OBJECTIVES: Insulin pump use among people with type 2 diabetes (T2D) is increasing. This work aims to identify predictors of insulin pump initiation among people with T2D.

METHODS: This was a retrospective, nested case-control study using the IBM^® Marketscan^® Commercial Database (2015-2020). People with T2D (18-64 years) with first prescription for bolus insulin (index date) on or after 1 January 2016 and no prior pump claims were included. From this cohort, cases (pump initiators) were identified and matched (1:4, with replacement) with controls using incidence density sampling for time at risk (time from index to the lookback date, which was set to pump initiation for cases) and index date (±6 months). Conditional logistic regression (CLR) and penalized CLR models assessed predictors of pump initiation and model choice was evaluated using the Akaike information criterion. A sensitivity analysis implemented matching without replacement.

RESULTS: From the cohort (N=32,104; mean age (standard deviation), 51.8 (9.5); 49.1% female), 726 cases were identified and matched to controls (n=2,904). In the 6 months prior to lookback, the top 5 predictors were CGM/FGM use (vs. BGM only; odds ratio, 5.4; 95% confidence interval, [3.8, 7.7]), presence of an endocrinologist visit (2.7 [2.2, 3.4]), acute metabolic complications (e.g., ketoacidosis, hypoglycaemia with coma; 2.5 [1.2, 5.4]), insulin use (no insulin vs. insulin; 0.2 [0.1-0.4]), and West region (vs. North Central; 0.6 [0.4, 0.9]). Results of the sensitivity analysis were similar to the primary analysis, with the exception that elevated HbA1c replaced region in the top 5 predictors.

CONCLUSIONS: Factors associated with pump initiation among people with T2D include technology use (CGM), diabetes engagement including endocrinologist visits, maintained insulin use, and acute metabolic complications.