- Email: [email protected]
Cardiovascular Medication Utilization and Adherence Among Heart Failure Patients in Rural and Urban Areas: A Retrospective Cohort Study
Canadian Journal of Cardiology 31 (2015) 341e347
Clinical Research
Cardiovascular Medication Utilization and Adherence Among Heart Failure Patients in Rural and Urban Areas: A Retrospective Cohort Study Gaetanne K. Murphy, MSc,a Finlay A. McAlister, MD, MSc,b and Dean T. Eurich, PhDa a b
School of Public Health, University of Alberta, Edmonton, Alberta, Canada
Division of General Internal Medicine, Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada
ABSTRACT
RESUM E
Background: Rural residence is a negative prognostic factor for heart failure (HF). The objective was to explore rural and urban differences in the utilization, adherence, and persistence with medications, and mortality among incident HF patients. Methods: Using administrative databases from Alberta (Canada), subjects > 65 years old with a first hospitalization for HF between 1999 and 2008 who survived 90 days after discharge were identified. Pharmacy claims for renin-angiotensin system (RAS) agents, b-blockers (BBs), digoxin, or spironolactone were identified. The association between rural and urban residence and medication utilization, adherence (optimal adherence defined as 80% adherence over 1 year), persistence, and 1-year mortality was assessed. Results: The cohort included 10,430 patients, with a mean age of 80.2 (SD, 7.7) years, 47% were male, and 25% were rural residents. Rural residents were less likely to receive RAS agents (74% vs 79%, adjusted odds ratio [aOR], 0.78; 95% confidence interval [CI], 0.690.89) or BBs (44% vs 54%; aOR, 0.83; 95% CI, 0.73-0.93) than urban residents, but had similar use of other medications. Although < 69% of patients who received RAS agents and 53% who received BBs had optimal adherence, few differences in adherence or persistence were detected among patients in rural vs urban areas. The 1-year mortality rate was significantly lower for patients who demonstrated optimal adherence to RAS agents or BBs (aOR, 0.78; 95% CI, 0.65-0.94) with no significant differences in the first 6 months between patients residing in rural vs urban areas.
sidence en zone rurale est un facteur pronostique Introduction : La re gatif de l’insuffisance cardiaque (IC). L’objectif e tait d’examiner les ne rences entre le milieu rural et le milieu urbain quant à l’utilisation, diffe ve rance dans la prise des me dicaments, ainsi l’observance et la perse chez les patients nouvellement atteints d’IC. que la mortalite thodes : À l’aide des banques de donne es administratives de Me l’Alberta (Canada), les sujets > 65 ans ayant eu une première hospitalisation pour une IC entre 1999 et 2008 qui survivaient 90 jours e taient releve s. Les demandes de remboursement pour après le conge nine-angiotensine (SRA), des bdes agents agissant sur le système re taient releve es. bloquants (BB), de la digoxine ou de la spironolactone e sidence en zone rurale et la re sidence en zone L’association entre la re dicaments, de l’observance urbaine pour ce qui est de l’utilisation des me finie comme e tant une observance 80 % (observance optimale de ve rance et de la mortalite à 1 an e tait e value e. durant 1 an), de la perse sultats : La cohorte de 10 430 patients dont l’âge moyen e tait de Re cart-type, 7,7) ans comptait 47 % d’hommes et 25 % de re sidents 80,2 (e tait moins probable que les re sidents de zones de zones rurales. Il e rurales reçoivent des agents agissant sur le SRA (74 % vs 79 %, ratio ajuste [RIAa], 0,78; intervalle de confiance [IC] à d’incidence approche 95 %, 0,69-0,89) ou des BB (44 % vs 54 %; RIAa, 0,83; IC à 95 %, 0,73sidents de zones urbaines, mais l’utilisation e tait simi0,93) que les re dicaments. Bien que < 69 % des patients qui laire pour les autres me recevaient des agents agissant sur le SRA et 53 % des patients qui recevaient des BB montraient une observance optimale, peu de
Angiotensin-converting enzyme inhibitors (ACEIs), angiotensin-receptor blockers (ARBs), hydralazine with longacting nitrates, b-blockers (BBs), and spironolactone have all been shown to decrease morbidity and mortality in patients with heart failure (HF).1 Despite this evidence, these medications are often underutilized because of a failure to initiate, persist with, or adhere to therapy.2-4
In patients with HF, medication nonadherence alters the clinical status of patients and is associated with poor outcomes.1,2,5 Although patient-related factors, such as resources, knowledge, and attitudes are often the focus of adherence, these represent just one dimension that affects adherence behaviour. There is increasing focus on the effect of geography on utilization and adherence to drug therapies and subsequent outcomes. Previous studies have observed rural vs urban variations in HF outcomes6 that might, at least partially, be related to barriers such as social isolation, financial constraints, lower education, limited health care facilities, distance to care, physician shortages, and lack of access to specialist care.7,8 Although it has been speculated that these barriers would cause harm through primary underuse of evidence-based
Received for publication August 26, 2014. Accepted November 25, 2014. Corresponding author: Dr Dean T. Eurich, 2-040 Li Ka Shing Center for Health Research Innovation, University of Alberta, Edmonton, Alberta T6G 2E1, Canada. Tel.: þ1-780-492-6333; fax: þ1-780-492-7455. E-mail: [email protected] See page 346 for disclosure information.
http://dx.doi.org/10.1016/j.cjca.2014.11.024 0828-282X/Ó 2015 Canadian Cardiovascular Society. Published by Elsevier Inc. All rights reserved.
342
Canadian Journal of Cardiology Volume 31 2015
Conclusions: Rural residents with HF were less likely to receive RAS agents or BBs, but few differences in adherence were noted compared with their urban counterparts. Suboptimal adherence with evidencebased HF therapy was associated with increased risk of mortality.
rences dans l’observance ou la perse ve rance e taient de tecte es diffe entre les patients des zones rurales vs les patients des zones urbaines. à 1 an e tait significativement plus faible chez les patients La mortalite montraient une observance optimale du traitement par les qui de agents agissant sur le SRA ou les BB (RIAa, 0,78; IC à 95 %, 0,65rence significative entre les patients re sidant en 0,94), et ce, sans diffe sidant en zone urbaine au cours des 6 zone rurale vs les patients re premiers mois. tait moins probable que les re sidents des zones Conclusions : Il e rurales qui souffrent d’IC reçoivent des agents agissant sur le SRA ou rences dans l’observance e taient note es des BB, mais peu de diffe gions urbaines. Une comparativement à leurs homologues des re sur des donne es observance sous-optimale du traitement de l’IC fonde tait associe e à l’augmentation du risque de mortalite . probantes e
medications, it is possible that these barriers might also negatively affect patient adherence to therapies, although data to support this premise are lacking. Previous studies have shown that lower levels of adherence to evidence-based medications among HF patients is associated with increased risk of hospitalization and death.9,10 However, we are not aware of any studies that have evaluated potential adherence differences according to rural vs urban residence. Thus, this study was undertaken to determine if, among newly diagnosed HF patients, the utilization, adherence, and persistence with HF-specific medications differs between those living in rural and urban areas, and whether these medication use patterns could explain 1-year mortality differences.
code of 428.x or I50.x).12,13 To ensure incident HF, we excluded those with a HF-related hospitalization (ICD 9/10 HF claim code in any diagnosis field) within 5 years before the index hospitalization. Because we were primarily interested in medication usage after diagnosis, to ensure subjects had sufficient opportunity to initiate therapies, we excluded all patients who died within 90 days of hospital discharge. In addition, those with an index hospitalization length of stay > 365 days, missing postal code data, and patients with no claims in any of the databases during the follow-up period were also excluded. All patients were followed until death or 365 days after discharge from their index HF hospitalization. As per previous methodology used by Statistics Canada and others,6,14 rural and urban residence was determined according to the postal code of each patient’s home address obtained from the registry file.
