Figure 3. Predicted Probability of Receiving a Diabetes Medication Intensification at Discharge, by Inpatient and Outpatient Glucose Control.

Preadmission hemoglobin A_1c was measured using the most recent laboratory value collected within 1 year preceding hospitalization. Inpatient blood glucose control was defined by the number of elevated blood glucose recordings as severely elevated (≥3 recordings of ≥300 mg/dL [to convert to millimoles per liter, multiply by 0.0555]), moderately elevated (≥3 recordings of ≥200 mg/dL without meeting criteria for severely elevated), or not elevated. Predicted probabilities were estimated following mixed-effect logistic regression accounting for age, sex, race/ethnicity, income, Charlson Comorbidity Index score, length of stay, primary discharge diagnosis, year, hospital training status, receipt of steroids during hospitalization, preadmission hemoglobin A_1c, inpatient blood glucose level, an interaction term for preadmission hemoglobin A_1c and inpatient blood glucose level, and random effects to account for clustering by Veterans Health Administration hospital. Error bars indicate 95% CIs. To convert hemoglobin A_1c to proportion of total hemoglobin, multiply by 0.01.