Aylin Altan, PhD

: Heart failure (HF) is a common chronic condition associated with substantial healthcare cost burden. This study aimed to characterize patients with HF, compare costs by receipt and timing of HF-related outpatient pharmacotherapy (HFRx), and determine associations between patient and clinical characteristics and costs.

: This was a retrospective administrative database study of adults in commercial and Medicare Advantage with Part D (MAPD) plans.

: Included patients had ≥2 non-inpatient or ≥1 inpatient claim(s) containing HF diagnosis codes during 2010-2011; the date of the earliest claim was the index date. Costs (medical [inpatient hospitalizations, ambulatory, emergency, other] and pharmacy) were calculated up to 24 months post index. Patient characteristics and HFRx timing were examined descriptively and assessed by multivariable analysis for associations with costs.

: A total of 117,911 patients were identified (mean age 71.3 years, 48.4% male, 75.0% MAPD enrollees); 28.7% had no evidence of HFRx within 60 days post index. Patients receiving HFRx only after 60 days had the highest Year 1 costs ($58,771 [SD $107,638]) with inpatient hospitalization contributing more than half of that total (55.0%). This trend continued into Year 2, although costs were lower. Having no HFRx within 60 days post index, higher comorbidity burden, and male gender were associated with higher costs in multivariable analysis.

: Patients who did not receive HFRx within 60 days of a HF-related encounter had higher healthcare costs. This trend remained after adjustment in multivariable analysis, suggesting an opportunity to reduce costs by better optimizing HF disease management.

Heart failure (HF) is a common disease with a debilitating course, poor prognosis, and high mortality. In the United States, HF affects approximately 5.7 million adults and its prevalence is expected to exceed 8 million by 2030.

 It is also the leading cause of hospitalization among Medicare enrollees.

 Approximately 20% of patients with HF die within 12 months of onset and 50% die within 5 years.

As prevalence of HF rises in the United States, annual direct costs are expected to exceed $69 billion by 2030.

 The main drivers of high costs are inpatient hospitalizations and re-hospitalizations,

 thus preventive care efforts to mitigate such expensive resource use hold importance.

Optimizing HF treatment with guideline-directed HF-related pharmacotherapy (HFRx),

 such as angiotensin-converting-enzyme inhibitors (ACE-Is), angiotensin-receptor blockers (ARBs), aldosterone receptor antagonists (AAs), and beta blockers (BBs), has been found to reduce HF disease burden by slowing the progression of disease, thus decreasing hospitalizations, costs, and mortality.

 Nevertheless, such HFRx remains relatively underutilized,

 suggesting many patients may not be optimally managed following a HF event.

Less well understood in HF patients is the relationship between HF treatment delay and other patient characteristics on their healthcare cost burden. This information would be useful to healthcare providers and payers and may have disease management and cost implications. This study aimed to a) describe a real-world sample of patients with HF by receipt of HFRx within 60 days of an index encounter; b) compare healthcare costs based on receipt of HFRx within varying post index time periods; and c) examine associations between HFRx timing and healthcare costs.

This is a retrospective cohort study of administrative claims data from the Optum™ Research Database (ORD) from January 1, 2009, to December 31, 2013. The ORD includes approximately 14 million enrollees in commercial plans and 500,000 enrollees in Medicare Advantage with Part D (MAPD) plans, and it provides a geographically diverse sample representative of the US population. Medical and pharmacy claims (including diagnosis, procedure, and facility codes and costs) and enrollment data were obtained from this de-identified database and merged with race/ethnicity information. Social Security Administration dates of death were obtained to censor data collection for patients who died prior to the end of the study period. The study was conducted in compliance with the Health Insurance Portability and Accountability Act.

Patients aged ≥18 years were included if a diagnosis code for HF (International Classification of Diseases, Ninth Edition, Clinical Modification [ICD-9-CM] codes 402.x1, 404.x1, 404.x3, 428.xx)

 was observed in any position for ≥1 inpatient stay or ≥2 non-inpatient encounters within the study identification period of January 1, 2010, through December 31, 2011 (Figure 1). The index date was assigned as date of the first qualifying claim. Continuous plan enrollment was required for 12 months prior to the index date and at least 1 month (30 days) from index until the earliest of end of study period, health plan disenrollment, date of death if before end of the study period, or 729 days from index date. Patients were excluded if: gender, geographic region, or health plan type were missing; age at index was ≥65 years among commercial enrollees; or any claim was dated >45 days following date of death. Outcomes were assessed for 4 non-mutually exclusive post index time periods including Months 1-2 (M1-2, ie days 1-60), Months 3-12 (M3-12, ie days 61-365), Year 1 (days 1-365), and Year 2 (days 366-729).

