Estimates of insulin needs and dispensation given wastage, alternative glycemic targets, and non-insulin therapies in US populations with type 2 diabetes mellitus: A microsimulation study.

Aims: Registries and health plans estimate insulin need for population health metrics. We sought to identify how such estimates affect population- and individual-level estimates of over- and under-treatment.

Methods: We developed a microsimulation comparing estimated insulin need to dispensation using the National Health and Nutrition Examination Survey (NHANES, 2005-2016, N= 2832) and Medical Expenditure Panel Survey (MEPS, 2005-2016, N = 29,615).

Results: From NHANES, ~21.6% of people with type 2 diabetes would require insulin to achieve a HbA1c target of 7% after maximum titration of two non-insulins (60.7 IU/person/day, or 84,629,833 vials of 1000 IU in the US). From MEPS, we observed 57.4 IU/person/day of insulin dispensed (81,585,842 vials). About 29% of people were dispensed at least two standard deviations less than their estimated need, and 22% at least two standard deviations more than estimated need. Population-level need estimates reduced 39.4% if liberalizing HbA1c targets to 8% for people ≥75 years old.

Conclusions: Estimated insulin needs of people with type 2 diabetes in the U.S. are consistent with their dispensed insulin at the population level, but are sensitive to HbA1c targets for older adults, and conceal under- and over-treated subpopulations.