Methods Setting A population-based cohort of patients with incident HF was assembled using deidentified administrative databases from Alberta Health. Alberta Health manages a single-payer government-funded health care system that provides universal access to hospital, emergency department, and physician services for all 3.7 million residents within the province of Alberta, Canada. Patient data from 5 demographic, vital statistic, and health care utilization databases were linked as described previously.11 Briefly, the Canadian Institute for Health Information Discharge Abstract Database supplied data on hospital admission dates, most responsible diagnosis, and up to 24 secondary diagnoses. The Alberta Blue Cross Medication Database provided outpatient prescription drug utilization data for all Alberta residents 65 years of age or older. The Alberta Health Care Insurance Plan Registry file provided demographic and vital statistics data, and the Ambulatory Care and Practitioner Claims Databases were used to obtain information on emergency department visits and office-based physician visits. The Health Ethics Research Board at the University of Alberta approved this study (Pro00033827). Study cohort We identified all subjects > 65 years of age who were discharged from hospital between April 1, 1999 and December 31, 2008 with the most responsible diagnosis of HF (International Classification of Diseases [ICD] 9 or 10
Outcomes The coprimary outcomes were the proportion of patients utilizing and adherent to HF-related medications. Utilization of HF-related medications was defined as at least 1 dispensation 7 days before and up to 1 year after discharge. Seven days before discharge was used to capture new prescriptions that might have been filled during the hospital stay or while patients were transitioning to out-of-hospital care. Adherence was defined as proportion of days with medication coverage (PDC) 0.8 for the medications of interest (ACEI, ARB, BB, digoxin, and spironolactone). By convention, the PDC was calculated as the sum of days with medication available divided by the total days of follow-up for patients with a minimum of 2 dispensations for that drug class.15 The primary analysis assumed medication was available on the dispensation date and for the estimated number of days supplied, based on the quantity dispensed and usual dosing frequency. An 80% adherence level has been associated with a reduced risk of death in HF5 and is the threshold commonly used in studies of cardiovascular medication adherence. Secondarily we evaluated medication persistence at 1 year. Persistence was defined as the availability of medication for at least 1 day during a 30-day window, at 3, 6, or 12 months after the first dispensation. Patients who died before the end of follow-up had their last observation carried forward. In addition, we evaluated whether adherence to HF-related medications within the first 6 months of discharge was associated with mortality in the subsequent 6 months. All outcomes were
Murphy et al. Rural and Urban Medication Use for HF Cohort
estimated for each drug class separately and for the reninangiotensin system (RAS) agents combined (ACEI or ARB). Statistical analysis Differences in baseline characteristics between urban and rural residents were compared using c2, Student t, or MannWhitney U tests, as appropriate. Multivariable logistic regression was used to evaluate the association between rural vs urban residence according to the proportion treated, adherent, and persistent to each drug class. For all-cause mortality, multivariable logistic regression was used to assess the effect of optimal adherence, defined as a BB or RAS agent available 80% of days in the first 6 months after discharge, on mortality for the subsequent 6 months, after adjustment for all other covariates. Covariates included in the models included variables known to be associated with medication adherence and clinical outcomes in patients with HF. To evaluate model overfitting, Hosmer-Lemeshow goodness of fit tests were conducted on all models with no obvious violations noted. In addition, we included the interaction between rural vs urban residence and adherence to determine any potential rural vs urban differences according to adherence on 1-year mortality. Covariates included age, sex, median neighbourhood income quartile, comorbidities (myocardial infarction, angina, diabetes, cerebrovascular disease, hypertension, valvular disease, arrhythmias, chronic pulmonary disease, peripheral vascular disease, neoplasm, dementia, peptic ulcer, or renal disease based on the index hospital admission and any admissions in the previous year), year of index hospitalization (ie, year of diagnosis), number of hospitalizations, physician or emergency department visits in the year before the index event, physician specialty (general practitioner, internal medicine, or cardiologist), and medications dispensed in the 90 days before the index hospitalization (BB, RAS, digoxin, statin, spironolactone, loop diuretic, amiodarone, or warfarin).12 First-order interactions between rural and urban status and age and sex were tested for all outcomes and no clinically important interactions were identified. The median neighbourhood income was estimated using postal codes and neighbourhood-level census data.11 All analyses were done using STATA version 12.1 (StataCorp LP, College Station, TX). To evaluate the robustness of our study results, we undertook several ancillary analyses. First, we adjusted the PDC for: (1) hospitalization (total days in hospital during follow-up were added to the days with drug coverage); (2) overlapping days supply (dispensation date of overlapping dispensations were moved forward to the end of supply date of the previous dispensation); and (3) medication discontinuation (follow-up was truncated at the date of the last dispensation for those who discontinued therapyddefined as a 30-day gap with no medication available, and no further dispensations). Second, analyses using alternate thresholds to define adherence (PDC 0.7 or 0.9) were conducted because studies9,10 have linked higher and lower adherence rate cut points to HFrelated outcomes. Third, we included all patients who were alive for at least 30 days after the index hospitalization. Fourth, we excluded early dispensations (those occurring within 7 days before discharge from the index hospitalization) from our analyses. Fifth, we repeated our mortality analysis using a Cox proportional hazards framework to account for
343
time to event. Last, we evaluated “new users” of our medications of interest by restricting analyses to those who had no dispensations for HF-related medications in the 90 days before the index hospitalization. Results Of 23,767 potentially eligible patients, we identified 10,430 patients with an incident HF hospitalization between April 1, 1999 and December 31, 2008 who met the cohort inclusion criteria. The reasons for exclusion were: previous HF-related hospitalization (n ¼ 8014); age < 66 years (n ¼ 2934), death during the index hospitalization (n ¼ 1417) or within 90 days of discharge (n ¼ 928), no health care claims during follow-up (n ¼ 40), length of stay > 365 days (n ¼ 3), and missing postal code (n ¼ 1). Mean age was 80.2 years (SD, 7.7), 4909 (47%) were male, and 2580 (25%) were rural residents (Table 1). At baseline, rural residents were younger, more likely to be female, had fewer comorbidities, lower neighbourhood income, had fewer outpatient physician visits (especially with specialists), but more hospitalizations and emergency department visits in the year before HF diagnosis compared with urban patients. In the 90 days before the index hospitalization, more urban than rural patients filled a prescription for a RAS agent or BB. Among the cohort who survived at least 90 days after the index hospitalization, the average follow-up was 343 days (SD, 61) days, and 359 (13.9%) rural and 1197 (15.2%) urban residents died (adjusted odds ratio [aOR], 0.95; 95% confidence interval [CI], 0.82-1.09). More rural patients were hospitalized (for any cause) during follow-up with 23% of rural patients hospitalized once, and 37% hospitalized 2 or more times, compared with 25% and 29% of urban patients, respectively. Medication treatment Overall, 8072 (77%) filled at least 1 prescription for a RAS agent (Table 2) with 43% of RAS users considered new users (ie, no RAS dispensations 90 days before the index hospitalization). A total of 5368 (51%) patients received a BB (47% were new users); most (n ¼ 4582; 44%) were prescribed bisoprolol, carvedilol, or metoprolol. Overall, 4706 (45%) filled prescriptions for a RAS agent and a BB concurrently. Although most patients (8362; 80%) filled at least 1 prescription for a loop diuretic, other HF-related medications were used in less than one-third of this cohort. Most patients filled their first prescription within 90 days of discharge (ACEI 92%, ARB 79%, BB 87%, digoxin 87%, spironolactone 79%). Rural patients were less likely to fill prescriptions for RAS agents (aOR, 0.78; 95% CI, 0.69-0.89), BB (aOR, 0.83; 95% CI, 0.73-0.93), or RAS agent and BB (aOR, 0.78; 95% CI, 0.70-0.88) in the first year after discharge than were urban residents, but had similar rates of use of loop diuretics, digoxin, or spironolactone. Medication adherence The proportion of patients who were adherent (PDC 0.8) was lowest for spironolactone (rural 56%, urban 50%; aOR, 1.18; 95% CI, 0.95-1.45) and highest for RAS agents