Patient characteristics examined for the pre-index period included age, gender, health plan type (commercial or MAPD), US geographic Census region,

 and race/ethnicity. Length of the post index period was also recorded. Clinical characteristics included pre-index HF, Quan-Charlson21 comorbidity score (grouped as 0, 1-2, 3-4, ≥5), comorbid atrial fibrillation (AF), comorbid diabetes, and receipt of HFRx.

Receipt of HFRx was identified from outpatient pharmacy claims within the 4 post index time periods: M1-2, M3-12, Year 1, and Year 2. The HFRx identified included ACE-Is, ARBs, AAs, BBs, and HFRx classified as Other (eg, diuretics, hydralazine + isosorbide dinitrate, digoxin) (eAppendix Table; eAppendices available at www.ajpb.com). Treatment patterns for receipt of HFRx were described as mono-, dual, or triple therapy, or no HFRx (eAppendix Table). Annual and per-patient-per-month (PPPM) costs were calculated by combining plan-paid and patient-paid amounts and categorized as total (medical plus pharmacy), medical (ambulatory [office and outpatient], emergency department, inpatient, and other), and pharmacy. All costs were adjusted to 2013 US dollars using the Medical Care Component of the Consumer Price Index.

Individuals in the study sample were stratified into 1 of 4 groups based on timing of treatment in the first year (Year 1) post index (eAppendix Figure). These were the “Treated” group, defined as patients receiving HFRx in both M1-2 and M3-12; the “Interrupted Treatment” group, defined as patients receiving HFRx in M1-2 but not in M3-12; the “Delayed Treatment” group, defined as patients not receiving HFRx in M1-2 but receiving HFRx in M3-12; and the “Not Treated” group, defined as patients not receiving HFRx in either M1-2 or M3-12. Annual costs for Year 2 were anchored on presence and timing of HFRx treatment in Year 1.

All study variables were described as means (SDs) or percentages. Pearson’s χ2 test was used for dichotomous and polychotomous variables. Independent samples t test and 1-way analysis of variance were used for continuous variables with 2 cohorts and >2 cohorts, respectively. A significance level of α = .05 was applied.

Multivariable generalized linear models (GLMs) with gamma distribution and log link were used to examine associations among select patient, clinical, and insurance characteristics and post index healthcare costs.23 PPPM cost was selected as the dependent variable in the regression models because of varying follow-up lengths of individuals in the study sample. Independent variables included M1-2 HFRx, age, gender, health plan type, race/ethnicity, geographic region, pre-index HF, Quan-Charlson comorbidity score, AF, and diabetes.

A total of 117,911 individuals meeting sample selection criteria were included in the final sample (Figure 1). Of these, 81,853 had complete 1-year follow-up while 61,035 had 2-year follow-up (eAppendix Figure). All results are presented for M1-2 and M3-12 or by treatment pattern during Year 1 of follow-up: Treated, Interrupted Treatment, Delayed Treatment, and Not Treated.

Demographic and Clinical Characteristics of Overall Sample and Treatment Groups

The study sample had a mean (SD) patient age of 71.3 (13.2) years, and was 51.6% female, 72.8% white, and 75.50% MAPD-enrolled (Table 1). Approximately one-third (32.8%) of individuals had HF diagnosis in the 12-month pre-index period, while 26.4% and 41.7% had evidence of AF and diabetes, respectively.

The distribution of individuals across the 4 treatment groups were as follows: 51.2% Treated, 1.2% Interrupted Treatment, 7.1% Delayed Treatment, and 9.9% Not Treated. Of the 4 treatment groups, the Treated group had the highest proportion of females (53.1%), African Americans (13.8%), patients with pre-index HF (34.7%), and patients with comorbid diabetes (43.3%). The Not Treated group had the highest proportion of MAPD enrollees (77.8%) and patients residing in the West region (10.9%). The Interrupted group had the lowest proportion of patients with AF (14.3%).

Nearly three-fourths (71.3%) of patients in the overall sample received HFRx during M1-2 (Table 2). Dual-therapy ACE-I/ARB plus BB (24.0%) was most common, followed by monotherapy BB (18.0%) and monotherapy ACE-I/ARB (12.7%). There was no evidence of prescription fills for any HFRx during M1-2 in 28.7% of patients.

Among patients followed through Year 1 (n = 81,853) who received any HFRx during M1-2 (71.3% of total), nearly all of them (97.6%) continued treatment the rest of the year, with 43.6% receiving dual therapy ACE-I/ARB plus BB, and 19.2% and 14.3% receiving monotherapy BB and monotherapy ACE-I/ARB, respectively. In contrast, of those who had no evidence of treatment in M1-2 (28.7% of total), about three-fifths (58.0%) of them had no evidence of prescription fills the remainder of Year 1.

