OBJECTIVES: To determine the predictive utility of admission random blood sugar (RBS) as a screening tool for undiagnosed diabetes and pre-diabetes among hospitalized patients, and to determine its optimal cutoff value that will best predict the presence of diabetes and pre-diabetes. Secondarily, we also determined the predictive utility as well as the optimal cut-off value for hemoglobin Ale (HBAlc) in screening for diabetes and pre-diabetes.

METHOD: This is a cross-sectional criterion-reference study, including all patients aged 20 to 75 years old, with no known history of diabetes, who were admitted to the general medical ward. Patients who were pregnant , on chronic corticosteroid therapy, or had renal insufficiency and hematological problems were excluded. Glycemic classification was based on the discharge diagnosis of diabetes, or results of FBS and 75 grams OGTT done 6 weeks after, among those not discharged with any anti-diabetic medications. Operating characteristics of each cutoff value were determined and the optimal cutoff value for admission RBS and HBAlc were derived through an ROC curve.

RESULTS: Based on the standard screening criteria, 84 (35 percent) out of the 240 subjects were diagnosed to have diabetes, while 29 subjects (12.1 percent) had pre-diabetes. The best cut-off value for admission RBS in the detection of unrecognized DM was derived at 150 mg/dl with a resultant sensitivity of 85% (as compared to 46% with the conventional cut-off value of 200mg%) and specificity of area under the curve shows excellent predictive ability (AUC=0.92). Glycosylated hemolobin at cut off value of 6.2 showed comparable predictive ability (AUC=0.94), with sensitivity of 90% and specificity of 86%. Both admissions RBS and HBA 1c however showed poor predictive ability for pre-diabetes. To our knowledge, this is the first study reporting on the role of admission RBS for diabetes and pre-diabetes case finding in a hospital setting.

CONCLUSIONS: Admission RBS was found to be an excellent screening tool for undiagnosed diabetes but not pre-diabetes among hospitalized patients. Lowering down the cuttoff value from the conventional 200mg% to 150mg%, greatly improves the sensitivity and still with adequate specificity for screening of diabetes. (Author)