|Home | About | Journals | Submit | Contact Us | Français|
Maren K. Olsen PhD; Durham VAMC (152), 508 Fulton St. Durham NC 27705
Janet M. Grubber MSPH; Durham VAMC (152), 508 Fulton St. Durham NC 27705
Alice M. Neary RN; Center for Health Services Research in Primary Care, Duke University, 2424
Erwin Road, Hock Plaza, Durham, NC 27703
Melinda M. Orr M.Ed; Durham VAMC (152), 508 Fulton St. Durham NC 27705
Benjamin J. Powers MD; Center for Health Services Research in Primary Care, Duke University, 2424 Erwin Road, Hock Plaza, Durham, NC 27703
Martha B. Adams, MD Box 3230 DUMC, Durham, NC 27710
Laura P. Svetkey MD, Duke Hypertension Center at Sarah W. Stedman Nutrition and Metabolism Center 3475 Erwin Rd, Suite 100 Durham, NC 27705
Shelby D. Reed PhD, Duke Clinical Research Institute, 2400 Pratt Street, Room 311, Durham, NC 27705
Yanhong Li MS, Duke Clinical Research Institute, 2400 Pratt Street, Room 311, Durham, NC 27705
Rowena J. Dolor, MD, MHS, Duke Clinical Research Institute, 2400 Pratt Street, Room 311, Durham, NC 27705
Eugene Z. Oddone MD, MHS; Durham VAMC (152), 508 Fulton St. Durham NC 27705
Less than 40% of Americans with hypertension have adequate blood pressure (BP) control.
To compare two self-management interventions for improving BP control among hypertensive patients.
A 2 by 2 randomized trial stratified by enrollment site and patient literacy status with two-year follow-up (5/2004-1/2008).
Two university-affiliated primary care clinics.
636 patients were randomized (31% recruitment rate) among the 2060 eligible hypertensive patients.
Research assistants randomized eligible patients via a centralized blinded and stratified randomization algorithm to receive either: 1) usual care; 2) bi-monthly tailored nurse-administered telephone intervention targeting hypertension-related behaviors; 3) BP monitoring consisting of measuring BP three times per week, or; 4) a combination of the two interventions.
The primary outcome was BP control evaluated at six-month intervals over 24 months. 475 (75%) completed the 24-month BP follow-up.
Improvements in proportion of BP control for the intervention groups relative to the usual care group at 24 months were: behavioral group, 4.3% (95% CI: −4.5%, 12.9); home BP monitoring group, 7.6% (95% CI: −1.9%, 17.0%); and, combined interventions, 11.0% (95% CI: 1.9%, 19.8%). For systolic BP, relative to usual care, the 24 month difference was, +0.6 mmHg (95% CI: −2.2, 3.4) for the behavioral intervention group, −0.6 mmHg (95% CI: −3.6, 2.3) for the home monitoring group, and −3.9 mmHg (95% CI: −6.9, −0.9) for the combined interventions. Similar patterns were observed for diastolic BP at 24 months.
Changes in medication use and diet were only monitored in intervention participants; 25% lacked 24 month outcome data; 73% had adequate BP control at baseline; the study setting was an academic health center, all factors that potentially limit generalizability.
Combined home BP monitoring and tailored behavioral telephone intervention improved BP control, systolic BP, and diastolic BP at 24 months relative to usual care.
Despite the importance of hypertension and availability of multiple treatment options, only a third of U.S. patients with hypertension have their blood pressure (BP) under effective control (1). Home BP monitoring is a potentially important aspect of successful hypertension management (2). Adopting a healthy lifestyle and adhering to provider recommendations also is important in managing hypertension. Few educational and BP monitoring studies provide long-term follow-up (>12 months) or implement a multidimensional intervention tailored to patients’ needs and delivered by telephone (3). A tailored telephone behavioral approach and home BP monitoring could potentially provide a cost-effective method for improving BP outcomes (4).
This study compares the effectiveness of BP self-monitoring alone or in combination with a nurse-administered tailored behavior self-management intervention to usual care.
Using a 2 by 2 design, patients were randomized to one of four groups: 1) usual care; 2) nursed administered telephone-based tailored behavioral intervention; 3) home BP monitoring or 4) a combination of the two interventions. Recruitment occurred from May 2004 to December 2005 with follow-up from November 2004 to January 2008. The study was approved by the Duke Institutional Review Board and all patients provided written informed consent.
Potentially eligible study patients were identified through weekly data extractions from the billing and appointment database for two Duke University Health System primary care clinics. Patients were cared for by 7 faculty general internists in one clinic and 85 residents under the supervision of faculty at the other.
