Health care resource utilization and costs associated with nonadherence and nonpersistence to antidepressants in major depressive disorder

BACKGROUND: Nonadherence and nonpersistence to antidepressants in major depressive disorder (MDD) are common and associated with poor clinical and functional outcomes and increased health care resource utilization (HCRU) and costs. However, contemporary real-world evidence on the economic effect of antidepressant nonadherence and nonpersistence is limited. OBJECTIVE: To assess the effect of nonadherence and nonpersistence to antidepressants on HCRU and costs in adult patients with MDD enrolled in U.S. commercial and Medicare supplemental insurance plans. METHODS: This was a retrospective new-user cohort study using administrative claims data from the IBM MarketScan Commercial and Medicare Supplemental databases from January 1, 2010, to December 31, 2018. We identified adult patients with MDD aged ≥ 18 years who initiated antidepressant therapy for a new MDD episode between January 1, 2011, and December 31, 2017. Twelve-month total all-cause HCRU and costs (2019 U.S. dollars) were characterized for patients who were adherent/nonadherent and persistent/nonpersistent to antidepressants at 6 months. Adherence was defined as having proportion of days covered (PDC) ≥ 80%, and persistence was defined as having continuous antidepressant therapy without a ≥ 30-day gap. Multivariable negative binomial regression and 2-part models adjusted for baseline characteristics were used to estimate incidence rate ratios (IRRs) for HCRU and incremental costs of nonadherence and nonpersistence, respectively. RESULTS: A total of 224,645 patients with MDD (commercial: n = 209,422; Medicare supplemental: n = 15,223) met all study inclusion criteria. Approximately half of patients were nonadherent (commercial: 48%; Medicare supplemental: 50%) or nonpersistent (commercial: 49%; Medicare supplemental: 52%) to antidepressants at 6 months. After controlling for baseline characteristics, nonadherent patients experienced significantly more inpatient hospitalizations (commercial, adjusted IRR [95% CI]: 1.34 [1.29 to 1.39]; Medicare supplemental: 1.19 [1.12 to 1.28]) and emergency room (ER) visits (commercial, adjusted IRR [95% CI]: 1.43 [1.40 to 1.45]; Medicare supplemental: 1.28 [1.21 to 1.36]) compared with adherent patients. Similar results were observed in nonpersistent patients. Adjusted mean differences revealed that nonadherent and nonpersistent patients accumulated significantly higher medical costs (commercial: $568 [95% CI: $354 to $764] and $491 [$284 to $703]; Medicare supplemental: $1,621 [$314 to $2,774] and $1,764 [$451 to $2,925]), inpatient costs (commercial: $650 [$490 to $801] and $564 [$417 to $716]; Medicare supplemental: $1,546 [$705 to $2,308] and $1,567 [$778 to $2,331]), and ER costs (commercial: $130 [$115 to $143] and $129 [$115 to $142]; Medicare supplemental: $82 [$23 to $150] and $80 [$18 to $150]), and incurred significantly lower pharmacy costs (commercial: −$561 [−$601 to −$521] and −$576 [−$616 to −$540]; Medicare supplemental: −$510 [−$747 to −$227] and −$596 [−$830 to −$325]) compared with adherent and persistent patients, respectively. CONCLUSIONS: This study found more hospitalizations and ER use and higher total medical costs among patients who were nonadherent and nonpersistent to antidepressants at 6 months. Strategies that promote better adherence and persistence may lower HCRU and medical costs in patients with MDD.


RESULTS:
A total of 224,645 patients with MDD (commercial: n = 209,422; Medicare supplemental: n = 15,223) met all study inclusion criteria. Approximately half of patients were nonadherent (commercial: 48%; Medicare supplemental: 50%) or nonpersistent (commercial: 49%; Medicare supplemental: 52%) to antidepressants at 6 months. After What is already known about this subject • Adherence and persistence to antidepressant therapy for major depressive disorder (MDD) are poor, with only 35%-55% of patients remaining adherent or persistent to antidepressant therapy at 6 months.
• Antidepressant nonadherence and nonpersistence are associated with increased health care resource utilization (HCRU), costs, and risk of relapse and recurrence.
• Contemporary real-world evidence on the effect of nonadherence and nonpersistence to antidepressants on HCRU and costs in patients with MDD is limited.

