THE MODEL SCREENING RISKS OF HYPOGLYCEMIA AMONG DIABETIC PATIENTS

Abstract

          The aim of study was developed a screening risks scoring system for hypoglycemia among diabetic patients. The study consisted of 2 phases as follows: Phase 1 was a case-control study designed to find potential risk factors that may have led to hypoglycemia among diabetic patients. The quality of the research tool was verified for validity and reliability. The sample of 360 patients met the criteria. The data was analyzed by calculating adjusted odds ratios, 95% CI of the ORs using multiple logistic regression models. Phase 2 was the subsequent creation of a screening tool based on the risk factors identified from the first phase. A population-based cohort of 1,200 participant, We observed individuals who were admitted with hypoglycemia at baseline for 10 months for the development of screening risks. Multivariate logistic regression model coefficients were used to assign a score to each category of the variable categorist. The screening risk score value was derived from a receiver operating characteristic curve and the 95% CI in order to find a suitable cut-off point. The evaluate of validity and precision for the screening risk score was considered sensitivity, specificity, accuracy, predictive value and Likelihood Ratio of the test. The validity of the score was tested in patients receiving treatment at the outpatient department by prospective follow-up of 10 months. The gold standard of the study was fasting blood sugar.

           Results: The first phase is Risk Factors for Hospitalization Due to hypoglycemia in Diabetic Patient in North Thailand. It was published in Journal of Diabetes Mellitus, 2014. For the second phase, screening tool for assessing risks of hypoglycemia among diabetic patients has been developed. The risk score of each independent variable or each item of the tool was obtained by using the beta coefficients of the equation deriving from the multiple logistic regression model. The risk score for each participant ranged from 0 to 12 points. The quality of the screening risk was estimated as follows: the area under the ROC curve was 0.814 (95%CI=0.783-0.846; P-value < 0.001), which was good, the most reasonable cut-off point was 5, sensitivity 82.1% (95%CI=0.74-0.91), specificity 75% (95%CI=0.74-0.77), PPV 43.8% (95%CI=0.41-0.47), NPV 94.8% (95%CI=0.94-0.95) and Likelihood Ratio Positive 3.41 and Accuracy of test 77.1%.