Using observational data to inform the design of a prospective effectiveness study for a novel insulin delivery device

Abstract

Michael Grabner,1 Yong Chen,2 Matthew Nguyen,3 Scott D Abbott,3 Ralph Quimbo1 1HealthCore, Inc., Wilmington, DE, USA; 2Merck and Co., Inc., Whitehouse Station, NJ, USA; 3Valeritas, Inc., Bridgewater, NJ, USA Objective: To inform the design and assess the feasibility of a prospective effectiveness study evaluating an insulin delivery device for patients with diabetes mellitus to be conducted within the membership of a large US commercial insurer. Methods: Providers who issued ≥1 insulin prescription between January 1, 2011 and September 30, 2011 were selected from administrative claims contained in the HealthCore Integrated Research DatabaseSM. Adult diabetes patients with visits to these providers were identified. Providers were dichotomized into high- [HVPs] and low-volume providers [LVPs] based on median number of diabetes patients per provider. Results: We identified 15,349 HVPs and 15,313 LVPs (median number of patients = 14). Most HVPs were located in the Midwest (6,291 [41.0%]) and South (5,092 [33.2%]), while LVPs were evenly distributed across regions. Over 80% (12,769) of HVPs practiced family or internal medicine; 6.4% (989) were endocrinologists. HVPs prescribed insulin to an average of 25% of patients. Patients of HVPs (522,527) had similar characteristics as patients of LVPs (80,669), except for geographical dispersion, which followed that of providers. Approximately 65% of patients were aged 21-64 years and 97% had type 2 diabetes. Among patients with ≥1 available HbA1C result during 2011 (103,992), 48.3% (50,193) had an average HbA1C ≥7.0%. Among patients initiating insulin, 79.6% (22,205) had an average HbA1C ≥7.0%. Conclusion: The observed provider and patient populations support the feasibility of the prospective study. Sampling of patients from HVPs is efficient while minimizing bias as patient characteristics are similar to those from LVPs. The study also highlights unmet needs for improved glycemic control since approximately half of patients with diabetes are not on goal. Keywords: diabetes mellitus, claims analysis, prospective study design, insulin delivery