Health Care Utilization and Costs After Initiating Budesonide/Formoterol Combination or Fluticasone/Salmeterol Combination Among COPD Patients New to ICS/LABA Treatment

Abstract

BACKGROUND:

Chronic obstructive pulmonary disease (COPD) affects approximately 15 million people in the United States and accounts for approximately $36 billion in economic burden, primarily due to medical costs. To address the increasing clinical and economic burden, the Global Initiative for Chronic Obstructive Lung Disease emphasizes the use of therapies that help prevent COPD exacerbations, including inhaled corticosteroid/long-acting beta2-agonist (ICS/LABA).

OBJECTIVE:

To evaluate health care costs and utilization among COPD patients newly initiating ICS/LABA combination therapy with budesonide/formoterol (BFC) or fluticasone/salmeterol (FSC) in a managed care system.

METHODS:

COPD patients aged 40 years and older who initiated BFC (160/4.5 μg) or FSC (250/50 μg) treatment between March 1, 2009, and March 31, 2012, were identified using claims data from major U.S. health plans. BFC and FSC patients were propensity score matched (1:1) on age, sex, prior asthma diagnosis, prior COPD-related health care utilization, and respiratory medication use. COPD-related, pneumonia-related, and all-cause costs and utilization were analyzed during the 12-month follow-up period. Post-index costs were assessed with generalized linear models (GLMs) with gamma distribution. Health care utilization data were analyzed via logistic regression (any event vs. none) and GLMs with negative binomial distribution (number of visits) and were adjusted for the analogous pre-index variable as well as pre-index characteristics that remained imbalanced after matching.

RESULTS:

After matching, each cohort had 3,697 patients balanced on age (mean 64 years), sex (female 52% BFC and 54% FSC), asthma and other comorbid conditions, prior COPD-related health care utilization, and respiratory medication use. During the 12-month follow-up, COPD-related costs averaged $316 less for BFC versus FSC patients ($4,326 vs. $4,846; P = 0.003), reflecting lower inpatient ($966 vs. $1,202; P < 0.001), pharmacy ($1,482 vs. $1,609; P = 0.002), and outpatient/office ($1,378 vs. $1,436; P = 0.048) costs, but higher emergency department ($257 vs. $252; P = 0.033) costs. Pneumonia-related health care costs were also lower on average for BFC patients ($2,855 vs. $3,605; P < 0.001). Similarly, initiating BFC was associated with lower all-use health care costs versus initiating FSC ($21,580 vs. $24,483; P < 0.001, respectively). No differences in health care utilization were found between the 2 groups.

CONCLUSIONS:

In this study, although no difference was observed in rates of health care utilization, COPD patients initiating BFC treatment incurred lower average COPD-related, pneumonia-related, and all-cause costs versus FSC initiators, which was driven by cumulative differences in inpatient, outpatient, and pharmacy costs.

What is already known about this subject

• Inhaled corticosteroid/long-acting beta2-agonist (ICS/LABA) therapies are common first-line treatment options for the treatment of chronic obstructive pulmonary disease (COPD) in patients with a history of exacerbations.

• Prior research on the health care cost differences between the ICS/LABA therapies budesonide/formoterol (BFC) and fluticasone/salmeterol (FSC) in the United States is extremely limited.

What this study adds

• This study adds to previous work analyzing differences in health care costs for patients initiating the 2 therapies by providing a longer follow-up period and the use of multivariable models to adjust for pre-index differences.

• This study demonstrated that COPD patients initiating BFC therapy had lower COPD-related, pneumonia-related, and all-cause costs than patients initiating FSC treatment.

• This study found no significant differences in health care utilization between patients initiating BFC therapy versus those initiating FSC therapy.

Chronic obstructive pulmonary disease (COPD) is an increasingly burdensome respiratory illness that affects about 6.5% of the adult population (approximately 15 million people) in the United States.^1-5 The disease is characterized by inflammation and declining lung function and is linked to increasing rates of morbidity and mortality. The severe effects of COPD are manifest in several aspects of patients’ lives and activities: 700,000 inpatient hospitalizations, 10.3 million outpatient visits, and 1.5 million emergency department (ED) visits in 2010,^2,6 along with considerable absenteeism, in the United States.⁷ COPD results in substantial economic burden ($36 billion), estimated at $32.1 billion for medical costs and $3.9 billion for worker absenteeism—about 16.4 million lost work days—in the United States in 2010. By 2020, annual COPD-related medical costs are projected to increase to $49 billion.⁶

In the face of such serious economic and clinical challenges, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) emphasized the use of therapies that help to prevent COPD exacerbations as a central element of its disease management recommendations.⁸ The use of controller medications is supported by the GOLD 2015 treatment guidelines, which endorsed inhaled corticosteroid/long-acting beta2-agonist (ICS/LABA) combination therapies as a first-line maintenance therapy for patients with a history of COPD exacerbation.^8,9 At the time of this study, 2 ICS/LABA combination agents, budesonide/formoterol (BFC, 160/4.5 μg)¹⁰ and fluticasone/salmeterol (FSC, 250/50 μg)¹¹ were indicated for maintenance treatment of airflow obstruction in COPD patients in the United States, and are included in this study. A third ICS/LABA, fluticasone/vilanterol, is now available for maintenance treatment of COPD¹²; however, this therapy was approved in May 2013, after our study period ended in March 2013, and thus was not included.

Previous studies have shown that patients initiating BFC have fewer exacerbations and pneumonia events compared with FSC patients^4,13,14; however, information on costs associated with these ICS/LABA combination treatments is limited. To our knowledge, only 1 U.S. study evaluated medical service costs for patients receiving BFC and FSC treatments.⁴ The authors found no difference in mean total costs between the COPD patient groups receiving BFC and FSC agents.

