Falling into the coverage gap: Part D drug costs and adherence for Medicare Advantage prescription drug plan beneficiaries with diabetes

Abstract

Objective: To compare drug costs and adherence among Medicare beneficiaries with the standard Part D coverage gap versus supplemental gap coverage in 2006.

Data sources: Pharmacy data from Medicare Advantage Prescription Drug (MAPD) plans.

Study design: Parallel analyses comparing beneficiaries aged 65+ with diabetes in an integrated MAPD with a gap versus no gap (n=28,780); and in a network-model MAPD with a gap versus generic-only coverage during the gap (n=14,984).

Principal findings: Drug spending was 3 percent (95 percent confidence interval [CI]: 1-4 percent) and 4 percent (CI: 1-6 percent) lower among beneficiaries with a gap versus full or generic-only gap coverage, respectively. Out-of-pocket expenditures were 189 percent higher (CI: 185-193 percent) and adherence to three chronic drug classes was lower among those with a gap versus no gap (e.g., odds ratio=0.83, CI: 0.79-0.88, for oral diabetes drugs). Annual out-of-pocket spending was 14 percent higher (CI: 10-17 percent) for beneficiaries with a gap versus generic-only gap coverage, but levels of adherence were similar.

Conclusions: Among Medicare beneficiaries with diabetes, having the Part D coverage gap resulted in lower total drug costs, but higher out-of-pocket spending and worse adherence compared with having no gap. Having generic-only coverage during the gap appeared to confer limited benefits compared with having no gap coverage.

Figure 1

Adjusted Drug Costs in Months before and after Beneficiaries Reached the Gap Threshold among Subjects with >U.S.$2,250 in Total Drug Costs in 2006 Note. These graphs present adjusted monthly total (top panel) and out-of-pocket (bottom panel) drug costs for all Part D drugs in up to 6 months before (−6) and 3 months after (+3) beneficiaries exceeded the coverage gap threshold in 2006 (U.S.$2,250 in total drug costs). Subjects were aligned by the month they exceeded the gap threshold (month 0). All comparisons are within individual health systems: in the Integrated MAPD, we compared having a coverage gap versus no gap (left panels); in the Network MAPD we compared having a coverage gap versus generic-only coverage during the gap (right panels). Costs were estimated using a one-part generalized linear model with log transformed costs and a generalized estimating approach; we adjusted for covariates using a propensity score. Error bars represent 95 percent confidence intervals and were calculated using the delta method. Costs spike at month zero because subjects are aligned by the month in which their drug costs reached U.S.$2,250. Results in tabular format are available in the supporting information Appendix SA1.

Figure 2

Adjusted Adherence in Months before and after Beneficiaries Reached the Gap Threshold among Subjects with >U.S.$2,250 in Total Drug Costs in 2006 Note. The Y-axis scale for all graphs is 50–90 percent. These graphs present the adjusted percent of subjects adherent to oral diabetes (top panels), hypertension (middle panels), and lipid (bottom panels) drugs in up to 6 months before and 3 months after beneficiaries exceeded the coverage gap threshold in 2006 (U.S.$2,250 in total drug costs). All comparisons are within individual health systems: in the Integrated MAPD, we compared having a coverage gap versus no gap (left panels); in the Network MAPD, we compared having a coverage gap versus generic-only coverage during the gap (right panels). Subjects were aligned by the month they exceeded the gap threshold (month 0). Odds of adherence were estimated using a generalized estimating approach with a logit link. We adjusted for covaritates using a propensity score; adjusted percentages were calculated treating all subjects as if they had a coverage gap or supplemented gap, respectively. Error bars represent 95 percent confidence intervals and were calculated using the delta method. Results in tabular format are available in thesupporting information Appendix SA1.