Factors affecting blood sugar changes in diabetic patients using a three-level model in analysis of longitudinal data

Abstract

Background: Diabetes, a currently threatening disease, has severe consequences for individuals’ health conditions. The present study aimed to investigate the factors affecting the changes in the longitudinal outcome of blood sugar using a three-level analysis with the presence of missing data in diabetic patients. Methods: A total of 526 diabetic patients were followed longitudinally selected from the annual data collected from the rural population monitored by Tonekabon health centers in the North of Iran during 2018-2019 from the Iranian Integrated Health System (SIB) database. In analyzing this longitudinal data, the three-level model (level 1: observation (time), level 2: subject, level 3: health center) was carried out with multiple imputations of possible missing values in longitudinal data. Results: Results of fitting the three-level model indicated that every unit of change in the body mass index (BMI) significantly increased the fasting blood sugar by an average of 0.5 mg/dl (p=0.024). The impact of level 1 (observations) was insignificant in the three-level model. Still, the random effect of level 3 (healthcare centers) showed a highly significant measure for health centers (14.62, p<0.001). Conclusion: The BMI reduction, the healthcare centers' socioeconomic status, and the health services provided have potential effects in controlling diabetes.

Keywords

• blood sugar change
• diabetic patients
• body mass index
• health care centers
• three-level model
• missing data
• fcs imputation algorithm
• longitudinal data