Impact of Co-Payment Discounts on Antidepressant Medication Adherence and Costs
A unique co-payment discount program for antidepressants was associated with higher antidepressant medication adherence, lower likelihood of discontinuation, and lower healthcare costs.
Major depressive disorder (MDD) is the leading cause of disability worldwide1 and is projected to represent the second-largest disease burden globally by 2020.2 In the United States, MDD affects roughly 35 million adults annually, with an estimated lifetime prevalence of 16%,3 and was the second-leading contributor to years lived with disability in 2010 a 13% increase from 2000.4 On average, there were 7.9 million ambulatory visits for MDD in 2006 and 2007,5 and 383,000 hospital discharges with primary MDD diagnoses in 2009, accounting for 2.6 million hospital days.6 The most recent study reporting overall costs associated with depression in the United States estimated that in 2000, direct costs were $26.1 billion and indirect costs were $51.5 billion.7
The 2010 American Psychiatric Association practice guideline for MDD includes antidepressant therapy, alone or with psychotherapy.8 Medications approved for MDD include selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants, bupropion, nefazodone, and mirtazapine. SSRIs and newer SNRIs, such as desvenlafaxine (DES) and extended-release venlafaxine (VENXR), are considered first-line treatment options.9
Research based on a variety of data suggest that adherence to antidepressants is often poor.10-14 Studies of antidepressant adherence using administrative claims data measured adherence with medication possession ratio or proportion of days covered (PDC) defined as the percentage of days subjects had their antidepressants on hand relative to the total days they were observed. Using 75% or 80% as the threshold for adherence, 34% to 51% of patients were adherent over 4 to 6 months.10-12 Poor adherence and early discontinuation of antidepressants are associated with higher risk of relapse,15,16 higher rates of short-term disability,17 and higher healthcare costs.10,18 Conversely, improved adherence to antidepressants is associated with prolonged remission from depression,19 increased productivity,20 and reduced medical costs.10,18 Cantrell and colleagues10 found yearly medical costs were $423 to $511 lower among adherent patients treated with antidepressants; Eaddy and colleagues18 reported that annual medical plus SSRI costs were $968 lower among more adherent patients.
Programs to improve medication adherence through enhanced patient interaction (eg, patient education, follow-up reminders, counseling) have been implemented for multiple medical conditions, including depression, with variable success.14,21,22 A systematic review of randomized controlled interventions to improve adherence found 49% (19 of 39) were associated with statistically significant increases in medication adherence, and 44% (17 of 39) reported statistically significant improvements in treatment outcomes.22
Health plans also use direct economic incentives to improve medication adherence, though literature examining the impact of such incentives on treatment for depression is lacking. Published research suggests that lowering prescription co-payments can increase adherence to medications for chronic conditions such as congestive heart failure, diabetes, hyperlipidemia, and hypertension.23-25 In these studies, intervention groups that received co-payment reductions for all fills of targeted medications had adherence 2 to 4 percentage points higher in the first year,23,25 and up to a 5-percentage-point improvement over 2 years24 compared with control groups that received smaller or no co-payment reductions.
We are unaware of research evaluating co-payment reduction based on adherence to antidepressants. Interventions documented in the literature apply co-payment reductions to all fills of the targeted medications based on formulary status regardless of adherence,23-25 and a co-payment discount contingent upon timing of refills has not been evaluated.
In October 2009, a large US health plan began a “Refill and Save Program” (RSP), which ties a financial incentive directly to adherence. The RSP provides co-payment discounts (ie, $20 for a 30-day supply from retail pharmacies, or $50 for a 90-day supply mail-order prescription, representing a 40% discount, on average) for selected medications, including branded DES and VENXR, when refilled within 30 days after the end of a previous SSRI, SNRI, or bupropion fill.26 The RSP provides a unique opportunity to examine the relationship between financial incentives and antidepressant adherence. The study objective was to evaluate the impact of co-payment discounts for DES or VENXR through the RSP on adherence, discontinuation, and healthcare costs by comparing outcomes between health plan members with and without RSP benefits.
METHODS
Study Design and Data Source
This retrospective cohort analysis used data from the Optum Research Database, which contains medical and pharmacy claims and enrollment information from a large, geographically diverse US health insurance plan.
Sample Identification and Cohort Definition
The study population included commercially insured health plan members 18 years or older with ≥1 pharmacy claim for DES or VENXR from October 1, 2009, through March 31, 2010 (identification [ID] period), corresponding to the first 6 months of the RSP. Duloxetine, an SNRI, was included in the RSP. However, duloxetine is also indicated for multiple chronic pain conditions, and members on duloxetine likely would have been clinically different from those on DES or VENXR27; therefore, these members were excluded from the study population.