To begin addressing this important evidence gap, we evaluated health care costs and utilization among COPD patients naïve to ICS/LABA fixed-dose combination medications who initiated BFC or FSC therapies in a large, real-world, managed care setting. To reduce the likelihood that a shorter follow-up time, as evidenced by a prior study,⁴ might have contributed to the lack of observed difference in mean costs between BFC and FSC cohorts, this study used a longer follow-up time of 12 months. Furthermore, in recognition of pneumonia as an important safety endpoint and cost driver among patients receiving ICS/LABA combination therapy, this study also evaluated mean pneumonia-related costs for the 2 cohorts.^15,16 The cost outcomes analyzed in this study were a secondary endpoint in a study primarily designed to compare the rates of COPD exacerbations in the year following initiation of the study medications (ClinicalTrials.gov identifier NCT01921127).¹⁷

Methods

Study Design and Population

In this retrospective, claims-based study, patients identified in the HealthCore Integrated Research Environment (HIRE) were followed for 12 months after initiating treatment with ICS/LABA combination agents BFC (160/4.5 μg) or FSC (250/50 μg). The primary objective of this study was to examine differences in COPD exacerbation rates between the 2 groups, the results of which have been published elsewhere.¹⁷ This analysis used the intent-to-treat approach and attributed all post-index treatment outcomes to the index medications of interest. The treatment index date was defined as the first date of treatment initiation corresponding to the first prescription fill of either BFC or FSC within the study intake period. All patients initiating BFC and FSC therapy between March 1, 2009, and March 31, 2012, were identified in HIRE. The HIRE contains enrollees in 14 major regional health plans and at the time of this study contained longitudinal claims data for more than 31 million enrollees from all U.S. census regions. This study was conducted in accordance with the Agency for Healthcare Research and Quality’s guidelines for conducting comparative effectiveness research.¹⁸ The management of all data and study materials conformed with all the relevant Health Insurance Portability and Accountability Act of 1996 (HIPAA) rules. A limited dataset, which excluded patient-identifying information, was used for all analyses, as defined by the HIPAA Privacy Rule. A national, independent institutional review board approved this study.

Inclusion/Exclusion Criteria.

Only patients who were naïve to ICS/LABA during the 12 months prior to initiating BFC (160/4.5 μg) or FSC (250/50 μg) were included in the study. The pharmacy claim date of treatment initiation was considered the index date. To be included, patients were required to be aged at least 40 years on the index date and have continuous active health plan enrollment, with both medical and pharmacy benefits, encompassing at least 12 months before (pre-index period) and 12 months after (post-index period) the index date. Furthermore, at least 1 inpatient visit with a primary diagnosis for COPD (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] diagnosis code 491.xx, 492.xx, or 496.xx), and/or at least 1 ED visit with a COPD diagnosis (either primary or secondary), and/or at least 2 other medical claims with a COPD diagnosis (either primary or secondary) during the pre-index period were required for inclusion in the study. Patients diagnosed with cancer, those who received ≥ 180 days of oral corticosteroids (OCS) in the 12-month pre-index period, and those who initiated both study medications on the same date were ineligible for the study.

Outcome Measures

Health Care Costs.

The costs attributable to inpatient hospitalizations, ED visits, outpatient/office visits, skilled nursing facility visits, and pharmacy prescriptions were compiled and reported as COPD-related, pneumonia-related, and all-cause (due to any condition). The total cost of care, including actual costs paid by the patient and those paid by the health plan, were included in all cost calculations. The sum of all costs was considered the total health care cost, and the sum of all non-pharmacy costs was considered medical costs. COPD-related costs include the costs of events that were linked to a COPD diagnosis code as a primary diagnosis for inpatient stays and any diagnosis for other services. A COPD-related pharmacy cost was defined as the cost of any of the following medications and medication classes, and includes the cost of the index drug: ICS/LABA, ICS, LABA, long-acting muscarinic antagonist (LAMA), short-acting beta2-agonist (SABA), short-acting muscarinic antagonist (SAMA), SABA/SAMA combination, OCS, theophylline, roflumilast, and antibiotics. Pneumonia-related costs did not require a primary pneumonia diagnosis and included visits with any pneumonia diagnosis (480.xx-486.xx), including viral, bacterial, and other/unspecified pneumonia. Pneumonia-related costs did not include any pharmacy costs, because of the inability to determine whether prescribed antibiotics (the typical medication received via a pharmacy for the treatment of pneumonia) were specific to pneumonia or for the treatment of another infection. All-cause health care costs include the costs of all health care encounters regardless of reason or diagnosis. For all cost data, the cost of procedures were included in their applicable place of service; for example, if a pulmonary function test was performed during a COPD-related office visit, the cost of that test would be included in the COPD-related office visit costs.

Health Care Utilization.

The presence of a visit (any vs. none) and the number of visits by place of service (inpatient hospitalization, intensive care unit [ICU], ED, outpatient/office visit) were captured. As with cost calculations, health care utilization was considered COPD related if the medical claim contained an applicable COPD diagnosis code; a primary diagnosis was required for inpatient stays, and a claim in any position was adequate for other service locations. Pneumonia-related utilizations included visits with a diagnosis of viral, bacterial, or other/unspecified pneumonia in any position. The diagnostic criteria using the claims database was validated against abstracted medical records with 80% positive predictive value.¹⁹

Other Variables of Interest.

Comorbid conditions were based on the presence of at least 1 medical claim with an ICD-9-CM diagnosis code for the condition of interest. Adherence to the index medication was measured by the proportion of days covered (PDC). The PDC was calculated as the ratio of the sum of days supply of all prescription fills during the 12 month follow-up period divided by 365 days.

Statistical Analysis

Propensity Score Matching.

Patients in the BFC and FSC cohorts were propensity score matched to reduce selection bias and create more comparable groups.^20,21 Although adequate clinical information was not available, propensity score matching, using proxies for clinical patient attributes, enabled the formation of comparable cohorts. Random forest methods²² were used to estimate the propensity scores as the probability of receiving BFC therapy, and patients were matched with the greedy nearest neighbor (1:1) matching technique without replacement.²¹ The outcome variable in the model was dichotomous, indicating whether a patient received BFC (1) or FSC (0) therapy. Treatment cohorts were considered well balanced when the P value for the difference between groups was > 0.05 for each of the prespecified variables listed below. The matching scheme specified a priori the balancing of age, sex, prior asthma diagnosis, COPD-related inpatient hospitalizations and ED visits, OCS fills, antibiotic fills, SABA or SABA/SAMA fills, LABA fills, and LAMA fills.

Statistical Models.

A generalized linear model (GLM) using the gamma distribution and a log-link function was used to model health care costs. Model adjusted mean differences, 95% confidence intervals, and P values are reported. All costs were adjusted for calendar year and are reported in 2012 dollars. To evaluate utilization outcomes, a GLM with a negative binomial distribution and log link was used for the number of visits and other count data, and logistic models were used for dichotomous outcomes. Adjusted mean differences (AMDs) in costs were calculated as the difference in predicted mean cost values between groups after adjusting for covariates in the statistical model.

