6F-01 - One Disease, Three Definitions: Unpacking the Impact of Cohort Entry Date Choice on Type 2 Diabetes Identification Using Claims

Background: Real-world burden of disease studies using medical claims often lack clarity and consistency in defining disease population. Specifically, the differences due to selection of cohort entry dates have not been described. This study addressed this gap by examining three methods for identifying individuals with type 2 diabetes (T2D) to illustrate the impact in a chronic disease.

Objectives: This study aimed to evaluate the impact of three cohort entry date definitions on the size and clinical characteristics of T2D cohorts.

Methods: This study used Optum claims data (2020-2023) to identify individuals with T2D, requiring one inpatient or two outpatient visits with ICD10 diagnosis code E11.xx. Additionally, individuals needed at least one year baseline of continuous enrollment (enrollment criterion).
The study applied three methods to define the cohort entry date and construct cohort. Method 1 (M1) defined cohort entry as the first date when the diagnostic criterion was met and then applied the enrollment criterion to construct the cohort. Method 2 (M2) set the cohort entry date as the first date when both the diagnostic and enrollment criteria were met, allowing dates without healthcare encounters. Method 3 (M3) was built on M2 but defined the cohort entry date as the first T2D healthcare encounter after meeting all criteria. All methods ensure the follow-up period starts only after meeting all inclusion criteria, avoiding immortal time bias.
The study followed individuals until insurance disenrollment or death. It described the size and key clinical characteristics of the three cohorts.

Results: The three cohorts had an average age of 68-69 years and 50-51% females.
M2 identified the largest cohort (N=3,472,189), while M1 yielded the smallest (N=2,307,731) due to implicitly excluding individuals with T2D claims in the baseline year, primarily those with existing T2D. M3 identified slightly fewer individuals than M2, with 3,077,784 individuals, as it required an additional healthcare encounter after meeting inclusion criteria.
M1 and M3 base all cohort entry dates on healthcare encounters, while 31% of M2's dates are not. Inpatient visits account for 16% of entry dates in M1, compared to 11% and 10% in M2 and M3. The length of follow-up (LFU) was longest in the cohort defined by M1, with a median of 809 days. M2 had a shorter LFU of 669 days, and M3 had the shortest of 649 days.
The mortality rate was lowest in the cohort defined by M1 (50.7%) compared to M2 (57.0%) and M3 (57.5%).

Conclusions: The choice of cohort entry date significantly impacts T2D cohort identification. M1 should be used with caution, as existing T2D cases with frequent healthcare use are excluded. Selecting an appropriate cohort entry date that aligns with the study's target population and objective is crucial.