Similarly, for patients whose post index period was 2 years (n = 61,035), nearly all (94.9%) who had evidence of Year 1 HFRx continued on HFRx in Year 2. Approximately two-fifths (40.9%) received dual therapy ACE-I/ARB plus BB, and 19.5% and 14.5% received monotherapy BB and monotherapy ACE-I/ARB, respectively. Of every 100 patients with follow-up for 1 year and not receiving treatment in M1-2, 58 did not receive treatment through M3-12 of Year 1, and of every 100 patients with follow-up for 2 years and not receiving treatment Year 1, 86 still did not receive treatment through the second year of follow-up.

Impact of Heart Failure-Related Pharmacotherapy Within the First 60 Days (M1-2) on PPPM Costs

The multivariable models examined the impact of the presence of HFRx fills within M1-2 of Year 1 of follow-up. Results are presented in Table 3. As compared with the presence of M1-2 monotherapy ACE-I/ARB (reference treatment), no evidence of M1-2 HFRx was associated with significantly higher all-cause (2.44; P <.001) and HF-related (3.66; P <.001) PPPM costs. Male gender, Quan-Charlson scores (≥3), and pre-index diabetes were each associated with higher all-cause and HF-related costs (all P <.001). By contrast, pre-index HF was associated with lower costs (Table 3).

Treatment Timing and Year 1 and Year 2 Annual Healthcare Costs

The multivariable findings point to the influence of HFRx within the period immediately following the index HF event; however, characterization of the specific treatment timing patterns remained of interest. Further descriptive analyses were conducted examining annual healthcare costs in Years 1 and 2, based upon timing of treatment in Year 1 (Table 4).

The Year 1 all-cause cost (mean [SD]) was $41,294 ($70,184), half of which was contributed by hospitalizations ($20,706 [$51,395]). Pharmacy costs accounted for 10.3% of total cost ($4270 [$7503]). Patients in the Delayed Treatment group had the highest costs ($58,772 [$107,638]), while those in the Interrupted Treatment group had slightly higher all-cause costs ($43,646 [$70,830]) than their Treated group counterparts ($40,289 [$62,323]). The Not Treated group had the lowest annual ($33,563 [$72,167]) and pharmacy ($1600 [$5891]) costs.

!--page-->The mean annual HF-related medical costs were $18,748 ($46,277), accounting for 45.4% of the all-cause total, with HF-related hospitalizations contributing the largest portion to those costs (83.8%) (Table 4). Similar to the pattern for all-cause costs, the Delayed Treatment patients had the highest HF-related medical costs ($28,048 [$77,007]), the Interrupted Treatment group had higher medical costs ($19,959 [$39,595] than the Treated group ($18,107 [39,944]), and those in the Not Treated group had the lowest costs ($15,185 [$47,161]).

Costs (all-cause) in Year 2 were markedly lower than in Year 1 ($24,397 [$47,951]) (Table 4). Although the hospitalization costs decreased substantially, they still contributed one-third of the total all-cause costs in Year 2. All-cause pharmacy costs were comparable in both years. Cost patterns in the second year tracked similarly to the pattern observed in Year 1: Delayed Treatment patients had the highest costs ($27,906 [$51,940]), followed by those in the Treated ($25,231 [$48,210]) and Interrupted Treatment ($19,812 [$40,910]) groups. Those in the Not Treated group had the lowest costs ($18,175 [$43,750]).

The HF-related costs also decreased considerably in Year 2 ($6554 [$26,958]); the bulk of the costs (73.8%) remained attributable to inpatient hospitalizations (Table 4). The cost patterns differed slightly in Year 2 with the Treated group exhibiting costs similar to Delayed Treatment group (Treated: $6986 [$28,127]; Delayed: $6803[$22,958]). The Not Treated group experienced lower costs than their Treated and Delayed Treatment counterparts, but had higher costs than those in the Interrupted Treatment group (Not Treated: $4523 [$24,273]; Interrupted: $3509 [$14,452]).

HF is a progressively debilitating disease that carries a high risk of morbidity4 and mortality.

 Optimizing HF-related outpatient pharmacotherapy has been associated with better outcomes

; however, the impact of timing of such treatment on disease burden is less understood. The current study adds to the literature by describing patients with HF, including type and timing of HF-related outpatient pharmacotherapy, and associations of these with healthcare costs.

 total medical costs for HF patients were high (Year 1 all-cause costs: $41,295; Year 2: $24,397) with a large proportion attributable to hospitalizations. The HF-related medical costs were impacted to an even larger degree due to hospitalizations. Comparison of HF-related costs with other cardiovascular encounters/events was hampered because the specific nature of the encounter could not be directly aligned with event definitions used in other research studies. Nevertheless, 1 recent study examining patients hospitalized for transient ischemic attack or ischemic stroke found costs of $26,201-$30,349 through 1 year, varying by pharmacotherapies prescribed at discharge.

 reported mean 2-year costs of $51,058 among patients who had a myocardial infarction and were taking hyperlipidemia medications for primary prevention. Specific examination of events and comorbid conditions may elucidate what exactly is driving high costs.