Covariates identified in the 12-month pre-index period that remained imbalanced after matching (Table 1) were included in all statistical models, and consisted of the sum of inpatient hospital stays > 5 days (yes vs. no), leukotriene receptor antagonist use (0, 1, ≥ 2 fills), geographic region (Northeast, Midwest, South, or West), peripheral vascular disease/atherosclerosis (yes vs. no), and index-prescribing physician specialty (pulmonologist, internal medicine, family medicine/general practitioner, cardiologist, allergist/immunologist, nonphysician, or other specialty). The models also incorporated the analogous pre-index variable; for example, when modeling post-index all-cause health care costs, the model controlled for the pre-index all-cause health care costs. Although P < 0.05 was considered statistically significant for all outcomes, effect estimates and confidence intervals were primarily used in the interpretation of results. No adjustment was made for multiple testing. The propensity scores used for matching were obtained using R,²³ and all other analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC).

TABLE 1.

Propensity Score Matching Results^a

Matching Criteria	Before Matching			After Matching
	BFC (n = 3,788)	FSC (n = 6,439)	P Value	BFC (n = 3,697)	FSC (n = 3,697)	P Valueb
Priority matching
Demographics
Mean ± SD age, years	63.7 ± 11.4	65.0 ± 12.0	< 0.001	63.7 ± 11.5	64.0 ± 11.8	0.314
Female, % (n)	52.3 (1,981)	54.9 (3,535)	0.011	52.3 (1,932)	53.8 (1,989)	0.184
Prior asthma diagnosis, % (n)	36.2 (1,371)	32.3 (2,078)	< 0.001	35.7 (1,320)	34.6 (1,280)	0.330
COPD severity
Mean ± SD prior COPD inpatient visits	0.13 ± 0.40	0.1 ± 0.42	0.025	0.13 ± 0.40	0.14 ± 0.43	0.561
Mean ± SD prior COPD ED visits	0.18 ± 0.52	0.18 ± 0.52	0.652	0.18 ± 0.52	0.19 ± 0.55	0.353
Mean ± SD prior OCS fills	1.26 ± 1.74	1.01 ± 1.39	< 0.001	1.22 ± 1.68	1.17 ± 1.54	0.182
Mean ± SD prior antibiotic fills	2.74 ± 2.75	2.48 ± 2.61	< 0.001	2.70 ± 2.72	2.70 ± 2.81	0.944
COPD medications
Mean ± SD prior SABA or SABA/SAMA fills	2.99 ± 4.45	2.35 ± 3.83	< 0.001	2.89 ± 4.37	2.88 ± 4.41	0.916
Mean ± SD prior LABA fills	0.24 ± 1.35	0.11 ± 0.92	< 0.001	0.22 ± 1.29	0.16 ± 1.10	0.117
Mean ± SD prior LAMA fills	1.19 ± 2.61	0.98 ± 2.32	< 0.001	1.15 ± 2.55	1.11 ± 2.53	0.522
Optional matching
Mean ± SD pre-index exacerbations	1.08 ± 1.25	0.94 ± 1.14	< 0.001	1.07 ± 1.24	1.05 ± 1.23	0.528
Pre-index exacerbations, 0 vs. 1 +, % (n)	61.7 (2,337)	58.9 (3,795)	0.006	61.3 (2,267)	61.7 (2,282)	0.720
Due to COPD-related inpatient visit	11.2 (426)	13.4 (863)	0.002	11.4 (420)	12.1 (447)	0.329
Due to COPD-related ED visit	14.1 (536)	14.6 (938)	0.562	14.1 (523)	15.0 (553)	0.323
Due to COPD outpatient/office visit + OCS or antibiotics	47.4 (1,797)	40.0 (2,575)	< 0.001	46.8 (1,732)	45.0 (553)	0.107
Pre-index pneumonia diagnosis, % (n)	22.4 (849)	24.3 (1,567)	0.027	22.3 (823)	23.5 (867)	0.223

^aAdditional balance was achieved on index month, quarter, and year; health plan type; prescribing physician a pulmonologist; hospitalizations due to cardiovascular disease, pneumonia, or asthma; long-term oxygen use; comorbid conditions (insomnia, allergic rhinitis, sinusitis, gastroesophageal reflux disease, anxiety, major depressive disorder, other depression, obesity, asthma, sleep apnea, pneumonia, bronchiectasis, cystic fibrosis, coal worker pneumoconiosis, asbestosis, pneumoconiosis [due to other silica, inorganic dust, inhalation of other dust, and pneumoconiosis unspecified], respiratory conditions due to chemical fumes and vapors or other unspecified external agents, post-inflammatory pulmonary fibrosis, other alveolar and parietoalveolar pneumonopathy, lung involvement in conditions classified elsewhere, other diseases of lung, chronic respiratory failure, extrinsic allergic alveolitis, tuberculosis, lipoid pneumonia, detergent asthma, osteoporosis, diabetes mellitus, dyslipidemia, hyperglycemia, hypertension, pulmonary hypertension, congestive heart failure, left ventricular heart failure, coronary artery disease, myocardial infarction, unstable angina, other coronary artery disease, stroke, transient ischemic attack, and other cerebrovascular disease, as well as the mean Deyo-Charlson Comorbidity Index score); prior medication use categorized as 0, 1, or 2+ fills (OCS, antibiotics, ICS, LABA, roflumilast, theophylline, SABA, SAMA, SABA/SAMA combination, any cardiovascular related, and statins); influenza vaccination; pneumococcal vaccination; and continuity of care. Of all the baseline measures examined, only the presence of inpatient hospital stays > 5 days, leukotriene receptor antagonist use (0, 1, 2+), geographic region, presence of peripheral vascular disease/atherosclerosis, and index prescribing physician specialty (other than pulmonologist) were found to be statistically different between the 2 groups, all of which were controlled for in the multivariable models.

^bP values compare BFC with FSC and were calculated by using t-tests (age), generalized linear models with Poisson distribution (counts of fills, visits, and exacerbations), and chi-square tests (dichotomous variables).

