Patient Characteristics Associated With Non-acceptance of Intermittently Scanned Continuous Glucose Monitoring...

Objective Intermittently scanned continuous glucose monitoring (isCGM) improves glycemic outcomes in patients with diabetes; however, some patients decline its use even when it is recommended by healthcare providers. We investigated patient characteristics associated with acceptance of isCGM in individuals with type 2 diabetes treated with insulin. Methods This single-center, retrospective study included outpatients with type 2 diabetes who were receiving insulin therapy and were advised to use isCGM by a single physician between June 2022 and December 2024. Patients with impaired decision-making capacity, clear reasons for declining device attachment (such as dermatological disease), definite type 1 diabetes, or gestational diabetes were excluded. The primary outcome was non-acceptance of isCGM after physician recommendation. Fisher's exact test and the Wilcoxon test were performed to compare the isCGM acceptance group with the non-acceptance group. Univariate and multivariable logistic regression analyses were performed using acceptance or non-acceptance of isCGM as the dependent variable. To address potential overfitting and correlations among predictors, elastic net penalized logistic regression with 10-fold cross-validation was performed. Lambda (λ) is the tuning parameter that controls the strength of the penalty term in the loss function. The optimal λ value was selected for model regularization, and predictors with non-zero coefficients at that value were selected. Bidirectional stepwise selection was used as a sensitivity analysis. Results A total of 142 patients were included, of whom 98 accepted isCGM and 44 did not. Compared to the acceptance group, the non-acceptance group had a higher total insulin dose, longer duration of insulin use, and higher body mass index (BMI). In univariate logistic regression, log-transformed total daily insulin dose, log-transformed duration of insulin use, and BMI were associated with lower odds of accepting isCGM, whereas no variable remained statistically significant in the multivariable model. In elastic net analysis, BMI, log-transformed total daily insulin dose, log-transformed duration of insulin use, and log-transformed HbA1c were selected. The area under the curve was 0.63, suggesting modest discriminative ability. Repeated elastic net analyses showed high selection frequencies for BMI, total insulin dose, duration of insulin use, and HbA1c. Age was not meaningfully associated with acceptance. Conclusion Treatment intensity, reflected by a higher insulin dose, longer duration of insulin use, and a higher number of insulin injections, may be associated with declining isCGM. Higher HbA1c may be associated with willingness to accept isCGM, whereas older age alone may not be a major barrier. These exploratory findings require confirmation in larger prospective studies.