The apparent influence of not having HFRx within the first 60 days (M1-2) on PPPM cost was notable (>2-fold increased risk of higher costs). The association of male gender and higher comorbidity burden to higher costs contrasted with older age and pre-index HF diagnoses, which were associated with lower costs. It is possible that longer history of disease may be accompanied by a more established treatment course resulting in lower costs. These findings suggest timing of guideline-informed treatment plays an important role in costs and could inform disease management approaches, particularly regarding acute events.

Further characterization of the timing of treatment patterns allowed a more detailed view of cost differences within the study sample. All-cause and HF-related costs in Year 1 were highest among those not having HFRx within the first 60 days but who did have subsequent HFRx (Delayed Treatment group). This may point to a critical window during which timely treatment may exert an impact on disease progression and resulting costs. During Year 2, the pattern of all-cause costs remained similar to Year 1. However, for HF-related medical costs in Year 2, the pattern of higher costs among those with delayed treatment reflected a closer symmetry to their treated counterparts. Those treated had slightly higher costs (<$200) than those with delayed treatment. This similarity may be attributable to the amount of time in treatment through Year 1 (at least 9 months for each group), possibly indicating a longer-term effect of total time in treatment in addition to timing of treatment initiation. Furthermore, patients who were not treated at all in Year 1 experienced lower costs overall but demonstrated higher HF-related medical costs in Year 2 than those with interrupted treatment. This may indicate that HFRx treatment timing may have been predicated on clinical severity or ejection fraction (EF) not measured in this study. The Year 2 findings seem to suggest that patients who were treated or had delayed treatment in Year 1 may be more similar to each other and may be distinct from those who experienced interrupted or no treatment in Year 1. Notably, the subset of study patients with no evidence of HF-related pharmacotherapy within 60 days post index (M1-2) but who did receive HFRx during the remainder of the Year 1 (Delayed Treatment) exhibited the worst cost outcomes across the 4 treatment timing groups.

Interestingly, more than a quarter (28.7%) of patients had no claims for HFRx within 60 days post index. Similarly, Deschaseaux et al found more than one-third of patients did not receive HFRx within 30 days of the initial diagnosis.

 These percentages appear little improved since 2005, when Lee et al demonstrated a ACE-I or ARB prescription rate of only 67% within 90 days of discharge among patients with HF at high risk for death; low prescribing rates continued up to 1 year.

 Similar claims-based studies found that 30% of patients did not receive ACE-I/ARB during 1 year following an index HF encounter

 and at least 46% of patients received no form of HFRx post discharge.

 The current study showed about half of patients without treatment for HF within 60 days after index HF diagnosis remained untreated over 2 years; this may highlight an opportunity to address an unmet need among patients with HF.

Because this study utilized administrative claims data, measures were subject to possible inaccuracy; however, error rates are not expected to vary by comparison groups. The evidence of a condition was established by ICD-9-CM codes, which reflect the claims submitted for reimbursement, rather than a confirmed diagnosis. Also, this study did not capture New York Heart Association Functional Classification, EF, laboratory results, blood pressure, smoking status, or body mass index, all of which might influence a physician’s decisions regarding when and what to prescribe. In particular, the lack of EF prevented the distinction of HF with reduced EF (HFrEF) versus HF with preserved EF (HFpEF). Recommended therapy for patients with HFpEF focuses on controlling systolic and diastolic blood pressure, and expert opinion is diverging regarding treatment with BBs, ACE-Is, and ARBs. Given also that HFpEF represents approximately half of patients with HF,

 it is entirely possible that the patients in this study who had no claims for HFRx within 2 months post index were largely patients with HFpEF. As such, these patients would not be considered undertreated according to guidelines, and treatment with an ACE-I, ARB, or BB may not have resulted in lower costs. Separately, specific indicators of severity are not available among claims data; the absence of this information potentially confounds the analysis. Furthermore, the claims data do not capture medication fills not processed through the primary insurer. In addition, it was not possible to account for patient barriers to filling prescriptions, or for nonadherence. Claims data may underestimate prescription fills and associated costs for generic medications commonly used to treat HF (eg, carvedilol, metoprolol, lisinopril). Finally, the data came from a managed care population; therefore, results may not be applicable to patients in non-managed care settings. However, the population largely affected by HF is >65 years of age, and the majority of patients in this study were in a MAPD healthcare plan, thus generalization to a large proportion of Medicare-insured individuals in the US is reasonable.

