[
Original Research Genetic and Developmental Disorders
]
Pulmonary Medication Adherence and Health-care Use in Cystic Fibrosis Alexandra L. Quittner, PhD; Jie Zhang, PhD; Maryna Marynchenko, MBA; Pooja A. Chopra, MS; James Signorovitch, PhD; Yana Yushkina, BA; and Kristin A. Riekert, PhD
Poor treatment adherence is common in cystic fibrosis (CF) and may lead to worse health outcomes and greater health-care use. This study evaluated associations of adherence to pulmonary medications, age, health-care use, and cost among patients with CF.
BACKGROUND:
METHODS: Patients with CF aged ⱖ 6 years were identified in a national commercial claims database. A 12-month medication possession ratio (MPR) was computed for each pulmonary medication and then averaged for a composite MPR (CMPR) for each patient. The CMPR was categorized as low (, 0.50), moderate (0.50-0.80), or high (ⱖ 0.80). Annual health-care use and costs were measured during the first and second year and compared across adherence categories by multivariable modeling.
Mean CMPR for the sample (N 5 3,287) was 48% ⫾ 31%. Age was inversely related to CMPR. In the concurrent year, more CF-related hospitalizations were observed among patients with low (event rate ratio [ERR], 1.35; 95% CI, 1.15-1.57) and moderate (ERR, 1.25; 95% CI, 1.05-1.48) vs high adherence; similar associations were observed for all-cause hospitalizations and CF-related and all-cause acute care use (hospitalizations 1 ED) in the concurrent and subsequent year. Rates of CF-related and all-cause outpatient visits did not differ by adherence. Low and moderate adherence predicted higher concurrent health-care costs by $14,211 ($5,557-$24,371) and $8,493 (2$1,691 to $19,709), respectively, compared with high adherence.
RESULTS:
Worse adherence to pulmonary medications was associated with higher acute health-care use in a national, privately insured cohort of patients with CF. Addressing adherence may reduce avoidable health-care use. CHEST 2014; 146(1):142-151
CONCLUSIONS:
Manuscript received August 16, 2013; revision accepted January 6, 2014; originally published Online First January 30, 2014. ABBREVIATIONS: CF 5 cystic fibrosis; CMPR 5 composite medication possession ratio; ICD-9-CM 5 International Classification of Diseases, Ninth Revision, Clinical Modification; MPR 5 medication possession ratio AFFILIATIONS: From the Department of Psychology (Dr Quittner), University of Miami, Coral Gables, FL; Novartis Pharmaceuticals Corp (Dr Zhang), East Hanover, NJ; Analysis Group, Inc (Mss Marynchenko, Chopra, and Yushkina and Dr Signorovitch), Boston, MA; and Division of Pulmonary and Critical Care Medicine (Dr Riekert), The Johns Hopkins University School of Medicine, Baltimore, MD.
142 Original Research
Parts of this study have been reported in abstract form at the 26th Annual North American Cystic Fibrosis Conference, October 11-13, 2012, Orlando, FL. FUNDING/SUPPORT: This study was supported by the Novartis Pharmaceuticals Corp, which manufactures inhaled tobramycin. CORRESPONDENCE TO: Alexandra L. Quittner, PhD, Department of Psychology, University of Miami, 5665 Ponce de Leon Blvd, Coral Gables, FL 33146; e-mail: [email protected] © 2014 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.13-1926
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
Recent advances in the treatment and aggressive management of cystic fibrosis (CF) have led to significant improvements in prognosis, with median predicted survival reaching 37 years.1 However, daily management requires a complex, time-consuming treatment regimen to address the multisystem effects of CF.2 Long-term use of one or more pulmonary medications is recommended,3 which can include inhaled antibiotics (eg, tobramycin, colistin, aztreonam) and oral azithromycin and nebulized mucolytic agents (eg, dornase alfa, hypertonic saline).3,4 The burden of treatment and poor adherence have emerged as critical issues in the daily management of CF.5,6 Prior analyses of objective adherence data have reported low rates of adherence to individual pulmonary medications among patients with CF, ranging from 31% to 79%.7-12 Most of these studies were limited by small sample sizes and limited age ranges. Furthermore, little is known about how rates of adherence to pulmonary medications vary across the life span from childhood through older adulthood, although there appears to be a significant decrease in adherence during adolescence.13,14 Adherence may also affect health-care use and costs. Two studies
Materials and Methods Data Source and Sample Selection The Thomson Reuters MarketScan Commercial Claims and Encounters Database (January 2005 to June 2011) includes medical and drug data for active employees and their dependents with primary coverage through employer-sponsored private health insurance throughout the United States.18 The MarketScan claims database offers the largest claims capture of physician office visits; hospital stays; retail, mail-order, and specialty pharmacies; and carve-out care. The database undergoes extensive validity and editing procedures to ensure high quality and consistency and has been presented in . 400 peer-reviewed articles. Patients included in this retrospective cohort study met the following criteria: (1) two or more independent CF diagnoses (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD9-CM], code 277.0X) during medical visits occurring ⱖ 30 days apart, (2) one or more prescription fills for a pulmonary medication (azithromycin, dornase alfa, hypertonic saline, or an inhaled antibiotic [aztreonam, colistin, or tobramycin]), and (3) continuous enrollment for ⱖ 180 days. For each patient meeting these criteria, the end date of the first 180-day continuous enrollment period and the end dates of subsequent 365-day periods were defined as potential index dates. An index date was randomly chosen among all potential index dates meeting the following additional inclusion criteria: (4) continuous enrollment for ⱖ 365 days after the index date, (5) a filled prescription for a pulmonary medication within the prior 90 days, and (6) age ⱖ 6 years. Measures For each long-term pulmonary medication filled ⱕ 90 days before the index date, adherence was calculated as the medication possession ratio (MPR), that is, the cumulative days’ supply of medication dispensed divided by 365 days after the index date (excluding hospitalized days). Values were truncated to 100%. Drug-specific MPRs were calculated for each pulmonary medication and then averaged to obtain a composite MPR (CMPR).
journal.publications.chestnet.org
found that worse adherence to inhaled tobramycin was associated with increased risk of hospitalization and health-care costs,15,16 whereas one-third found no association between dornase alfa adherence and inpatient respiratory exacerbations but a trend toward higher overall health-care charges.17 Lower overall pulmonary medication adherence (average adherence to dornase alfa, hypertonic saline, azithromycin, and inhaled tobramycin) was also associated with greater occurrence of pulmonary exacerbations requiring IV antibiotics and lung function testing in a retrospective medical record review of patients with CF.5 To our knowledge, no study has assessed the associations between overall adherence to long-term pulmonary medications and age, health-care use, and health-care costs in CF. Using retrospective claims data for privately insured patients with CF, this study measured drug-specific and overall adherence to pulmonary medications and assessed variations in overall adherence across demographic subgroups and their associations with healthcare use. We hypothesized that poorer overall medication adherence would be associated with older age and higher health-care use and costs.
CMPR was converted to a categorical variable as follows: low (, 0.50), moderate (0.5 ⱕ CMPR , 0.80), and high (ⱖ 0.80).5 Additional details for this calculation appear in e-Appendix 1 and e-Tables 1 and 2. Health-care use and costs were measured for two time periods to assess the prospective relationship with adherence: (1) the first year (365 days) after the index date to assess concurrent associations with adherence and (2) the second year after the index date (ie, 365 days after index date 1 365 days). Use included hospitalizations and ED, outpatient, and acute care (hospitalizations 1 ED) visits. Costs included the sum of hospitalizations, ED, and outpatient charges adjusted for inflation to 2011 US dollars. All-cause and CF-related health-care use and costs were evaluated; medical visits were considered CF related if associated with a CF diagnosis (ICD-9-CM code 277.0X). Pharmacy costs were not included. Baseline characteristics included demographics, index year, comorbidities, number of medications, and medication copayment. Comorbidities were identified on the basis of at least one claim associated with the corresponding ICD-9-CM code and required to be present during the 180-day baseline period. Statistical Analysis Mean CMPRs were compared across patient subgroups, including age, sex, and number of pulmonary medications dispensed ⱕ 90 days preceding the index date. Between-group differences were assessed with Wilcoxon rank sum tests. Health-care use was compared across CMPR levels using unadjusted and adjusted event rate ratios estimated from univariable and multivariable negative binomial regression models, respectively. Unadjusted differences in costs were tested with Wilcoxon rank sums. Multivariable generalized linear models with a g-distribution and a log link were used to estimate adjusted cost differences, with statistical significance assessed through nonparametric bootstrap resampling with 499 iterations. All multivariable models were adjusted for age, sex, comorbidities, index year, number of medications, and medication copayment at baseline. All analyses were conducted with SAS 9.2 (SAS Institute Inc) statistical software.