What this study adds
• This study provides a comprehensive update on the contemporary economic effect of antidepressant nonadherence and nonpersistence in a broad U.S. adult MDD population enrolled in commercial and Medicare supplemental plans.
• Our findings underscore the need for new strategies that promote better adherence and persistence to reduce health care burden among patients with MDD.
Major depressive disorder (MDD) affects more than 17 million Americans in the United States and is a leading cause of disability worldwide. 1,2 MDD is associated with substantial clinical and economic burden; in 2010, MDD was estimated to account for $210 billion in the United States, with total costs split approximately equally between direct medical costs and indirect workplace costs (i.e., absenteeism and presenteeism costs). 3 American Psychiatric Association guidelines recommend a minimum of 6-12 weeks of initial treatment with antidepressants or other treatment modalities to induce remission of symptoms (i.e., acute phase), which is followed by 4-9 months of continuous treatment to prevent relapse (i.e., continuation phase) and maintenance treatment that may be required indefinitely to prevent recurrence for patients with 3 or more major depressive episodes, chronic depression, or risk factors for recurrence. 4 The National Committee for Quality Assurance (NCQA) continues to endorse various Antidepressant Medication Management (AMM) Healthcare Effectiveness Data and Information Set (HEDIS) measures for effective acute and continuation phase treatment with antidepressants to improve the quality of depression care. 5 Although clinical guidelines emphasize the importance of adherence and persistence to antidepressants, adherence and persistence to antidepressants are suboptimal, with only 35%-55% of patients remaining adherent or persistent to antidepressant therapy at 6 months. [6][7][8][9][10] Commonly cited reasons for nonadherence or nonpersistence to antidepressants include lack of efficacy, 11 intolerability, 11 regimen complexity, 12 out-ofpocket costs, 12,13 and patient beliefs about antidepressant medications. 14,15 Both antidepressant nonadherence and nonpersistence are significant barriers to depression treatment effectiveness and subsequent remission, 4 and failure to achieve adequate response or remission with antidepressant therapy is associated with poorer clinical and functional outcomes, 16,17 increased health care resource utilization (HCRU) and costs, [16][17][18][19][20] and work productivity loss. [16][17][18] Previous research has demonstrated that antidepressant nonadherence is associated with increased risk of relapse and recurrence, 21-23 more hospitalizations and emergency room (ER) visits, 7,8,24 and higher health care costs. 8,9,25,26 In addition, antidepressant nonpersistence, which has been assessed in previous studies as discontinuation in the first 30-180 days of antidepressant therapy, is independently associated with increased risk of relapse and recurrence 10 and higher all-cause and psychiatric HCRU and costs in patients with MDD. 7,9,[27][28][29] Nevertheless, contemporary literature on the effect of antidepressant nonadherence and nonpersistence on HCRU and costs in a broad MDD population is limited. Newer antidepressants have been introduced over the past decade (i.e., vilazodone, vortioxetine, levomilnacipran), and much of the existing literature on the association between antidepressant nonadherence 8,25,26 or nonpersistence 25,28,29 and HCRU and costs has been published before 2010. 8,25,26,28,29 Due to the high prevalence of MDD and burden associated with poor treatment outcomes in MDD, understanding the economic effect of antidepressant nonadherence and nonpersistence is of continued interest to health systems and managed care organizations. Additionally, there are inconsistent definitions and thresholds of antidepressant adherence and persistence used in the existing literature, 30,31 and studies using standardized objective measures of adherence and persistence are needed. 30 Furthermore, much of the previous literature has focused on privately insured commercial populations, 7,8,[25][26][27][28][29] and MDD is prevalent in other insured populations, including the Medicare population where MDD affects approximately 15% of beneficiaries aged 65 years and older. 32 Therefore, the objective of this study was to provide a contemporary update on the effect of antidepressant nonadherence and nonpersistence on HCRU and costs in adult patients with MDD enrolled in commercial and Medicare supplemental insurance plans. CONCLUSIONS: This study found more hospitalizations and ER use and higher total medical costs among patients who were nonadherent and nonpersistent to antidepressants at 6 months. Strategies that promote better adherence and persistence may lower HCRU and medical costs in patients with MDD.
This study did not meet the definition of human subjects research and therefore did not require University of Washington Institutional Review Board approval, as the IBM MarketScan databases are deidentified and compliant with the Health Insurance Portability and Accountability Act of 1996.