Pharmacy claims included RSP benefit and co-payment reduction indicators. Members with ≥1 pharmacy claim for DES or VENXR with an RSP benefit indicator during the ID period were assigned to the RSP cohort; their index date was the date of the first DES or VENXR claim with an RSP indicator. Members without an RSP indicator on any DES and VENXR claim during the ID period were assigned to the non-RSP cohort; their index date was the date of the first observed DES or VENXR claim during the ID period. The antidepressant filled on the index date was the index antidepressant.
Members were continuously enrolled with medical and pharmacy benefits for 6 months pre-index (baseline period, starting as early as April 1, 2009) and 9 months post index (follow-up period, starting on the index date and ending no later than December 31, 2010). Members with schizophrenia (International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis code 295. xx) were excluded, as were non-RSP cohort members with any pharmacy claims indicating an RSP benefit during baseline.
Members were further stratified by baseline index antidepressant use: 1) “naïve” users had no index antidepressant claims during baseline; and 2) “current” users had ≥1 index antidepressant claim during baseline. Members whose RSP status changed during follow-up (ie, from the non-RSP cohort to the RSP cohort, or vice versa) were also identified.
Measures
Member characteristics and health plan features. Baseline member characteristics included age, sex, US census region, index month, Charlson comorbidity index (CCI) score,28 and binary variables indicating use of index and non-index antidepressants. Health plan features included mental health coverage administered as a carve-out separate from the medical benefit; health plan affiliation via administrative services only (ASO; vs fully insured); and 2 indicators of high-deductible health plans (HDHP). HDHP indicators identified members with consistent index antidepressant co-payment values of $0 (vs ≥1 claim with co-payment value >$0); and members with any cost amount in the deductible field on pharmacy or medical claims during follow-up (vs no amount in the deductible field). When the co-payment values were consistently $0, we assumed members paid the entire cost of their drugs until their plan deductibles were met.
Outcomes. The primary outcomes were adherence to and discontinuation of the index antidepressant during follow-up. Adherence was measured as PDC, computed as the percentage of cumulative days’ supply of the index antidepressant out of total days in the follow-up period. If members were hospitalized during follow-up while on their index antidepressant, it was assumed the medication was provided by the facility during the inpatient stay. The RSP program required a refill within 30 days of the run-out of a previous antidepressant fill in order to receive the co-payment discount; therefore, discontinuation was defined as a gap of ≥30 days in index antidepressant use.
The secondary outcome measure was total healthcare costs, representing the sum of costs for pharmacy claims and medical services, including inpatient, emergency department, ambulatory, and other medical services (ie, those rendered in other venues). Total costs were computed as the sum of health plan and member-paid amounts. Costs were adjusted to 2010 dollars using the annual medical care component of the Consumer Price Index.29
Statistical Analyses
Unadjusted differences between the RSP and non-RSP cohorts were evaluated with t tests for means accounting for unequal variances as appropriate—and with Pearson χ2 tests for percentages. The relationship between RSP and PDC was modeled using ordinary least-squares regression, and the likelihood of discontinuation was modeled with Cox proportional hazards regression. Costs were modeled using a generalized linear model with a gamma distribution and log link.30 All descriptive and regression analyses were performed on the whole population, and separately for the naïve- and current-user strata. Covariates were potential confounders of the relationship between the RSP and study outcomes. All models were adjusted for cohort, age, sex, US census region, index month, index antide-pressant, mental health coverage indicator, ASO, 2 indicators of HDHPs, use of non-index antidepressants during baseline, and baseline PDC for all antidepressants. Models for the whole population were also adjusted for naïve- or current-user status. As a sensitivity analysis, regression models were estimated for members without an RSP status change during follow-up. All analyses were performed with SAS, version 9.2 (SAS Institute, Cary, North Carolina).
RESULTS
Member Sample
The study population was composed of 46,138 members, 28,925 (62.7%) and 17,213 (37.3%) in the RSP and non-RSP cohorts, respectively (
Figure
).
The mean age of all members was 48 ± 12 years, and 75% were female (
Table 1
); 21% were naïve to their index antidepressant. The RSP cohort had a slightly higher percentage of males (P <.001) and ASO health plan membership (P <.001) compared with the non-RSP cohort. Both cohorts had low CCI scores, with a significant but small difference between cohorts in the current-user strata. Ten percent (N = 4621) of the study population had an RSP status change during follow-up; 4% in the RSP cohort lost the RSP benefit and 20% in the non-RSP cohort gained the RSP benefit (data not shown).