143
Results Patient Characteristics
The study sample included 3,287 patients with CF (Table 1). Mean age was 22.8 ⫾ 13.0 years, and female patients comprised 49% of the sample (Table 2). The demographic characteristics of the present sample were comparable to the general CF population as reported in the Cystic Fibrosis Registry1; however, the present sample included more adults (age ⱖ 18 years, 56% vs 48%) and more patients with Pseudomonas aeruginosa infection (63.0% vs 51%). A subgroup of 1,420 patients had continuous eligibility for 730 days following their index dates and, thus, were included in analyses of second-year health-care use and cost outcomes. Adherence to Pulmonary Medications
Mean MPR was highest for dornase alfa (57%) followed by inhaled tobramycin (51%), azithromycin (50%), inhaled aztreonam (47%), inhaled colistin (42%), and hypertonic saline (40%) (Fig 1). The average CMPR was 48%; 51% of patients had a CMPR , 0.50, whereas only 20% of patients had a CMPR ⱖ 0.80. MPRs for individual drugs were significantly correlated with CMPR, with Spearman r ranging from 0.70 for inhaled colistin to 0.86 for azithromycin (all P , .001). Age was related to CMPR, with the highest adherence among patients aged 6 to 10 years (59%) compared with the other age categories (all P , .001) (Figs 2-3). This finding was consistent across all drug-specific MPRs (e-Tables 3-8). Male patients had higher CMPRs than female patients (50% vs 45%, respectively; P , .001); however, post hoc analyses showed that the sex difference was statistically significant only for azithromycin
(P , .001) (e-Tables 3-8). Patients who filled more than one pulmonary medication had significantly higher adherence than those who filled one medication only (all P , .01) (mean CMPR, 41%, 51%, 56%, 60%, and 62% for one, two, three, four, and five medications filled, respectively). This association between adherence and CF prescription burden was replicated for drug-specific MPRs (e-Tables 3-8). Association With Health-care Use
In the regression-adjusted analysis of first-year healthcare use, low adherence was associated with significantly TABLE 2
] Baseline Characteristics
Characteristica
Value
No. patients
3,287
Demographics Female sex
1,604 (48.8)
Age category 6-10 y
528 (16.1)
11-17 y
900 (27.4)
18-25 y
775 (23.6)
26-35 y
511 (15.6)
ⱖ 36 y
573 (17.4)
Baseline comorbidities Charlson comorbidity index
0.7 ⫾ 1.0
Pseudomonas aeruginosa infection
2,071 (63.0)
Sinusitis
599 (18.2)
Diabetes mellitus
459 (14.0)
Asthma
442 (13.5)
Gastritis
186 (5.7)
Depression
181 (5.5)
Osteoporosis
153 (4.7)
Failure to thrive/growth failure
117 (3.6)
No. medications TABLE 1
] Sample Selection, 2005 to 2011
Inclusion Criterion Two or more CF diagnoses
CF pulmonary medicationsb
No. Patients 11,590
One or more prescriptions filled for a pulmonary medication
6,665
Continuous enrollment for ⱖ 180 d
6,459
Continuous enrollment for ⱖ 365 d following any potential index date
4,730
One or more prescriptions filled for a pulmonary medication within 90 d before any potential index date
3,565
Aged ⱖ 6 y as of any potential index date
3,287
CF 5 cystic fibrosis.
144 Original Research
2.1 ⫾ 1.0
Other CF-related medications
c
Total medications
2.0 ⫾ 1.5 10.1 ⫾ 5.6
Total medication copayment, 2011 USD/mo
107.1 ⫾ 202.7
Data are presented as No. (%) or mean ⫾ SD unless otherwise indicated. USD 5 US dollars. See Table 1 legend for expansion of other abbreviation. aThe baseline period is defined as the 180-d period prior to the index date. bIndicates treatment with one of the following pulmonary medications: inhaled tobramycin, inhaled colistin, inhaled aztreonam, oral azithromycin, nebulized dornase alfa, and nebulized hypertonic saline. cIndicates treatment with one or more of the following medications: leukotriene modifiers, inhaled bronchodilators, inhaled steroids, allergy medications, antidiabetic medications, digestives, vitamins, enzymes, ibuprofen, and prednisone.
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
Figure 1 – Mean MPRs for various long-term pulmonary medications used in cystic fibrosis. The bottom, midline, and top of each box represent the lower quartile, median, and upper quartile, respectively. The end points of the vertical lines represent the minimum and maximum values (l indicates the mean value). The composite MPR is the average of the individual drug MPRs.
more all-cause and CF-related hospitalizations and ED visits compared with high adherence (Table 3). Similarly, moderate levels of adherence were associated with
significantly more all-cause and CF-related hospitalizations compared with high adherence. Low adherence was also associated with more all-cause and CF-related
Figure 2 – CMPRs by age category. The bottom, midline, and top of each box represent the lower quartile, median, and upper quartile, respectively. The end points of the vertical lines represent the minimum and maximum values (l indicates the mean value). The CMPR is the average of the individual drug medication possession ratios.
journal.publications.chestnet.org
145
Figure 3 – CMPRs by number of medications prescribed. The bottom, midline, and top of each box represent the lower quartile, median, and upper quartile, respectively. The end points of the vertical lines represent the minimum and maximum values (l indicates the mean value). The CMPR is the average of the individual drug medication possession ratios.
acute care visits compared with high adherence, whereas moderate adherence was associated with more CFrelated acute care visits. There were no statistically significant differences in the frequency of outpatient visits between adherence groups. These analyses were stratified by age-group and repeated, and effects of similar magnitude were found (e-Tables 9-13).
associated with higher CF-related costs of $14,211 compared with high adherence (P , .001), and moderate adherence was associated with additional costs of $8,493 compared with high adherence (P 5 .10). During the second year of the study period, after adjusting for patient-level covariates, there were no statistically significant group differences in health-care costs.
Similar relationships were observed between first-year adherence levels and second-year health-care use, with low and moderate levels of adherence predicting significantly higher second-year frequencies of CF-related hospitalizations relative to high levels of adherence. Low adherence was also significantly associated with higher second-year rates of all-cause hospitalization and allcause ED visits vs high adherence, and low and moderate levels of adherence were associated with more all-cause and CF-related acute care visits vs high adherence. For both first- and second-year health-care use outcomes, unadjusted results were largely consistent with regression-adjusted findings (e-Table 14).
Discussion
In the first year, patients with low and moderate adherence incurred higher unadjusted all-cause health-care use costs compared with patients with high adherence (Table 4). CF-related health-care costs differed significantly between moderate- and high-adherence groups. In the regression-adjusted analysis, low adherence was 146 Original Research
This retrospective analysis used administrative claims data from a large, nationwide database to evaluate the relationship between overall long-term pulmonary medication adherence and demographic variables, healthcare use, and health-care costs among patients with CF. About one-half of the sampled patients filled less than one-half of their medications, which is consistent with adherence rates seen in other chronic illnesses.19 Adherence over a 1-year period, categorized as low (CMPR , 0.50), moderate (0.50 ⱕ CMPR , 0.80), and high (CMPR ⱖ 0.80), was concurrently and prospectively associated with health-care use. Rates of adherence decreased across age-groups from childhood through older adulthood. Low levels of adherence to long-term pulmonary medications were associated with significantly more all-cause and CF-related hospitalizations and ED visits relative to high adherence during both the first and the second years of the study; low
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
TABLE 3
] Health-care Use Stratified by Composite Pulmonary Medication Adherence Adjusted ERRb,c (95% CI)
Annual Event Rate per Patienta Health-care Use
Low CMPR
Moderate CMPR
High CMPR
Low vs High CMPR
Moderate vs High CMPR
…
…
During the first year after study index dated No. patients
1,675
949
663
All cause Hospitalizations 1 ED visits
1.44e
1.22e
0.99
1.40 (1.23-1.60)e
1.10 (0.95-1.27)
Hospitalizations
e
0.86
0.79
0.60
1.39 (1.19-1.62)
1.20 (1.02-1.42)e
ED visits
0.58e
0.43
0.39
1.40 (1.14-1.73)e
0.96 (0.76-1.21)
24.35
22.99
0.95 (0.89-1.02)
0.98 (0.91-1.06)
Outpatient visits
22.84
e
e
CF related Hospitalizations 1 ED visits
1.00e
0.94e
0.70
1.37 (1.18-1.58)e
1.22 (1.04-1.44)e
Hospitalizations
0.78e
0.76e
0.56
1.35 (1.15-1.57)e
1.25 (1.05-1.48)e
ED visits
0.22
0.18
0.14
1.44 (1.05-1.98)
1.16 (0.82-1.64)
12.71
0.92 (0.85-1.01)
1.04 (0.95-1.15)
…
…
Outpatient visits
e
12.29
14.85e
e
During the second year after study index datef No. patients
753
380
287
All cause Hospitalizations 1 ED visits
1.31e
1.22e
0.89
1.46 (1.19-1.79)e
1.26 (1.01-1.57)e
Hospitalizations
0.77
0.83e
0.62
1.29 (1.03-1.63)e
1.23 (0.95-1.58)
ED visits
0.54
0.39
0.28
1.80 (1.30-2.50)
1.29 (0.90-1.86)
25.07
24.75
23.04
1.02 (0.91-1.14)
0.95 (0.83-1.07)
Outpatient visits
e
e
CF related Hospitalizations 1 ED visits
0.83
0.99e
0.67
1.34 (1.07-1.69)e
1.44 (1.12-1.84)e
Hospitalizations
0.68
0.81
0.55
1.32 (1.04-1.68)
1.39 (1.07-1.80)e
ED visits
0.15
0.18
0.11
1.50 (0.87-2.56)
1.55 (0.86-2.77)
13.57
14.41
12.92
0.97 (0.84-1.12)
0.96 (0.82-1.13)
Outpatient visits
e
e
CMPR 5 composite medication possession ratio; ERR 5 event rate ratio; MPR 5 medication possession ratio. See Table 1 and 2 legends for expansion of other abbreviations. aPatients with CF were grouped into three categories based on their CMPRs, namely, low (, 0.50), moderate (0.50 ⱕ CMPR , 0.80), and high (0.80 ⱕ CMPR ⱕ 1.00). bERRs, 95% CIs, and P values were estimated using a negative binomial regression. ERR . 1 indicates that the incidence of the event is higher in patients with a low or moderate MPR than in those with a high MPR. cAdjusted ERR was estimated after adjusting for the following covariates: age, sex, comorbidities (Charlson comorbidity index, P aeruginosa, growth failure, depression, osteoporosis, sinusitis, asthma, gastritis/gastroesophageal reflux disease), index year, number of medications, and copayment at baseline. dAssociations between current annual health-care use and current annual CMPR were measured during the 365-d period following the index date. eP , .001 compared with the high-adherence group. fAssociations were measured between second-year annual health-care use and current-year CMPR.
