Study design
We used a retrospective matched cohort study design to analyze de-identified medical and pharmacy claims data and enrollment information from Optum’s de-identified Clinformatics Data Mart database, spanning January 1, 2013 to December 31, 2018.
Commercial and Medicare Advantage health plan enrollees who initiated treatment with UMEC/VI or TIO (dry powder inhaler) between January 1, 2014 and December 31, 2017 (identification period; Fig. 1) were identified. The date of each patient’s first pharmacy claim for either UMEC/VI or TIO during the identification period was used as their index date. Patient characteristics were assessed during the pre-index period, which covered the 12-month period before the index date. Study outcomes were evaluated during the 12-month post-index period, censored at the end of eligibility, end of data availability, or death (whichever occurred earliest). For analyses of COPD exacerbations and medical costs, the on-treatment period spanned from the index date until the earliest of discontinuation of the index medication (defined as a ≥ 45-day gap [115-day gap for mail orders] in days of supply between the end of a dispensing and the next fill, or between the end of the last dispensing and the end of data availability), a switch to a non-index medication, end of eligibility, end of data availability, or death (whichever occurred earliest). Each patient was included in one of two mutually exclusive cohorts based on whether they were receiving UMEC/VI or TIO at index.
Patients
Patients were eligible for inclusion in the study if they had ≥ 1 pharmacy claim for fixed-dose UMEC/VI or TIO during the identification period, and had continuous enrollment with medical and pharmacy coverage during the 12-month pre- and post-index periods. Additional inclusion criteria were ≥ 1 medical claim with a COPD diagnosis code (International Classification of Diseases 9th Edition [ICD-9; before October 1, 2015] or 10th Edition [ICD-10; on or after October 1, 2015]; Additional file 1: Table 1) in any position on the index date or during the pre-index period, and age ≥ 40 years as of the index date. Patients with a moderate or severe COPD-related exacerbation on the index date were excluded, whereas patients with exacerbations at any other time during the pre-index period remained eligible. In addition, patients who had any of the following were excluded from the study: a medical claim with an ICD-9 or ICD-10 diagnosis code for asthma (Additional file 1: Table 1) during the eligibility period; a pharmacy claim for any maintenance medication containing an inhaled corticosteroid, LABA, or LAMA during the pre-index period; a pharmacy claim for a non-index maintenance medication (including single or multiple inhaler triple therapy) on the index date; or a pharmacy claim for both UMEC/VI and TIO on the index date.
Outcomes
Patient demographics, clinical characteristics, respiratory medication history, COPD-related exacerbations, and COPD-related HCRU and medical costs were assessed during the 12-month pre-index period and on the index date.
On-treatment Kaplan–Meier (KM) rates of time-to-first (TTF) exacerbation and rates of overall, moderate, and severe COPD-related exacerbations were reported for the 12-month post-index period. Exacerbations with a start date on or before the index date were considered part of the pre-index period and were not included in the analysis of post-index outcomes. Moderate COPD-related exacerbations were outpatient or emergency room (ER) visits with a diagnosis code for COPD in the primary position (Additional file 1: Table 1) and ≥ 1 dispensing/administration of systemic corticosteroids or guideline-recommended antibiotics in the 5 days prior to or following the visit. Severe COPD-related exacerbations were defined as an inpatient hospitalization with a diagnosis code for COPD in the primary position (Additional file 1: Table 1). If a patient had ≥ 2 exacerbations within 14 days, these were considered as one exacerbation and categorized based on the most severe event.
Medication adherence measures included the mean proportion of days covered (PDC) and the proportion of adherent patients (PDC ≥ 0.8). PDC was calculated as the number of days a patient had available index medication, identified using filled prescriptions, divided by the number of days between the index date and the end of the 12-month post-index period. For patients who refilled a medication before their previous fill ran out (ie, overlapping dispensings), the refill date was considered to be the end of the previous prescription’s days of supply.
Mean COPD-related medical costs were identified as any claim with a primary or secondary diagnosis of COPD (Additional file 1: Table S1) and evaluated during the on-treatment period. Total COPD-related medical costs were stratified by hospitalization, ER visits, outpatient visits, and other visit costs.
Data analyses
Patients were matched 1:1 between the UMEC/VI and TIO cohorts using propensity scores (PS) estimated from a multivariable logistic regression model, as previously described [14]. Patients were also exact-matched on pre-index COPD-related moderate or severe exacerbations.
Patient characteristics were evaluated during the 12-month pre-index period and compared between the UMEC/VI and TIO cohorts. Descriptive statistics including mean, standard deviation (SD), and median values for continuous variables, and relative frequencies and proportions for categorical variables were reported for pre-index characteristics in both the unmatched and matched cohorts. Standardized differences between cohorts of < 10% were not considered statistically relevant.
TTF on-treatment COPD-related exacerbation was evaluated using KM survival analysis, with KM exacerbation rates reported at 3, 6, 9, and 12 months post-index and compared between matched cohorts using hazard ratios (HR) from Cox proportional hazards regression with robust standard errors.
Rates of on-treatment COPD-related moderate and severe exacerbations were reported per 100 person-days and compared between matched cohorts with rate ratios estimated using generalized estimating equations Poisson regression. The 95% confidence intervals (CIs) and p-values were generated from non-parametric bootstrap procedures.
Mean PDC and the proportion of adherent patients (PDC ≥ 0.8) were compared between matched cohorts using generalized estimating equations and conditional logistic regression models, respectively.
On-treatment COPD-related medical costs were calculated per patient per month (PPPM), and inflation-adjusted to 2019 $US based on the medical care component of the Consumer Price Index. Statistical significance of cost differences between matched cohorts was evaluated using non-parametric bootstrap procedures.