Abstract
Poor medication adherence is a significant problem in hypertensive African Americans. Although motivational interviewing (MINT) is effective for adoption and maintenance of health behaviors in patients with chronic diseases, its effect on medication adherence remains untested in this population.
This randomized controlled trial tested the effect of a practice-based MINT counseling vs. usual care (UC) on medication adherence and blood pressure (BP) in 190 hypertensive African Americans (88% women; mean age 54 years). Patients were recruited from two community-based primary care practices in New York City. The primary outcome was adherence measured by electronic pill monitors; the secondary outcome was within-patient change in office BP from baseline to 12 months.
Baseline adherence was similar in both groups (56.2 and 56.6% for MINT and UC, respectively, P = 0.94). Based on intent-to-treat analysis using mixed-effects regression, a significant time × group interaction with model-predicted posttreatment adherence rates of 43 and 57% were found in the UC and MINT groups, respectively (P = 0.027), with a between-group difference of 14% (95% confidence interval, −0.2 to −27%). The between-group difference in systolic and diastolic BP was −6.1 mm Hg (P = 0.065) and −1.4 mm Hg (P = 0.465), respectively, in favor of the MINT group.
A practice-based MINT counseling led to steady maintenance of medication adherence over time, compared to significant decline in adherence for UC patients. This effect was associated with a modest, nonsignificant trend toward a net reduction in systolic BP in favor of the MINT group.
African Americans have higher prevalence of hypertension and poorer hypertension-related outcomes than whites.1,2,3 Poor medication adherence may explain the poor BP control in African Americans.4,5 Behavioral counseling strategies are effective in improving medication adherence of chronic diseases patients,6,7 with little data in hypertensive African Americans. Successful interventions include those that are emotionally supportive; involve patients in their care; address patients' beliefs about medications; and enhance patients' confidence in their ability to overcome barriers to adherence.8,9,10,11,12,13 Motivational interviewing (MINT), a counseling approach that has gained increased popularity in primary care practices,14 encompasses several of these characteristics. MINT is defined as directive, patient-centered counseling designed to motivate patients for change by helping them recognize and resolve the discrepancy between their behavior, personal goals, and values.15 In patients with chronic diseases, MINT is effective in facilitating the adoption and maintenance of recommended health behaviors including weight loss, smoking cessation, and dietary habits.16,17,18,19,20 Despite its proven efficacy, the effect of MINT on medication adherence remains untested in hypertensive African Americans in primary care settings.
In this randomized controlled trial, we tested the effect of a practice-based MINT vs. usual care (UC) on medication adherence and blood pressure (BP), among hypertensive African Americans. We hypothesized a greater effect of MINT on medication adherence compared to UC at 12 months and that this effect would be associated with a significant reduction in clinic BP.
Methods
Setting and patients. As described elsewhere,21 patients were recruited from two community-based primary care practices affiliated with New York Presbyterian Hospital Ambulatory Care Network. Eligible patients were identified via electronic medical records (EMRs) and asked to participate during routine office visits by trained research assistants (RAs). Eligibility criteria included self-identification as African American or black; age ≥ 18 years; diagnosis of hypertension; taking at least one antihypertensive medication; uncontrolled BP on two successive office visits before screening (BP ≥140/90 mm Hg or ≥130/80 mm Hg for those with kidney disease or diabetes);22 and fluency in English. Patients who agreed to participate provided written informed consent. Cornell and Columbia University Institutional Review Boards approved the study.
Baseline assessment and follow-up visits. At baseline, trained RAs assessed patients' demographics and clinical history. EMRs of patients were reviewed for clinic BP, and history of comorbidity was documented using Charlson Comorbidity Index.23 Each patient was provided and taught how to use an electronic pill cap equipped with the Medication Events Monitoring System (MEMS) (Aprex, Fremont, CA). Patients were instructed to bring their MEMS to all study visits. Follow-up assessments were conducted every 3 months during which patients' adherence data were downloaded from their MEMS pill caps, and their clinic BP readings were retrieved from their EMRs. Final assessment was conducted at 12 months. All patients were reimbursed $25 after each study visit.