STUDY POPULATION
Inclusion Criteria. The study population consisted of adult patients with MDD who were identified between January 1, 2011, and December 31, 2017, and initiated an antidepressant for the treatment of a new MDD episode. Patients with MDD were defined as having either of the following: ≥ 1 inpatient medical claim with a primary International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9-CM/ICD-10-CM) diagnosis code for MDD or ≥ 1 inpatient or outpatient medical claim with a secondary ICD-9-CM or ICD-10-CM diagnosis code for MDD, followed by a second confirmatory inpatient or outpatient claim with an ICD-9-CM or ICD-10-CM diagnosis code for MDD in any position within 6 months after the initial medical claim (Supplementary Table 1, available in online article). The date of the first qualifying MDD diagnosis between January 1, 2011, and December 31, 2017, was defined as the initial MDD diagnosis date. Antidepressants for MDD included selective serotonin reuptake inhibitors, selective norepinephrine reuptake inhibitors, monoamine oxidase inhibitors, tricyclic

STUDY DESIGN AND DATA SOURCE
This was a retrospective new-user cohort study using administrative claims data from the IBM MarketScan Commercial and Medicare Supplemental databases from January 1, 2010, to December 31, 2018 ( Figure 1). The IBM MarketScan Commercial database consists of employer and health plan medical and pharmacy claims data for more than 40 million commercial plan members annually. 33 Enrollees include employees, their spouses, and their dependents who are covered by employer-sponsored private health insurance. 33 A variety of fee-for-service, fully capitated, and partially capitated health plans are represented in the IBM MarketScan Commercial database, including point-of-service, exclusive provider organizations, preferred provider organization plans, indemnity plans, health maintenance organizations, consumer-driven health plans, and highdeductible health plans. 33 The IBM MarketScan Medicare Supplemental database contains data on the Medicarecovered portion of payment (i.e., coordination of benefits amount) and employer-and out-of-pocket patient expenses for health care services provided in inpatient and outpatient settings for U.S. Medicare beneficiaries. 33 Medical claims are linked to outpatient prescription drug claims and person-level enrollment information. 33 HCRU = health care resource utilization; MDD = major depressive disorder.  43 and reflected all payments made to providers by insurers (i.e., plan and coordination of benefits) and patients (i.e., copay, co-insurance, deductible). Patients with evidence of capitated claims or negative total costs in any cost subcategory were excluded from cost analyses. Table 2, available in online article). To ensure antidepressant claims were for the treatment of MDD, the index antidepressant date was defined as the date of the first antidepressant pharmacy claim within 60 days after a qualifying MDD diagnosis. 6 Patients were required to be aged ≥ 18 years at index antidepressant date and have continuous enrollment in medical and pharmacy benefits for 12 months before the initial MDD diagnosis date (baseline period) and 12 months after the index antidepressant date (post-index period).