Initial inclusion criteria were: 1) diagnosis of hypertension at least 12 months prior to the data pull date (ICD9 code of 401.0, 401.1, or 401.9), 2) enrollment with a primary care physician at the included clinic for at least 12 months prior to data extraction, 3) self-report currently taking anti-hypertensive medication, 4) scheduled non-lab primary care provider appointment during the next 30 days, and 5) resident in one of 32 specified zip codes in the areas surrounding Duke University Health System.
Exclusion criteria applied prior to randomization were: 1) diagnosis of dementia, Parkinson’s disease, atrial fibrillation, or end stage renal disease; 2) patient of a study investigator or physician not expected to remain at the practice during the entire study period; 3) resident in nursing home or receiving home health care; 4) hospitalization for stroke or heart attack, surgery for blocked arteries, or diagnosis of metastatic cancer in the prior 3 months; 5) poor vision or difficulty hearing on the telephone; 6) difficulty understanding English on the telephone; 7) participant in another BP study; 8) spouse participating in current study; 9) arm circumference >17 inches and wrist circumference > 8.5 inches; 10) pregnant or planning to become pregnant in next 2 years; and 11) does not receive most of medical care from Duke clinics.
Patients were also excluded at post-randomization time periods for the following reasons: 1) no longer receiving medical care at the Duke clinics; 2) initiating dialysis; 3) receiving an organ transplant in prior 6 months; 4) residing in a nursing home or receiving home health care; 5) having no phone; or 6) being diagnosed with pulmonary hypertension in prior 6 months.
Study members mailed qualifying patients a letter from the patient’s primary care provider giving information and stating that the study team may contact them to participate in the study. Research assistants made weekly screening phone calls to patients from a randomly ordered list of eligible patients with upcoming clinic appointments.
Overall, 2060 letters were mailed to patients inviting them to participate in the study. Research assistants attempted to contact 1728 potential participants by phone. Six hundred and fifty-six were enrolled and consented; the remaining 1072 were not consented. Twenty additional patients were excluded at the time of the baseline interview (see Figure 1 for study flow).
The remaining 636 eligible patients were randomized to one of four groups: usual care, tailored behavioral phone intervention alone, home blood pressure monitor alone, or both interventions. Randomization was stratified at baseline by enrollment site (two primary care clinics) and literacy status (≥ 9th vs. < 9th grade as determined by the Rapid Estimate of Adult Literacy in Medicine (REALM (5))). Within each stratum, patients were randomized using consecutively numbered envelopes. Randomization blocks of size 16 were computer generated by the study statistician and used to ensure ongoing balanced enrollment across the four groups. The research assistants were blinded to block size, and patient randomization sequences were maintained by the study statistician in a separate office away from the clinics. At completion of the baseline interview, a research assistant opened an envelope and disclosed the patient’s randomization status. Participants were reimbursed $25 for the baseline visit and for each of the four subsequent 6-month blood pressure measurements ($125 total).
Patients randomized to usual care received their hypertension care from their primary care provider. They were not provided home BP monitors, nor did they have access to the nurse-administered behavioral intervention. They underwent the same 6 month outcome assessment measurements as other groups.
Patient factors targeted in the tailored intervention included perceived risk of hypertension, memory, literacy, social support, patients’ relationships with their health care providers, and side effects of anti-hypertension medication. In addition, the intervention focused on improving adherence to the following hypertension recommendations: the Dietary Approaches to Stop Hypertension (DASH) dietary pattern, (6–9) weight loss, (10, 11) reduced sodium intake, (11, 12) regular moderate-intensity physical activity, (13, 14) smoking cessation, and moderation of alcohol intake. (15)
The intervention was delivered by a single nurse during bi-monthly telephone calls. All information was presented in an easily understood format with a Flesch-Kincaid readability (16) score of <9th grade. Each encounter included a core group of modules potentially implemented during each call (e.g., medication and side effects) plus additional modules activated at specific intervals (e.g., diet, social support).(17)
Patients randomized to the home BP monitor intervention received an Omron HEM 773AC arm monitor if arm circumference was ≤ 17 inches. If arm circumference was > 17 inches, individuals received an Omron HEM 637 wrist monitor (if wrist < 8.5 inches). Two research assistants trained patients in proper use of the home BP devices. At each 6-month outcome assessment, patients were retrained if their BP assessment procedure was incorrect. Patients were asked to take their BP three times per week on three separate days, at the same time of day, and record their values in a log. Patients were asked to mail their logs in every 2 months using study-provided pre-addressed, stamped envelopes.
