Approximately half of all elderly patients have elevated blood pressure, and proper treatment of this disorder leads to decreased cardiovascular morbidity in patients 65 and older. This study examined the effect of initial drug choice and comorbidity on medication compliance. We conducted a retrospective follow-up of 8643 outpatients aged 65 to 99 with newly prescribed antihypertensive therapy (AHT) from 1982 to 1988 in the New Jersey Medicaid and Medicare programs. Compliance was measured in terms of the number of days in which AHT was available to the patient during the 12 months following the initiation of therapy. Odds ratios (OR) and 95% confidence intervals (CI) for the outcome of good compliance (≥80%) were calculated.
In a logistic regression model, good compliance (≥80%) was significantly associated with use of newer agents such as angiotensin converting enzyme inhibitors (OR 1.9, 95% CI 1.6 to 2.2) and calcium channel blockers (OR 1.7, 95% CI 1.5 to 2.1) as compared to thiazides, the presence of comorbid cardiac disease (OR 1.2, 95% CI 1.1 to 1.2), and multiple physician visits (OR 2.2, 95% CI 1.8 to 2.5). Good compliance was inversely associated with use of multiple pharmacies (OR 0.4, 95% CI 0.4 to 0.5) and number of medications prescribed overall (OR 0.8, 95% CI 0.7 to 0.9).
Drug choice, comorbidity, and health services utilization were significantly associated with AHT compliance and represent important considerations in the management of high blood pressure. Noncompliance may be an important cause of treatment failure in elderly hypertensives. Am J Hypertens 1997;10:697–704 © 1997 American Journal of Hypertension, Ltd.
Approximately half of patients over the age of 65 will have elevated blood pressure.1 Managing this disorder is a major concern in this vulnerable population group, as the elderly are at particularly high risk for the adverse consequences of elevated blood pressure including stroke and myocardial infarction.2 Treatment costs of hypertension and related factors approach $60 billion yearly.3
Recent trials including the Systolic Hypertension in the Elderly Program (SHEP) study,4 Swedish Trial in Older Patients with Hypertension (STOP-Hypertension),5 and Medical Research Council (MRC) trial6 have confirmed the utility of treating elevated blood pressure in preventing these adverse long-term events. Yet despite the evidence from these clinical trials, normotension is often not achieved in the population. According to the recent Third National Health and Nutrition Examination Survey (NHANES III) report, only 25% of those treated for hypertension (>140/90 versus >160/95 in NHANES II) in a community-based sample have adequate blood pressure control.7 One likely explanation for this discrepancy between the efficacy of antihypertensive medications in clinical trials and population-based findings is noncompliance, the “hidden” epidemic. Risk factors for noncompliance have varied from study to study but generally include the number of drugs in the regimen8 as well as the dosing frequency of each medication.9 Age has not been a consistent predictor of compliance.10,11 In addition, older patients with hypertension often have other cardiac and noncardiac comorbidity as well as multiple medications used for treating these disorders.
In a previous analysis, we found compliance rates of only 49% in a 1-year follow-up study of an elderly population newly initiated on antihypertensive medications.12 This study examined the effect of initial drug choice and comorbidity on antihypertensive therapy compliance.
Sources of Data
The state of New Jersey's Medicaid and Medicare programs from 1982 to 1988 were used to identify the study population. Patient enrollment in these programs was ascertained through the Medicaid enrollment file, which identifies eligibility for Medicaid benefits, dates of coverage, and demographic characteristics including age, gender, and race as well as information on nursing home residency and date of death. Data on inpatient and outpatient physician visits as well as information on all filled medication prescriptions were also identified.
Selection of Study Subjects
Study subjects were Medicaid enrollees aged 65 or older who had filed their first claim for an antihypertensive drug during the study period: new users could have no claim for any antihypertensive agent in the 12 months prior to their first (index) claim. In addition, these subjects were required to be active users of the Medicaid program, documented by their having filled at least one health service claim in each 4-month period in the year before and the year after their first antihypertensive prescription.
The initial screen identified 9,468 new users of at least one antihypertensive drug. Of these potential study subjects, we excluded persons in the hospital on or about the time of their first antihypertensive medication prescription (N = 52) as well as those in the institutionalized setting (N = 773) to yield a final cohort of 8643 study subjects.
Definitions of Compliance
To calculate compliance rates, we extracted complete data on all prescriptions filled after the first antihypertensive medication claim. Compliance measurements were based on quantity dispensed and days supply data in order to measure the number of days a patient had antihypertensive medication available during the 12-month period following the first index prescription (“days covered”). The method used to calculate compliance figures has been previously described elsewhere.13,14 Although the use of free samples is not measured in this claims database, physicians are much less likely to give free samples to elderly on Medicaid, given the fact that these patients receive their medications at no cost through the Medicaid program.