BFC = budesonide/formoterol; COPD = chronic obstructive pulmonary disease; ED = emergency department; FSC = fluticasone/salmeterol; ICS = inhaled corticosteroid; LABA = long-acting beta2-agonist; LAMA = long-acting muscarinic antagonist; OCS = oral corticosteroids; SABA = short-acting beta2-agonist; SAMA = short-acting muscarinic antagonist; SD = standard deviation.

Results

Cohort Matching

A total of 3,788 BFC and 6,439 FSC initiators met all inclusion/exclusion criteria prior to matching. Following propensity score matching, a total of 3,697 patients from each cohort were included. Table 1 presents the balance of demographics and pre-index variables before and after matching. Descriptive statistics for patients who were and were not matched for each group can be found in the Appendix (available in online article).

Demographic and Clinical Characteristics for Matched Patients

On average, patients were aged 64 years in each treatment group, and the groups had a similar proportion of females (52% BFC and 54% FSC). In addition, there were similar rates of previously diagnosed pneumonia (22% BFC, 23% FSC), asthma (36% BFC, 35% FSC), and hypertension (69% BFC, 68% FSC), among other conditions, during the 12-month pre-index period. Proxies for COPD severity were also well balanced, including the mean number of prior COPD-related inpatient visits (0.13 BFC, 0.14 FSC), COPD-related ED visits (0.18 BFC, 0.19 FSC), and antibiotic fills (2.7 in each group). The use of prior respiratory medications at baseline was also similar for the 2 cohorts, as reflected by proportions of patients with at least 1 fill of OCS (56% BFC, 55% FSC), ICS (12% BFC, 11% FSC), SABA (59% BFC, 60% FSC), and SABA/SAMA combination (23% for both). During the 12-month follow-up period, adherence to the index medication was similar in each group, as evidenced by a PDC of 33% in the BFC cohort and 34% for FSC patients.

Post-index Outcomes

Health Care Utilization.

Overall, the rates of COPD-related and all-cause utilization were similar for patients in the 2 cohorts during the 12-month follow-up period, as shown in Table 2. The proportion of patients with COPD-related encounters were similar for the 2 cohorts, including inpatient hospitalizations (6.2% BFC, 6.9% FSC), ICU stays (0.6% and 0.9%), ED visits (12.3% and 11.8%), and outpatient/office visits (82.4% and 83.1%). No difference was seen in the average length of inpatient hospitalization or the average number of visits for any place of service between the 2 groups.

TABLE 2.

Post-initiation Health Care Utilization

Health Care Utilization	BFC (n = 3,697)a			FSC (n = 3,697)a			95% CLs
	Mean	SD	Median	Mean	SD	Median	Estimateb	Lower	Upper	P Value
All-cause health care resource utilization
All-cause inpatient hospitalizations
Patients with ≥ 1 event, % (n)	30.6 (1,131)			31.9 (1,178)			0.98	0.89	1.09	0.742
Length of stay per patientc	9.3	14.6	5.0	11.2	17.2	5.0	-0.10	-0.44	0.25	0.553
Number of eventsc	1.9	1.5	1.0	2.0	1.7	1.0	0.04	-0.06	0.14	0.468
All-cause ICU stays
Patients with ≥ 1 event, % (n)	6.0 (220)			6.5 (241)			0.95	0.78	1.15	0.563
Length of stay per patientc	2.0	5.5	1.0	1.9	4.2	1.0	0.00	-0.22	0.26	0.991
Number of eventsc	1.2	0.6	1.0	1.3	0.7	1.0	0.00	-0.18	0.21	0.991
All-cause ED visits
Number of patients with ≥ 1 event, % (n)	27.4 (1,012)			27.1 (1,003)			1.05	0.94	1.16	0.407
Number of eventsc	1.6	1.4	1.0	1.7	1.7	1.0	0.02	-0.08	0.14	0.678
All-cause outpatient/office visits
Patients with ≥ event, % (n)	99.5 (3,679)			99.2 (3,669)			1.59	0.88	2.88	0.127
Number of eventsc	33.1	30.7	25.0	34.9	32.3	27.0	0.03	-0.44	0.52	0.893
COPD-related health care resource utilization
COPD-related inpatient hospitalizations
Patients with ≥ event, % (n)	6.2 (228)			6.9 (255)			0.91	0.76	1.10	0.347
Length of stay per patientc	7.0	9.1	4.0	7.2	10.5	4.0	0.04	-0.96	1.21	0.949
Number of eventsc	1.3	1.1	1.0	1.2	0.7	1.0	0.10	-0.08	0.32	0.302
COPD-related ICU stays
Patients with ≥ event, % (n)	0.6 (24)			0.9 (32)			0.73	0.43	1.25	0.256
Length of stay per patientc	1.8	2.0	1.0	2.7	8.5	1.0	0.16	-0.66	1.73	0.766
Number of eventsc	1.3	0.7	1.0	1.1	0.3	1.0	0.13	-0.33	0.85	0.639
COPD-related ED visits
Patients with ≥ event, % (n)	12.3 (453)			11.8 (438)			1.08	0.93	1.24	0.312
Number of eventsc	1.4	0.9	1.0	1.4	0.9	1.0	0.02	-0.11	0.18	0.746
COPD-related outpatient/office visits
Patients with ≥ event, % (n)	82.4 (3,047)			83.1 (3,072)			0.93	0.82	1.05	0.250
Number of eventsc	9.6	11.9	5.0	10.0	12.8	5.0	-0.26	-0.63	0.14	0.197
Pneumonia-related health care resource utilization
Pneumonia-related inpatient hospitalizations
Patients with ≥ event, % (n)	8.92 (329)			10.3 (382)			0.87	0.75	1.02	0.094
Length of stay per patientc	11.6	20.4	6.0	12.7	19.3	7.0	-0.28	-1.81	1.48	0.737
Number of eventsc	1.4	0.9	1.0	1.4	1.0	1.0	0.03	-0.13	0.20	0.718
Pneumonia-related ED visits
Patients with ≥ event, % (n)	1.0 (36)			1.3 (47)			0.80	0.51	1.23	0.305
Number of eventsc	1.1	0.2	1.0	1.1	0.5	1.0	-0.01	-0.40	0.58	0.956
Pneumonia-related outpatient/office visits
Patients with ≥ event, % (n)	12.0 (443)			12.6 (466)			0.97	0.84	1.12	0.640
Number of eventsc	3.9	5.7	2.0	4.1	6.5	2.0	-0.11	-0.54	0.38	0.654

^aValues are mean, SD, and median unless otherwise specified.