344
Canadian Journal of Cardiology Volume 31 2015
Table 1. Baseline characteristics Rural Characteristic Total (total n ¼ 10,430) Age, years* Male sex Median neighbourhood household income < $50,149 $50,150-$69,434 $69,435-$89,005 $89,006 Missing Drug treatment 90 days before index hospitalization RAS agents ACEI ARB BB Amiodarone Digoxin Hydralazine Long-acting nitrates Loop diuretic Spironolactone Statin Warfarin Number dispensations (any drug)y Number of drugsy Healthcare services in previous year General practitioner visit Internal medicine visit Cardiologist visit Number of physician visitsy Number of hospitalizationsy Number of emergency department visitsy Comorbidities Charlson comorbidity score* Ischemic heart disease Myocardial infarction Angina Diabetes Cerebrovascular disease Hypertension Valvular disease Cardiac arrhythmias Chronic pulmonary disease Peripheral vascular disease Neoplasm Dementia Peptic ulcer disease Renal disease
Urban
n
%
n
%
2580 79.8 (7.9) 1281
24.7
75.3
49.7
7850 80.3 (7.3) 3628
1077 912 326 84 181
41.7 35.3 12.6 3.3 7.0
2998 1841 1528 1188 295
38.2 23.5 19.5 15.1 3.8
1169 923 321 707 57 326 10 269 941 140 466 415 8 (11) 6 (5)
45.3 35.8 12.4 27.4 2.2 12.6 0.4 10.4 36.5 5.4 18.1 16.1
3826 2881 1204 2595 200 976 45 1011 2916 373 1687 1558 8 (10) 6 (6)
48.7 36.7 15.3 33.1 2.5 12.4 0.6 12.9 37.1 4.8 21.5 19.8
2541 1130 412 16 (16) 0 (1) 2 (3)
98.5 43.8 16.0
7686 5618 2899 19 (17) 0 (1) 2 (2)
97.9 71.6 36.9
2.3 (1.5) 840 388 158 755 142 1070 267 782 719 74 141 122 16 225
32.6 15.0 6.1 29.3 5.5 41.5 10.3 30.3 27.9 2.9 5.5 4.7 0.6 8.7
2.6 (1.6) 3614 1710 571 2399 552 4352 1558 3214 2246 345 371 555 73 1075
46.2
46.0 21.8 7.3 30.6 7.0 55.4 19.8 40.9 28.6 4.4 4.7 7.1 0.9 13.7
P 0.005 0.002 < 0.001
<
0.002 0.4 0.001 0.001 0.34 0.79 0.26 0.001 0.54 0.17 0.001 0.001 0.30 0.001
< < < < <
0.065 0.001 0.001 0.001 0.001 0.001
< <
< <
< 0.001 < 0.001 < 0.001 0.047 0.21 0.007 < 0.001 < 0.001 < 0.001 0.47 0.001 0.13 < 0.001 0.14 < 0.001
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, b-blocker; RAS, renin-angiotensin system. * Mean (SD). y Median (interquartile range).
(rural 66%, urban 70%; aOR, 0.88; 95% CI, 0.78-1.00; P ¼ 0.049; Table 3). Rural patients were less likely to be adherent to RAS agents than urban patients, although the differences were small. For BB, digoxin, and spironolactone, no significant rural vs urban differences in adherence were observed in adjusted analyses.
(Supplemental Fig. S1). The differences between rural and urban patients were not significant except for ARBs at 3 months (77% vs 80%; aOR, 0.75; 95% CI, 0.57-0.98) and RAS agents at 12 months (74% vs 77%; aOR, 0.86; 95% CI, 0.75-0.98) in the adjusted analyses; however, the clinical importance of these differences is questionable.
Medication persistence
Mortality
Persistence declined after discharge with the greatest reduction occurring in the first 3 months after initiating treatment. By 1 year, or until death or censoring, persistence was highest for RAS agents (76%), lowest for spironolactone (57%), and similar (approximately 70%) for BB or digoxin
Of the 7911 patients who survived 180 days after discharge, 158 (8.5%) of rural and 538 (8.9%) of urban residents died by 1 year (aOR, 1.04; 95% CI, 0.85-1.28). Patients who demonstrated optimal adherence (had a BB, ACEI, or ARB available for at least 80% of days) within the
Murphy et al. Rural and Urban Medication Use for HF Cohort
345
Table 2. Evidence-based medication utilization during follow-up Rural Drug class RAS agent ACEI ARB RAS agent or hydralazine with longacting nitrate BB BB and RAS agent Digoxin Spironolactone
Urban
Rural vs urban, unadjusted OR (95% CI)
Rural vs urban, adjusted OR (95% CI)
P
P
n
%
n
%
1903 1627 494 1907
73.8 63.1 19.1 73.9
6169 5236 1764 6197
78.6 66.7 22.5 78.9
0.77 0.85 0.82 0.76
(0.69-0.85) (0.78-0.94) (0.73-0.91) (0.68-0.84)
< 0.001 0.001 < 0.001 < 0.001
0.78 0.84 0.93 0.78
(0.69-0.89) (0.75-0.93) (0.82-1.05) (0.69-0.88)
< 0.001 0.001 0.24 < 0.001
1124 953 758 688
43.6 36.9 29.4 26.7
4244 3753 2323 1979
54.1 47.8 29.6 25.2
0.66 0.64 0.99 1.08
(0.60-0.72) (0.58-0.70) (0.90-1.09) (0.98-1.19)
< 0.001 < 0.001 0.84 0.14
0.83 0.78 0.98 1.10
(0.73-0.93) (0.70-0.88) (0.87-1.11) (0.98-1.23)
0.001 < 0.001 0.76 0.12
Total patients receiving drug therapy, n ¼ 10,430. ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, b-blocker; CI, confidence interval; OR, odds ratio; RAS, reninangiotensin system.
first 6 months had a lower risk of death at 1 year than those who were suboptimally adherent (aOR, 0.78; 95% CI, 0.650.94). We did not detect any significant differences in 1-year mortality for rural vs urban patients who were adherent (aOR, 1.09; 95% CI, 0.86-1.38) or who were nonadherent (aOR, 0.95; 95% CI, 0.66-1.35; P ¼ 0.50 for interaction). Sensitivity analyses Analyses accounting for time in hospital, early discontinuation, overlapping supply, excluding early dispensations (in 7 days before discharge from the index hospitalization), alternate adherence definitions ( 0.7 or 0.9), including all patients who survived at least 30 days after discharge, and analyses restricted to “new users,” were generally consistent with our primary analysis (Supplemental Tables S1-S4). Analysis of mortality using a Cox proportional hazards model produced results similar to the logistic model (Supplemental Table S5).
Discussion We found that clinically important differences exist between rural and urban HF patients in the use of evidencebased therapies, with rural patients being less likely to receive RAS agents and BBs, however, few differences in adherence were noted between urban and rural patients. Importantly, a lower rate of adherence to RAS agents or BBs within the first 6 months of hospital discharge was associated with an increased risk of mortality in rural and urban patients.
Optimal use of medication was relatively low with approximately 25% of patients stopping RAS agents or BBs by the end of follow-up. Moreover, < 69% of patients who received RAS agents and 53% of those who received BBs had medication available for at least 80% of days. In general, the adherence and persistence values we observed were similar to other HF studies; suggesting that more concerted efforts need to be made to improve the utilization and adherence to medications known to affect clinical outcomes in HF patients.2-4 We have extended this literature to show that rural patients might not be substantially different compared with urban patients in this regard. Overall, a substantial proportion of patients demonstrate suboptimal medication use irrespective of geography. More importantly, good adherence was associated with a lower mortality risk, and despite the differences we observed between the rural and urban patients in comorbidities, baseline health care service utilization, and socioeconomic status, we found no significant differences in 1-year mortality between adherent rural and urban patients after adjustment. In an earlier study, we had demonstrated no significant mortality differences between rural and urban HF patients, although urban patients were more likely to have office-based physician visits after discharge and exhibited 30% lower rates of hospitalization and emergency department visits after discharge compared with rural patients.6 Despite our large population-based sample, several limitations must be acknowledged. First, our data lack detailed clinical information, such as ejection fraction or symptom status, and other data that might be used to identify patients with contraindications or intolerance to medications. Although the diagnostic codes we used to identify the incident
Table 3. Adherence to evidence-based medications Drug class Adherent (PDC 0.8) RAS agent ACEI ARB BB Digoxin Spironolactone
Rural
Urban
n
%
n
%
Total n
1161 959 242 510 476 329
65.6 64.6 55.8 50.0 70.0 56.0
4059 3307 946 2103 1491 833
69.8 69.0 60.1 54.2 70.7 49.6
7587 6274 2009 4902 2790 2267
Rural vs urban, unadjusted OR (95% CI) 0.82 0.82 0.84 0.85 0.97 1.29
(0.74-0.92) (0.72-0.93) (0.68-1.04) (0.74-0.97) (0.80-1.17) (1.07-1.56)
P 0.001 0.001 0.11 0.019 0.74 0.008
Rural vs urban, adjusted OR (95% CI) 0.88 0.88 0.86 0.88 0.98 1.18
(0.78-1.00) (0.77-1.00) (0.68-1.10) (0.75-1.03) (0.79-1.21) (0.95-1.45)
P 0.049 0.059 0.24 0.11 0.86 0.13
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, b-blocker; CI, confidence interval; OR, odds ratio; PDC, proportion of days covered; RAS, renin-angiotensin system.