We examined the patterns of filled prescriptions for each study subject, including the prescription dates, quantity dispensed, and days supply to calculate the number of days each patient had medication throughout the year. Data from the days supply and quantity dispensed fields were closely correlated with the standard dosing regimens described in reference works.15,16 To avoid mislabeling patients as noncompliant if their antihypertensive therapy (AHT) was switched from one agent to another, such switches in treatment counted towards compliance. Days spent in the hospital counted as days of full compliance. In the case of overlapping prescriptions, the duration of use of a prescription was not duplicated if another prescription was also available on that day. We categorized subjects into two groups: ≥80% versus <80% days covered during the study year (compliant v noncompliant).
Factors Associated with Compliance
We examined demographic data, hospitalizations, physician visits, and pharmacy records for the 120 days prior to the first filled antihypertensive drug prescription (index date) as potential factors associated with compliance. Specific variables of interest included age, gender, race, year of initiation of antihypertensive therapy, number of antihypertensive medications, class of antihypertensive medication, number of nonstudy medications, number of physician visits, and number of pharmacies used. Because severity of illness is an important possible confounder when examining compliance with antihypertensive medication, we measured the presence of diagnoses, medications, and procedures indicative of coronary artery disease (CAD) or congestive heart failure (CHF). In addition, because adverse effects associated with thiazides are believed to be dose related,17 we converted all thiazide doses into hydrochlorothiazide-milligram equivalents15 to test for a dose-response relationship between thiazide exposure and compliance.
Associations between the variables of interest and ≥80% compliance with the antihypertensive medication regimen were assessed using logistic regression18 using SAS CATMOD software (SAS Institute Inc., Cary, NC).19 Potential factors associated with compliance included age (65 to 74, 75 to 84, 85 and older), race (white, black, other), gender, year of initiation of therapy (1982 to 1984, 1985 to 1986, 1987 to 1988), antihypertensive medication class (thiazides, β-adrenergic blockers, calcium channel blockers (CCB), angiotensin converting enzyme (ACE) inhibitors, multiple classes, or other), presence of CAD or CHF, and number of pharmacies used (1, >1). Intensity of medical care in the 120 days before the first antihypertensive prescription was measured by the number of prescriptions filled for any drug (0 to 3, 4 to 7, ≥8) and number of physician visits (0 to 3, 4 to 7, ≥8).
We next undertook other analyses to test the robustness of the study findings. Because the presence of other cardiovascular illnesses is potentially related to both AHT drug choice and compliance, we studied patients with either CAD or CHF. We also evaluated the subset of patients who filled at least two AHT prescriptions, eliminating those with the poorest compliance and those who used antihypertensive medications for other short-term indications. Finally, we studied the subgroup of patients who used the same category of antihypertensive medication throughout the follow-up period, in order to determine the role of switching AHT on compliance.
Binary variables were calculated with a 0/1 classification in all testing.20 Confidence intervals (CIs) for the estimated odds ratios and significance tests for differences from the null value were calculated using the estimated standard errors.21 Tests for possible interactions among independent variables were performed.22
The characteristics of the study population (N = 8,643) are summarized in Table 1. Subjects had a mean age of 76.5 (±8.1) years, and 19% of the population were 85 and older. The population was predominantly women (77%) and white (67%).
|Year of therapy|
|Comorbid cardiac disease|
|Health care use in the 120 days before antihypertensive therapy|
|Number of prescriptions|
|Number of physician visits|
|Number of pharmacies|
|Year of therapy|
|Comorbid cardiac disease|
|Health care use in the 120 days before antihypertensive therapy|
|Number of prescriptions|
|Number of physician visits|
|Number of pharmacies|
These patients were active users of the health care system in the 120 days before the first antihypertensive prescription. The average number of different prescriptions filled was approximately seven, and over a fifth used more than one pharmacy in filling these prescriptions. One in seven patients had a hospital stay during this period.
Measurement of Compliance
Diuretics were the most commonly prescribed initial agents, accounting for 50% of first prescriptions, followed by β-adrenergic blockers (12%), calcium channel blockers (12%), angiotensin converting enzyme (ACE) inhibitors (5%), and others (17%); 4% of the population was initiated on multiple agents. In the 12 months following an initial antihypertensive prescription, the average patient had antihypertensive medication available for only 161 days of the year (43%), a pattern similar to that in a smaller cohort described previously.12 In addition, a fifth (21%) of the study cohort did not fill a second prescription for any antihypertensive medication during the follow-up period. Of patients started on therapy, only 20% achieved good levels of compliance, defined as 80% or greater based on physicians' instructions for use.