antidepressants, and other antidepressants (Supplementary
Exclusion Criteria. To avoid potential bias from other neuropsychiatric conditions, patients were excluded if they had diagnoses for bipolar/manic disorder, mood disorders other than MDD, Alzheimer disease, Parkinson disease, or dementia during the study period. Additionally, patients with diagnoses of schizophrenic disorder, psychosis-related disorders, drug-induced depression, or depressive-type psychosis in the 12-month baseline period before the initial MDD diagnosis date were excluded. Patients were also excluded if they had evidence of pregnancy, childbirth, or breastfeeding at any time during the study period, as these conditions may affect antidepressant adherence and persistence. To ensure antidepressants were prescribed for a new MDD episode, additional exclusions included any MDD claims or claims for antidepressants or combination antidepressant/antipsychotic agents (i.e., amitriptyline/ perphenazine, fluoxetine/olanzapine) in the 12-month baseline period and evidence of combination antidepressant therapy or augmentation therapy on or within 30 days after the index antidepressant date. Combination antidepressant therapy was defined as 1 of the following: initiating ≥ 2 different antidepressants on the index antidepressant date, initiating an additional antidepressant with a different active ingredient within 30 days after the index antidepressant date 34 that overlaps with the index antidepressant for ≥ 30 days during the first 60 days of the index date (overlap did not need to be consecutive), or having more than 3 antidepressants within the first 30 days after the index date. 34 Augmentation therapy was defined as having any claims for an antipsychotic, mood stabilizer, or combination antidepressant/antipsychotic agent on or within 30 days after the index antidepressant date (Supplementary Table 3, available in online article). 34

STUDY MEASURES AND OUTCOMES
Baseline Demographic and Clinical Characteristics. Baseline patient demographic and clinical characteristics were assessed during the 12-month baseline period before the initial MDD diagnosis date. Baseline demographic characteristics included age, sex, plan type, geographic region, pulmonary disease (Table 1). In both populations, the most commonly prescribed index antidepressants were selective serotonin reuptake inhibitors (commercial: 73%; Medicare supplemental: 74%), followed by other antidepressant classes (commercial: 16%; Medicare supplemental: 15%). At 6 months, 48% and 49% of patients in the commercial population and 50% and 52% of patients in the Medicare supplemental population were nonadherent and nonpersistent to antidepressants, respectively.

ALL-CAUSE HEALTH CARE COSTS
Nonadherent and nonpersistent patients in the commercial and Medicare supplemental populations incurred significantly higher adjusted mean medical, inpatient, and ER costs and significantly lower adjusted mean pharmacy costs (

STATISTICAL ANALYSIS
Descriptive statistics were used to describe demographic and clinical characteristics during the 12-month baseline period and to calculate unadjusted 12-month post-index all-cause HCRU and costs. Unadjusted differences in HCRU and costs between nonadherent/adherent and nonpersistent/persistent patients were assessed using chi-square tests for categorical variables and unpaired t-tests with unequal variances for continuous variables. Additionally, multivariable negative binomial models were used to calculate adjusted incidence rate ratios (IRRs) to assess the association between nonadherence or nonpersistence to antidepressants and 12-month post-index HCRU. Adjusted incremental mean differences in health care costs associated with nonadherence and nonpersistence were assessed using multivariable 2-part models (logistic regression to determine the probability of having non-zero costs, followed by generalized linear models using a log-link function and gamma distribution). A nonparametric bootstrapping procedure with 500 iterations was used to estimate 95% CIs for the mean incremental cost differences, which were obtained using the 2.5th and 97.5th percentiles of the bootstrapped distribution. Multivariable HCRU and cost models were adjusted for the following variables: age, sex, geographic region, plan type, index year, Quan-CCI, index antidepressant class, and 12-month baseline total all-cause health care costs. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). P values ≤ 0.05 were considered statistically significant.