adherence was also associated with higher concurrent health-care costs. A similar trend was observed for allcause and CF-related hospitalizations in patients with moderate adherence. In contrast, the frequency of outpatient visits was similar across adherence levels. Adherence to individual pulmonary medications, measured using MPR, has been reported previously in CF,5,7-9,12,15,17 albeit in much smaller samples. Consistent with these previous reports, patients in the present study had poor adherence to all pulmonary therapies, journal.publications.chestnet.org
with average drug-specific adherence ranging from 40% to 57%. However, because MPRs are based on pharmacy refill histories, these data should be considered an upper bound on patients’ true level of adherence. To the extent that patients may fill prescriptions but not take the medication, true adherence is likely to be even lower than the observed MPRs. The MPR across all pulmonary medications was poor and ranged from 0% to 100%. Subgroup analyses of adherence highlighted two correlates of higher adherence: younger age and greater 147
148 Original Research
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
5,465 (1,270-27,728)
11,274 (2,884-34,603)
33,055 ⫾ 65,837e
16,801 (4,717-50,722)
44,824 ⫾ 85,135e
380
9,036 (2,594-32,652)
33,210 ⫾ 73,861e
16,261 (4,767-43,425)
45,239 ⫾ 97,007e
949
Moderate CMPR, USD
6,062 (2,262-27,732)
24,003 ⫾ 49,620
9,863 (3,773-38,293)
37,067 ⫾ 81,845
287
5,695 (2,116-22,215)
23,525 ⫾ 54,879
9,733 (3,963-31,589)
34,432 ⫾ 77,088
663
High CMPR, USD
…
5,702 (23,662; 14,093)
…
7,610 (23,869; 20,102)
…
…
7,287 (557; 14,880)e
…
14,211 (5,557; 24,371)e
…
Low to High CMPR
…
3,959 (24,742; 13,195)
…
2,558 (29,633; 16,494)
…
…
7,464 (2123; 15,989)
…
8,493 (21,691; 19,709)
…
Moderate to High CMPR
Adjusted Cost Difference,b,c USD
Patients with CF were grouped into three categories based on their CMPR, namely, low (, 0.50), moderate (0.50 ⱕ CMPR , 0.80), and high (0.80 ⱕ CMPR ⱕ 1.00). Average health-care costs were compared between the cohorts using Wilcoxon rank sum tests. IQR 5 interquartile range. See Table 1-3 legends for expansion of other abbreviations. aTotal health-care costs 5 hospitalization 1 outpatient 1 ED costs. Costs were adjusted for inflation and expressed in 2011 USD. bAdjusted incremental cost differences between the two study cohorts were estimated using generalized linear models with a log link and g-distribution or two-part models, when appropriate, while adjusting for the following covariates: age, sex, comorbidities (Charlson comorbidity index, P aeruginosa, growth failure, depression, osteoporosis, sinusitis, asthma, gastritis/gastroesophageal reflux disease), index year, number of medications, and copayment at baseline. cP values were estimated using nonparametric bootstrap resampling techniques of 499 iterations. dAssociations between current annual health-care costs and current annual CMPR were measured during the 365-d period following the index date. eP , .001 compared with the high-adherence group. fAssociations were measured between second-year health-care costs and current-year CMPR.
Median (IQR)
Mean ⫾ SD
31,183 ⫾ 75,462
14,384 (3,651-47,292)
Median (IQR)
CF related
46,379 ⫾ 97,170
753
5,758 (1,458-29,703)
Mean ⫾ SD
All cause
No. patients
During the second year after study index datef
Median (IQR)
Mean ⫾ SD
36,605 ⫾ 107,876
13,947 (3,566-49,440)
Median (IQR)
CF related
54,190 ⫾ 143,267
1,675
Low CMPR, USD
Mean ⫾ SD
All cause
No. patients
During the first year after study index dated
a
] Health-care Costs Stratified by Composite Pulmonary Medication Adherence
Health-care Costs
TABLE 4
regimen complexity (ie, number of pulmonary medications). Adherence was notably higher among children relative to all other age groups, which is consistent with previous studies showing a decline in adherence during the transition from childhood to adolescence.12,20 Because the present study examined both children and adults with CF, it is the first in our knowledge to demonstrate no evidence of an uptick of adherence in adulthood. This observed difference in MPR may be attributable to parental supervision and assistance with adherence tasks for younger patients21 and reflects the challenges parents and care teams face in helping patients to assume full responsibility for their daily CF care. These results strongly suggest that engaging parents and other sources of social support (eg, spouses of married patients) be a high priority for interventions that improve adherence in patients with CF. Contrary to expectations, greater regimen complexity was associated with higher adherence. CMPR and all individual drug MPRs increased with the number of pulmonary medications dispensed. CF care providers may be hesitant to prescribe additional medications to patients who find their current regimen challenging to follow. Alternatively, a more complex regimen may be a proxy for illness severity, with sicker patients adhering more to therapies. Of course, patient willingness to accept additional treatments may be an indicator of other factors associated with better adherence, such as ability to afford drug copayments, motivation to integrate treatments into one’s routine, and desire to aggressively treat illnesses.22 The increasing costs of medications and copays are well-recognized barriers to adherence, including the willingness to initiate new therapy.23 Additional research is needed to understand the causal relationship between regimen complexity and adherence because it has important implications for decision-making in CF care. This study found that low adherence was associated with higher health-care use and costs in the concurrent and subsequent year, suggesting that interventions targeting improvements in adherence in CF may result in better health outcomes and reduced acute care use. Additional research is needed to identify cost-effective interventions that are successful in changing adherence behaviors in real-world clinical settings and to objectively measure these changes. For physicians using electronic health systems, e-prescribing records have been proposed to quantitatively assess adherence during routine outpatient visits.5 Pharmacy costs were not included
journal.publications.chestnet.org
because their expense may not offset the benefits of higher adherence. Future cost-effectiveness studies should look beyond financial costs and incorporate other outcomes of nonadherence, such as loss of productivity and quality of life. Given the observed large drop in adherence during adolescence and adulthood, effective strategies for promoting better adherence will likely involve collaborations among clinicians, patients, and families to identify and address barriers to adherence.21,22,24,25 Lack of time, forgetting, and embarrassment about taking medications in public have been cited as common barriers to adherence in CF.22,24 High rates of depression and anxiety have also been reported by both patients with CF and parent caregivers,26,27 with evidence that these psychologic symptoms affect adherence.13 Interventions that can help to overcome these barriers include reducing financial burden, improving physician-patient interactions, and using behavioral strategies, such as structured routines, positive reinforcement, and problem-solving.23,28-30 The fact that the number of outpatient visits was similar across all levels of adherence suggests that there may be opportunity to use outpatient encounters to provide routine, yet brief adherence counseling. This study was subject to the limitations inherent in observational claims-based research. Claims databases cannot provide critical information such as individually prescribed regimens; thus, we cannot provide an estimate of primary nonadherence (ie, a patient fails to obtain a single fill of a prescribed medication). Additionally, both oral and IV antibiotics pose a challenge for adherence calculations because they are prescribed for both short- and long-term use. Unfortunately, with no prescribed treatment plan there is no consistent pattern of dispensing that clarifies the intended use of the medication. Similarly, important demographic and clinical markers and outcomes are not available for use as covariates or health outcomes, such as socioeconomic status, lung function, and health-related quality of life. Thus, although the analyses adjusted for proxy measures of severity, such as baseline prescriptions and comorbidities, the observed associations could be confounded by other factors. In addition, MPR calculations summarize adherence over a long period and, thus, do not capture patients’ daily medication-taking patterns. A recent article demonstrated that children with CF were more adherent with their nebulized medications on weekdays than on 149
weekends and when school was in session than during vacation.31 As devices that monitor daily patterns of inhaled medication use become more available, the opportunities to understand adherence patterns will lead to more efficacious and tailored interventions. Finally, the MarketScan database does not access records for services reimbursed by state Medicaid programs or a secondary insurance plan not captured by MarketScan; consequently, the results may underestimate adherence and health-care use and costs for patients. The results also reflect MPRs among patients with private insurance, a marker of higher socioeconomic status, and may not be generalizable to other populations. Calculations of composite adherence as well as the observed associations between adherence and health-care use and costs should be
Acknowledgments Author contributions: J. S. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. A. L. Q., J. Z., M. M., P. A. C., J. S., and K. A. R. contributed to the study conception and design; A. L. Q., J. Z., M. M., P. A. C., J. S., and K. A. R. contributed to data interpretation; M. M., P. A. C., J. S., and Y. Y. contributed to data analysis; A. L. Q., M. M., P. A. C., and K. A. R. contributed to the writing of the manuscript; and A. L. Q., J. Z., M. M., P. A. C., J. S., Y. Y., and K. A. R. contributed to the review and final approval of the manuscript. Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Drs Quittner and Riekert were paid consultants to Novartis Pharmaceuticals Corp for this project. The terms of this arrangement are being managed by the University of Miami and The Johns Hopkins University in accordance with its conflict of interest policies. Dr Zhang is a current employee of Novartis Pharmaceuticals Corp. Mss Marynchenko, Chopra, and Yushkina and Dr Signorovitch are current employees of Analysis Group, Inc, which has received consultancy fees from Novartis Pharmaceuticals Corp. Role of sponsors: The study sponsor (Novartis Pharmaceuticals Corp) did not place any restrictions on statements made in the final version of the manuscript. Additional information: The e-Appendix and e-Tables can be found in the Supplemental Materials section of the online article.