346
HF cohort are used frequently4,6 and validated in our province,12,13 in one study it was found that the diagnostic accuracy of hospital abstract data was lower for rural than urban hospitals in a different Canadian province.16 Thus, differences in diagnostic accuracy could potentially explain some of the observed rural-urban differences in medication utilization if a greater proportion of rural patients were misclassified as having HF. However, differential diagnostic accuracy is unlikely to bias measures of adherence or persistence, because these analyses were all conducted in treated patients. Second, estimates of adherence based on prescription refill data might overestimate actual consumption, and cannot provide information on the prevalence of primary nonadherence (failure to fill a prescription).15 That being said, we used accepted methods for measuring adherence from population-based administrative data.15 Further, we selected a cohort that survived at least 90 days after discharge to minimize bias by ensuring patients had an adequate opportunity to fill prescriptions, and that adherence was not biased upward by patients with a short follow-up time; however, we acknowledge that this population is likely a more stable population. We also calculated persistence until death or censoring to account for differences in the length of follow-up. In addition, we conducted several sensitivity analyses to confirm our findings were robust. However, we do not know if patients were truly nonadherent or nonpersistent, or if the medications were stopped because of clinical or tolerance issues. Third, lack of clinical data might have led to unmeasured confounding despite adjustment for variables known to predict outcomes in HF which might also have lead to difficult to identify channeling or confounding by indication biases. Adherence is affected by numerous interacting elements that include socioeconomic, condition, therapy, patient, and health-system related factors, and many of these factors are not available or might be incompletely measured in administrative data. Fourth, because patients were classified based on their residential address, it is possible that care might have occurred in any rural or urban hospital. Although urban patients can receive care in rural setting, the likelihood is low; however, it might be reasonable to assume that patients with a severe chronic illness from a rural location might be more likely to receive care in an urban centre. This would have served to reduce any discrepancy between rural patients and urban patients; thus, although we found differences in RAS agents and BB use, these differences might, in fact, be underestimated because of this misclassification bias. Finally, our results might not be generalizable to the overall HF population because our cohort was limited to those > 65 years of age and diagnosed in-hospital. Patients diagnosed in-hospital have a substantially poorer prognosis than those diagnosed in the emergency department or outpatient clinics.17 However, hospitalized patients are more homogeneous at baseline and the administrative data case definition performs best for hospital data. Rural residents with a first diagnosis of HF were less likely to receive RAS agents or BBs after discharge from the hospital. Rural patients exhibited similar adherence compared with urban patients for most evidence-based HF therapies. Irrespective of geographic locale, adherence and persistence to proven efficacious HF therapies are suboptimal for rural and urban patients, leading to increased risk of mortality. Future
Canadian Journal of Cardiology Volume 31 2015
interventions should aim to improve utilization of evidencebased therapies for HF, particularly among rural residents, but regardless of geography, more intensive follow-up might be warranted to ensure therapies are being adhered to and are not discontinued inappropriately by patients. Acknowledgements The authors thank Dr Jeffrey A. Bakal and Dr JohnMichael Gamble for data management and statistical advice. This study is based in part on data provided by Alberta Health. The interpretation and conclusions contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta. Neither the Government of Alberta nor Alberta Health express any opinion in relation to this study. Funding Sources Dr Eurich receives salary support from Alberta Innovates Health Solutions and the Canadian Institutes of Health Research. Dr McAlister is a senior health scholar with Alberta Innovates Health Solutions and holds the Capital Health Chair in Cardiovascular Outcomes. Disclosures The authors have no conflicts of interest to disclose. References 1. Hunt SA, Abraham WT, Chin MH, et al. 2009 Focused update incorporated into the ACC/AHA 2005 guidelines for the diagnosis and management of heart failure in adults. A report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines Developed in Collaboration with the International Society for Heart and Lung Transplantation. J Am Coll Cardiol 2009;53: e1-90. 2. Gislason GH, Rasmussen JN, Abildstrom SZ, et al. Persistent use of evidence-based pharmacotherapy in heart failure is associated with improved outcomes. Circulation 2007;116:737-44. 3. Setoguchi S, Choudhry NK, Levin R, Shrank WH, Winkelmayer WC. Temporal trends in adherence to cardiovascular medications in elderly patients after hospitalization for heart failure. Clin Pharmacol Ther 2010;88:548-54. 4. Lamb DA, Eurich DT, McAlister FA, et al. Changes in adherence to evidence-based medications in the first year after initial hospitalization for heart failure: observational cohort study from 1994 to 2003. Circ Cardiovasc Qual Outcomes 2009;2:228-35. 5. Fitzgerald AA, Powers JD, Ho PM, et al. Impact of medication nonadherence on hospitalizations and mortality in heart failure. J Card Fail 2011;17:664-9. 6. Gamble JM, Eurich DT, Ezekowitz JA, et al. Patterns of care and outcomes differ for urban versus rural patients with newly diagnosed heart failure, even in a universal healthcare system. Circ Heart Fail 2011;4: 317-23. 7. Goins RT, Williams KA, Carter MW, Spencer SM, Solovieva T. Perceived barriers to health care access among rural older adults: a qualitative study. J Rural Health 2005;21:206-13. 8. Pong R, DesMeules M, Heng D, et al. Patterns of health service utilization in rural Canada. Chronic Dis Can 2011;31(suppl 1):1-36.
Murphy et al. Rural and Urban Medication Use for HF Cohort 9. Karve S, Cleves MA, Helm M, et al. Good and poor adherence: optimal cut-point for adherence measures using administrative claims data. Curr Med Res Opin 2009;25:2303-10. 10. Wu JR, Moser DK, De Jong MJ, et al. Defining an evidence-based cutpoint for medication adherence in heart failure. Am Heart J 2009;157:285-91. 11. Ezekowitz JA, van Walraven C, McAlister FA, Armstrong PW, Kaul P. Impact of specialist follow-up in outpatients with congestive heart failure. Can Med Assoc J 2005;172:189-94. 12. Lee DS, Donovan L, Austin PC, et al. Comparison of coding of heart failure and comorbidities in administrative and clinical data for use in outcomes research. Med Care 2005;43:182-8. 13. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9. 14. Du Plessis V, Beshiri R, Bollman RD, Clemenson H. Definitions of Rural. Rural and Small Town Canada Analysis Bulletin. Vol. 3, No. 3. Ottawa: Statistics Canada, 2001.
347
15. Peterson AM, Nau DP, Cramer JA, et al. A checklist for medication compliance and persistence studies using retrospective databases. Value Health 2007;10:3-12. 16. Blackburn DF, Schnell G, Lamb DA, et al. Coding of heart failure diagnoses in Saskatchewan: a validation study of hospital discharge abstracts. J Popul Ther Clin Pharmacol 2011;18:e407-15. 17. Ezekowitz JA, Kaul P, Bakal JA, Quan H, McAlister FA. Trends in heart failure care: has the incident diagnosis of heart failure shifted from the hospital to the emergency department and outpatient clinics? Eur J Heart Fail 2011;13:142-7.