Relationship Between Drug Choices, Comorbidity, and Compliance
In the logistic regression model measuring ≥80% compliance, using thiazide users as the referent group, ACE inhibitor users (OR 1.9, 95% CI 1.6 to 2.1), CCB users (OR 1.7, 95% CI 1.5 to 2.1), and β-blocker users (OR 1.4, 95% CI 1.2 to 1.7) were significantly more likely to be compliant (Fig. 1). Patients on multiple agents were as likely as those on single agents to be compliant (OR 0.8, 95% CI 0.7 to 0.9). Furthermore, there was no relationship between thiazide dose and compliance (OR 0.9 for >50 hydrochlorothiazide [HCTZ] mg equivalents versus 1 to 25 HCTZ mg equivalents, P = .33).
Initial drug choice as a factor related to antihypertensive therapy (AHT) compliance, adjusted for age, gender, race, and year of therapy. The y axis depicts the odds ratio for good compliance (≥80% of days covered) for patients beginning antihypertensive therapy. MULTIPLE, initiation of therapy with two or more drugs; THIAZIDES, thiazide or related diuretics; B-BLOCKERS, β-adrenergic blockers; OTHER, central adrenergic agonists, peripheral adrenergic antagonists, or vasodilators; CCBs, calcium-channel blockers; and ACE INHIB, angiotensin converting enzyme inhibitors. For each variable, the odds ratio is indicated by the horizontal line and the 95% confidence interval is indicated by the vertical line.
Fig. 2 shows that compliance was significantly better in patients with ischemic heart disease or congestive heart failure (OR 1.2, 95% CI 1.1 to 1.3) and patients who had more physician visits (OR 2.2 for eight or more recent visits, 95% CI 1.8 to 2.5). However, compliance was significantly lower in patients prescribed multiple other medications (OR 0.8 for eight or more prescriptions, 95% CI 0.7 to 0.9) and patients who used more than one pharmacy (OR 0.4, 95% CI 0.4 to 0.5) (Fig. 3). All relationships were adjusted for effects of age, gender, race, and year of initiation of antihypertensive therapy.
Number of other medications and number of pharmacies as factors related to AHT compliance, adjusted for age, gender, race, and year of therapy. The y axis depicts the odds ratio for good compliance (≥80% of days covered) for patients beginning antihypertensive therapy. Number of other medications and number of pharmacies used were measured in the 120 days preceding the initial antihypertensive prescription. For each variable, the odds ratio is indicated by the horizontal line and the 95% confidence interval is indicated by the vertical line.
Number of physician visits and comorbid cardiac disease as factors related to AHT compliance, adjusted for age, gender, race, and year of therapy. The y axis depicts the odds ratio for good compliance (≥80% of days covered) for patients beginning antihypertensive therapy. Physician visits and presence of heart disease were measured in the 120 days preceding the initial antihypertensive prescription. For each variable, the odds ratio is indicated by the horizontal line and the 95% confidence interval is indicated by the vertical line.
We next studied those patients (N = 6,828) who filled two or more AHT prescriptions during the 12-month follow-up period. The overall compliance rate in this cohort increased as expected to 54%, with slightly greater than a third of patients achieving good compliance rates (≥80%). Nonetheless, the same factors as those found in the larger cohort emerged as significantly related to compliance. Similarly, those patients who filled prescriptions for the same antihypertensives throughout the study (N = 6,747) showed the same predictors of compliance as seen in the primary analysis.
It is possible that some of the drugs studied may have been used primarily to treat coronary artery disease or congestive heart failure rather than hypertension. We therefore repeated the analysis separately among AHT patients with any evidence (N = 3,244) and without any evidence (N = 5,399) of diagnoses or other drug treatment for either of these two disorders. Compliance rates were similar in both groups: patients with coexisting CAD or CHF had AHT available for 46% of days during the follow-up year, and patients without any evidence of either of these conditions had AHT available for 40% of days. Factors associated with compliance in the two subgroups were virtually identical to those seen in the entire cohort of antihypertensive users; however, CCB users had differing patterns of compliance in the two subgroups in that a compliance advantage over thiazide users was seen only in patients with coexisting cardiovascular disease. For each model tested, there was no evidence of interactions among the independent variables.
These findings indicate that drug choice, comorbidity, and health services utilization are all strongly related to good versus poor compliance in newly treated hypertensives. Compliance was significantly higher in patients started on ACE inhibitors or CCBs compared to thiazides, in patients with preexisting cardiovascular disease such as CAD and CHF, and in patients with more frequent physician visits. Conversely, patients using multiple other medications or more than one pharmacy were significantly less likely to be compliant. These results were robust across the various study cohort definitions used. Several of the factors identified in this study suggest avenues for intervention to improve compliance in this vulnerable population.