References 1. Cystic Fibrosis Foundation Patient Registry. 2011 Annual Data Report.
150 Original Research
2.
3.
4.
5.
6.
7.
8.
9.
replicated in other representative samples to provide further insight.
Conclusions Achieving and maintaining adherence to long-term pulmonary therapy is critically important to the effective management of CF. This study found that rates of adherence to long-term CF pulmonary medications, using 12-month CMPRs, were significant predictors of subsequent health-care use and costs in a national sample of patients with CF. Adherence to long-term pulmonary medications was quite low (, 50%) and differed systematically across age and sex. Additional research is needed to understand the major causes of poor adherence in patients with CF and to develop effective strategies to improve adherence and, consequently, quality of life.
Bethesda, MD: Cystic Fibrosis Foundation; 2012. Sawicki GS, Ren CL, Konstan MW, Millar SJ, Pasta DJ, Quittner AL; Investigators and Coordinators of the Epidemiologic Study of Cystic Fibrosis. Treatment complexity in cystic fibrosis: trends over time and associations with site-specific outcomes. J Cyst Fibros. 2013;12(5):461-467. Mogayzel PJ Jr, Naureckas ET, Robinson KA, et al; Pulmonary Clinical Practice Guidelines Committee. Cystic fibrosis pulmonary guidelines. Chronic medications for maintenance of lung health. Am J Respir Crit Care Med. 2013;187(7):680-689. Pressler T, Henig NR, Malyszczak W. Aztreonam for inhalation solution, an aerosolized antibiotic for the treatment of pseudomonal airway infection in patients with cystic fibrosis. US Res Dis. 2011;7(2):107-112. Eakin MN, Bilderback A, Boyle MP, Mogayzel PJ, Riekert KA. Longitudinal association between medication adherence and lung health in people with cystic fibrosis. J Cyst Fibros. 2011;10(4):258-264. Sawicki GS, Rasouliyan L, McMullen AH, et al. Longitudinal assessment of healthrelated quality of life in an observational cohort of patients with cystic fibrosis. Pediatr Pulmonol. 2011;46(1):36-44. Burrows JA, Bunting JP, Masel PJ, Bell SC. Nebulised dornase alpha: adherence in adults with cystic fibrosis. J Cyst Fibros. 2002;1(4):255-259. Elkins MR, Robinson M, Rose BR, et al; National Hypertonic Saline in Cystic Fibrosis (NHSCF) Study Group. A controlled trial of long-term inhaled hypertonic saline in patients with cystic fibrosis. N Engl J Med. 2006;354(3):229-240. Modi AC, Lim CS, Yu N, Geller D, Wagner MH, Quittner AL. A multimethod assessment of treatment adher-
10.
11.
12.
13.
14.
15.
16.
17.
18.
ence for children with cystic fibrosis. J Cyst Fibros. 2006;5(3):177-185. Quinn J, Latchford G, Duff A, Conner M, Pollard K, Morrison L. Measuring, predicting and improving adherence to inhalation in patients with CF: a randomised controlled study of motivational interviewing [abstract]. Pediatr Pulmonol. 2004;38(S27):360. Modi AC, Quittner AL. Utilizing computerized phone diary procedures to assess health behaviors in family and social contexts. Child Health Care. 2006;35(1):29-45. Zindani GN, Streetman DD, Streetman DS, Nasr SZ. Adherence to treatment in children and adolescent patients with cystic fibrosis. J Adolesc Health. 2006;38(1):13-17. Smith BA, Modi AC, Quittner AL, Wood BL. Depressive symptoms in children with cystic fibrosis and parents and its effects on adherence to airway clearance. Pediatr Pulmonol. 2010;45(8):756-763. Modi AC, Cassedy AE, Quittner AL, et al. Trajectories of adherence to airway clearance therapy for patients with cystic fibrosis. J Pediatr Psychol. 2010;35(9):1028-1037. Briesacher BA, Quittner AL, Saiman L, Sacco P, Fouayzi H, Quittell LM. Adherence with tobramycin inhaled solution and health care utilization. BMC Pulm Med. 2011;11:5. Wertz DA, Chang CL, Stephenson JJ, Zhang J, Kuhn RJ. Economic impact of tobramycin in patients with cystic fibrosis in a managed care population. J Media Econ. 2011;14(6):759-768. Nasr SZ, Chou W, Villa KF, Chang E, Broder MS. Adherence to dornase alfa treatment among commercially insured patients with cystic fibrosis. J Media Econ. 2013;16(6):801-808. Hansen LG, Chang S. White Paper: Health Research Data for the Real World:
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
The MarketScan Databases. Ann Arbor, MI: Truven Health Analytics; 2012. 19. DiMatteo MR, Giordani PJ, Lepper HS, Croghan TW. Patient adherence and medical treatment outcomes: a metaanalysis. Med Care. 2002;40(9): 794-811. 20. Fitzgerald D. Non-compliance in adolescents with chronic lung disease: causative factors and practical approach. Paediatr Respir Rev. 2001;2(3):260-267. 21. Modi AC, Marciel KK, Slater SK, Drotar D, Quittner AL. The influence of parental supervision on medical adherence in adolescents with cystic fibrosis: developmental shifts from pre to late adolescence. Child Health Care. 2008;37(1):78-92. 22. George M, Rand-Giovannetti D, Eakin MN, Borrelli B, Zettler M, Riekert KA. Perceptions of barriers and facilitators: self-management decisions by older ado-
journal.publications.chestnet.org
23.
24.
25.
26.
27.
lescents and adults with CF. J Cyst Fibros. 2010;9(6):425-432. Eaddy MT, Cook CL, O’Day K, Burch SP, Cantrell CR. How patient cost-sharing trends affect adherence and outcomes: a literature review. P T. 2012;37(1):45-55. Modi AC, Quittner AL. Barriers to treatment adherence for children with cystic fibrosis and asthma: what gets in the way? J Pediatr Psychol. 2006;31(8):846-858. Abbott J, Havermans T, Hart A. Adherence to the medical regimen: clinical implications of new findings. Curr Opin Pulm Med. 2009;15(6):597-603. Riekert KA, Bartlett SJ, Boyle MP, Krishnan JA, Rand CS. The association between depression, lung function, and health-related quality of life among adults with cystic fibrosis. Chest. 2007;132(1):231-237. Goldbeck L, Besier T, Hinz A, Singer S, Quittner AL; TIDES Group. Prevalence of symptoms of anxiety and depression
in German patients with cystic fibrosis. Chest. 2010;138(4):929-936. 28. Zolnierek KB, Dimatteo MR. Physician communication and patient adherence to treatment: a meta-analysis. Med Care. 2009;47(8):826-834. 29. Kahana S, Drotar D, Frazier TJ. Metaanalysis of psychological interventions to promote adherence to treatment in pediatric chronic health conditions. J Pediatr Psychol. 2008;33(6): 590-611. 30. McDonald HP, Garg AX, Haynes RB. Interventions to enhance patient adherence to medication prescriptions: scientific review. JAMA. 2002;288(22):2868-2879. 31. Ball R, Southern KW, McCormack P, Duff AJ, Brownlee KG, McNamara PS. Adherence to nebulised therapies in adolescents with cystic fibrosis is best on week-days during school term-time. J Cyst Fibros. 2013;12(5):440-444.