Outcomes
Adherence (PDC). The mean unadjusted PDC was 67.4% among all users and was higher in the RSP cohort (68.9%) compared with the non-RSP cohort (65%; P <.001;
Table 2
). Mean PDC was also higher for the RSP versus non-RSP cohort in the naïve and current strata. The co-variate-adjusted mean PDC (
Table 3
; complete model is included in
eAppendix Table 1
, available at www.ajmc. com) was 6.3 percentage points higher in the RSP versus the non-RSP cohort (P <.001). Adjusted mean PDC was 7.7 and 6.1 percentage points higher in the RSP cohort within the naïve- and current-user strata, respectively (all P <.001, Table 3).
Discontinuation. Overall, 57.1% of members discontinued their index antidepressants with no significant differences between the RSP and non-RSP cohorts (Table 2). Among naïve users, a higher percentage of the non-RSP cohort discontinued (unadjusted: 79.3% vs 72.6% RSP co-hort, P <.001), while among current users discontinuation was higher in the RSP cohort (unadjusted: 53.4% vs 49.9% non-RSP, P <.001). The adjusted likelihood of discontinu-ation (Table 3; complete model is included in
eAppendix Table 2
) was 23.2% lower in the RSP cohort versus the non-RSP cohort, on average (hazard ratio [HR], 0.768; 95% CI, 0.746-0.790). This result was robust within the naïve-user (HR, 0.774; 95% CI, 0.737-0.814) and current-user (HR, 0.742; 95% CI, 0.716-0.769) strata. Testing of the Schoenfeld residuals indicated that the HRs were non-proportional, so the results reflect the average effect during follow-up.31 Inclusion of interactions between the model covariates and time to discontinuation provided similar HRs for the RSP indicator (results not shown).
Total Healthcare Costs
Unadjusted mean (SD) total healthcare costs during follow-up (Table 2) were $9111 ($17,335) in the RSP cohort and $9834 ($21,413) in the non-RSP cohort (P <.001). Among naïve users, total costs were $8867 ($17,943) for the RSP cohort versus $9710 ($22,057) for the non-RSP cohort (P = .046); among current users, total costs were $9174 ($17,175) for the RSP cohort versus $9871 ($21,219) for the non-RSP cohort (P = .001).
In general, unadjusted pharmacy costs were higher and unadjusted medical costs were lower in the RSP cohort versus the non-RSP cohort (Table 2). Ambulatory costs and inpatient costs were the main components of the medical costs, representing 65% and 26% of medical costs, respectively (data not shown).
Differences in covariate-adjusted total healthcare costs between the RSP and non-RSP cohorts were also statistically significant (Table 3; complete model is included in
eAppendix Table 3
). The RSP cohort had 7.6% lower mean healthcare costs compared with the non-RSP cohort (cost ratio [CR], 0.924; 95% CI, 0.886-0.962), with similar results among naïve users (CR, 0.911; 95% CI, 0.832-0.998) and current users (CR, 0.922; 95% CI, 0.882-0.963).
Sensitivity Analysis
The sensitivity analyses of members whose RSP status did not change during follow-up (data not shown) were consistent with the primary analysis. After adjustment for covariates, PDC was 1.3 (95% CI, 0.7-1.9) percentage points higher, likelihood of discontinuation was 6.1% lower (HR, 0.939; 95% CI, 0.910-0.970), and healthcare costs were 5.1% lower (CR, 0.949; 95% CI, 0.906-0.994) in the RSP cohort versus the non-RSP cohort.
In the naïve- and current-user strata, the adjusted estimates were in the same direction as the primary analysis; however, RSP versus non-RSP comparisons were not statistically significant for PDC among current users or for total healthcare costs among naïve users. Among naïve users, mean PDC was 4.1 (95% CI, 2.6-5.6) percentage points higher in the RSP cohort; however, differences between cohorts were not statistically significant among current users (estimate: 0.6; 95% CI, −0.1 to 1.3). Likelihood of discontinuation in the RSP cohort was lower among both naïve users (HR, 0.863; 95% CI, 0.817-0.911) and current users (HR, 0.957; 95% CI, 0.919-0.997). Differences in healthcare costs between cohorts were not statistically significant (CR, 0.954; 95% CI, 0.863-1.056) among naïve users, but among current users, costs were lower in the RSP cohort (CR, 0.941; 95% CI, 0.896-0.988).