Supplementary Material To access the supplementary material accompanying this article, visit the online version of the Canadian Journal of Cardiology at www.onlinecjc.ca and at http://dx.doi.org/10. 1016/j.cjca.2014.11.024.
Clinical Research
Cardiovascular Medication Utilization and Adherence Among Heart Failure Patients in Rural and Urban Areas: A Retrospective Cohort Study Gaetanne K. Murphy, MSc,a Finlay A. McAlister, MD, MSc,b and Dean T. Eurich, PhDa a b
School of Public Health, University of Alberta, Edmonton, Alberta, Canada
Division of General Internal Medicine, Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada
ABSTRACT
RESUM E
Background: Rural residence is a negative prognostic factor for heart failure (HF). The objective was to explore rural and urban differences in the utilization, adherence, and persistence with medications, and mortality among incident HF patients. Methods: Using administrative databases from Alberta (Canada), subjects > 65 years old with a first hospitalization for HF between 1999 and 2008 who survived 90 days after discharge were identified. Pharmacy claims for renin-angiotensin system (RAS) agents, b-blockers (BBs), digoxin, or spironolactone were identified. The association between rural and urban residence and medication utilization, adherence (optimal adherence defined as 80% adherence over 1 year), persistence, and 1-year mortality was assessed. Results: The cohort included 10,430 patients, with a mean age of 80.2 (SD, 7.7) years, 47% were male, and 25% were rural residents. Rural residents were less likely to receive RAS agents (74% vs 79%, adjusted odds ratio [aOR], 0.78; 95% confidence interval [CI], 0.690.89) or BBs (44% vs 54%; aOR, 0.83; 95% CI, 0.73-0.93) than urban residents, but had similar use of other medications. Although < 69% of patients who received RAS agents and 53% who received BBs had optimal adherence, few differences in adherence or persistence were detected among patients in rural vs urban areas. The 1-year mortality rate was significantly lower for patients who demonstrated optimal adherence to RAS agents or BBs (aOR, 0.78; 95% CI, 0.65-0.94) with no significant differences in the first 6 months between patients residing in rural vs urban areas.
sidence en zone rurale est un facteur pronostique Introduction : La re gatif de l’insuffisance cardiaque (IC). L’objectif e tait d’examiner les ne rences entre le milieu rural et le milieu urbain quant à l’utilisation, diffe ve rance dans la prise des me dicaments, ainsi l’observance et la perse chez les patients nouvellement atteints d’IC. que la mortalite thodes : À l’aide des banques de donne es administratives de Me l’Alberta (Canada), les sujets > 65 ans ayant eu une première hospitalisation pour une IC entre 1999 et 2008 qui survivaient 90 jours e taient releve s. Les demandes de remboursement pour après le conge nine-angiotensine (SRA), des bdes agents agissant sur le système re taient releve es. bloquants (BB), de la digoxine ou de la spironolactone e sidence en zone rurale et la re sidence en zone L’association entre la re dicaments, de l’observance urbaine pour ce qui est de l’utilisation des me finie comme e tant une observance 80 % (observance optimale de ve rance et de la mortalite à 1 an e tait e value e. durant 1 an), de la perse sultats : La cohorte de 10 430 patients dont l’âge moyen e tait de Re cart-type, 7,7) ans comptait 47 % d’hommes et 25 % de re sidents 80,2 (e tait moins probable que les re sidents de zones de zones rurales. Il e rurales reçoivent des agents agissant sur le SRA (74 % vs 79 %, ratio ajuste [RIAa], 0,78; intervalle de confiance [IC] à d’incidence approche 95 %, 0,69-0,89) ou des BB (44 % vs 54 %; RIAa, 0,83; IC à 95 %, 0,73sidents de zones urbaines, mais l’utilisation e tait simi0,93) que les re dicaments. Bien que < 69 % des patients qui laire pour les autres me recevaient des agents agissant sur le SRA et 53 % des patients qui recevaient des BB montraient une observance optimale, peu de
Angiotensin-converting enzyme inhibitors (ACEIs), angiotensin-receptor blockers (ARBs), hydralazine with longacting nitrates, b-blockers (BBs), and spironolactone have all been shown to decrease morbidity and mortality in patients with heart failure (HF).1 Despite this evidence, these medications are often underutilized because of a failure to initiate, persist with, or adhere to therapy.2-4
In patients with HF, medication nonadherence alters the clinical status of patients and is associated with poor outcomes.1,2,5 Although patient-related factors, such as resources, knowledge, and attitudes are often the focus of adherence, these represent just one dimension that affects adherence behaviour. There is increasing focus on the effect of geography on utilization and adherence to drug therapies and subsequent outcomes. Previous studies have observed rural vs urban variations in HF outcomes6 that might, at least partially, be related to barriers such as social isolation, financial constraints, lower education, limited health care facilities, distance to care, physician shortages, and lack of access to specialist care.7,8 Although it has been speculated that these barriers would cause harm through primary underuse of evidence-based
Received for publication August 26, 2014. Accepted November 25, 2014. Corresponding author: Dr Dean T. Eurich, 2-040 Li Ka Shing Center for Health Research Innovation, University of Alberta, Edmonton, Alberta T6G 2E1, Canada. Tel.: þ1-780-492-6333; fax: þ1-780-492-7455. E-mail: [email protected] See page 346 for disclosure information.
http://dx.doi.org/10.1016/j.cjca.2014.11.024 0828-282X/Ó 2015 Canadian Cardiovascular Society. Published by Elsevier Inc. All rights reserved.
342
Canadian Journal of Cardiology Volume 31 2015
Conclusions: Rural residents with HF were less likely to receive RAS agents or BBs, but few differences in adherence were noted compared with their urban counterparts. Suboptimal adherence with evidencebased HF therapy was associated with increased risk of mortality.
rences dans l’observance ou la perse ve rance e taient de tecte es diffe entre les patients des zones rurales vs les patients des zones urbaines. à 1 an e tait significativement plus faible chez les patients La mortalite montraient une observance optimale du traitement par les qui de agents agissant sur le SRA ou les BB (RIAa, 0,78; IC à 95 %, 0,65rence significative entre les patients re sidant en 0,94), et ce, sans diffe sidant en zone urbaine au cours des 6 zone rurale vs les patients re premiers mois. tait moins probable que les re sidents des zones Conclusions : Il e rurales qui souffrent d’IC reçoivent des agents agissant sur le SRA ou rences dans l’observance e taient note es des BB, mais peu de diffe gions urbaines. Une comparativement à leurs homologues des re sur des donne es observance sous-optimale du traitement de l’IC fonde tait associe e à l’augmentation du risque de mortalite . probantes e
medications, it is possible that these barriers might also negatively affect patient adherence to therapies, although data to support this premise are lacking. Previous studies have shown that lower levels of adherence to evidence-based medications among HF patients is associated with increased risk of hospitalization and death.9,10 However, we are not aware of any studies that have evaluated potential adherence differences according to rural vs urban residence. Thus, this study was undertaken to determine if, among newly diagnosed HF patients, the utilization, adherence, and persistence with HF-specific medications differs between those living in rural and urban areas, and whether these medication use patterns could explain 1-year mortality differences.
code of 428.x or I50.x).12,13 To ensure incident HF, we excluded those with a HF-related hospitalization (ICD 9/10 HF claim code in any diagnosis field) within 5 years before the index hospitalization. Because we were primarily interested in medication usage after diagnosis, to ensure subjects had sufficient opportunity to initiate therapies, we excluded all patients who died within 90 days of hospital discharge. In addition, those with an index hospitalization length of stay > 365 days, missing postal code data, and patients with no claims in any of the databases during the follow-up period were also excluded. All patients were followed until death or 365 days after discharge from their index HF hospitalization. As per previous methodology used by Statistics Canada and others,6,14 rural and urban residence was determined according to the postal code of each patient’s home address obtained from the registry file.