Randomization. After baseline assessment, patients were randomly assigned to either UC or MINT group by the study statistician, using sealed envelopes. Separate randomization schedules were developed from a computerized random-number generator, balanced at set intervals, using permutated blocks, to assure equal numbers in each group. Due to the nature of the behavioral intervention, neither the patients nor the RAs were blinded to the intervention. However, the clinic staff who recorded the BP data were blinded to patient assignment. It is important to note that medication adherence data were downloaded automatically into the computer from the MEMS caps through a reader. Thus, both the RAs and patients could not affect MEMS adherence outcome.
Intervention. Details of the intervention are described elsewhere.21 Briefly, patients in the MINT group received UC plus behavioral counseling about medication adherence using MINT techniques. Each patient received a 30–40 min MINT counseling session at 3, 6, 9, and 12 months. All sessions were conducted by trained RAs using a structured MINT counseling script.21 All sessions were audiotaped, and fidelity of the RAs to MINT techniques was assessed regularly by a trained MINT rater, who provided feedback to the RAs based on the recordings.24 Because the use of MEMS pill caps is associated with increased adherence,25 the intervention was delivered 3 months postrandomization in order to allow for habituation to the use of MEMS and assessment of patients' baseline medication adherence level.
Patients in the UC group did not receive MINT counseling, but completed all assessments at the same time intervals as the MINT group.
Outcomes and measurements. The primary outcome was medication adherence between 10 and 12 months assessed with MEMS pill caps, the accepted “gold standard” for adherence assessment.26 Patients were required to monitor adherence to one antihypertensive medication taken once daily. For each patient, there is a value from 0 to k, where k is the number of times MEMS recorded an opening for each day the patient was in the study. The values for each day were converted to a binary record, 0 = MEMS not opened, whereas 1 = MEMS opened once/day. This is a proxy for pill usage; that is, we operationalized adherence as opening the bottle. This metric (known as taking adherence), was the proportion of days in which the patient took his or her medication as prescribed (once daily in this case).27 Patients were assigned missing values for the days their MEMS were not used (i.e., “drug holidays”). There is no way to distinguish or ascertain whether or not patients took their pills if they opened their pill bottles. Aside from malfunctioning MEMS caps, the other conditions when we treated the data as missing are periods when patients do not have their MEMS caps, e.g., during hospitalization, emergency room visits, and if their medications were discontinued by their physicians in preparation for medical procedures. In these cases, we do not know whether the patients would have opened their MEMS or not, so we treated the data as missing (and not as 0).
The secondary outcome was within-patient change in systolic and diastolic BP from baseline to 12 months. Because of the desire to mimic real-world primary care practice, we did not influence the BP measurement protocol used by the individual practices. All BP data were extracted from patients' EMR log of BP taken during patient's routine office visits. Thus, BP measurements were carried out by nurses or certified medical assistants with mercury sphygmomanometers.21 For this purpose, patients were required to be seated with their arms bare, and at least one BP reading was taken for each patient (as indicated in the EMRs).
Statistical analysis. The sample size was determined by a power analysis using a moderate change in adherence rates as the effect size, power of 0.80, and significance level of α = 0.05. This analysis suggested a sample size of 86 patients per group, but 190 patients were randomized (95 per group). The primary outcome was MEMS adherence between 10 and 12 months. However, MEMS data were categorized as missing for those patients who dropped out of the study and those who did not return their MEMS caps. To conform to intent-to-treat principles, mixed-effects regression models were used to test the time × group interaction for the primary analysis. For this purpose, all available data for all patients was used to estimate and test the intervention effect. However, the description of the results and follow-up contrasts focused on the group differences between 10 and 12 months. In addition, we carried out extensive evaluation of the missing value patterns to determine whether data were missing at random before testing the intervention effects. For the 30 patients (16%) for whom there were no MEMS data due to damage of the MEMS pill caps, we evaluated the extent to which these patients differed from other patients on all available data and undertook sensitivity analyses to assess the possible effects of the missing data on the study results.
