SHM Converge Scientific Abstract ePoster Gallery-Development of a Prediction Model to Identify Non-Critically Ill Inpatients at Low Risk of Hyperglycemia for Targeted De-Escalation of Low-Value Glucose Monitoring

Development of a Prediction Model to Identify Non-Critically Ill Inpatients at Low Risk of Hyperglycemia for Targeted De-Escalation of Low-Value Glucose Monitoring

Back to course

Pdf Summary

This study aimed to reduce low-value inpatient glucose monitoring by identifying non-critically ill hospitalized patients at low risk for persistent hyperglycemia. Current guidelines recommend insulin treatment for persistent hyperglycemia, defined as two glucose values above 180 mg/dL, but there is no clear guidance on how often to monitor patients who do not meet this threshold. As a result, many patients receive frequent point-of-care glucose (POC-G) checks that may not be necessary.<br /><br />Researchers performed a retrospective secondary analysis of data from five Johns Hopkins hospitals between January 2015 and May 2019. The cohort included adults with at least four POC-G measurements during hospitalization and excluded patients with type 1 diabetes, ICU admission, very high admission glucose, early hypoglycemia, or use of certain diabetes medications/insulin. They built logistic regression models using demographic and clinical variables available within the first 48 hours of admission to predict whether patients would later develop persistent hyperglycemia.<br /><br />The main finding was that among patients who had 0–1 glucose values above 180 mg/dL in the first 48 hours, 79% did not later meet criteria for persistent hyperglycemia. Simple models using average early POC-G values and the presence of a diabetes diagnosis were able to identify low-risk patients with a negative predictive value of 87–89%, suggesting these patients could safely have glucose monitoring de-escalated. Models including hemoglobin A1C were not favored because of substantial missing data, though A1C may improve future models.<br /><br />The authors conclude that early inpatient glucose data, especially mean POC-G and diabetes diagnosis, can help distinguish patients unlikely to develop persistent hyperglycemia. They recommend external validation in a separate patient population before developing a decision-support tool.

Asset Subtitle

Niloofar Latifi

Meta Tag

Author List Daniel J. Brotman, Jianqiao Ma, Megan Tschudy, Mohammed Abusamaan, Nestoras ? Mathioudakis, Niloofar Latifi

Category Research

Concept Glucose Monitoring

Concept Hyperglycemia

Concept Persistent Hyperglycemia

Concept Negative Predictive Value

Concept Noncritically Ill Hospitalized Patient

Distinguished Non-Finalist

Presenter Organization Johns Hopkins Medicine

Presenting Author Niloofar Latifi

Track Value in Hospital Medicine

Keywords

inpatient glucose monitoring

persistent hyperglycemia

point-of-care glucose

low-value care

hospitalized patients

logistic regression

negative predictive value

diabetes diagnosis

hemoglobin A1C

retrospective study

Glucose Monitoring

Hyperglycemia

Persistent Hyperglycemia

Negative Predictive Value

Noncritically Ill Hospitalized Patient