Abstract
PURPOSE: To investigate the impact of clinical and socio-economic factors on glycaemic control and construct statistical models to predict optimal glycaemic control (OGC) after implementing intermittently scanned continuous glucose monitoring (isCGM) systems. METHODS: This retrospective study included 1072 type 1 diabetes patients (49.0% female) from three centres using isCGM systems. Clinical data and net income from the census tract were collected for each individual. OGC was defined as time in range > 70%, with time below 70 mg/dL < 4%. The sample was randomly split in two equal parts. Logistic regression models to predict OGC were developed in one of the samples, and the best model was selected using the Akaike information criterion and adjusted for Pearson's and Hosmer-Lemeshow's statistics. Model reliability was assessed via external validation in the second sample and internal validation using bootstrap resampling. RESULTS: Out of 2314 models explored, the most effective predictor model included annual net income per person, sex, age, diabetes duration, pre-isCGM HbA1c, insulin dose/kg, and the interaction between sex and HbA1c. When applied to the validation cohort, this model demonstrated 72.6% specificity, 67.3% sensitivity, and an area under the curve (AUC) of 0.736. The AUC through bootstrap resampling was 0.756. Overall, the model's validity in the external cohort was 80.4%. CONCLUSIONS: Clinical and socio-economic factors significantly influence OGC in type 1 diabetes. The application of statistical models offers a reliable means of predicting the likelihood of achieving OGC following isCGM system implementation.