These analyses revealed that receipt of guideline-directed HFRx soon after an HF-related encounter influenced healthcare costs through 2 post index years. Costs were primarily driven by hospitalizations, and were higher for patients without evidence of HFRx within 60 days, which points to the importance of early intervention following an HF event/encounter. This study also revealed evidence of suboptimal prescribing of HFRx within the 60 days following the HF event. These findings together suggest an opportunity for providers to address an unmet need in terms of pharmacotherapy and for payers to influence costs for patients with HF.

The authors thank Chun-Lan Chang, PhD, for contributions to data interpretation and manuscript development. Medical writing support was provided by Caroline Jennermann, MS, of Optum, funded by Novartis Pharmaceuticals Corporation (NPC).

: Concept and design (SJT, JPS, ENO, PAR, AA); acquisition of data (SJT, JPS); analysis and interpretation of data (SJT, JPS, ENO, PAR, AA); drafting of the manuscript (SJT, JPS, ENO, PAR); critical revision of the manuscript for important intellectual content (SJT, JPS, ENO, PAR, AA); statistical analysis (JPS, PAR).

: Health Economics & Outcomes Research, Novartis Pharmaceuticals Corporation (SJT, PAR), East Hanover, NJ; Health Economics & Outcomes Research, Optum, Inc (JPS, AA), Eden Prairie, MN; Institute for Health Outcomes, Policy & Economics, Rutgers University (ENO), Piscataway, NJ.

 Jason P. Swindle, PhD, MPH, Optum, Inc, 11000 Optum Circle, Eden Prairie, MN 55344; Tel: 312-348-6337; E-mail: 

1. Mozaffarian D, Benjamin EJ, Go AS, et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics--2015 update: a report from the American Heart Association [published correction appears in Circulation. 2016;133(8):e417]. Circulation. 2015;131(4):e29-e322. doi: 10.1161/CIR.0000000000000152. 

2. Go AS, Mozaffarian D, Roger VL, et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Executive summary: heart disease and stroke statistics--2014 update: a report from the American Heart Association. Circulation. 2014;129(3):399-410. doi: 10.1161/01.cir.0000442015.53336.12. 

3. Heidenreich PA, Albert NM, Allen LA, et al; American Heart Association Advocacy Coordinating Committee; Council on Arteriosclerosis, Thrombosis and Vascular Biology; Council on Cardiovascular Radiology and Intervention; Council on Clinical Cardiology; Council on Epidemiology and Prevention; Stroke Council. Forecasting the impact of heart failure in the United States: a policy statement from the American Heart Association. Circ Heart Fail. 2013;6(3):606-619. doi: 10.1161/HHF.0b013e318291329a.

4. Roger VL, Go AS, Lloyd-Jones DM, et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics--2011 update: a report from the American Heart Association. Circulation. 2011;123(4):e18-e209. doi: 10.1161/CIR.0b013e3182009701. 

Articles

The findings of a retrospective study suggest that timing of heart failure (HF)-related pharmacotherapy following a HF-related encounter has potential implications on subsequent healthcare costs.

Pharmacotherapy Timing in Heart Failure and Impact on Healthcare Costs

: To compare adherence to and persistence with chronic obstructive pulmonary disease (COPD) maintenance medications between Medicare Advantage Prescription Drug (MAPD) plan enrollees and Medicare prescription drug (PD)-only plan enrollees.

: This was a retrospective, cross-sectional study using Medicare health plan eligibility and pharmacy claim information collected from the Optum Research Database. The study period was January 1, 2011, to June 30, 2013.

: Descriptive and multivariable analyses of adherence and persistence were performed. Adherence was quantified as the proportion of days covered (PDC) by any COPD maintenance medication, and persistence was appraised as the duration of therapy.

: The study population included 130,305 patients, with 21% and 79% enrolled in MAPD and PD-only plans, respectively. PD-only plan enrollees were significantly more adherent (PDC ≥80%) than MAPD plan enrollees (40.5% vs 38.7%; P <.001), based on descriptive and multivariable analysis (odds ratio, 1.09; 95% CI, 1.03-1.15; P = .002). PD-only plan enrollees had a similar likelihood of discontinuation compared with MAPD plan enrollees based on multivariable analysis (hazard ratio, 1.00; 95% CI, 0.96-1.05; P = .877).

: MAPD plan enrollees do not have better adherence to medication regimens or persistence with COPD-maintenance medication than PD-only plan enrollees, despite financial incentive of MAPD health plans to improve medication adherence. Understanding the basis for this will be relevant for policy researchers and decision makers in developing strategies to improve patient outcomes through persistent medication use.