Results
We screened 529 patients, of whom 330 were eligible, and we enrolled 190 into the trial (Figure 1). Their mean age was 54 years, 88% were female, 17% were married, 77% had high school or college education, with more than a half unemployed. The baseline mean systolic BP was 144 mm Hg (s.d. = 19.2), with a mean diastolic BP of 86.6 mm Hg (s.d. = 11.4). Of the patients, 45% had a Charlson comorbidity score ≥3 with one-third reporting diabetes, 8% had heart failure, and 4% had kidney disease. As shown in Table 1, there were no significant differences between both groups at baseline, and their adherence rates were also similar (56.2 ± 35.5% for MINT vs. 56.6 ± 34.1% for UC, P = 0.948).
Consort flow diagram.
Medication adherence
Completer analysis. A total of 190 MEMS pill caps were distributed to patients. Of these, 160 caps were returned (84%), and 30 patients (16%) had no usable MEMS data due to damage. Of the 160 patients with MEMS data, 111 (70%) had complete MEMS data. Among these 111 patients, the MINT group had a higher adherence rate compared to UC (60% vs. 47%, respectively, P = 0.054) with a between-group difference of 13% (95% confidence interval, −0.2–27%).
Intent-to-treat analysis. The 160 patients with MEMS data contributed a total of 520 observations to the analyses. If all 160 patients had complete data, there would have been 640 (four assessment periods × 160 patients) MEMS observations. Thus, the available data represents 520/640 = 81% of the possible MEMS data. The primary analysis was mixed-effects regression using all 520 observations. A critical assumption for using mixed-effects regression with incomplete data is that the data were missing at random. Based on Little's MCAR test, we could not conclude that the data were MCAR (χ2 = 26.7, degrees of freedom (df) = 15, P = 0.031). Thus, we created a variable indicating whether or not a patient had missing data, and correlated this variable with adherence rates at the four assessment periods. The correlations between missingness and these adherence rates were small and nonsignificant. Most important, the correlation with the final posttreatment adherence rates at 10–12 months was −0.015, P = 0.872, suggesting that the data can be treated as missing at random.
Overall, there was a significant reduction in adherence throughout the study period that corresponded to a reduction in adherence rate of 4% per quarter, and the test of this overall downward linear trend was significant (t with 127.5 df = −2.77, P = 0.006). To test whether these trends differ between the MINT and UC groups, models containing a time (quarterly trend) × intervention interaction were fitted. The parameter for the critical time × intervention interaction of.0456 was statistically significant (t with 121.4 df = 2.24, P = 0.027). To interpret this interaction, the parameter estimate from the mixed-effects regression model was used to predict the pre- and postintervention adherence rates for both groups. Using mixed regression analysis, a significant time × group interaction with model-predicted postintervention adherence rates of 43% and 57% was found for the UC and MINT groups, respectively (P = 0.027), see Table 2.
In order to describe the observed differences in the adherence pattern across time for both groups, we fitted a LOWESS smoother to the data obtained (Figure 2). Then we evaluated the pre-post differences in adherence separately for the UC and MINT groups in follow-up contrasts, using mixed regression but with time (assessment period) specified as a four-level factor. In this analysis, there was a significant drop between the baseline vs. postintervention adherence rate for the UC group (−12.9%, t = −2.89 with 159 df, P = 0.004) and a slight increase between the baseline vs. postintervention adherence rate for the MINT group (1.1%, t =.24 with 159 df, P = 0.810). This analysis indicates that the primary effect of MINT was to prevent the decline in medication adherence observed in the UC group. Based on the parameter estimates from the mixed regression analysis, there was a significant overall drop in adherence of about 4% per quarter, but this effect was offset with an additional gain of about 0.5% per quarter for the MINT group. Thus, there was a significant decrease in adherence for the UC group from baseline to 12 months and an overall nonsignificant increase in adherence for the MINT group; this difference is reflected in the significant group × contrast interaction.
