MIXED-EFFECTS MODEL FOR LONGITUDINAL STUDY OF FASTING BLOOD SUGAR LEVEL AMONG TYPE 2 DIABETES PATIENTS

Abstract

The main purpose of the study was to determine the pattern of change in Fasting Blood Sugar (FBS) level as well as to obtain the best model for predicting patient's FBS level based on some covariates. A historic data is obtained from the diabetes unit in Ketu-South Municipal Hospital, on 26 (32.5%) males and 54 (67.5%) females of the 80 type 2 diabetes patients on treatment for a 2-year period. Their FBS level, body weight and blood pressure were regularly monitored and thus generating repeated measures. Profile analysis was use to study the pattern of change in the FBS level with respect to the covariates. Linear mixed effect model was use for modeling the FBS level of the patients. The outcome showed that, the trend of the FBS level over time follows cubic function, indicating that initially the FBS level usually increases and then eventually declines only to rise again. The duration of treatment, the body weight, the blood pressure (systolic and diastolic), and the educational status were the factors that significantly influenced the FBS level of patients on treatment. A stepwise selection method was use to fit a reduced regression prediction model from a full prediction model. The trend equation was developed to estimate the rate of change in FBS level. Diagnostic test confirmed that the regression and trend models based on the covariates are adequate for predicting and estimating FBS level of diabetic patients on treatment. It is therefore, recommended that further research should be done to include other risk factors of diabetes in order to improve upon the regression model for the strategic intervention and management of diabetes.