151
Original Research Genetic and Developmental Disorders
]
Pulmonary Medication Adherence and Health-care Use in Cystic Fibrosis Alexandra L. Quittner, PhD; Jie Zhang, PhD; Maryna Marynchenko, MBA; Pooja A. Chopra, MS; James Signorovitch, PhD; Yana Yushkina, BA; and Kristin A. Riekert, PhD
Poor treatment adherence is common in cystic fibrosis (CF) and may lead to worse health outcomes and greater health-care use. This study evaluated associations of adherence to pulmonary medications, age, health-care use, and cost among patients with CF.
BACKGROUND:
METHODS: Patients with CF aged ⱖ 6 years were identified in a national commercial claims database. A 12-month medication possession ratio (MPR) was computed for each pulmonary medication and then averaged for a composite MPR (CMPR) for each patient. The CMPR was categorized as low (, 0.50), moderate (0.50-0.80), or high (ⱖ 0.80). Annual health-care use and costs were measured during the first and second year and compared across adherence categories by multivariable modeling.
Mean CMPR for the sample (N 5 3,287) was 48% ⫾ 31%. Age was inversely related to CMPR. In the concurrent year, more CF-related hospitalizations were observed among patients with low (event rate ratio [ERR], 1.35; 95% CI, 1.15-1.57) and moderate (ERR, 1.25; 95% CI, 1.05-1.48) vs high adherence; similar associations were observed for all-cause hospitalizations and CF-related and all-cause acute care use (hospitalizations 1 ED) in the concurrent and subsequent year. Rates of CF-related and all-cause outpatient visits did not differ by adherence. Low and moderate adherence predicted higher concurrent health-care costs by $14,211 ($5,557-$24,371) and $8,493 (2$1,691 to $19,709), respectively, compared with high adherence.
RESULTS:
Worse adherence to pulmonary medications was associated with higher acute health-care use in a national, privately insured cohort of patients with CF. Addressing adherence may reduce avoidable health-care use. CHEST 2014; 146(1):142-151
CONCLUSIONS:
Manuscript received August 16, 2013; revision accepted January 6, 2014; originally published Online First January 30, 2014. ABBREVIATIONS: CF 5 cystic fibrosis; CMPR 5 composite medication possession ratio; ICD-9-CM 5 International Classification of Diseases, Ninth Revision, Clinical Modification; MPR 5 medication possession ratio AFFILIATIONS: From the Department of Psychology (Dr Quittner), University of Miami, Coral Gables, FL; Novartis Pharmaceuticals Corp (Dr Zhang), East Hanover, NJ; Analysis Group, Inc (Mss Marynchenko, Chopra, and Yushkina and Dr Signorovitch), Boston, MA; and Division of Pulmonary and Critical Care Medicine (Dr Riekert), The Johns Hopkins University School of Medicine, Baltimore, MD.
142 Original Research
Parts of this study have been reported in abstract form at the 26th Annual North American Cystic Fibrosis Conference, October 11-13, 2012, Orlando, FL. FUNDING/SUPPORT: This study was supported by the Novartis Pharmaceuticals Corp, which manufactures inhaled tobramycin. CORRESPONDENCE TO: Alexandra L. Quittner, PhD, Department of Psychology, University of Miami, 5665 Ponce de Leon Blvd, Coral Gables, FL 33146; e-mail: [email protected] © 2014 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.13-1926
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
Recent advances in the treatment and aggressive management of cystic fibrosis (CF) have led to significant improvements in prognosis, with median predicted survival reaching 37 years.1 However, daily management requires a complex, time-consuming treatment regimen to address the multisystem effects of CF.2 Long-term use of one or more pulmonary medications is recommended,3 which can include inhaled antibiotics (eg, tobramycin, colistin, aztreonam) and oral azithromycin and nebulized mucolytic agents (eg, dornase alfa, hypertonic saline).3,4 The burden of treatment and poor adherence have emerged as critical issues in the daily management of CF.5,6 Prior analyses of objective adherence data have reported low rates of adherence to individual pulmonary medications among patients with CF, ranging from 31% to 79%.7-12 Most of these studies were limited by small sample sizes and limited age ranges. Furthermore, little is known about how rates of adherence to pulmonary medications vary across the life span from childhood through older adulthood, although there appears to be a significant decrease in adherence during adolescence.13,14 Adherence may also affect health-care use and costs. Two studies
Materials and Methods Data Source and Sample Selection The Thomson Reuters MarketScan Commercial Claims and Encounters Database (January 2005 to June 2011) includes medical and drug data for active employees and their dependents with primary coverage through employer-sponsored private health insurance throughout the United States.18 The MarketScan claims database offers the largest claims capture of physician office visits; hospital stays; retail, mail-order, and specialty pharmacies; and carve-out care. The database undergoes extensive validity and editing procedures to ensure high quality and consistency and has been presented in . 400 peer-reviewed articles. Patients included in this retrospective cohort study met the following criteria: (1) two or more independent CF diagnoses (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD9-CM], code 277.0X) during medical visits occurring ⱖ 30 days apart, (2) one or more prescription fills for a pulmonary medication (azithromycin, dornase alfa, hypertonic saline, or an inhaled antibiotic [aztreonam, colistin, or tobramycin]), and (3) continuous enrollment for ⱖ 180 days. For each patient meeting these criteria, the end date of the first 180-day continuous enrollment period and the end dates of subsequent 365-day periods were defined as potential index dates. An index date was randomly chosen among all potential index dates meeting the following additional inclusion criteria: (4) continuous enrollment for ⱖ 365 days after the index date, (5) a filled prescription for a pulmonary medication within the prior 90 days, and (6) age ⱖ 6 years. Measures For each long-term pulmonary medication filled ⱕ 90 days before the index date, adherence was calculated as the medication possession ratio (MPR), that is, the cumulative days’ supply of medication dispensed divided by 365 days after the index date (excluding hospitalized days). Values were truncated to 100%. Drug-specific MPRs were calculated for each pulmonary medication and then averaged to obtain a composite MPR (CMPR).
journal.publications.chestnet.org
found that worse adherence to inhaled tobramycin was associated with increased risk of hospitalization and health-care costs,15,16 whereas one-third found no association between dornase alfa adherence and inpatient respiratory exacerbations but a trend toward higher overall health-care charges.17 Lower overall pulmonary medication adherence (average adherence to dornase alfa, hypertonic saline, azithromycin, and inhaled tobramycin) was also associated with greater occurrence of pulmonary exacerbations requiring IV antibiotics and lung function testing in a retrospective medical record review of patients with CF.5 To our knowledge, no study has assessed the associations between overall adherence to long-term pulmonary medications and age, health-care use, and health-care costs in CF. Using retrospective claims data for privately insured patients with CF, this study measured drug-specific and overall adherence to pulmonary medications and assessed variations in overall adherence across demographic subgroups and their associations with healthcare use. We hypothesized that poorer overall medication adherence would be associated with older age and higher health-care use and costs.
CMPR was converted to a categorical variable as follows: low (, 0.50), moderate (0.5 ⱕ CMPR , 0.80), and high (ⱖ 0.80).5 Additional details for this calculation appear in e-Appendix 1 and e-Tables 1 and 2. Health-care use and costs were measured for two time periods to assess the prospective relationship with adherence: (1) the first year (365 days) after the index date to assess concurrent associations with adherence and (2) the second year after the index date (ie, 365 days after index date 1 365 days). Use included hospitalizations and ED, outpatient, and acute care (hospitalizations 1 ED) visits. Costs included the sum of hospitalizations, ED, and outpatient charges adjusted for inflation to 2011 US dollars. All-cause and CF-related health-care use and costs were evaluated; medical visits were considered CF related if associated with a CF diagnosis (ICD-9-CM code 277.0X). Pharmacy costs were not included. Baseline characteristics included demographics, index year, comorbidities, number of medications, and medication copayment. Comorbidities were identified on the basis of at least one claim associated with the corresponding ICD-9-CM code and required to be present during the 180-day baseline period. Statistical Analysis Mean CMPRs were compared across patient subgroups, including age, sex, and number of pulmonary medications dispensed ⱕ 90 days preceding the index date. Between-group differences were assessed with Wilcoxon rank sum tests. Health-care use was compared across CMPR levels using unadjusted and adjusted event rate ratios estimated from univariable and multivariable negative binomial regression models, respectively. Unadjusted differences in costs were tested with Wilcoxon rank sums. Multivariable generalized linear models with a g-distribution and a log link were used to estimate adjusted cost differences, with statistical significance assessed through nonparametric bootstrap resampling with 499 iterations. All multivariable models were adjusted for age, sex, comorbidities, index year, number of medications, and medication copayment at baseline. All analyses were conducted with SAS 9.2 (SAS Institute Inc) statistical software.