This study compared adherence and persistence patterns between Medicare Advantage Prescription Drug (MAPD) plan and prescription drug (PD)-only plan enrollees for chronic obstructive pulmonary disease maintenance medication. Findings presented will be relevant for policy makers to develop strategies for improving patient compliance to medications.

Adherence was significantly higher for PD-only plan enrollees compared with MAPD plan enrollees (40.5% vs 38.7%; P <.001).

Higher adherence of PD-only plan enrollees was also noted in multivariable analysis (odds ratio, 1.09; 95% CI, 1.03-1.15; P = .002)

A similar likelihood of discontinuation between PD-only and MAPD plan enrollees was observed, based on multivariable analysis (hazard ratio, 1.00; 95% CI, 0.96-1.05; P = .877).

Chronic obstructive pulmonary disease (COPD) represents a substantial disease burden. Approximately 14.8 million Americans have a COPD diagnosis, and COPD is currently the third-leading cause of death in the United States.

 COPD also has significant economic consequences, with 2010 medical costs estimated at $32.1 billion, which is projected to increase to $49 billion by 2020.

A large proportion of patients with COPD are 65 years or older and, therefore, are Medicare beneficiaries in many cases.

 In 2013, approximately 68% of Medicare beneficiaries were enrolled in Part D, an optional prescription drug coverage program available to all individuals with Medicare.

Among those with Part D coverage, two-thirds have stand-alone prescription drug (PD) plans that are available to patients with the traditional fee-for-service Medicare plan. The remaining one-third have Medicare Advantage Prescription Drug (MAPD) plans, which are offered by insurance companies and cover hospital (Medicare Part A), medical (Medicare Part B), and prescription (Medicare Part D) costs.

All Part D plans must meet standard criteria, but may offer varied premiums, benefit designs, gap coverage, formularies, and management rules, as well as enhanced benefits.

 Based on 2011 data from CMS, premiums are an average of $25 higher on PD-only than on MAPD plans.

 The vast majority of PD-only and MAPD plans operate a tiered cost-sharing structure during the initial coverage period, differentiating among generic, preferred, and nonpreferred brand-name drugs and specialty drugs.

In 2011, PD plans generally charged greater co-pays than MAPD plans for generic and preferred brands, but lower co-pays for nonpreferred or specialty brands.

 By law, PD plans must cover 7% of the costs of generics during the coverage gap, whereas pharmaceutical companies are obliged to offer a 50% price discount on brand-name drugs during the coverage gap. In addition to this legislation, some plans offer additional gap coverage, reducing total out-of-pocket costs incurred by patients during the coverage gap, although this is mostly for generics only.

A greater share of MAPD plans offer gap coverage than PD-only plans (67% vs 6%).6 Furthermore, PD-only and MAPD plans may offer enrollees the option of an “enhanced plan,” which can widen the range of drugs covered, remove the annual deductible, or lower co-pay rates.

Since MAPD plans cover both medical services and prescription drugs, these health plans have an incentive to control total cost of care; therefore, they often expend greater effort than PD-only plans to ensure adherence, with the assumption of greater offsets in medical care for chronic conditions, including COPD.

 However, in a study of MAPD plan enrollees, patients with COPD had one of the lowest observed rates of adherence and some of the highest pharmacy costs compared with patients with other chronic diseases (PD-only plan enrollees were not evaluated in this study).

In an effort to understand factors that affect medication adherence, differences in the structure of MAPD and PD-only plans represent an intriguing policy variable—one previously explored for statin usage.

 These factors are particularly relevant since patient adherence to COPD maintenance medication is associated with fewer hospitalizations and lower Medicare costs.

Identifying whether components of health management offered by an insurer could be a driver of adherence in patients with COPD could ultimately lead to improved medication use in elderly patients with chronic conditions. Therefore, this study was undertaken to evaluate adherence to and persistence with COPD maintenance medications among patients enrolled in MAPD or PD-only plans.

This study had 2 objectives: 1) to compare adherence to and persistence with COPD maintenance medications in MAPD plan enrollees versus PD-only plan enrollees; and 2) to identify covariates that are significantly associated with medication adherence, especially those relevant to Medicare policy (eg, low-income subsidy [LIS], coverage gap).

This was a retrospective, cross-sectional database study using data from a large, geographically diverse, US managed care plan. Eligibility and pharmacy claims data were extracted from the Optum Research Database, which included approximately 2.1 million annual MAPD plan enrollees with medical and pharmacy coverage, and approximately 5.6 million annual PD-only plan enrollees. In accordance with the Health Insurance Portability and Accountability Act, the use of de-identified data did not require institutional review board approval or waiver of authorization.