^bVariables reporting number and percentage are dichotomous, analyzed via logistic regression, and odds ratios are the reported estimate, whereas variables reporting mean, SD, and median are continuous, analyzed via negative binomial models, and mean differences are the reported estimate. Statistical comparisons are comparing BFC with FSC. Model covariates include sum of inpatient hospital stays > 5 days (0 vs. 1), leukotriene receptor antagonist use (0, 1, 2+), geographic region, peripheral vascular disease/atherosclerosis (0 vs. 1), index prescribing physician specialty, and analogous pre-index variable.

^cIncluding only patients with at least 1 event; length of stay defined as the number of days from admission to discharge. Same date admission and discharge are counted as 1.

BFC = budesonide/formoterol; CL = confidence limit; COPD = chronic obstructive pulmonary disease; ED = emergency department; FSC = fluticasone/salmeterol; ICU = intensive care unit; SD = standard deviation.

Pneumonia-related encounters were also similar for patients in the 2 cohorts (Table 2). Inpatient hospitalizations were reported for 8.9% and 10.3%, ED visits for 1% and 1.3%, and outpatient/office visits for 12% and 12.6% of the patients in the BFC and FSC cohorts, respectively. The average number of visits for each service type and the length of inpatient hospitalization visits were also similar between the 2 groups.

Similar proportions of patients in the BFC and FSC cohorts, respectively, had at least 1 all-cause inpatient hospitalization (30.6% and 31.9%), ICU stay (6% and 6.5%), ED visit (27.4% and 27.1%), and outpatient/office visit (99.5% and 99.2%). When examining those that had at least 1 event, the average number of visits was similar between the 2 cohorts, including the number of inpatient hospitalizations (1.9 vs. 2.0), ICU stays (1.2 vs. 1.3), ED visits (1.6 vs. 1.7), and outpatient/office visits (33.1 vs. 34.9). Although not statistically significant, those initiating BFC had an average length of inpatient hospitalization stays nearly 2 days fewer than those who initiated FSC (9.3 days vs. 11.2 days).

Health Care Costs.

Although similar rates of health care utilization were observed between the 2 cohorts, BFC patients incurred $316 less in COPD-related costs than FSC patients during the first year after treatment initiation ($4,326 vs. $4,846, AMD = -$316, P = 0.003, respectively). The lower mean costs in the BFC relative to the FSC cohort were driven by lower COPD-related inpatient costs ($966 vs. $1,202, AMD = -$248, P < 0.001), outpatient/office visit costs ($1,378 vs. $1,436, AMD = -$97, P = 0.048), and pharmacy costs ($1,482 vs. $1,609, AMD = -$121, P = 0.002), despite higher COPD-related ED visit costs ($257 vs. $252, AMD = $26, P = 0.033), as shown in Table 3.

TABLE 3.

Post-initiation Health Care Costs

Health Care Costs	BFC, $ (n = 3,697)			FSC (Reference), $ (n = 3,697)			AMD,a $	95% CLs, $		P Value
	Mean	SD	Median	Mean	SD	Median		Lower	Upper
All-cause health care costsb
Total all-cause costs	21,580	46,376	10,255	24,483	45,324	11,186	-1,884	-2,685	-1,047	< 0.001
All-cause medical costs	16,943	45,139	5,574	19,619	44,030	6,487	-1,729	-2,513	-898	< 0.001
Inpatient hospitalization	8,519	38,072	0	10,314	34,708	0	-1,242	-2,082	-288	0.012
Emergency department	621	1,903	0	675	2,306	0	-12	-65	47	0.670
Outpatient/office visit	7,310	16,440	3,827	7,921	17,873	3,911	-464	-778	-135	0.006
Other medicalc	494	3,147	0	710	3,602	0	34	-3	75	0.076
All-cause pharmacy costs	4,637	6,141	3,205	4,864	5,710	3,465	-67	-200	71	0.339
COPD-related health care costsb
Total COPD-related costs	4,326	9,267	2,165	4,846	9,627	2,433	-316	-512	-108	0.003
COPD-related medical costs	2,843	8,927	508	3,236	9,274	584	-252	-432	-57	0.012
Inpatient hospitalization	966	6,915	0	1,202	7,688	0	-248	-337	-147	< 0.001
Emergency department	257	1,120	0	252	1,119	0	26	2	52	0.033
Outpatient/office visit	1,378	3,322	396	1,436	2,852	434	-97	-186	-1	0.048
Other medicalc	243	1,815	0	346	2,167	0	-15	-32	3	0.101
COPD-related pharmacy costs	1,482	1,634	947	1,609	1,712	1,105	-121	-193	-45	0.002
Pneumonia-related health care costsb,d
Pneumonia-related medical costs	2,855	32,237	0	3,605	23,095	0	-716	-967	-431	< 0.001
Inpatient hospitalization	2,728	32,033	0	3,409	22,880	0	-640	-894	-349	< 0.001
Emergency department	20	291	0	26	287	0	0	-2	2	0.691
Outpatient/office visit	57	346	0	78	557	0	-11	-14	-7	< 0.001
Other medicalc	51	951	0	93	1,310	0	0	-4	4	0.894

^aAdjusted mean differences are from gamma regression models (link = log). Statistical comparisons are comparing BFC with FSC (reference group); i.e., mean difference = mean (BFC)–mean (FSC). Model covariates include sum of inpatient hospital stays > 5 days (0 vs. 1), leukotriene receptor antagonist use (0, 1, 2+), geographic region, peripheral vascular disease/atherosclerosis (0 vs. 1), index prescribing physician specialty, and analogous pre-index cost variable (log-transformed to normalize).

^b$0 costs were manipulated to $1 for the gamma regression model; all costs were adjusted for calendar year, reported in 2012 U.S. dollars.

^cOther medical costs include skilled nursing facility costs. dICD-9-CM diagnosis codes used to identify pneumonia include viral pneumonia (480.xx), bacterial pneumonia (481.xx-485.xx), and other/unspecified pneumonia (486.xx).

AMD = adjusted mean difference; BFC = budesonide/formoterol; CL = confidence limit; COPD = chronic obstructive pulmonary disease; FSC = fluticasone/salmeterol; ICD-9-CM = International Classification of Diseases, Ninth Revision, Clinical Modification; SD = standard deviation.

