Baiju R Shah1, Sakina Walji2, Alexander Kiss3, Jacqueline E James4, Julia M Lowe5. 1. Department of Medicine, University of Toronto, Toronto, Ontario, Canada; Institute for Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada; Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada. Electronic address: baiju.shah@ices.on.ca. 2. Department of Family and Community Medicine, University of Toronto, Toronto, Ontario, Canada; Mount Sinai Hospital, Toronto, Ontario, Canada. 3. Institute for Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada; Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada. 4. Department of Medicine, University of Toronto, Toronto, Ontario, Canada; Mount Sinai Hospital, Toronto, Ontario, Canada. 5. Department of Medicine, University of Toronto, Toronto, Ontario, Canada; Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada.
Abstract
OBJECTIVES: We sought to develop the Hypoglycemia During Hospitalization (HyDHo) scoring system, to predict the risk for hypoglycemia during hospitalization in patients with diabetes at the time of admission to a general medical unit. METHODS: We randomly selected 300 patients with diabetes who had been admitted to a medical inpatient unit at a teaching hospital. Hypoglycemia was defined as any point-of-care glucose test result ≤3.9 mmol/L. Demographic and clinical predictors of hypoglycemia were identified through review of the hospitalization record. Bivariate associations between each predictor variable and hypoglycemia were used to choose variables for a logistic regression model. Model coefficients were converted into an integer points score. The selected model was applied to a validation dataset from 300 similar randomly selected patients admitted to a different teaching hospital. RESULTS: In the derivation cohort, 72 (25%) patients experienced hypoglycemia during their hospitalizations. The final selected model included 5 variables: age, emergency department visit 6 months prior, insulin use, use of oral agents that do not induce hypoglycemia, and severe chronic kidney disease. With a score of ≥9, sensitivity was 86% and specificity was 32%. The model had adequate discrimination and good calibration in the validation cohort. CONCLUSIONS: A parsimonious risk prediction model that uses 5 key clinical variables predicts hypoglycemia during hospitalization at the time of admission. More than one-quarter of patients at low risk for hypoglycemia had scores below the threshold. They could be identified at the time of admission by applying the HyDHo scoring system and may need less intensive glucose monitoring while in hospital.
RCT Entities:
OBJECTIVES: We sought to develop the Hypoglycemia During Hospitalization (HyDHo) scoring system, to predict the risk for hypoglycemia during hospitalization in patients with diabetes at the time of admission to a general medical unit. METHODS: We randomly selected 300 patients with diabetes who had been admitted to a medical inpatient unit at a teaching hospital. Hypoglycemia was defined as any point-of-care glucose test result ≤3.9 mmol/L. Demographic and clinical predictors of hypoglycemia were identified through review of the hospitalization record. Bivariate associations between each predictor variable and hypoglycemia were used to choose variables for a logistic regression model. Model coefficients were converted into an integer points score. The selected model was applied to a validation dataset from 300 similar randomly selected patients admitted to a different teaching hospital. RESULTS: In the derivation cohort, 72 (25%) patients experienced hypoglycemia during their hospitalizations. The final selected model included 5 variables: age, emergency department visit 6 months prior, insulin use, use of oral agents that do not induce hypoglycemia, and severe chronic kidney disease. With a score of ≥9, sensitivity was 86% and specificity was 32%. The model had adequate discrimination and good calibration in the validation cohort. CONCLUSIONS: A parsimonious risk prediction model that uses 5 key clinical variables predicts hypoglycemia during hospitalization at the time of admission. More than one-quarter of patients at low risk for hypoglycemia had scores below the threshold. They could be identified at the time of admission by applying the HyDHo scoring system and may need less intensive glucose monitoring while in hospital.