Results of LOWESS Smootehr Curve Testing the Time × Treatment Interaction. All the individual data points are shown as 0 = UC, × = motivational interviewing (MI). The fitted line is a LOWESS smoother. Period 1 = preintervention average; 2 = average after first MI session; 3 = average after second MI session; 4 = average after third MI session.
Sensitivity analyses of patients without MEMS data
As stated previously, 30 patients had no usable MEMS data. These 30 patients were evenly split between the MINT and UC groups (N = 14 and 16, respectively) and there was no difference between the 160 patients who had MEMS data and the 30 who had none on any of the baseline measures. Nonetheless, the extent to which the results of the primary analyses would change under various scenarios of hypothetical data for these 30 patients was evaluated. Under the most extreme conservative case where the adherence rates for these 30 patients is the opposite of those found in the 160 patients with data (i.e., the 14 UC subjects had higher adherence rates than the 16 MINT subjects), the critical time × treatment effect is still in the direction that supports a treatment effect, but decreases from.046 to.028 and is no longer conventionally significant (P = 0.098). Under a less extreme scenario where the effect on these 30 subjects does not reverse, but simply averages the same in both groups the effect becomes smaller (0.037 vs. 046, and is marginally significant, P = 0.067). These results are based on very conservative assumptions about what might have happened with the 30 randomized subjects who had no data. Perhaps, the most reasonable assumption is that these 30 subjects would have behaved like their counterparts in the UC and MINT groups who did provide data; under this assumption, the results would be even stronger than those reported.
BP
Mixed regression analysis was used to compare the effects of MINT vs. UC on systolic and diastolic BP. Parameter estimates from the mixed regression analysis indicated a significant overall drop in systolic BP of 5.1 mm Hg across 12 months for both groups (t with 145.1, df = −2.26, P = 0.026), and the MINT group showed an additional drop (time × treatment interaction) of 6.1 mm Hg (t with 150.5 df = −1.86, P = 0.065; Table 2). For diastolic BP, there was also a significant overall drop of 3.5 mm Hg across the 12 months (t with 151.2, df = −2.61, P = 0.01), but the MINT group did not show an additional drop (time × treatment interaction, t with 156.7, df = −0.73, P = 0.465; Table 2). If we restrict data analysis to only those patients with BP data at baseline and 12 months, there are 61 UC patients and 52 MINT patients with complete data. For SBP, the average within-subject change for the UC group was −7.7 ± 19.8 mm Hg (t with 60 df = 3.04, P = 0.003), whereas the change for MINT group was −9.7 ± 22.2 mm Hg (t with 51 df = −3.14, P = 0.003) For DBP, the change for UC group was −4.25 ± 10.6 mm Hg (t with 60 df = 3.13, P = 0.003) and for the MINT group it was −4.60 ± 14.1 mm Hg (t with 51 df = 2.36, P = 0.022)
Discussion
In this practice-based trial, MINT counseling led to steady maintenance of medication adherence over 12 months, compared to a significant decline noted in the UC group. This effect was associated with a modest, nonsignificant trend toward a net reduction in systolic BP in favor of the MINT group. To our knowledge, this is the only randomized trial that tested the long-term effect of practice-based MINT on medication adherence and BP control in hypertensive African Americans. We are not aware of any practice-based behavioral intervention targeted at medication adherence in this patient population. Another major strength of our study is the long duration of medication adherence assessment with the objective MEMS. Previous adherence intervention trials suffered from the lack of objective measures of adherence, short-term duration of the outcomes assessed, and low minority participation.6,7,9,28
MINT is an increasingly popular patient-centered approach to behavioral management of chronic diseases in primary care settings.29,30 One mechanism through which MINT exerts its positive effects on health behaviors is enhancement of self-motivation,14,31 which increases patients' readiness to change and confidence in their ability to overcome barriers necessary to achieve a desired outcome. Although results from recent reviews indicate the positive effect of MINT on psychological, physiological, and lifestyle-change outcomes in patients with chronic diseases,29,30 only one pilot study assessed the effect of MINT on self-reported adherence to medications in HIV patients.32 In that study, patients randomized to the intervention group received three MINT sessions delivered by nurses whereas the control group patients received UC. Although not significant, at 2 months, patients in the MINT group reported higher self-reported adherence scores and fewer missed doses. Although this study provides preliminary evidence for the role of MINT in improving medication adherence, the lack of an objective measure of adherence and short duration limits the interpretation of the true effect of MINT on medication adherence. In our study, we assessed medication adherence for a much longer duration (12 months) with an objective and accepted gold standard for adherence assessment.
