Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Health Outcomes and Policy Research


Health Policy

Research Advisor

James E. Bailey, MD


Justin Gatwod, PhD; Cameron Kaplan, PhD; Satya Surbhi, PhD; Elizabeth A. Tolley, PhD


Cost-effectiveness, Healthcare utilization, Insulin, Low-income population, Medication adherence


Background: Despite being a lifesaving medicine for many patients with diabetes, insulin cost is growing out of reach for many. Of the two broad classes of insulin available in the US, analog insulin is newer, costlier, and more widely used than human insulin. The new mode of delivery with pen devices has shown promise in improving adherence and achieving glycemic control and gaining popularity. But comparing adherence to different insulins is challenging due to the lack of a validated insulin adherence method. Very little is known about the comparative benefits of different insulin types and different modes of delivery in the US. Objectives: The main goal of this project was to compare the real-world benefits of different insulin types. To accomplish this, three studies were conducted with the following aims: (1) to identify an effective method for measuring insulin adherence by its ability to predict change in hemoglobin A1c (A1c); (2) to examine the differences in 2a) improvement in a1c, 2b) adherence, and 2c) healthcare utilization between users of the analog vial, analog pen, and human insulin; and (3) to compare the cost-effectiveness of the analog vial, analog pen, and human insulin from the payer’s perspective in terms of 3a) cost per unit A1c reduction and 3b) cost per ER visit and hospitalization avoided. Methods: This project involved retrospective cohort studies using administrative claims data from Tennessee Medicaid and electronic health records from a regional diabetes registry in the Mid-South. Participants included were US adults with type 2 diabetes who filled at least 2 prescriptions for insulin between 1/1/2016 and 6/30/2018. For the first study, medication possession ratio (MPR), proportion of days covered (PDC), and medication user rate (MUR) were compared by the ability of the three measures to predict change in A1c using multiple linear regression and receiver operating characteristic curve. Each patient’s A1c records and body-mass index were extracted from the registry. All other information including insulin use was collected from the Medicaid claims data. The second study employed multivariate logistic regression to compare the improvement in A1c and adherence, and negative binomial regression to compare the rates of emergency room (ER) visits and hospitalizations between analog vial, analog pen, and human insulin users. Cost-effectiveness was analyzed for the third study using the claims and registry data. Adjusted mean cost, mean A1c reduction, and mean ER visit and hospitalization were determined using generalized linear models with gamma distribution, multiple linear regression, and negative binomial regression. Cost-effectiveness was computed by the incremental cost-effectiveness ratio (ICER). Results: The cohort for studies 1, 2a, and 3a consisted of 533 patients and was a subset of the cohort for studies 2b, 2c, and 3b with 2,763 patients. In both cohorts and all insulin groups, the majority were female and African Americans and resided in health-professional-shortage areas. Adherence to insulin was poor across all insulin groups ranging between 53\% to 60\%. We found MPR and PDC to be strong predictors of change in A1c and identified both as effective methods for measuring insulin adherence. The second study used MPR and PDC to compare adherence between insulin groups. Analog vial users were less likely to be adherent compared to human insulin users (OR: 0.686; 95\% CI: 0.531 – 0.885). No difference in adherence was observed between analog pen and human insulin. No significant association between insulin type and A1c improvement was detected, indicating neither type of insulin was more likely than another to show improvement in A1c. Analog vial users were more likely to visit the ER compared to human insulin users (OR: 1.275; 95\% 1.027 – 1.582) and analog pen users (OR: 1.528; 95\% CI: 1.289 – 1.812) and be hospitalized compared to analog pen users (OR: 1.820; 95\% CI: 1.315 – 20519). Analog pen users were less likely to be hospitalized compared with human insulin users (OR: 0.637; 95\% CI: 0.407 – 0.997). In the third study, human insulin was dominant over analog vial in cost per unit A1c reduction and cost per ER visit and hospitalization. Analog pen was more cost-effective than analog vial with small ICER values. Conclusion: Our study demonstrated that analog insulin when delivered in a vial, had worse adherence, higher rates of ER visits and hospitalizations, and less cost-effectiveness in comparison with human insulin. Users of analog insulin via pen devices had better adherence and this type of insulin was more cost-effective. Physicians, payers, and policymakers should consider the real-world benefits of analog vs. human insulin when choosing or prioritizing insulin for patients. Future research is warranted to compare the benefits of all insulin types and modes of delivery in other populations.

Declaration of Authorship

Declaration of Authorship is included in the supplemental files.




2023-016-Shuvo-DOA.pdf (104 kB)
Declaration of Authorship

Available for download on Saturday, May 10, 2025