daniel J Rubin*1, Elizabeth A Handorf2 and Marie E McDonnell3
1Temple University School of Medicine, Philadelphia, PA, 2Fox Chase Cancer Center, Philadelphia, PA, 3Boston Medical Center, Boston, MA
Hospital readmissions within 30 days of discharge (early readmissions, ER) are a high-priority healthcare quality measure and target for cost reduction. ER contribute $18 B to the $91 B annual hospital costs of diabetic patients in the US.A better understanding of the factors contributing to ER risk is necessary to identify high-risk diabetic patients and develop ER risk reduction interventions.

We conducted a retrospective cohort study of adult patients with diabetes discharged from an urban academic medical center between 1/1/2004 and 12/31/2010. Diabetes was defined by ICD-9-CM code 250 or documentation of a diabetes-specific medication upon admission. For patients with >1 hospitalization, each discharge was considered an index discharge. Index discharges were excluded for patient age <18 years, transfer to another hospital, inpatient death, outpatient death within 30 days of discharge, or discharge from an obstetric service. We constructed a predictive model for all-cause readmission by multivariate logistic regression. Half the sample was randomly selected as a development set to build the model, and the other half was used as a validation set to test the model.

In the sample of 14,845 patients, there were 7,758 ER (21.0% of 36,988 discharges). Out of 39 sociodemographic and clinical variables examined, there were 15 significant predictors of ER in the model (P<0.01). The top 5 largest effect sizes were found with a discharge within 30 days before the index admission (OR 2.15, 95%CI 1.95-2.37), discharge against medical advice (OR 1.75, 95%CI 1.34-2.27), being disabled and unemployed (OR 1.69, 95%CI 1.40-2.04), 4 distinct diagnoses of macrovascular complications (OR 1.60, 95%CI 1.20-2.13), and receiving inpatient nutrition support (OR 1.45, 95%CI 1.18-1.80). Other significant predictors were gender, race, education level, medical insurance, BMI, pre-admission insulin use, a history of depression or gastroparesis, serum creatinine, and hematocrit. Nonsignificant variables included age and length of stay. There is excellent agreement in ER risk prediction between the development and validation samples. The model has good discrimination (C-statistic 0.69 in development sample and 0.68 in validation sample).

This model is valid for predicting ER risk of patients with diabetes and may be useful to target discharge support resources to those at higher risk. Knowledge of ER causes and risk factors may inform the development of ER risk reduction interventions.

Nothing to Disclose: DJR, EAH, MEM

*Please take note of The Endocrine Society's News Embargo Policy at http://www.endo-society.org/endo2013/media.cfm

Sources of Research Support: Temple University School of Medicine Department of Medicine Faculty Development Research Award