Because diagnosis codes on medical claims cannot be used to identify the presence of COPD among PD-only plan enrollees, and because medications to treat COPD are used in asthma as well, an algorithm to identify patients with COPD for study inclusion based on pharmacy claims alone was required. This approach was previously successfully evaluated by Mapel and colleagues (2010) using outpatient pharmacy data.

The algorithm was developed and validated using the MAPD population. The sensitivity and specificity of various combinations of rescue and maintenance medications were examined for their ability to identify patients with COPD. This process resulted in the following algorithm for identifying COPD patients: a) filled 1 or more prescriptions for roflumilast (an alpha1-proteinase inhibitor), tiotropium, aclidinium, arformoterol, formoterol, or indacaterol; or b) filled 2 or more prescriptions for ipratropium or ipratropium-albuterol combination.

Patients who met the criteria identified in the algorithm between January 1, 2011, and June 30, 2012, were identified as patients with COPD. An index date was defined as the first date that a prescription for a COPD maintenance medication was filled between January 1, 2012, and June 30, 2012. The COPD medication filled on the index date was designated as the index medication.

Patients were required to have been eligible for Medicare in the year prior to the index date (aged ≥65 years); have COPD based on the pharmacy claims—based algorithm described above; have received at least 1 pharmacy claim for COPD maintenance medication from January 1, 2012, through June 30, 2012; and have been continuously enrolled with pharmacy coverage in an MAPD or PD-only plan for 12 or more months prior to, and 60 days or more after, the index date.

Patients were excluded if they received a prescription for a leukotriene receptor antagonist at any time during the study period, if they switched between MAPD and PD-only plans during the study period, or if their gender was unknown as of the index date. Outcomes, including adherence and persistence, were measured during a 365-day post index period (inclusive of the index date), or until death or disenrollment, in which case the post index period ranged from 60 to 365 days.

Adherence was measured as the proportion of days covered (PDC) by any COPD maintenance medication. PDC was calculated as the number of days the patient had any type of COPD maintenance medication during the post index period, divided by the postindex period. Number of days of coverage was calculated using "days supplied" fields from pharmacy claims. The "days supplied" amounts indicate the number of days of therapy covered by the prescription fill.

This approach is commonly employed when pharmacy claims are used for adherence and persistence measurement. PDC values range from 0 to 1, with 1 indicating perfect adherence. Patients with a PDC value of less than 0.80 were defined as nonadherent with therapy.

 This approach is similar to that used in Medicare star ratings.

Persistence was measured as the duration of therapy (DOT).

This was defined as the number of days between the index date and the run-out date of the last COPD maintenance medication claim preceding a gap in therapy of 60 days or more. For patients without at least a 60-day gap in fills, the DOT was equivalent to the length of their post index period. Adherence and persistence were calculated using all COPD maintenance medications; switching between medications was considered as maintaining persistence.

Demographics (including age, gender, geographic location, and socioeconomic factors) and clinical characteristics were captured within the 12-month pre-index period. Socioeconomic factors (including median household income, probability of having a college education, likelihood of being nonwhite, and rural/urban residence) were obtained by linking the enrollee’s zip code to 2010 US Census data.

Pharmacy benefit factors included pharmacy benefit type, as well as LIS status, which was obtained from enrollment information. True out-of-pocket costs were calculated from January 1, 2012, to the index date, for patient payments for deductibles and co-pays, payments by other payers, and reductions in payments due to LIS.

The clinical and medication-related characteristics evaluated, included coverage phase of index prescription (eg, coinsurance, gap, catastrophic); gap coverage (as defined in Table 16,14); prescribing physician specialty for index prescription; Medicare health-risk assessment level; fills for short-acting beta2-agonists, corticosteroids, or chronic conditions; continuing user status; and total drug spend (patient and plan spend) during the pre-index period.

The Medicare health-risk assessment level was based on the pre-index prescription drug hierarchical condition categories (Rx-HCC) score provided by Medicare.

 The Rx-HCC model uses demographic and disease information from a base year to predict a patient’s drug costs in the next year.

Power analysis indicated that a sample size of 4288 patients with a ratio of 1:3 for enrollees in MAPD to PD-only plan cohorts, respectively, was needed to detect a 5% difference in PDC between cohorts with 80% power. Descriptive analyses were used to compare study variables, including covariates and outcome measures, between the MAPD and PD-only plan cohorts.

Adherence was modeled between MADP and PD-only plan cohorts using logistic regression analysis. Statistical significance was determined using either 

2 or Student’s t test. Kaplan-Meier analyses were conducted to compare the time to discontinuation of therapy for MAPD versus PD-only plans. A multivariable Cox proportional hazards model was employed to determine hazard ratio (HR) and P values for time to discontinuation. Fixed covariates utilized in the model included pre-index subject demographics, socioeconomic variables, clinical characteristics, and pharmacy benefit factors.