BASELINE CHARACTERISTICS
A total of 5,745,208 adult patients with MDD were identified in IBM MarketScan Commercial and Medicare Supplemental databases between January 1, 2011, and December 31, 2017 ( Figure 2). Of these, 224,645 adult patients with MDD (commercial: n = 209,422; Medicare supplemental: n = 15,223) met all study inclusion criteria. The mean (SD) ages for the commercial and Medicare supplemental populations were 40 (14) and 75 (8) years, and the majority of patients were female (commercial: 61%; Medicare supplemental: 62%; Table 1). Most patients were enrolled in preferred provider organization plans in the commercial population (59%), whereas most patients were enrolled in comprehensive plans in the Medicare supplemental population (44%). Compared with adherent and persistent patients, nonadherent and nonpersistent patients had a higher prevalence of anxiety, alcohol-use disorder, substance-use disorder, and chronic a MDD ICD-9-CM codes: 296.20-296. 24 44 which was similarly reflected in our study; Medicare supplemental patients with MDD in our study had a markedly higher prevalence of comorbid coronary heart disease, congestive heart failure, chronic pulmonary disease, and diabetes. There is also evidence that older patients with MDD have a poorer depression prognosis compared with younger patients with MDD. 44 These factors may therefore contribute to higher health care spending compared with younger, healthier patients with depression. Total health care costs were not significantly different between nonadherent/adherent and nonpersistent/ persistent patients in this study, as medical cost savings due to reductions in HCRU among adherent and persistent patients may not sufficiently offset the higher pharmacy costs due to improved antidepressant adherence and persistence in the short term. It is plausible that the potential benefits of improved antidepressant adherence and persistence may not be fully realized in the short term, as various Our findings support and extend previous work on the association between nonadherence and nonpersistence to antidepressants and all-cause HCRU and costs in patients with MDD. Similar to previous studies, our study found greater all-cause inpatient hospitalizations and ER use among nonadherent and nonpersistent patients compared with adherent and persistent patients, respectively. 7,8,27 Also consistent with previous research, 8,9,25,26,28 nonadherence and nonpersistence were associated with significantly higher adjusted medical costs in both the commercial and Medicare supplemental populations, which were largely driven by approximately 20%-40% higher inpatient hospitalization and ER costs among nonadherent and nonpersistent patients. Given the association between nonremission and poor depression treatment outcomes, the higher medical costs and rates of hospitalizations and ER visits among antidepressant nonadherent and nonpersistent patients may, in part, reflect higher rates of nonremission among these patients.
Of note, Medicare supplemental patients who were nonadherent and nonpersistent to antidepressants incurred  other comorbid conditions may be associated with longerterm cost savings.
Overall, the high rates of antidepressant nonadherence and nonpersistence in our study, which were associated with greater all-cause hospitalizations and ER use, suggest inadequate depression treatment for a large proportion of patients with MDD and, therefore, highlight a continued need for effective strategies to improve antidepressant adherence and persistence, which may improve outcomes and reduce the health care burden associated with MDD. Interventions to improve antidepressant adherence and persistence may have important implications for health systems and managed care organizations, given the potential cost savings attributable to improved depression treatment outcomes, 16,17,19 as well as continued efforts to incentivize real-world observational studies have found a significantly lower risk of relapse or recurrence (e.g., evidence of mental health-related ER visit, psychiatric hospitalization, suicide attempt, evidence of electroconvulsive therapy (ECT), or re-initiation of MDD pharmacotherapy after a gap of 6 months) over follow-up periods of up to 2 years among patients who were adherent to antidepressants during the first 6 months of treatment compared with those who were nonadherent to antidepressants. [21][22][23] Furthermore, adherence and persistence to antidepressant therapy for MDD are associated with better medication adherence and improved clinical outcomes for comorbid conditions, including diabetes, chronic obstructive pulmonary disease, and cardiovascular disease. 9,24,26,45 Similarly, the effect of improved antidepressant adherence and persistence on Fourth, although various baseline characteristics were adjusted for in multivariable analyses, additional factors that may contribute to adherence, persistence, HCRU, and health care costs in MDD, such as depression severity, 13,62 prescriber specialty, 63,64 specific treatment patterns (e.g., combination antidepressant therapy, augmentation therapy, switching, dose escalation), 28,29,34,[65][66][67] or other clinical characteristics, were not assessed or adequately captured and, therefore, could not be controlled for. For example, patients who require treatment switches, dose escalation, combination antidepressant therapy, or augmentation therapy may represent more complex, difficult-to-treat patients, regardless of adherence or persistence.
Finally, the study population was limited to adult patients with MDD who initiated antidepressant therapy for a new MDD episode and were enrolled in commercial and Medicare supplemental insurance plans. Therefore, these findings may not be generalizable to patients with MDD who have different characteristics from our study population, including patients with treatment-resistant depression, patients with insurance coverage outside of commercial or Medicare supplemental insurance plans, and uninsured individuals.
Despite these limitations, there are several strengths of this study. First, this study provides a comprehensive update to the existing literature on the economic effect of antidepressant nonadherence and nonpersistence in MDD; to our knowledge, this study is the largest real-world observational study to assess the association between antidepressant nonadherence and nonpersistence and HCRU and costs in adult MDD patients to date, with more than 200,000 adult patients with MDD included in this study. Our patient population was representative of patients with MDD from diverse geographic regions across the United States and enrolled in various insurance plans (i.e., commercial and Medicare supplemental insurance plans) and plan types (i.e., capitated and noncapitated plans). Additionally, our study includes newer antidepressants that have been introduced since 2010, so this study serves to provide a contemporary update on the current landscape of MDD medication adherence and persistence and their effect on HCRU and costs in patients with MDD.

