Data Source and Study Design
This was a retrospective matched cohort study using healthcare insurance claims from Optum’s de-identified Clinformatics Data Mart database, with data ranging from January 1, 2013, to December 31, 2018. The database comprises claims data from a geographically diverse population of 12–14 million United Health Group members, including both commercial and Medicare Advantage health plan data, and contains historical data on patient demographics, insurance coverage, dates of eligibility and death, claims for inpatient and outpatient visits, costs of services, and filled prescriptions.
The study was designed and conducted in accordance with the principles of the 1964 Declaration of Helsinki and its later amendments. The study used fully de-identified retrospective claims data, which were compliant with the patient requirements of the Health Insurance Portability and Accountability Act of 1996 (HIPAA), and as such was not classified as research involving human participants. Therefore, institutional review board approval was not required.
The index date was defined as the date of the first pharmacy claim for UMEC/VI or TIO during the identification period, which spanned from January 1, 2014, to December 31, 2018 (Fig. 1). Patients were classified into mutually exclusive cohorts based on their index medication (UMEC/VI or TIO). The on-treatment period was defined as the period from the index date to the date of a pharmacy fill for a non-index maintenance medication (i.e., inhaled corticosteroid [ICS], LABA, or LAMA), discontinuation of the index medication, end of continuous eligibility, end of data availability, or death. Discontinuation was defined as a ≥ 45-day gap (≥ 115 days for mail orders) between the end of a dispensing and the next fill, or between the end of the last dispensing and the end of data availability. For the analysis of inpatient readmissions, the follow-up period was defined as the 30-day or 90-day period after the index discharge date (defined as the discharge date for the first COPD-related inpatient admission during the on-treatment period) (Fig. 1).
Patients included in the primary analysis of inpatient admissions were ≥ 40 years of age at index, and were required to have: ≥ 1 pharmacy claim for UMEC/VI or TIO (via dry-powder inhaler) during the identification period; ≥ 12 months of continuous medical and pharmacy coverage during the pre-index period; and ≥ 1 medical claim with a diagnosis for COPD during the 12-month pre-index period, or on the index date. A complete list of International Statistical Classification of Diseases and Related Health Problems (ICD) codes used to identify claims for COPD and COPD exacerbations is included in Table S1 (in the electronic supplementary material). Patients were excluded if they had: ≥ 1 pharmacy claim for maintenance medication containing ICS, LABA, or LAMA during the 12-month pre-index period; ≥ 1 pharmacy claim on the index date for both UMEC/VI and TIO, a non-index controller medication, or single- or multiple-inhaler triple therapy (ICS/LAMA/LABA); or ≥ 1 medical claim with a diagnosis for asthma any time prior to or after the index date (ICD 9th Edition Clinical Modification codes: 493.xx; ICD 10th Edition Clinical Modification codes: J45.3x, J45.4x, J45.5x, J45.9xx).
Patients included in the readmission analysis were also required to have ≥ 1 COPD-related admission during the on-treatment period, to be alive at the index discharge date, and to have continuous medical and pharmacy coverage for ≥ 30 or ≥ 90 days after the index discharge date. Readmissions were also evaluated among a subgroup of patients with a short initial COPD-related admission (length of stay [LOS] of 1–3 days).
The primary outcome was time to first COPD-related inpatient admission during the on-treatment period, defined as an inpatient admission with a primary or secondary diagnosis of COPD (Table S1 in the electronic supplementary material). Secondary outcomes included the rate of COPD-related inpatient admissions during the on-treatment period (up to 12 months post-index) and the rate of all-cause and COPD-related inpatient readmissions within 30 and 90 days of the index discharge date. Readmission rates were assessed regardless of whether patients were on treatment at the time of readmission; medication changes after the initial inpatient admission were reported.
Patients in the UMEC/VI and TIO cohorts were matched (1:1) using propensity score matching based on probability estimates from a logistic regression model including the following baseline covariates: age, sex, region, insurance type, year and quarter of index date, Quan-Charlson Comorbidity Index (CCI) score, proportion of patients with pre-index moderate and severe exacerbations, respiratory medications (including systemic corticosteroids, short-acting β2-agonists [SABA], short-acting muscarinic antagonists [SAMA], and SABA/SAMA combinations), HRU, medical costs, and Elixhauser  and Diagnostic and Statistical Manual-V (DSM-V) comorbidities with a prevalence ≥ 5% in either cohort (Table S2 in the electronic supplementary material).
Descriptive statistics for patient characteristics during the pre-index period included means and standard deviations (SD) for continuous variables and relative frequencies for categorical variables. The primary outcome was assessed using Kaplan–Meier survival analysis and compared between cohorts using hazard ratios (HR) from Cox proportional hazards regression models. COPD-related inpatient admission rates were reported per 100 person-days and compared using rate ratios (RR) from Poisson regression models; confidence intervals (CI) and p values were calculated using non-parametric bootstrap procedures. The proportions of patients in either cohort who switched medication after the initial inpatient admission were reported. The proportions of patients with 30-day and 90-day readmissions were reported, and readmission rates were calculated as the total number of events per patient (not censored after the first event) divided by the total number of patients.