Methods Setting A population-based cohort of patients with incident HF was assembled using deidentified administrative databases from Alberta Health. Alberta Health manages a single-payer government-funded health care system that provides universal access to hospital, emergency department, and physician services for all 3.7 million residents within the province of Alberta, Canada. Patient data from 5 demographic, vital statistic, and health care utilization databases were linked as described previously.11 Briefly, the Canadian Institute for Health Information Discharge Abstract Database supplied data on hospital admission dates, most responsible diagnosis, and up to 24 secondary diagnoses. The Alberta Blue Cross Medication Database provided outpatient prescription drug utilization data for all Alberta residents 65 years of age or older. The Alberta Health Care Insurance Plan Registry file provided demographic and vital statistics data, and the Ambulatory Care and Practitioner Claims Databases were used to obtain information on emergency department visits and office-based physician visits. The Health Ethics Research Board at the University of Alberta approved this study (Pro00033827). Study cohort We identified all subjects > 65 years of age who were discharged from hospital between April 1, 1999 and December 31, 2008 with the most responsible diagnosis of HF (International Classification of Diseases [ICD] 9 or 10
Outcomes The coprimary outcomes were the proportion of patients utilizing and adherent to HF-related medications. Utilization of HF-related medications was defined as at least 1 dispensation 7 days before and up to 1 year after discharge. Seven days before discharge was used to capture new prescriptions that might have been filled during the hospital stay or while patients were transitioning to out-of-hospital care. Adherence was defined as proportion of days with medication coverage (PDC) 0.8 for the medications of interest (ACEI, ARB, BB, digoxin, and spironolactone). By convention, the PDC was calculated as the sum of days with medication available divided by the total days of follow-up for patients with a minimum of 2 dispensations for that drug class.15 The primary analysis assumed medication was available on the dispensation date and for the estimated number of days supplied, based on the quantity dispensed and usual dosing frequency. An 80% adherence level has been associated with a reduced risk of death in HF5 and is the threshold commonly used in studies of cardiovascular medication adherence. Secondarily we evaluated medication persistence at 1 year. Persistence was defined as the availability of medication for at least 1 day during a 30-day window, at 3, 6, or 12 months after the first dispensation. Patients who died before the end of follow-up had their last observation carried forward. In addition, we evaluated whether adherence to HF-related medications within the first 6 months of discharge was associated with mortality in the subsequent 6 months. All outcomes were
Murphy et al. Rural and Urban Medication Use for HF Cohort
estimated for each drug class separately and for the reninangiotensin system (RAS) agents combined (ACEI or ARB). Statistical analysis Differences in baseline characteristics between urban and rural residents were compared using c2, Student t, or MannWhitney U tests, as appropriate. Multivariable logistic regression was used to evaluate the association between rural vs urban residence according to the proportion treated, adherent, and persistent to each drug class. For all-cause mortality, multivariable logistic regression was used to assess the effect of optimal adherence, defined as a BB or RAS agent available 80% of days in the first 6 months after discharge, on mortality for the subsequent 6 months, after adjustment for all other covariates. Covariates included in the models included variables known to be associated with medication adherence and clinical outcomes in patients with HF. To evaluate model overfitting, Hosmer-Lemeshow goodness of fit tests were conducted on all models with no obvious violations noted. In addition, we included the interaction between rural vs urban residence and adherence to determine any potential rural vs urban differences according to adherence on 1-year mortality. Covariates included age, sex, median neighbourhood income quartile, comorbidities (myocardial infarction, angina, diabetes, cerebrovascular disease, hypertension, valvular disease, arrhythmias, chronic pulmonary disease, peripheral vascular disease, neoplasm, dementia, peptic ulcer, or renal disease based on the index hospital admission and any admissions in the previous year), year of index hospitalization (ie, year of diagnosis), number of hospitalizations, physician or emergency department visits in the year before the index event, physician specialty (general practitioner, internal medicine, or cardiologist), and medications dispensed in the 90 days before the index hospitalization (BB, RAS, digoxin, statin, spironolactone, loop diuretic, amiodarone, or warfarin).12 First-order interactions between rural and urban status and age and sex were tested for all outcomes and no clinically important interactions were identified. The median neighbourhood income was estimated using postal codes and neighbourhood-level census data.11 All analyses were done using STATA version 12.1 (StataCorp LP, College Station, TX). To evaluate the robustness of our study results, we undertook several ancillary analyses. First, we adjusted the PDC for: (1) hospitalization (total days in hospital during follow-up were added to the days with drug coverage); (2) overlapping days supply (dispensation date of overlapping dispensations were moved forward to the end of supply date of the previous dispensation); and (3) medication discontinuation (follow-up was truncated at the date of the last dispensation for those who discontinued therapyddefined as a 30-day gap with no medication available, and no further dispensations). Second, analyses using alternate thresholds to define adherence (PDC 0.7 or 0.9) were conducted because studies9,10 have linked higher and lower adherence rate cut points to HFrelated outcomes. Third, we included all patients who were alive for at least 30 days after the index hospitalization. Fourth, we excluded early dispensations (those occurring within 7 days before discharge from the index hospitalization) from our analyses. Fifth, we repeated our mortality analysis using a Cox proportional hazards framework to account for
343
time to event. Last, we evaluated “new users” of our medications of interest by restricting analyses to those who had no dispensations for HF-related medications in the 90 days before the index hospitalization. Results Of 23,767 potentially eligible patients, we identified 10,430 patients with an incident HF hospitalization between April 1, 1999 and December 31, 2008 who met the cohort inclusion criteria. The reasons for exclusion were: previous HF-related hospitalization (n ¼ 8014); age < 66 years (n ¼ 2934), death during the index hospitalization (n ¼ 1417) or within 90 days of discharge (n ¼ 928), no health care claims during follow-up (n ¼ 40), length of stay > 365 days (n ¼ 3), and missing postal code (n ¼ 1). Mean age was 80.2 years (SD, 7.7), 4909 (47%) were male, and 2580 (25%) were rural residents (Table 1). At baseline, rural residents were younger, more likely to be female, had fewer comorbidities, lower neighbourhood income, had fewer outpatient physician visits (especially with specialists), but more hospitalizations and emergency department visits in the year before HF diagnosis compared with urban patients. In the 90 days before the index hospitalization, more urban than rural patients filled a prescription for a RAS agent or BB. Among the cohort who survived at least 90 days after the index hospitalization, the average follow-up was 343 days (SD, 61) days, and 359 (13.9%) rural and 1197 (15.2%) urban residents died (adjusted odds ratio [aOR], 0.95; 95% confidence interval [CI], 0.82-1.09). More rural patients were hospitalized (for any cause) during follow-up with 23% of rural patients hospitalized once, and 37% hospitalized 2 or more times, compared with 25% and 29% of urban patients, respectively. Medication treatment Overall, 8072 (77%) filled at least 1 prescription for a RAS agent (Table 2) with 43% of RAS users considered new users (ie, no RAS dispensations 90 days before the index hospitalization). A total of 5368 (51%) patients received a BB (47% were new users); most (n ¼ 4582; 44%) were prescribed bisoprolol, carvedilol, or metoprolol. Overall, 4706 (45%) filled prescriptions for a RAS agent and a BB concurrently. Although most patients (8362; 80%) filled at least 1 prescription for a loop diuretic, other HF-related medications were used in less than one-third of this cohort. Most patients filled their first prescription within 90 days of discharge (ACEI 92%, ARB 79%, BB 87%, digoxin 87%, spironolactone 79%). Rural patients were less likely to fill prescriptions for RAS agents (aOR, 0.78; 95% CI, 0.69-0.89), BB (aOR, 0.83; 95% CI, 0.73-0.93), or RAS agent and BB (aOR, 0.78; 95% CI, 0.70-0.88) in the first year after discharge than were urban residents, but had similar rates of use of loop diuretics, digoxin, or spironolactone. Medication adherence The proportion of patients who were adherent (PDC 0.8) was lowest for spironolactone (rural 56%, urban 50%; aOR, 1.18; 95% CI, 0.95-1.45) and highest for RAS agents