DISCUSSION
Commercial health plan members with RSP adherence-based co-payment discounts had greater adherence to and less discontinuation of their index antidepressants than did members without the RSP; the RSP cohort also had lower healthcare costs. The sensitivity analysis of members with no evidence of RSP status change during follow-up was consistent with respect to discontinuation; PDC among naïve users; and total healthcare costs among current users.
The co-payment discount was positively associated with adherence consistent with studies of medication co-payment discounts for other chronic conditions24,25 but to our knowledge, this is the first study to focus on antidepressants. Adherence in the RSP cohort was 6.3 percentage points higher than in the non-RSP cohort. This effect is larger compared with prior studies and may be attributable to 2 factors. First, the RSP reduced co-payments for branded DES and VENRX by $20 to $50 per fill compared with the elimination of generic medication co-payments, representing approximately $11 savings per fill in other studies. Second, the RSP program tied the financial incentive to sustained adherence by requiring refills within 30 days after the end of a previous fill, whereas other programs provided co-payment discounts on all medication fills, irrespective of adherence.
It is possible that bias resulted from the cohort assignment criteria. First, members with lower adherence may have been assigned systematically to the non-RSP cohort (members in the non-RSP cohort may have discontinued their antidepressants before becoming eligible for the RSP). If bias were introduced, however, we would expect to observe a stronger RSP effect among current users; instead, the cohort effect was larger among naïve users, supporting the benefit of the RSP. Second, members could have changed RSP status during follow-up; this potential source of bias was addressed with the sensitivity analysis. Total healthcare costs were 7.6% lower in the RSP cohort. Total pharmacy costs were higher, on average, in the RSP cohort than in the non-RSP cohort. This was likely attributable, in part, to higher adherence in the RSP cohort. Although pharmacy costs contribute to total healthcare costs and typically are higher among more adherent patients, the increment appears to be offset by reduced expenditures for other services. While our study and others did not measure clinical outcomes, the relationship between improved adherence and lower healthcare costs is noteworthy, and the specific drivers of these cost differences warrant further research. The benefits of improved adherence to antidepressants may extend beyond depression treatment outcomes; inadequate treatment of depression may cascade into poor medication adherence for other chronic conditions, increasing healthcare costs related to these conditions. For example, depression has been linked to poor adherence with antidiabetic agents32-34; across a broad range of chronic conditions, depressed patients incurred higher non—mental health-related costs than did their counterparts without depression.35
Improving adherence to antidepressants may result in improved health outcomes and work productivity, and reduced healthcare resource utilization and costs. Consequently, interventions that facilitate adherence to antidepressants may improve patient outcomes and reduce depression-related economic burden. As cost sharing in-creases in health insurance plans, it is imperative to ensure that adherence to medications is not adversely affected.
Limitations
There are several limitations to consider. Adherence and discontinuation were measured using pharmacy claims and may not reflect whether members took antidepressants as prescribed. Generic extended-release venlafaxine became available during follow-up and may have impacted adherence among members with index VENXR; consequently, we computed a second PDC that included generic VENXR and found similar results (data not shown).
We adjusted for available member and health plan characteristics, but the inability to adjust for unmeasured factors, such as socioeconomic characteristics (eg, race, income, education) or potentially systematic differences in medical or pharmacy benefits between health plans that did and did not participate in the RSP, may have biased the results. RSP participation and co-payment reduction were recorded inconsistently on less than 1% of claims during the second and third months of the program, which may have led to cohort misclassification for a small number of members. Although pharmacy claims data are complete for all members, it is possible that mental health—related medical claims were incomplete for the 6% of members without the mental health coverage indicator. For example, 17.9% of members without the mental health indicator had ≥1 depression-related ambulatory visits during follow-up compared with 24.2% of those with the indicator (P <.001, data not shown). It is unclear whether these differences reflect variation in coverage. To address this limitation, the analysis controlled for carved-out mental health coverage. Because health plan benefit data were unavailable, we inferred enrollment in high-deductible health plans and may have misclassified members, leading to incomplete adjustment for these covariates.
The data used for this study came from a managed care population of commercially insured members with at least 15 months of continuous health plan enrollment, and as such, our results may not be generalizable to other populations such as Medicaid or Medicare beneficiaries, or the uninsured. Finally, only 9 months of RSP experience were available for this study. Longer-term evaluations are needed to determine if the benefits observed in this study are sustained over time.
CONCLUSIONS
The RSP cohort had higher index antidepressant PDC, lower likelihood of discontinuation, and lower mean healthcare costs compared with the non-RSP cohort. These results suggest that co-payment discounts dependent on time of refill may have a positive impact on adherence and reduce costs among individuals treated with antidepressants.