143
Results Patient Characteristics
The study sample included 3,287 patients with CF (Table 1). Mean age was 22.8 ⫾ 13.0 years, and female patients comprised 49% of the sample (Table 2). The demographic characteristics of the present sample were comparable to the general CF population as reported in the Cystic Fibrosis Registry1; however, the present sample included more adults (age ⱖ 18 years, 56% vs 48%) and more patients with Pseudomonas aeruginosa infection (63.0% vs 51%). A subgroup of 1,420 patients had continuous eligibility for 730 days following their index dates and, thus, were included in analyses of second-year health-care use and cost outcomes. Adherence to Pulmonary Medications
Mean MPR was highest for dornase alfa (57%) followed by inhaled tobramycin (51%), azithromycin (50%), inhaled aztreonam (47%), inhaled colistin (42%), and hypertonic saline (40%) (Fig 1). The average CMPR was 48%; 51% of patients had a CMPR , 0.50, whereas only 20% of patients had a CMPR ⱖ 0.80. MPRs for individual drugs were significantly correlated with CMPR, with Spearman r ranging from 0.70 for inhaled colistin to 0.86 for azithromycin (all P , .001). Age was related to CMPR, with the highest adherence among patients aged 6 to 10 years (59%) compared with the other age categories (all P , .001) (Figs 2-3). This finding was consistent across all drug-specific MPRs (e-Tables 3-8). Male patients had higher CMPRs than female patients (50% vs 45%, respectively; P , .001); however, post hoc analyses showed that the sex difference was statistically significant only for azithromycin
(P , .001) (e-Tables 3-8). Patients who filled more than one pulmonary medication had significantly higher adherence than those who filled one medication only (all P , .01) (mean CMPR, 41%, 51%, 56%, 60%, and 62% for one, two, three, four, and five medications filled, respectively). This association between adherence and CF prescription burden was replicated for drug-specific MPRs (e-Tables 3-8). Association With Health-care Use
In the regression-adjusted analysis of first-year healthcare use, low adherence was associated with significantly TABLE 2
] Baseline Characteristics
Characteristica
Value
No. patients
3,287
Demographics Female sex
1,604 (48.8)
Age category 6-10 y
528 (16.1)
11-17 y
900 (27.4)
18-25 y
775 (23.6)
26-35 y
511 (15.6)
ⱖ 36 y
573 (17.4)
Baseline comorbidities Charlson comorbidity index
0.7 ⫾ 1.0
Pseudomonas aeruginosa infection
2,071 (63.0)
Sinusitis
599 (18.2)
Diabetes mellitus
459 (14.0)
Asthma
442 (13.5)
Gastritis
186 (5.7)
Depression
181 (5.5)
Osteoporosis
153 (4.7)
Failure to thrive/growth failure
117 (3.6)
No. medications TABLE 1
] Sample Selection, 2005 to 2011
Inclusion Criterion Two or more CF diagnoses
CF pulmonary medicationsb
No. Patients 11,590
One or more prescriptions filled for a pulmonary medication
6,665
Continuous enrollment for ⱖ 180 d
6,459
Continuous enrollment for ⱖ 365 d following any potential index date
4,730
One or more prescriptions filled for a pulmonary medication within 90 d before any potential index date
3,565
Aged ⱖ 6 y as of any potential index date
3,287
CF 5 cystic fibrosis.
144 Original Research
2.1 ⫾ 1.0
Other CF-related medications
c
Total medications
2.0 ⫾ 1.5 10.1 ⫾ 5.6
Total medication copayment, 2011 USD/mo
107.1 ⫾ 202.7
Data are presented as No. (%) or mean ⫾ SD unless otherwise indicated. USD 5 US dollars. See Table 1 legend for expansion of other abbreviation. aThe baseline period is defined as the 180-d period prior to the index date. bIndicates treatment with one of the following pulmonary medications: inhaled tobramycin, inhaled colistin, inhaled aztreonam, oral azithromycin, nebulized dornase alfa, and nebulized hypertonic saline. cIndicates treatment with one or more of the following medications: leukotriene modifiers, inhaled bronchodilators, inhaled steroids, allergy medications, antidiabetic medications, digestives, vitamins, enzymes, ibuprofen, and prednisone.
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
Figure 1 – Mean MPRs for various long-term pulmonary medications used in cystic fibrosis. The bottom, midline, and top of each box represent the lower quartile, median, and upper quartile, respectively. The end points of the vertical lines represent the minimum and maximum values (l indicates the mean value). The composite MPR is the average of the individual drug MPRs.
more all-cause and CF-related hospitalizations and ED visits compared with high adherence (Table 3). Similarly, moderate levels of adherence were associated with
significantly more all-cause and CF-related hospitalizations compared with high adherence. Low adherence was also associated with more all-cause and CF-related
Figure 2 – CMPRs by age category. The bottom, midline, and top of each box represent the lower quartile, median, and upper quartile, respectively. The end points of the vertical lines represent the minimum and maximum values (l indicates the mean value). The CMPR is the average of the individual drug medication possession ratios.
journal.publications.chestnet.org
145
Figure 3 – CMPRs by number of medications prescribed. The bottom, midline, and top of each box represent the lower quartile, median, and upper quartile, respectively. The end points of the vertical lines represent the minimum and maximum values (l indicates the mean value). The CMPR is the average of the individual drug medication possession ratios.
acute care visits compared with high adherence, whereas moderate adherence was associated with more CFrelated acute care visits. There were no statistically significant differences in the frequency of outpatient visits between adherence groups. These analyses were stratified by age-group and repeated, and effects of similar magnitude were found (e-Tables 9-13).
associated with higher CF-related costs of $14,211 compared with high adherence (P , .001), and moderate adherence was associated with additional costs of $8,493 compared with high adherence (P 5 .10). During the second year of the study period, after adjusting for patient-level covariates, there were no statistically significant group differences in health-care costs.
Similar relationships were observed between first-year adherence levels and second-year health-care use, with low and moderate levels of adherence predicting significantly higher second-year frequencies of CF-related hospitalizations relative to high levels of adherence. Low adherence was also significantly associated with higher second-year rates of all-cause hospitalization and allcause ED visits vs high adherence, and low and moderate levels of adherence were associated with more all-cause and CF-related acute care visits vs high adherence. For both first- and second-year health-care use outcomes, unadjusted results were largely consistent with regression-adjusted findings (e-Table 14).
Discussion
In the first year, patients with low and moderate adherence incurred higher unadjusted all-cause health-care use costs compared with patients with high adherence (Table 4). CF-related health-care costs differed significantly between moderate- and high-adherence groups. In the regression-adjusted analysis, low adherence was 146 Original Research
This retrospective analysis used administrative claims data from a large, nationwide database to evaluate the relationship between overall long-term pulmonary medication adherence and demographic variables, healthcare use, and health-care costs among patients with CF. About one-half of the sampled patients filled less than one-half of their medications, which is consistent with adherence rates seen in other chronic illnesses.19 Adherence over a 1-year period, categorized as low (CMPR , 0.50), moderate (0.50 ⱕ CMPR , 0.80), and high (CMPR ⱖ 0.80), was concurrently and prospectively associated with health-care use. Rates of adherence decreased across age-groups from childhood through older adulthood. Low levels of adherence to long-term pulmonary medications were associated with significantly more all-cause and CF-related hospitalizations and ED visits relative to high adherence during both the first and the second years of the study; low
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
TABLE 3
] Health-care Use Stratified by Composite Pulmonary Medication Adherence Adjusted ERRb,c (95% CI)
Annual Event Rate per Patienta Health-care Use
Low CMPR
Moderate CMPR
High CMPR
Low vs High CMPR
Moderate vs High CMPR
…
…
During the first year after study index dated No. patients
1,675
949
663
All cause Hospitalizations 1 ED visits
1.44e
1.22e
0.99
1.40 (1.23-1.60)e
1.10 (0.95-1.27)
Hospitalizations
e
0.86
0.79
0.60
1.39 (1.19-1.62)
1.20 (1.02-1.42)e
ED visits
0.58e
0.43
0.39
1.40 (1.14-1.73)e
0.96 (0.76-1.21)
24.35
22.99
0.95 (0.89-1.02)
0.98 (0.91-1.06)
Outpatient visits
22.84
e
e
CF related Hospitalizations 1 ED visits
1.00e
0.94e
0.70
1.37 (1.18-1.58)e
1.22 (1.04-1.44)e
Hospitalizations
0.78e
0.76e
0.56
1.35 (1.15-1.57)e
1.25 (1.05-1.48)e
ED visits
0.22
0.18
0.14
1.44 (1.05-1.98)
1.16 (0.82-1.64)
12.71
0.92 (0.85-1.01)
1.04 (0.95-1.15)
…
…
Outpatient visits
e
12.29
14.85e
e
During the second year after study index datef No. patients
753
380
287
All cause Hospitalizations 1 ED visits
1.31e
1.22e
0.89
1.46 (1.19-1.79)e
1.26 (1.01-1.57)e
Hospitalizations
0.77
0.83e
0.62
1.29 (1.03-1.63)e
1.23 (0.95-1.58)
ED visits
0.54
0.39
0.28
1.80 (1.30-2.50)
1.29 (0.90-1.86)
25.07
24.75
23.04
1.02 (0.91-1.14)
0.95 (0.83-1.07)
Outpatient visits
e
e
CF related Hospitalizations 1 ED visits
0.83
0.99e
0.67
1.34 (1.07-1.69)e
1.44 (1.12-1.84)e
Hospitalizations
0.68
0.81
0.55
1.32 (1.04-1.68)
1.39 (1.07-1.80)e
ED visits
0.15
0.18
0.11
1.50 (0.87-2.56)
1.55 (0.86-2.77)
13.57
14.41
12.92
0.97 (0.84-1.12)
0.96 (0.82-1.13)
Outpatient visits
e
e
CMPR 5 composite medication possession ratio; ERR 5 event rate ratio; MPR 5 medication possession ratio. See Table 1 and 2 legends for expansion of other abbreviations. aPatients with CF were grouped into three categories based on their CMPRs, namely, low (, 0.50), moderate (0.50 ⱕ CMPR , 0.80), and high (0.80 ⱕ CMPR ⱕ 1.00). bERRs, 95% CIs, and P values were estimated using a negative binomial regression. ERR . 1 indicates that the incidence of the event is higher in patients with a low or moderate MPR than in those with a high MPR. cAdjusted ERR was estimated after adjusting for the following covariates: age, sex, comorbidities (Charlson comorbidity index, P aeruginosa, growth failure, depression, osteoporosis, sinusitis, asthma, gastritis/gastroesophageal reflux disease), index year, number of medications, and copayment at baseline. dAssociations between current annual health-care use and current annual CMPR were measured during the 365-d period following the index date. eP , .001 compared with the high-adherence group. fAssociations were measured between second-year annual health-care use and current-year CMPR.