344
Canadian Journal of Cardiology Volume 31 2015
Table 1. Baseline characteristics Rural Characteristic Total (total n ¼ 10,430) Age, years* Male sex Median neighbourhood household income < $50,149 $50,150-$69,434 $69,435-$89,005 $89,006 Missing Drug treatment 90 days before index hospitalization RAS agents ACEI ARB BB Amiodarone Digoxin Hydralazine Long-acting nitrates Loop diuretic Spironolactone Statin Warfarin Number dispensations (any drug)y Number of drugsy Healthcare services in previous year General practitioner visit Internal medicine visit Cardiologist visit Number of physician visitsy Number of hospitalizationsy Number of emergency department visitsy Comorbidities Charlson comorbidity score* Ischemic heart disease Myocardial infarction Angina Diabetes Cerebrovascular disease Hypertension Valvular disease Cardiac arrhythmias Chronic pulmonary disease Peripheral vascular disease Neoplasm Dementia Peptic ulcer disease Renal disease
Urban
n
%
n
%
2580 79.8 (7.9) 1281
24.7
75.3
49.7
7850 80.3 (7.3) 3628
1077 912 326 84 181
41.7 35.3 12.6 3.3 7.0
2998 1841 1528 1188 295
38.2 23.5 19.5 15.1 3.8
1169 923 321 707 57 326 10 269 941 140 466 415 8 (11) 6 (5)
45.3 35.8 12.4 27.4 2.2 12.6 0.4 10.4 36.5 5.4 18.1 16.1
3826 2881 1204 2595 200 976 45 1011 2916 373 1687 1558 8 (10) 6 (6)
48.7 36.7 15.3 33.1 2.5 12.4 0.6 12.9 37.1 4.8 21.5 19.8
2541 1130 412 16 (16) 0 (1) 2 (3)
98.5 43.8 16.0
7686 5618 2899 19 (17) 0 (1) 2 (2)
97.9 71.6 36.9
2.3 (1.5) 840 388 158 755 142 1070 267 782 719 74 141 122 16 225
32.6 15.0 6.1 29.3 5.5 41.5 10.3 30.3 27.9 2.9 5.5 4.7 0.6 8.7
2.6 (1.6) 3614 1710 571 2399 552 4352 1558 3214 2246 345 371 555 73 1075
46.2
46.0 21.8 7.3 30.6 7.0 55.4 19.8 40.9 28.6 4.4 4.7 7.1 0.9 13.7
P 0.005 0.002 < 0.001
<
0.002 0.4 0.001 0.001 0.34 0.79 0.26 0.001 0.54 0.17 0.001 0.001 0.30 0.001
< < < < <
0.065 0.001 0.001 0.001 0.001 0.001
< <
< <
< 0.001 < 0.001 < 0.001 0.047 0.21 0.007 < 0.001 < 0.001 < 0.001 0.47 0.001 0.13 < 0.001 0.14 < 0.001
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, b-blocker; RAS, renin-angiotensin system. * Mean (SD). y Median (interquartile range).
(rural 66%, urban 70%; aOR, 0.88; 95% CI, 0.78-1.00; P ¼ 0.049; Table 3). Rural patients were less likely to be adherent to RAS agents than urban patients, although the differences were small. For BB, digoxin, and spironolactone, no significant rural vs urban differences in adherence were observed in adjusted analyses.
(Supplemental Fig. S1). The differences between rural and urban patients were not significant except for ARBs at 3 months (77% vs 80%; aOR, 0.75; 95% CI, 0.57-0.98) and RAS agents at 12 months (74% vs 77%; aOR, 0.86; 95% CI, 0.75-0.98) in the adjusted analyses; however, the clinical importance of these differences is questionable.
Medication persistence
Mortality
Persistence declined after discharge with the greatest reduction occurring in the first 3 months after initiating treatment. By 1 year, or until death or censoring, persistence was highest for RAS agents (76%), lowest for spironolactone (57%), and similar (approximately 70%) for BB or digoxin
Of the 7911 patients who survived 180 days after discharge, 158 (8.5%) of rural and 538 (8.9%) of urban residents died by 1 year (aOR, 1.04; 95% CI, 0.85-1.28). Patients who demonstrated optimal adherence (had a BB, ACEI, or ARB available for at least 80% of days) within the
Murphy et al. Rural and Urban Medication Use for HF Cohort
345
Table 2. Evidence-based medication utilization during follow-up Rural Drug class RAS agent ACEI ARB RAS agent or hydralazine with longacting nitrate BB BB and RAS agent Digoxin Spironolactone
Urban
Rural vs urban, unadjusted OR (95% CI)
Rural vs urban, adjusted OR (95% CI)
P
P
n
%
n
%
1903 1627 494 1907
73.8 63.1 19.1 73.9
6169 5236 1764 6197
78.6 66.7 22.5 78.9
0.77 0.85 0.82 0.76
(0.69-0.85) (0.78-0.94) (0.73-0.91) (0.68-0.84)
< 0.001 0.001 < 0.001 < 0.001
0.78 0.84 0.93 0.78
(0.69-0.89) (0.75-0.93) (0.82-1.05) (0.69-0.88)
< 0.001 0.001 0.24 < 0.001
1124 953 758 688
43.6 36.9 29.4 26.7
4244 3753 2323 1979
54.1 47.8 29.6 25.2
0.66 0.64 0.99 1.08
(0.60-0.72) (0.58-0.70) (0.90-1.09) (0.98-1.19)
< 0.001 < 0.001 0.84 0.14
0.83 0.78 0.98 1.10
(0.73-0.93) (0.70-0.88) (0.87-1.11) (0.98-1.23)
0.001 < 0.001 0.76 0.12
Total patients receiving drug therapy, n ¼ 10,430. ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, b-blocker; CI, confidence interval; OR, odds ratio; RAS, reninangiotensin system.
first 6 months had a lower risk of death at 1 year than those who were suboptimally adherent (aOR, 0.78; 95% CI, 0.650.94). We did not detect any significant differences in 1-year mortality for rural vs urban patients who were adherent (aOR, 1.09; 95% CI, 0.86-1.38) or who were nonadherent (aOR, 0.95; 95% CI, 0.66-1.35; P ¼ 0.50 for interaction). Sensitivity analyses Analyses accounting for time in hospital, early discontinuation, overlapping supply, excluding early dispensations (in 7 days before discharge from the index hospitalization), alternate adherence definitions ( 0.7 or 0.9), including all patients who survived at least 30 days after discharge, and analyses restricted to “new users,” were generally consistent with our primary analysis (Supplemental Tables S1-S4). Analysis of mortality using a Cox proportional hazards model produced results similar to the logistic model (Supplemental Table S5).
Discussion We found that clinically important differences exist between rural and urban HF patients in the use of evidencebased therapies, with rural patients being less likely to receive RAS agents and BBs, however, few differences in adherence were noted between urban and rural patients. Importantly, a lower rate of adherence to RAS agents or BBs within the first 6 months of hospital discharge was associated with an increased risk of mortality in rural and urban patients.
Optimal use of medication was relatively low with approximately 25% of patients stopping RAS agents or BBs by the end of follow-up. Moreover, < 69% of patients who received RAS agents and 53% of those who received BBs had medication available for at least 80% of days. In general, the adherence and persistence values we observed were similar to other HF studies; suggesting that more concerted efforts need to be made to improve the utilization and adherence to medications known to affect clinical outcomes in HF patients.2-4 We have extended this literature to show that rural patients might not be substantially different compared with urban patients in this regard. Overall, a substantial proportion of patients demonstrate suboptimal medication use irrespective of geography. More importantly, good adherence was associated with a lower mortality risk, and despite the differences we observed between the rural and urban patients in comorbidities, baseline health care service utilization, and socioeconomic status, we found no significant differences in 1-year mortality between adherent rural and urban patients after adjustment. In an earlier study, we had demonstrated no significant mortality differences between rural and urban HF patients, although urban patients were more likely to have office-based physician visits after discharge and exhibited 30% lower rates of hospitalization and emergency department visits after discharge compared with rural patients.6 Despite our large population-based sample, several limitations must be acknowledged. First, our data lack detailed clinical information, such as ejection fraction or symptom status, and other data that might be used to identify patients with contraindications or intolerance to medications. Although the diagnostic codes we used to identify the incident
Table 3. Adherence to evidence-based medications Drug class Adherent (PDC 0.8) RAS agent ACEI ARB BB Digoxin Spironolactone
Rural
Urban
n
%
n
%
Total n
1161 959 242 510 476 329
65.6 64.6 55.8 50.0 70.0 56.0
4059 3307 946 2103 1491 833
69.8 69.0 60.1 54.2 70.7 49.6
7587 6274 2009 4902 2790 2267
Rural vs urban, unadjusted OR (95% CI) 0.82 0.82 0.84 0.85 0.97 1.29
(0.74-0.92) (0.72-0.93) (0.68-1.04) (0.74-0.97) (0.80-1.17) (1.07-1.56)
P 0.001 0.001 0.11 0.019 0.74 0.008
Rural vs urban, adjusted OR (95% CI) 0.88 0.88 0.86 0.88 0.98 1.18
(0.78-1.00) (0.77-1.00) (0.68-1.10) (0.75-1.03) (0.79-1.21) (0.95-1.45)
P 0.049 0.059 0.24 0.11 0.86 0.13
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, b-blocker; CI, confidence interval; OR, odds ratio; PDC, proportion of days covered; RAS, renin-angiotensin system.