Although MINT was not used, similar effects of behavioral counseling approach on adherence in patients with other chronic diseases have been reported.33,34,35,36,37,38 However, majority of these studies had short duration of adherence monitoring (typically 3–6 months) and were not practice based.34,35,36,37,38 In one of the few MEMS studies that extended beyond a 6-month monitoring period, Weber et al. also reported steady maintenance of MEMS adherence in 60 HIV-positive patients randomized to monthly cognitive-behavioral therapy compared to UC for a 12-month period.33 Although there was no significant worsening of adherence in the intervention arm, the UC arm showed significant reduction of 8.7% per year in MEMS adherence (P = 0.006).33
Although not statistically significant, the net reduction in systolic BP of 5 mm Hg over 12 months in favor of the MINT group achieved in our study was clinically meaningful. Woollard et al. reported a similar finding in an 18-week trial that tested the effect of MINT vs. UC among 166 hypertensive patients followed in a general practice in Australia.39 In that study, patients randomized to six face-to-face, monthly MINT sessions, including one low-intensity, face-to-face session and five telephone MINT sessions delivered by trained nurses, showed net significant decreases in systolic BP at 18 weeks (−4 and −2 mm Hg, respectively, P < 0.05), compared with UC. In traditional antihypertensive drug trials, the magnitude of systolic BP reduction noted in our trial has been associated with significant cardiovascular risk reduction and mortality.22
We should note two limitations of this study. First, majority of patients were low-income women, which limits generalizability of our findings to the broader African-American population. Second, BP control is a multifactorial problem that is often linked to physician factors such as clinical uncertainty and treatment intensification.40,41 Thus, the BP reduction noted in both groups may be a reflection of greater treatment intensity that is often noted in hypertensive African Americans.42,43,44 More important, it is quite possible that despite the significant decline in medication adherence noted in the UC group, they may have experienced greater intensity in medication adjustment than the MINT group. But we did not collect this data and as such cannot make any statement about this assertion. Despite these limitations, our findings suggest that MINT counseling delivered every 3 months in a practice-based setting led to steady maintenance of medication adherence in hypertensive African Americans compared to UC. This effect was associated with a modest, nonsignificant trend toward a net reduction in systolic BP in favor of the MINT group. These findings set the stage for future studies to assess the cost-effectiveness of this approach for maintenance of adherence to prescribed antihypertensive medications in this high-risk population. Future studies should explore the integration of MINT into standard practice for this patient population, especially given its widespread use for self-management behaviors in patients with chronic diseases.
Disclosure:
The authors declared no conflict of interest.
Acknowledgements
The study was supported by grants R01 HL69408, RO1 HL078566, and R24 HL76857 from the National Heart, Lung, and Blood Institute and R03 TW007452 from the National Institutes of Health, Bethesda, MD, USA. The funding agency played no role in the design, conduct, or reporting of the study, or in the decision to submit this manuscript for publication.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