Pneumonia-related costs followed a pattern similar to that of COPD-related costs. Patients in the BFC group incurred significantly lower pneumonia-related medical costs compared with patients in the FSC cohort ($2,855 vs. $3,605, AMD = -$716, P < 0.001), which was mainly driven by lower pneumonia-related inpatient costs ($2,728 vs. $3,409, AMD = -$640, P < 0.001).

In all, those initiating BFC incurred lower all-cause health care costs versus FSC patients during the first year after treatment initiation ($21,580 vs. $24,483, AMD = -$1,884, P < 0.001). The difference was driven mainly by lower all-cause inpatient costs ($8,519 vs. $10,314, AMD = -$1,242, P = 0. 012), and lower outpatient/office visit costs ($7,310 vs. $7,921, AMD = -$464, P = 0.006), among BFC versus FSC patients, respectively.

While heavily skewed, as cost data tend to be, differences in total all-cause costs were not influenced by outliers. The BFC group had the only observation that could be considered an outlier, with a cost more than triple the amount of the next highest cost in the cohort, whereas the FSC group had no such outliers. COPD-related costs contained no major outliers, whereas pneumonia-related costs had 1 large outlier in the BFC group with a value of more than 3 times greater than the second highest cost, and the FSC group had 1 moderate outlier with a cost more than 75% higher than the next highest.

Discussion

In this study, well-balanced cohorts of COPD patients receiving BFC or FSC treatments had similar all-cause, COPD-related, and pneumonia-related utilization rates for inpatient hospitalizations, ED visits, and outpatient/office visits. There were, however, statistically significant differences in all-cause, COPD-related, and pneumonia-related costs, reflecting the cumulative effect of the marginal differences between each utilization category. Overall, patients receiving BFC incurred lower COPD-related costs relative to patients receiving FSC therapy because of lower mean inpatient hospitalization, pharmacy, and outpatient/office visit costs, which more than offset the slightly higher per patient ED visit costs. This trend was repeated for all-cause costs driven by lower inpatient costs and lower outpatient/office visit costs incurred by BFC patients relative to the FSC cohort. Lower mean pneumonia-related inpatient costs among patients in the BFC cohort were the main contributor to the overall lower pneumonia-related costs relative to the FSC group. Although no statistical differences were found for measures of health care utilization, the FSC cohort had more patients with an ICU stay (6.5% vs. 6.0%) and had longer inpatients lengths of stay on average (mean days: 11.2 vs. 9.3), both of which are associated with high costs; thus, small nonstatistical differences in their utilization may be responsible for statistically significant differences in overall health care costs.

In contrast to these findings, a study by Roberts et al. (2011) reported no significant differences in COPD-related costs for medical services during the 3- to 6-month follow-up period.⁴ The authors reported that pharmacy costs were $43 greater for FSC relative to BFC patients during the limited follow-up period (compared with $121 over a 12-month follow-up in this study), which were attributed to the greater mean medication possession ratio among patients receiving FSC treatment.⁴ However, in the current study, adherence to the index medication, as measured by the PDC, was similar between the 2 groups. The Roberts et al. study was similar to our study in a number of key respects: patients were health plan enrollees, and sample size and age and sex distributions were generally similar. However, the difference in follow-up time, 6 months versus 1 year, was a salient difference.⁴ Although evidence from prior studies is not currently available, it is not implausible that as COPD patients continue to receive controller medications, the rate at which they require other services, especially costly inpatient hospitalization, could change and impact their overall costs. Although no differences in the number of hospitalizations and other services were detected, small, nonstatistically significant differences, including the length of inpatient hospitalization and prevalence of ICU stays, were seen between the 2 groups. Further evidence, including additional real-world effectiveness studies or randomized trials, is needed to clarify potential additional differences between these 2 agents.

A possible difference in costs may exist if different health plans offer different levels of coverage for the 2 medications being compared. However, in this study it is believed that both drugs, which are of the same class, were covered similarly, and the most recent health plan formularies show no difference in coverage between the 2 medications being studied.^24-26

A key strength of this study was the use of a robust design that incorporated propensity score matching to create cohorts with solid bases for comparison over the 12-month follow-up period. Furthermore, the random forest method utilized in the propensity score algorithm has been shown to be more effective in balancing treatment cohorts with less bias than traditional logistic regression techniques.²² A subset of pneumonia diagnoses in this study was validated via medical chart abstraction and review, which found a positive predictive value of 80%.¹⁹

The other important observation in the study of costs is that COPD-related costs are just a portion of total health care costs, although in this study they accounted for a significant fraction, with more than 20% of all-cause costs related to COPD. Additionally, because the presence of COPD is associated with a number of comorbid conditions, better COPD control may plausibly decrease total health care spending directly and indirectly. Further study is needed to determine whether a strategy of improved ICS/LBA compliance might lead to lower total health care costs while marginally increasing pharmacy costs.

Limitations

The results of this study must be viewed against important limitations, including those associated with a reliance on secondary data.^27,28 They may be generalizable only to commercially insured populations in the United States. Patients were not required to take the index medication during follow-up. As a result, they filled index medication an average of roughly 4 times during follow-up, which resulted in an average of just 33%-34% of post-index days covered with supply of the index medication. Nonetheless, the results of this intent-to-treat analysis were evaluated as the outcomes from the initiation of the study medications without prior ICS/LABA use.

There was a lack of detailed clinical information, including smoking status, pulmonary function assessment, and patient reported outcomes, to adjust for COPD disease severity and activity or to confirm COPD diagnoses, because such information is absent from the primary data source, administrative claims. However, the study matched on prior COPD exacerbations, which is the best predictor of future exacerbations.⁸

Pharmacy medications administered during inpatient hospitalizations were not included because that information was not available in these data. It was not possible to determine the primary reason for outpatient and ED visits via claims data; as a result, even when a COPD diagnosis code was present, the visit might have been for routine follow-up or non-COPD-related reasons.²⁹ Thus, the results of this study might not be generalized to BFC or FSC users with COPD overall.

Conclusions

No difference was found in rates of health care utilization between the 2 groups; however, this study demonstrated an association between COPD patients initiating BFC therapy and lower COPD-related costs relative to patients initiating FSC treatment. This result represented the cumulative effect of lower mean costs for inpatient hospitalizations, pharmacy, and outpatient/office visits counterbalancing slightly higher average ED visit costs. BFC patients also had lower all-cause costs, which were driven by lower hospitalization and outpatient/office visit costs incurred by BFC versus FSC patients, and a similar trend was observed for pneumonia-related costs. This is the first study comparing resource utilization and costs of BFC and FSC administered to ICS/LABA-naïve patients for a full year following treatment initiation. Additional studies are needed to evaluate other key aspects of the economic outcomes of BFC and FSC use among COPD patients.