346
HF cohort are used frequently4,6 and validated in our province,12,13 in one study it was found that the diagnostic accuracy of hospital abstract data was lower for rural than urban hospitals in a different Canadian province.16 Thus, differences in diagnostic accuracy could potentially explain some of the observed rural-urban differences in medication utilization if a greater proportion of rural patients were misclassified as having HF. However, differential diagnostic accuracy is unlikely to bias measures of adherence or persistence, because these analyses were all conducted in treated patients. Second, estimates of adherence based on prescription refill data might overestimate actual consumption, and cannot provide information on the prevalence of primary nonadherence (failure to fill a prescription).15 That being said, we used accepted methods for measuring adherence from population-based administrative data.15 Further, we selected a cohort that survived at least 90 days after discharge to minimize bias by ensuring patients had an adequate opportunity to fill prescriptions, and that adherence was not biased upward by patients with a short follow-up time; however, we acknowledge that this population is likely a more stable population. We also calculated persistence until death or censoring to account for differences in the length of follow-up. In addition, we conducted several sensitivity analyses to confirm our findings were robust. However, we do not know if patients were truly nonadherent or nonpersistent, or if the medications were stopped because of clinical or tolerance issues. Third, lack of clinical data might have led to unmeasured confounding despite adjustment for variables known to predict outcomes in HF which might also have lead to difficult to identify channeling or confounding by indication biases. Adherence is affected by numerous interacting elements that include socioeconomic, condition, therapy, patient, and health-system related factors, and many of these factors are not available or might be incompletely measured in administrative data. Fourth, because patients were classified based on their residential address, it is possible that care might have occurred in any rural or urban hospital. Although urban patients can receive care in rural setting, the likelihood is low; however, it might be reasonable to assume that patients with a severe chronic illness from a rural location might be more likely to receive care in an urban centre. This would have served to reduce any discrepancy between rural patients and urban patients; thus, although we found differences in RAS agents and BB use, these differences might, in fact, be underestimated because of this misclassification bias. Finally, our results might not be generalizable to the overall HF population because our cohort was limited to those > 65 years of age and diagnosed in-hospital. Patients diagnosed in-hospital have a substantially poorer prognosis than those diagnosed in the emergency department or outpatient clinics.17 However, hospitalized patients are more homogeneous at baseline and the administrative data case definition performs best for hospital data. Rural residents with a first diagnosis of HF were less likely to receive RAS agents or BBs after discharge from the hospital. Rural patients exhibited similar adherence compared with urban patients for most evidence-based HF therapies. Irrespective of geographic locale, adherence and persistence to proven efficacious HF therapies are suboptimal for rural and urban patients, leading to increased risk of mortality. Future
Canadian Journal of Cardiology Volume 31 2015
interventions should aim to improve utilization of evidencebased therapies for HF, particularly among rural residents, but regardless of geography, more intensive follow-up might be warranted to ensure therapies are being adhered to and are not discontinued inappropriately by patients. Acknowledgements The authors thank Dr Jeffrey A. Bakal and Dr JohnMichael Gamble for data management and statistical advice. This study is based in part on data provided by Alberta Health. The interpretation and conclusions contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta. Neither the Government of Alberta nor Alberta Health express any opinion in relation to this study. Funding Sources Dr Eurich receives salary support from Alberta Innovates Health Solutions and the Canadian Institutes of Health Research. Dr McAlister is a senior health scholar with Alberta Innovates Health Solutions and holds the Capital Health Chair in Cardiovascular Outcomes. Disclosures The authors have no conflicts of interest to disclose. References 1. Hunt SA, Abraham WT, Chin MH, et al. 2009 Focused update incorporated into the ACC/AHA 2005 guidelines for the diagnosis and management of heart failure in adults. A report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines Developed in Collaboration with the International Society for Heart and Lung Transplantation. J Am Coll Cardiol 2009;53: e1-90. 2. Gislason GH, Rasmussen JN, Abildstrom SZ, et al. Persistent use of evidence-based pharmacotherapy in heart failure is associated with improved outcomes. Circulation 2007;116:737-44. 3. Setoguchi S, Choudhry NK, Levin R, Shrank WH, Winkelmayer WC. Temporal trends in adherence to cardiovascular medications in elderly patients after hospitalization for heart failure. Clin Pharmacol Ther 2010;88:548-54. 4. Lamb DA, Eurich DT, McAlister FA, et al. Changes in adherence to evidence-based medications in the first year after initial hospitalization for heart failure: observational cohort study from 1994 to 2003. Circ Cardiovasc Qual Outcomes 2009;2:228-35. 5. Fitzgerald AA, Powers JD, Ho PM, et al. Impact of medication nonadherence on hospitalizations and mortality in heart failure. J Card Fail 2011;17:664-9. 6. Gamble JM, Eurich DT, Ezekowitz JA, et al. Patterns of care and outcomes differ for urban versus rural patients with newly diagnosed heart failure, even in a universal healthcare system. Circ Heart Fail 2011;4: 317-23. 7. Goins RT, Williams KA, Carter MW, Spencer SM, Solovieva T. Perceived barriers to health care access among rural older adults: a qualitative study. J Rural Health 2005;21:206-13. 8. Pong R, DesMeules M, Heng D, et al. Patterns of health service utilization in rural Canada. Chronic Dis Can 2011;31(suppl 1):1-36.
Murphy et al. Rural and Urban Medication Use for HF Cohort 9. Karve S, Cleves MA, Helm M, et al. Good and poor adherence: optimal cut-point for adherence measures using administrative claims data. Curr Med Res Opin 2009;25:2303-10. 10. Wu JR, Moser DK, De Jong MJ, et al. Defining an evidence-based cutpoint for medication adherence in heart failure. Am Heart J 2009;157:285-91. 11. Ezekowitz JA, van Walraven C, McAlister FA, Armstrong PW, Kaul P. Impact of specialist follow-up in outpatients with congestive heart failure. Can Med Assoc J 2005;172:189-94. 12. Lee DS, Donovan L, Austin PC, et al. Comparison of coding of heart failure and comorbidities in administrative and clinical data for use in outcomes research. Med Care 2005;43:182-8. 13. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9. 14. Du Plessis V, Beshiri R, Bollman RD, Clemenson H. Definitions of Rural. Rural and Small Town Canada Analysis Bulletin. Vol. 3, No. 3. Ottawa: Statistics Canada, 2001.
347
15. Peterson AM, Nau DP, Cramer JA, et al. A checklist for medication compliance and persistence studies using retrospective databases. Value Health 2007;10:3-12. 16. Blackburn DF, Schnell G, Lamb DA, et al. Coding of heart failure diagnoses in Saskatchewan: a validation study of hospital discharge abstracts. J Popul Ther Clin Pharmacol 2011;18:e407-15. 17. Ezekowitz JA, Kaul P, Bakal JA, Quan H, McAlister FA. Trends in heart failure care: has the incident diagnosis of heart failure shifted from the hospital to the emergency department and outpatient clinics? Eur J Heart Fail 2011;13:142-7.
Supplementary Material To access the supplementary material accompanying this article, visit the online version of the Canadian Journal of Cardiology at www.onlinecjc.ca and at http://dx.doi.org/10. 1016/j.cjca.2014.11.024.