Many studies have documented the delay in starting insulin, while this isn’t the major point 0f this study, it does provide a nice demonstration . The authors identified 15 studies in which basal insulin ws started in insulin naïve people with type 2 diabetes, and this table shows the starting A1Cs of these 2,837 pwd, segmented by the ending insulin dosage.
By definition, this population was willing to start insulin, since they freely volunteered for a clinical trial, yet the average starting A1C in the 2 groups was 8.6% and 9.0%. The question of why insulin wasn’t started sooner doesn’t seem to have a simple answer. See Abstract