APPENDIX. Characteristics of Unmatched and Matched FSC and BFC Patients

Characteristics	Unmatched FSC a (n = 2,742)		Matched FSC a (n = 3,697)		Matched BFC a (n = 91)		Unmatched BFC a (n = 3,697)
	N	%	N	%	N	%	N	%
Demographics
Mean ± SD age, years	66.4 ± 12.2		64.0 ± 11.8		63.2 ± 9.5		63.7 ± 11.46
40-49	206	7.5	388	10.5	8	8.8	392	10.6
50-59	678	24.7	1,063	28.8	24	26.4	1,065	28.8
60-64	409	14.9	627	17.0	24	26.4	629	17.0
65+	1,449	52.8	1,619	43.8	35	38.5	1,611	43.6
Female	1,546	56.4	1,989	53.8	49	53.9	1,932	52.3
Health plan type
HMO	828	30.2	1,153	31.2	28	30.8	1,149	31.1
PPO	1,722	62.8	2,264	61.2	58	63.7	2,283	61.8
CDHP	102	3.7	189	5.1	3	3.3	168	4.5
Other commercial	90	3.3	91	2.5	2	2.2	97	2.6
Geographic region
Northeast	714	26.0	767	20.8	10	11.0	629	17.0
Midwest	1,000	36.5	1,555	42.1	45	49.5	1,690	45.7
South	591	21.6	881	23.8	28	30.8	965	26.1
West	437	15.9	494	13.4	8	8.8	413	11.2
Prescriber specialty
Pulmonologist	600	21.9	1,042	28.2	49	53.9	1,094	29.6
Internal medicine	968	35.3	1,066	28.8	11	12.1	1,002	27.1
Family/general practitioner	693	25.3	960	26.0	23	25.3	1,022	27.6
Cardiologist	33	1.2	31	0.8	0	0.0	13	0.4
Allergist/immunologist	10	0.4	29	0.8	2	2.2	45	1.2
Nonphysician	191	7.0	248	6.7	4	4.4	253	6.8
Other specialty	106	3.9	128	3.5	0	0.0	114	3.1
Unknown	141	5.1	193	5.2	2	2.2	154	4.2
Index year
2009	1,022	37.3	1,185	32.1	24	26.4	1,109	30.0
2010	933	34.0	1,231	33.3	32	35.2	1,208	32.7
2011	617	22.5	982	26.6	21	23.1	1,057	28.6
2012	170	6.2	299	8.1	14	15.4	323	8.7
Index quarter
2009
Q1 (only March)	122	4.5	139	3.8	0	0.0	103	2.8
Q2 (Apr-Jun)	328	12.0	336	9.1	5	5.5	299	8.1
Q3 (Jul-Sep)	285	10.4	311	8.4	10	11.0	297	8.0
Q4 (Oct-Dec)	287	10.5	399	10.8	9	9.9	410	11.1
2010
Q1 (Jan-Mar)	302	11.0	376	10.2	10	11.0	384	10.4
Q2 (Apr-Jun)	253	9.2	321	8.7	10	11.0	313	8.5
Q3 (Jul-Sep)	212	7.7	255	6.9	4	4.4	258	7.0
Q4 (Oct-Dec)	166	6.1	279	7.6	8	8.8	253	6.8
2011
Q1 (Jan-Mar)	180	6.6	274	7.4	6	6.6	295	8.0
Q2 (Apr-Jun)	157	5.7	288	7.8	4	4.4	287	7.8
Q3 (Jul-Sep)	119	4.3	198	5.4	9	9.9	246	6.7
Q4 (Oct-Dec)	161	5.9	222	6.0	2	2.2	229	6.2
2012
Q1 (Jan-Mar)	170	6.2	299	8.1	14	15.4	323	8.7
COPD severity
Mean ± SD pre-index exacerbations	0.80 ± 1.00		1.05 ± 1.23		1.78 ± 1.56		1.07 ± 1.24
Due to COPD-related inpatient visit	0.16 ± 0.40		0.13 ± 0.39		0.07 ± 0.25		0.13 ± 0.37
Due to COPD-related ED visit	0.16 ± 0.44		0.18 ± 0.48		0.15 ± 0.39		0.17 ± 0.47
Due to COPD outpatient/office visit +OCS and/or antibiotics	0.47 ± 0.84		0.74 ± 1.06		1.56 ± 1.43		0.77 ± 1.08
COPD severity
Pre-index exacerbations, 1 or more event	1,513	55.2	2,282	61.7	70	76.9	2,267	61.3
Due to COPD-related inpatient visit	416	15.2	447	12.1	6	6.6	420	11.4
Due to COPD-related ED visit	385	14.0	553	15.0	13	14.3	523	14.2
Due to COPD outpatient/office visit +OCS and/or antibiotics	913	33.3	1,663	45.0	66	72.5	1,732	46.9
Comorbid conditions
Deyo-Charlson Comorbidity Index score
0	684	25.0	1,018	27.5	20	22.0	996	26.9
1-2	1,279	46.6	1,792	48.5	57	62.6	1,866	50.5
3-4	512	18.7	632	17.1	11	12.1	604	16.3
5 +	267	9.7	255	6.9	3	3.3	231	6.3
Mean ± SD		1.87 ± 1.85		1.63 ± 1.71		1.4 ± 1.3		1.61 ± 1.7
Individual comorbidities
Insomnia	174	6.4	244	6.6	6	6.6	247	6.7
Allergic rhinitis	261	9.5	535	14.5	22	24.2	581	15.7
Sinusitis	408	14.9	750	20.3	37	40.7	804	21.8
GERD	570	20.8	818	22.1	29	31.9	820	22.2
Anxiety	310	11.3	524	14.2	16	17.6	547	14.8
Major depressive disorder	126	4.6	189	5.1	6	6.6	179	4.8
Other depression	396	14.4	570	15.4	13	14.3	582	15.7
Obesity	305	11.1	470	12.7	10	11.0	472	12.8
Asthma	798	29.1	1,280	34.6	51	56.0	1,320	35.7
Sleep apnea	353	12.9	588	15.9	17	18.7	592	16.0
Pneumonia	700	25.5	867	23.5	26	28.6	823	22.3
Bronchiectasis	59	2.2	85	2.3	1	1.1	103	2.8
Cystic fibrosis	2	0.1	1	0.0	0	0.0	2	0.1
Coal worker pneumoconiosis	4	0.2	9	0.2	1	1.1	11	0.3