Medicaid and Medicare data offer major advantages in evaluating drug surveillance, including the ability to document all health care service use without recall bias or incomplete history information. Previous studies have demonstrated the high validity of the pharmacy claims in Medicaid files.23,24 However, the limitations of claims-based information must be recognized.25,26 Misclassification could result if a patient was started on one of the target medications without a diagnosis of hypertension: several studies have examined use of drug markers as a proxy for diabetes27 and hypertension28 with good correlations noted. All subjects will have demonstrated active use of Medicaid, so it is unlikely that prescription medications or medical services would have been obtained outside of these programs. It is possible that patients continued to fill their AHT prescriptions on a regular basis yet subsequently did not take these medications; however, we find this scenario as well as fraud and diversion unlikely.
We were surprised to find that users of ACE inhibitors and CCB in this population were almost twice as likely as thiazide users to be compliant. Although a recent study focusing on general practices in the United Kingdom noted similar severe rates of noncompliance (40% to 50%), the investigators found no differences among the four major drug classes studied during a 6-month follow-up period.29 One possible reason for our finding is that, in our cohort, patients with more severe hypertension or greater comorbidity may have been more likely to be started on ACE inhibitors and CCB, whereas those with milder hypertension were started on diuretic therapy. According to the Health Belief Model, sicker patients are more likely to take medications as directed because of the perceived need to treat their condition effectively.30 Second, if patients with mild hypertension were started on thiazide diuretics, physicians may not have been as firm in promoting compliance in the mild hypertensive group.
A third possibility is that diuretics may be associated with less compliance than the ACE inhibitors or CCB because of side effects: several studies have suggested31–34 and others disputed35–37 this relationship. However, we found no compliance differences as a function of thiazide dose. It is also possible that thiazide use was for transient nonhypertensive indications, such as pedal edema or fluid retention in the absence of a CHF diagnosis or treatment. If this were true, the lower level of compliance seen with thiazides would be a result of differing indications for use. Our findings persisted, however, when we considered patients who filled more than one AHT prescription. It is also unlikely that regimen changes account for the observed findings: blood pressure control has been shown to be equivalent among all agents.38,39 The results from large-scale clinical trials differ interestingly from those noted in this study. Specifically, compliance rates in several large-scale trials have found discontinuance rates at 15% to 20% for each of the major agents studied.40,41 Yet these findings may not reflect medication utilization by the general population because of selection basis: these large scale investigations were performed on patients volunteering to be in a study, who were followed closely under clinical trial conditions. Many studies of compliance have combined patients both newly initiated on therapy with patients currently receiving therapy to form the study population. This design causes the population studied to disproportionately represent “survivors,” those who stay with their regimens over time. Other investigations have enrolled patients receiving care at university-affiliated centers as well as patients given their medications free of charge as part of the study protocol. Therefore, these patients, their compliance patterns, and their physicians may not be representative of typical community-dwelling elderly.
The inverse relationship between multiple medications and compliance, as seen in our study, has been noted in other investigations,8 although the relationship between multiple pharmacy use and poor compliance has not been previously demonstrated to our knowledge. The use of multiple pharmacies can lead to the loss of the surveillance and counseling that are possible if a patient receives all his or her prescriptions from a single-pharmacy option. Patients with multiple physician visits and comorbid cardiovascular disease were more likely to achieve good compliance, findings also consistent with the Health Belief Model noted earlier.30 In addition, multiple physicians visits may have allowed for the added opportunity for education and promotion around compliance.
It is important to consider the generalizability of the study results to elderly patients not on Medicaid. Although patients with higher socioeconomic status may comply more faithfully, the opposite is also possible. Because medications are not covered by Medicare or by most other insurance plans, drugs are also one of the highest out-of-pocket health care expenses for the elderly42; noncompliance may therefore be an even greater concern in patients without access to free medications.
Additional studies are needed to further define and explain compliance. The observational nature of this study does not make it possible to say that better compliance is caused by use of ACE inhibitors/CCB, but rather the study design emphasizes association. Nonetheless, it is possible that managed care options for the elderly, with the potential for coordinated care among providers and a centralized pharmacy system, could decrease the frequency of the noncompliance we observed. Intervention studies are also needed, with a focus on strategies to improve medication use. Enhanced compliance may offer benefits to both patients and society in terms of improved blood pressure control43 as well as decreased morbidity44 and cost.45
We thank Rita Bloom and Sharon Hawley for their assistance in the preparation of the manuscript.
- angiotensin-converting enzyme inhibitors
- antihypertensive agents
- calcium channel blockers
- heart diseases
- cardiovascular system
- health services
- treatment failure
- antihypertensive therapy
- medication adherence