Clinical Prediction Tool To Identify Adults With Type 2 Diabetes at Risk for Persistent Adverse Glycemia in Hospital.

Area of Science:

• Endocrinology
• Clinical Medicine
• Diabetes Management

Background:

• Hospitalized patients with diabetes frequently experience hyperglycemia and hypoglycemia.
• Current tools for predicting adverse glycemia (AG) in hospital settings are lacking.
• Persistent AG poses significant risks to patient outcomes.

Purpose of the Study:

• To develop and validate a clinical prediction tool for early identification of hospitalized patients at risk for persistent adverse glycemia (AG).
• To identify key clinical factors available early in admission associated with persistent AG.

Main Methods:

• Analysis of a cohort of 594 adult inpatients with type 2 diabetes.
• Logistic regression modeling to construct a prediction tool using early admission clinical factors.
• Internal validation using a split-sample approach.

Main Results:

• Persistent AG occurred in 26% of patients.
• Factors associated with persistent AG included admission dysglycemia, high glycated hemoglobin, sulfonylurea or insulin treatment, and glucocorticoid use.
• The prediction tool demonstrated good accuracy (AUC = 0.806) with 84% sensitivity and 66% specificity.

Conclusions:

• A clinical prediction tool using readily available admission data can identify patients at high risk for persistent AG.
• This tool can facilitate early, targeted management by inpatient diabetes teams.
• Improved prediction of persistent AG can lead to better patient outcomes in hospital settings.