adherence was also associated with higher concurrent health-care costs. A similar trend was observed for allcause and CF-related hospitalizations in patients with moderate adherence. In contrast, the frequency of outpatient visits was similar across adherence levels. Adherence to individual pulmonary medications, measured using MPR, has been reported previously in CF,5,7-9,12,15,17 albeit in much smaller samples. Consistent with these previous reports, patients in the present study had poor adherence to all pulmonary therapies, journal.publications.chestnet.org
with average drug-specific adherence ranging from 40% to 57%. However, because MPRs are based on pharmacy refill histories, these data should be considered an upper bound on patients’ true level of adherence. To the extent that patients may fill prescriptions but not take the medication, true adherence is likely to be even lower than the observed MPRs. The MPR across all pulmonary medications was poor and ranged from 0% to 100%. Subgroup analyses of adherence highlighted two correlates of higher adherence: younger age and greater 147
148 Original Research
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
5,465 (1,270-27,728)
11,274 (2,884-34,603)
33,055 ⫾ 65,837e
16,801 (4,717-50,722)
44,824 ⫾ 85,135e
380
9,036 (2,594-32,652)
33,210 ⫾ 73,861e
16,261 (4,767-43,425)
45,239 ⫾ 97,007e
949
Moderate CMPR, USD
6,062 (2,262-27,732)
24,003 ⫾ 49,620
9,863 (3,773-38,293)
37,067 ⫾ 81,845
287
5,695 (2,116-22,215)
23,525 ⫾ 54,879
9,733 (3,963-31,589)
34,432 ⫾ 77,088
663
High CMPR, USD
…
5,702 (23,662; 14,093)
…
7,610 (23,869; 20,102)
…
…
7,287 (557; 14,880)e
…
14,211 (5,557; 24,371)e
…
Low to High CMPR
…
3,959 (24,742; 13,195)
…
2,558 (29,633; 16,494)
…
…
7,464 (2123; 15,989)
…
8,493 (21,691; 19,709)
…
Moderate to High CMPR
Adjusted Cost Difference,b,c USD
Patients with CF were grouped into three categories based on their CMPR, namely, low (, 0.50), moderate (0.50 ⱕ CMPR , 0.80), and high (0.80 ⱕ CMPR ⱕ 1.00). Average health-care costs were compared between the cohorts using Wilcoxon rank sum tests. IQR 5 interquartile range. See Table 1-3 legends for expansion of other abbreviations. aTotal health-care costs 5 hospitalization 1 outpatient 1 ED costs. Costs were adjusted for inflation and expressed in 2011 USD. bAdjusted incremental cost differences between the two study cohorts were estimated using generalized linear models with a log link and g-distribution or two-part models, when appropriate, while adjusting for the following covariates: age, sex, comorbidities (Charlson comorbidity index, P aeruginosa, growth failure, depression, osteoporosis, sinusitis, asthma, gastritis/gastroesophageal reflux disease), index year, number of medications, and copayment at baseline. cP values were estimated using nonparametric bootstrap resampling techniques of 499 iterations. dAssociations between current annual health-care costs and current annual CMPR were measured during the 365-d period following the index date. eP , .001 compared with the high-adherence group. fAssociations were measured between second-year health-care costs and current-year CMPR.
Median (IQR)
Mean ⫾ SD
31,183 ⫾ 75,462
14,384 (3,651-47,292)
Median (IQR)
CF related
46,379 ⫾ 97,170
753
5,758 (1,458-29,703)
Mean ⫾ SD
All cause
No. patients
During the second year after study index datef
Median (IQR)
Mean ⫾ SD
36,605 ⫾ 107,876
13,947 (3,566-49,440)
Median (IQR)
CF related
54,190 ⫾ 143,267
1,675
Low CMPR, USD
Mean ⫾ SD
All cause
No. patients
During the first year after study index dated
a
] Health-care Costs Stratified by Composite Pulmonary Medication Adherence
Health-care Costs
TABLE 4
regimen complexity (ie, number of pulmonary medications). Adherence was notably higher among children relative to all other age groups, which is consistent with previous studies showing a decline in adherence during the transition from childhood to adolescence.12,20 Because the present study examined both children and adults with CF, it is the first in our knowledge to demonstrate no evidence of an uptick of adherence in adulthood. This observed difference in MPR may be attributable to parental supervision and assistance with adherence tasks for younger patients21 and reflects the challenges parents and care teams face in helping patients to assume full responsibility for their daily CF care. These results strongly suggest that engaging parents and other sources of social support (eg, spouses of married patients) be a high priority for interventions that improve adherence in patients with CF. Contrary to expectations, greater regimen complexity was associated with higher adherence. CMPR and all individual drug MPRs increased with the number of pulmonary medications dispensed. CF care providers may be hesitant to prescribe additional medications to patients who find their current regimen challenging to follow. Alternatively, a more complex regimen may be a proxy for illness severity, with sicker patients adhering more to therapies. Of course, patient willingness to accept additional treatments may be an indicator of other factors associated with better adherence, such as ability to afford drug copayments, motivation to integrate treatments into one’s routine, and desire to aggressively treat illnesses.22 The increasing costs of medications and copays are well-recognized barriers to adherence, including the willingness to initiate new therapy.23 Additional research is needed to understand the causal relationship between regimen complexity and adherence because it has important implications for decision-making in CF care. This study found that low adherence was associated with higher health-care use and costs in the concurrent and subsequent year, suggesting that interventions targeting improvements in adherence in CF may result in better health outcomes and reduced acute care use. Additional research is needed to identify cost-effective interventions that are successful in changing adherence behaviors in real-world clinical settings and to objectively measure these changes. For physicians using electronic health systems, e-prescribing records have been proposed to quantitatively assess adherence during routine outpatient visits.5 Pharmacy costs were not included
journal.publications.chestnet.org
because their expense may not offset the benefits of higher adherence. Future cost-effectiveness studies should look beyond financial costs and incorporate other outcomes of nonadherence, such as loss of productivity and quality of life. Given the observed large drop in adherence during adolescence and adulthood, effective strategies for promoting better adherence will likely involve collaborations among clinicians, patients, and families to identify and address barriers to adherence.21,22,24,25 Lack of time, forgetting, and embarrassment about taking medications in public have been cited as common barriers to adherence in CF.22,24 High rates of depression and anxiety have also been reported by both patients with CF and parent caregivers,26,27 with evidence that these psychologic symptoms affect adherence.13 Interventions that can help to overcome these barriers include reducing financial burden, improving physician-patient interactions, and using behavioral strategies, such as structured routines, positive reinforcement, and problem-solving.23,28-30 The fact that the number of outpatient visits was similar across all levels of adherence suggests that there may be opportunity to use outpatient encounters to provide routine, yet brief adherence counseling. This study was subject to the limitations inherent in observational claims-based research. Claims databases cannot provide critical information such as individually prescribed regimens; thus, we cannot provide an estimate of primary nonadherence (ie, a patient fails to obtain a single fill of a prescribed medication). Additionally, both oral and IV antibiotics pose a challenge for adherence calculations because they are prescribed for both short- and long-term use. Unfortunately, with no prescribed treatment plan there is no consistent pattern of dispensing that clarifies the intended use of the medication. Similarly, important demographic and clinical markers and outcomes are not available for use as covariates or health outcomes, such as socioeconomic status, lung function, and health-related quality of life. Thus, although the analyses adjusted for proxy measures of severity, such as baseline prescriptions and comorbidities, the observed associations could be confounded by other factors. In addition, MPR calculations summarize adherence over a long period and, thus, do not capture patients’ daily medication-taking patterns. A recent article demonstrated that children with CF were more adherent with their nebulized medications on weekdays than on 149
weekends and when school was in session than during vacation.31 As devices that monitor daily patterns of inhaled medication use become more available, the opportunities to understand adherence patterns will lead to more efficacious and tailored interventions. Finally, the MarketScan database does not access records for services reimbursed by state Medicaid programs or a secondary insurance plan not captured by MarketScan; consequently, the results may underestimate adherence and health-care use and costs for patients. The results also reflect MPRs among patients with private insurance, a marker of higher socioeconomic status, and may not be generalizable to other populations. Calculations of composite adherence as well as the observed associations between adherence and health-care use and costs should be
Acknowledgments Author contributions: J. S. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. A. L. Q., J. Z., M. M., P. A. C., J. S., and K. A. R. contributed to the study conception and design; A. L. Q., J. Z., M. M., P. A. C., J. S., and K. A. R. contributed to data interpretation; M. M., P. A. C., J. S., and Y. Y. contributed to data analysis; A. L. Q., M. M., P. A. C., and K. A. R. contributed to the writing of the manuscript; and A. L. Q., J. Z., M. M., P. A. C., J. S., Y. Y., and K. A. R. contributed to the review and final approval of the manuscript. Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Drs Quittner and Riekert were paid consultants to Novartis Pharmaceuticals Corp for this project. The terms of this arrangement are being managed by the University of Miami and The Johns Hopkins University in accordance with its conflict of interest policies. Dr Zhang is a current employee of Novartis Pharmaceuticals Corp. Mss Marynchenko, Chopra, and Yushkina and Dr Signorovitch are current employees of Analysis Group, Inc, which has received consultancy fees from Novartis Pharmaceuticals Corp. Role of sponsors: The study sponsor (Novartis Pharmaceuticals Corp) did not place any restrictions on statements made in the final version of the manuscript. Additional information: The e-Appendix and e-Tables can be found in the Supplemental Materials section of the online article.