Asbestosis	10	0.4	17	0.5	1	1.1	10	0.3
Pneumoconiosis due to other silica	2	0.1	1	0.0	1	1.1	1	0.0
Pneumoconiosis due to inorganic dust	2	0.1	1	0.0	0	0.0	0	0.0
Pneumoconiosis due to inhalation of other dust	1	0.0	0	0.0	0	0.0	0	0.0
Pneumoconiosis unspecified	3	0.1	1	0.0	1	1.1	1	0.0
Respiratory conditions due to chemical fumes and vapors or other unspecified external agents	6	0.2	10	0.3	0	0.0	4	0.1
Post-inflammatory pulmonary fibrosis	152	5.5	208	5.6	3	3.3	193	5.2
Other alveolar and parietoalveolar pneumonopathy	50	1.8	51	1.4	0	0.0	43	1.2
Lung involvement in conditions classified elsewhere	7	0.3	4	0.1	0	0.0	11	0.3
Other diseases of lung	947	34.5	1,228	33.2	32	35.2	1,209	32.7
Chronic respiratory failure	108	3.9	171	4.6	5	5.5	184	5.0
Extrinsic allergic alveolitis	4	0.2	4	0.1	1	1.1	6	0.2
Tuberculosis	10	0.4	6	0.2	0	0.0	10	0.3
Lipoid pneumonia	0	0.0	0	0.0	91	100.0	3,697	100.0
Detergent asthma	1	0.0	1	0.0	84	92.3	3,388	91.6
Osteoporosis	331	12.1	341	9.2	7	7.7	309	8.4
Diabetes mellitus	732	26.7	912	24.7	15	16.5	959	25.9
Dyslipidemia	1,633	59.6	2,159	58.4	39	42.9	2,171	58.7
Hyperglycemia	162	5.9	206	5.6	7	7.7	198	5.4
Hypertension	1,988	72.5	2,511	67.9	58	63.7	2,535	68.6
Pulmonary hypertension	184	6.7	208	5.6	6	6.6	197	5.3
Congestive heart failure	691	25.2	696	18.8	11	12.1	654	17.7
Left ventricular heart failure	62	2.3	74	2.0	0	0.0	74	2.0
Peripheral vascular disease/atherosclerosis	486	17.7	548	14.8	8	8.8	485	13.1
Individual comorbidities
Coronary artery disease	981	35.8	1,152	31.2	19	20.9	1,156	31.3
Myocardial infarction	241	8.8	277	7.5	4	4.4	276	7.5
Unstable angina	119	4.3	121	3.3	3	3.3	147	4.0
Other coronary artery disease	947	34.5	1,117	30.2	16	17.6	1,124	30.4
Stroke	108	3.9	108	2.9	0	0.0	89	2.4
TIA and other cerebrovascular	521	19.0	493	13.3	10	11.0	463	12.5
Medication use
OCS use
0 fills	1,456	53.1	1,652	44.7	21	23.1	1,642	44.4
1 fill	785	28.6	940	25.4	12	13.2	953	25.8
2 + fills	501	18.3	1,105	29.9	58	63.7	1,102	29.8
ICS use
0 fills	2,601	94.9	3,299	89.2	55	60.4	3,243	87.7
1 fill	71	2.6	165	4.5	8	8.8	190	5.1
2 + fills	70	2.6	233	6.3	28	30.8	264	7.1
LAMA use
0 fills	1,858	67.8	2,460	66.5	40	44.0	2,499	67.6
1 fill	564	20.6	619	16.7	14	15.4	545	14.7
2 + fills	320	11.7	618	16.7	37	40.7	653	17.7
LABA use
0 fills	2,713	98.9	3,561	96.3	74	81.3	3,539	95.7
1 fill	10	0.4	44	1.2	1	1.1	35	1.0
2 + fills	19	0.7	92	2.5	16	17.6	123	3.3
SABA use
0 fills	1,273	46.4	1,474	39.9	20	22.0	1,532	41.4
1 fill	816	29.8	930	25.2	7	7.7	850	23.0
2 + fills	653	23.8	1,293	35.0	64	70.3	1,315	35.6
SAMA use
0 fills	2,642	96.4	3,480	94.1	71	78.0	3,486	94.3
1 fill	66	2.4	125	3.4	11	12.1	123	3.3
2 + fills	34	1.2	92	2.5	9	9.9	88	2.4
SABA/SAMA use
0 fills	2,263	82.5	2,860	77.4	65	71.4	2,859	77.3
1 fill	314	11.5	411	11.1	9	9.9	383	10.4
2 + fills	165	6.0	426	11.5	17	18.7	455	12.3
Antibiotic use
0 fills	642	23.4	720	19.5	10	11.0	712	19.3
1 fill	663	24.2	769	20.8	8	8.8	774	20.9
2 + fills	1,437	52.4	2,208	59.7	73	80.2	2,211	59.8
Roflumilast
0 fills	2,741	100.0	3,696	100.0	91	100.0	3,695	100.0
1 fill	1	0.0	1	0.0	0	0.0	2	0.1
2 + fills	0	0.0	0	0.0	0	0.0	0	0.0
Theophylline use
0 fills	2,699	98.4	3,600	97.4	80	87.9	3,590	97.1
1 fill	17	0.6	18	0.5	1	1.1	24	0.7
2 + fills	26	1.0	79	2.1	10	11.0	83	2.3

^aValues are number and percentage unless otherwise specified.

BFC = budesonide/formoterol; CDHP = consumer-directed health plan; COPD = chronic obstructive pulmonary disease; FSC = fluticasone/salmeterol; GERD = gastroesophageal reflux disease; HMO = health maintenance organization; ICS = inhaled corticosteroid; LABA = long-acting beta2-agonist; OCS = oral corticosteroids; PPO = preferred provider organization; SABA = short-acting beta2-agonist; SAMA = short-acting muscarinic antagonist; SD = standard deviation; TIA = transient ischemic attack.