Developing a Model to Predict Inpatient Hypoglycemic Episodes

Developing a Model to Predict Inpatient Hypoglycemic Episodes

This study will create a new model to predict hypoglycemic events in diabetic inpatients on anti-hyperglycemic therapy using retrospective data with the goal of developing a model that will accurately predict hypoglycemic episodes in the patient population - piloting the risk score that was developed in the context of EndoTool being rolled out at the institution, to determine the feasibility and acceptability of viewing the risk score in the Electronic Health Record

Design a model to predict hypoglycemic episodes in real time, rather than retrospectively identifying high risk patients who experienced a hypoglycemic event at some point during their hospitalization, as has been done prior. Potential predictors of hypoglycemia identified in prior work and additional predictors identified by the clinical team to develop a discrete-time multinomial logistic regression model to predict hypoglycemic events in real time. The risk score then will be piloted in the context of our institution's EndoTool Subcutaneous implementation determine the feasibility and acceptability of viewing the risk score in the Electronic Health Record

Inclusion Criteria:

• Providers taking care of adult patients (18 years and older) on the selected medical floors who are receiving anti-hyperglycemic medications on the chosen services (EndoTool and risk score implementation) and comparison groups (one group with EndoTool implementation and the other group without EndoTool implementation)

Exclusion Criteria:

• Providers not on the chosen services