References 1. Cystic Fibrosis Foundation Patient Registry. 2011 Annual Data Report.
150 Original Research
2.
3.
4.
5.
6.
7.
8.
9.
replicated in other representative samples to provide further insight.
Conclusions Achieving and maintaining adherence to long-term pulmonary therapy is critically important to the effective management of CF. This study found that rates of adherence to long-term CF pulmonary medications, using 12-month CMPRs, were significant predictors of subsequent health-care use and costs in a national sample of patients with CF. Adherence to long-term pulmonary medications was quite low (, 50%) and differed systematically across age and sex. Additional research is needed to understand the major causes of poor adherence in patients with CF and to develop effective strategies to improve adherence and, consequently, quality of life.
Bethesda, MD: Cystic Fibrosis Foundation; 2012. Sawicki GS, Ren CL, Konstan MW, Millar SJ, Pasta DJ, Quittner AL; Investigators and Coordinators of the Epidemiologic Study of Cystic Fibrosis. Treatment complexity in cystic fibrosis: trends over time and associations with site-specific outcomes. J Cyst Fibros. 2013;12(5):461-467. Mogayzel PJ Jr, Naureckas ET, Robinson KA, et al; Pulmonary Clinical Practice Guidelines Committee. Cystic fibrosis pulmonary guidelines. Chronic medications for maintenance of lung health. Am J Respir Crit Care Med. 2013;187(7):680-689. Pressler T, Henig NR, Malyszczak W. Aztreonam for inhalation solution, an aerosolized antibiotic for the treatment of pseudomonal airway infection in patients with cystic fibrosis. US Res Dis. 2011;7(2):107-112. Eakin MN, Bilderback A, Boyle MP, Mogayzel PJ, Riekert KA. Longitudinal association between medication adherence and lung health in people with cystic fibrosis. J Cyst Fibros. 2011;10(4):258-264. Sawicki GS, Rasouliyan L, McMullen AH, et al. Longitudinal assessment of healthrelated quality of life in an observational cohort of patients with cystic fibrosis. Pediatr Pulmonol. 2011;46(1):36-44. Burrows JA, Bunting JP, Masel PJ, Bell SC. Nebulised dornase alpha: adherence in adults with cystic fibrosis. J Cyst Fibros. 2002;1(4):255-259. Elkins MR, Robinson M, Rose BR, et al; National Hypertonic Saline in Cystic Fibrosis (NHSCF) Study Group. A controlled trial of long-term inhaled hypertonic saline in patients with cystic fibrosis. N Engl J Med. 2006;354(3):229-240. Modi AC, Lim CS, Yu N, Geller D, Wagner MH, Quittner AL. A multimethod assessment of treatment adher-
10.
11.
12.
13.
14.
15.
16.
17.
18.
ence for children with cystic fibrosis. J Cyst Fibros. 2006;5(3):177-185. Quinn J, Latchford G, Duff A, Conner M, Pollard K, Morrison L. Measuring, predicting and improving adherence to inhalation in patients with CF: a randomised controlled study of motivational interviewing [abstract]. Pediatr Pulmonol. 2004;38(S27):360. Modi AC, Quittner AL. Utilizing computerized phone diary procedures to assess health behaviors in family and social contexts. Child Health Care. 2006;35(1):29-45. Zindani GN, Streetman DD, Streetman DS, Nasr SZ. Adherence to treatment in children and adolescent patients with cystic fibrosis. J Adolesc Health. 2006;38(1):13-17. Smith BA, Modi AC, Quittner AL, Wood BL. Depressive symptoms in children with cystic fibrosis and parents and its effects on adherence to airway clearance. Pediatr Pulmonol. 2010;45(8):756-763. Modi AC, Cassedy AE, Quittner AL, et al. Trajectories of adherence to airway clearance therapy for patients with cystic fibrosis. J Pediatr Psychol. 2010;35(9):1028-1037. Briesacher BA, Quittner AL, Saiman L, Sacco P, Fouayzi H, Quittell LM. Adherence with tobramycin inhaled solution and health care utilization. BMC Pulm Med. 2011;11:5. Wertz DA, Chang CL, Stephenson JJ, Zhang J, Kuhn RJ. Economic impact of tobramycin in patients with cystic fibrosis in a managed care population. J Media Econ. 2011;14(6):759-768. Nasr SZ, Chou W, Villa KF, Chang E, Broder MS. Adherence to dornase alfa treatment among commercially insured patients with cystic fibrosis. J Media Econ. 2013;16(6):801-808. Hansen LG, Chang S. White Paper: Health Research Data for the Real World:
[
1 4 6 # 1 C H E S T J U LY 2 0 1 4
]
The MarketScan Databases. Ann Arbor, MI: Truven Health Analytics; 2012. 19. DiMatteo MR, Giordani PJ, Lepper HS, Croghan TW. Patient adherence and medical treatment outcomes: a metaanalysis. Med Care. 2002;40(9): 794-811. 20. Fitzgerald D. Non-compliance in adolescents with chronic lung disease: causative factors and practical approach. Paediatr Respir Rev. 2001;2(3):260-267. 21. Modi AC, Marciel KK, Slater SK, Drotar D, Quittner AL. The influence of parental supervision on medical adherence in adolescents with cystic fibrosis: developmental shifts from pre to late adolescence. Child Health Care. 2008;37(1):78-92. 22. George M, Rand-Giovannetti D, Eakin MN, Borrelli B, Zettler M, Riekert KA. Perceptions of barriers and facilitators: self-management decisions by older ado-
journal.publications.chestnet.org
23.
24.
25.
26.
27.
lescents and adults with CF. J Cyst Fibros. 2010;9(6):425-432. Eaddy MT, Cook CL, O’Day K, Burch SP, Cantrell CR. How patient cost-sharing trends affect adherence and outcomes: a literature review. P T. 2012;37(1):45-55. Modi AC, Quittner AL. Barriers to treatment adherence for children with cystic fibrosis and asthma: what gets in the way? J Pediatr Psychol. 2006;31(8):846-858. Abbott J, Havermans T, Hart A. Adherence to the medical regimen: clinical implications of new findings. Curr Opin Pulm Med. 2009;15(6):597-603. Riekert KA, Bartlett SJ, Boyle MP, Krishnan JA, Rand CS. The association between depression, lung function, and health-related quality of life among adults with cystic fibrosis. Chest. 2007;132(1):231-237. Goldbeck L, Besier T, Hinz A, Singer S, Quittner AL; TIDES Group. Prevalence of symptoms of anxiety and depression
in German patients with cystic fibrosis. Chest. 2010;138(4):929-936. 28. Zolnierek KB, Dimatteo MR. Physician communication and patient adherence to treatment: a meta-analysis. Med Care. 2009;47(8):826-834. 29. Kahana S, Drotar D, Frazier TJ. Metaanalysis of psychological interventions to promote adherence to treatment in pediatric chronic health conditions. J Pediatr Psychol. 2008;33(6): 590-611. 30. McDonald HP, Garg AX, Haynes RB. Interventions to enhance patient adherence to medication prescriptions: scientific review. JAMA. 2002;288(22):2868-2879. 31. Ball R, Southern KW, McCormack P, Duff AJ, Brownlee KG, McNamara PS. Adherence to nebulised therapies in adolescents with cystic fibrosis is best on week-days during school term-time. J Cyst Fibros. 2013;12(5):440-444.
151