Demographic and Pre-Index Characteristics

A total of 130,305 patients were included in the study: 26,867 (21%) and 103,438 (79%) were enrolled in MAPD and PD-only plans, respectively, from January 1, 2011, to June 30, 2013. On average, MAPD plan enrollees were younger than PD-only plan enrollees, with a significantly lower proportion of patients being 80 years or older (30.0% vs 37.7%;

The largest difference between the 2 cohorts was the pharmacy benefit type in which they enrolled, with 66.5% of MAPD plan enrollees participating in an enhanced plan compared with 3.5% of PD-only plan enrollees (P <.001). A lower percentage of PD-only plan enrollees had LIS status than MAPD plan enrollees (12.7% vs 18.2%; P <.001), although a similar number had Medicare and Medicaid dual coverage (LIS dual: 17.5% vs 17.6%; P = .578).

There was a higher percentage of MAPD plan enrollees compared with PD-only plan enrollees, who lived in zip codes with residents having an annual household income less than $35,000 (59.7% vs 52.6%; P <.001). A lower percentage of MAPD plan enrollees lived in zip codes with residents having a college degree (P <.001).

A modestly lower percentage of MAPD plan enrollees were continuing users of COPD maintenance medications (as opposed to new users at the index fill) compared with PD-only plan enrollees (84.1% vs 85.6%; P <.001). MAPD plan enrollees were more likely than PD-only plan enrollees to receive a prescription for their index medication from a primary care physician (59.6% vs 54.1%; P <.001), while a higher percentage of PD-only plan enrollees obtained this prescription from a pulmonologist compared with MAPD plan enrollees (22.8% vs 19.0%; P <.001).

Furthermore, a significantly larger percentage of PD-only plan enrollees had an Rx-HCC score that was categorized as low compared with MAPD plan enrollees (78.8% vs 71.8%; P <.001). For both cohorts, over 90% of patients were in the coinsurance phase of their coverage.

. Mean PDC was higher for PD-only plan enrollees compared with MAPD plan enrollees (0.643 vs 0.626; P <.001). PD-only plan enrollees were significantly more adherent (PDC ≥80%) than MAPD plan enrollees (40.5% vs 38.7%; P <.001). Likewise, in the multivariable analysis, adherence was significantly higher for PD-only plan enrollees compared with MAPD plan enrollees (odds ratio [OR], 1.09; 95% CI, 1.03-1.15; P = .002) (Table 3).

Continuing users of COPD maintenance medications were more likely to be adherent than new users (OR, 3.82; 95% CI, 3.66-3.99; P <.001). For every 1 year of increased age, there was a 0.8% drop in the likelihood of being adherent when all other variables remained the same (P <.001).

In the assessment of pharmacy benefit factors, patients with an enhanced pharmacy benefit plan were shown to be less likely to be adherent than those with a standard plan (OR, 0.92; 95% CI, 0.86-0.98; P <.001). Patients with LIS dual coverage were less adherent compared with those who received only an LIS, but were more adherent compared with those who did not receive an LIS (OR, 1.44; 95% CI, 1.38-1.50; P <.001).

Patients who filled an index prescription in the coverage gap phase or catastrophic phase were more likely to be adherent than patients who were in the coinsurance phase (OR, 1.40; 95% CI, 1.26-1.56; and OR, 3.26; 95% CI, 2.72-3.91, respectively [P <.001]). In addition, although only 3% of the patients had gap coverage, those patients were more likely to be adherent than those who did not have gap coverage (P <.001). Gender, rural/urban residence, and median household income were not associated with adherence.

. In unadjusted analysis, PD-only plan enrollees had a longer mean DOT than MAPD plan enrollees (235.9 days vs 230.7 days; P <.001). However, the Cox proportional hazard model indicated a similar likelihood of discontinuation of COPD maintenance medication between MAPD and PD-only plan enrollees (HR, 1.00; 95% CI, 0.96-1.05; P = .877) (Table 4).

Continuing user status was one of the strongest predictors of therapy discontinuation, with continuing users of COPD therapy being less likely to discontinue than users who were new to therapy at index date (HR, 0.38; 95% CI, 0.37-0.39; P <.001). Patients who filled their index prescription in the coverage gap phase or catastrophic phase were also more likely to be persistent than patients who did so in the coinsurance phase (P <.001).

The authors evaluated the significance of Medicare prescription plan choice on chronic obstructive pulmonary disease maintenance medication adherence and persistence using Optum Research Database information. 