Conclusions
The results of our study suggest that greater hospitalization and ER use and total medical costs are incurred among patients who are nonadherent and nonpersistent to antidepressant therapy compared with those who are adherent and persistent, which was observed in both the commercial and Medicare supplemental populations. Higher medical quality depression care in health care organizations. [46][47][48] Such efforts are evidenced by the continued adoption of NCQA HEDIS AMM measures for effective acute and continuation phase antidepressant treatment in provider pay-for-performance initiatives and merit-based incentive programs, 46 as well as the emergence of the NCQA HEDIS Depression Remission or Response outcomes measures, which evaluate the percentage of health plan members achieving treatment response and remission, 49 in various value-based payment arrangements. 47,50 Previous research has highlighted the clinical and economic benefits of interventions aimed at improving the quality of depression care in diverse practice settings, including collaborative care programs [51][52][53][54][55] and psychosocial interventions, 56,57 which have been found to improve antidepressant adherence and longer-term depression outcomes. 53,54 In the context of continued industry efforts to improve the quality of depression care, these findings therefore underscore the contemporary importance of strategies that promote better antidepressant adherence and persistence to improve depression care, and various measures, such as pay-for-performance initiatives, expansion of mental health benefits, and coverage of disease management and collaborative care programs, have been suggested as effective strategies to improve antidepressant adherence/ persistence and depression treatment. 13,48,55,58,59 LIMITATIONS Several limitations of this study should be noted. First, the calculation of adherence and persistence to antidepressants was based on pharmacy claims data, which do not confirm whether patients consumed their medication. The PDC and persistence calculations used in this study also did not account for the effect of hospitalization on adherence and persistence estimates. 60 Second, as with other retrospective claims database studies, this study is subject to limitations inherent to administrative claims databases, including data coding errors, omissions, and misclassification.
Third, this study did not consider indirect costs such as absenteeism or presenteeism costs, which have been found to be substantial in MDD and estimated to account for up to 50% of the economic burden of MDD in the United States. 3 Burton et al. (2007) found that employees who were nonadherent to acute and continuation phase antidepressant therapy were 39%-46% more likely to have short-term disability claims compared with adherent employees. 61 Therefore, the costs associated with nonadherence and nonpersistence to antidepressants are likely underestimated in this study. costs were largely driven by significantly higher inpatient hospitalization and ER costs among nonadherent and nonpersistent patients. Although clinical guidelines emphasize the importance of adherence and persistence to achieve remission and prevent depressive relapse and recurrence, adherence and persistence to existing MDD treatments remain suboptimal. The low rates of antidepressant adherence and persistence observed in this study reflect a continued unmet need for effective strategies that promote better adherence and persistence, which may increase the likelihood of achieving long-term recovery from depression and reduce the health care burden associated with MDD.

DISCLOSURES
This study was sponsored by Allergan, which was involved in the study design; data collection, analysis, and interpretation of data; and decision to present these results. Ta was supported by a training grant provided to the University of Washington by Allergan at the time this study was conducted. Tung and Gillard are employees of Allergan. Oliveri is an employee of Genesis Research. Sullivan and Devine have no financial disclosures.
This study was presented as a poster at AMCP 2020 (Virtual Meeting), April 21-24, 2020.