Predictors of all-cause 30 day readmission among Medicare patients with type 2 diabetes.

OBJECTIVE: Readmission is costly among patients with type 2 diabetes (T2DM) in Medicare Advantage Prescription Drug Plans; identifying high-risk patients is necessary for targeting reduction programs. The objective of this study was to develop a claims-based algorithm to predict all-cause 30 day readmission among patients with T2DM.

METHODS: This study used administrative data from 1 January 2012 through 31 January 2014. The cohort included hospitalized T2DM patients, aged 18-90 with ≥12 months' continuous enrollment before an unplanned hospital admission and ≥1 month of enrollment post-discharge, excluding patients in long-term care >30 days pre-index. Multivariate logistic regression predicted the likelihood of readmission following hospitalization in 2013. The analytic file was randomly split into training and test datasets to build and validate the model. Candidate variables included physician and patient demographics, baseline clinical conditions, and healthcare utilization metrics. Clinical conditions were classified using the Healthcare Cost and Utilization Project clinical classification system for ICD-9-CM.

RESULTS: Of 63,237 individuals, 17.1% experienced a readmission. Of nearly 200 candidate variables, 14 were predictors of readmission, including total cumulative number of days for inpatient stays and the number of emergency department visits in the baseline period. Male gender, older age, and certain comorbidities were associated with higher likelihood of readmission. The final model demonstrated good discriminant ability (c-statistic = 0.82).

CONCLUSIONS: This study provided evidence that certain patient characteristics and healthcare utilization are predictive of readmission. An algorithm with good discriminant ability was developed which could be used to target readmission reduction programs. Physician gender, specialty, and ownership status did not appear to influence the likelihood of readmission.