Patients randomized to the combined intervention received a home BP monitor, training on its use, and the bi-monthly nurse-administered behavioral self-management intervention. The nurse did not examine home BP values and did not use the home BP values to adjust the intervention.
Patient demographic information and diagnosis of diabetes was obtained from patients during a face-to-face baseline interview. Inadequate income was assessed by asking patients to report whether they had enough money to pay bills by cutting back on things; or difficulty paying bills(18). Health literacy was evaluated as a dichotomous variable with low literacy defined as The Rapid Estimate of Adult Literacy in Medicine (REALM) (19) score 0–60 (<9th grade level) and adequate literacy defined as REALM score 61–66 (≥9th grade level). (20) Body mass index were obtained form patients’ medical records.
The primary outcome was the proportion of participants with adequate BP control at each time point (baseline, and 6, 12, 18, and 24 months) over the 24 months. At each time point, BP control was defined based on Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure guidelines (21) (systolic BP < 140 and diastolic BP < 90 mmHg for patients without diabetes, and systolic BP <130 and diastolic BP <80 mmHg for patients with diabetes). Secondary outcomes were systolic and diastolic BPs at each time point over 24 months. BP was measured at each time point using a digital sphygmomanometer (BPTRU Automated Non-invasive BP monitor, Model BPM-100). The readings were taken by a research assistant who was blinded to the patient’s randomization assignment. Two BP readings were taken; the first after patients were seated and rested, for at least 5 minutes, and the second was 30 seconds after the first. The mean of the two values was used as the outcome for that time point.
We collected data on secondary outcomes including hypertension knowledge and self efficacy; however these results are not reported in this manuscript.
Electronic data representing medical resource use within the Duke University Health System was obtained for all enrolled patients representing the period from randomization through 24 months. These data included the cost estimates for all health care services.
The cost of each patient intervention encounter was based on an estimate of annual fixed costs (e.g. office space, telephone, etc.) and the nurse’s salary, and assuming the completion of seven encounters per six-hour day. This estimate was assigned to each completed patient encounter. In addition, our cost estimates included variable costs which included a relatively small amount for patient education materials (e.g., paper, toner). Variable costs for home BP monitoring costs included the cost of the BP monitor and batteries.
We did not anticipate any specific adverse effects of the intervention. We monitored for cardiovascular adverse events such as myocardial infarctions, mortality, and hospitalizations via medical record and patient report. The study had a Data Safety and Monitoring Board which met annually to review all adverse events.
Sample size estimation was based on the primary hypothesis that patients randomized to an intervention group would have improved BP control over the 24-month study period as compared to usual care. A linear change in BP control was assumed, so the comparison was a difference in slopes (i.e., the treatment by time interaction in a logistic mixed-effects regression model (22)). Sample size and power estimates were generated empirically via a simulation study using PROC NLMIXED in SAS v 9.1, Cary NC. From previous studies (23, 24), baseline BP control was estimated as 40%, and the 24-month dropout rate as 15%. The random intercept variance component was assumed to be 0.7 (equivalent to a patient interclass correlation=0.18) (22). To detect a difference in slopes resulting in a 10% improvement in BP control at 24 months with 80% power and a type-I error rate of 5%, 570 patients were needed; however, to account for dropout, we enrolled 636 individuals.
A logistic mixed-effects regression model (22) was used to estimate change in BP control over time for each of the intervention groups relative to usual care (i.e., a difference in slopes). Observed means and exploratory analyses indicated that, on average, BP control improved in a linear trend over time and there were equal correlations between patients’ repeated measurements. The fixed effects for the model were coded to reflect the 2×2 factorial design and included a common intercept (25), the two stratification variables, time (in months), and the following interaction terms: behavioral-by-month, home monitoring-by-month, and behavioral-by-home monitoring-by-month. A random effect was included to account for the correlation of patients’ repeated measurements over time. Estimates for the proportion in BP control for each intervention and usual care group at 12 and 24 months were converted to marginalized probabilities (22). One thousand bootstrap samples were used to derive confidence intervals around the estimated differences of the marginalized probabilities between each of the intervention groups relative to usual care at 12 and 24 months (26).
For systolic and diastolic BPs, general linear models (PROC MIXED in SAS v9.1) were used to estimate changes in BP over time and to test for BP differences in the intervention groups relative to usual care at 12 and 24 months. Exploratory analyses indicated that both systolic and diastolic BPs had quadratic shapes in which improvements in BPs were greatest during the first half of the study. Predictors in the model were coded to reflect the 2×2 factorial design. The final model included a common intercept, the two stratification variables, and the following interaction terms: behavioral-by-month, home monitoring-by-month, behavioral-by-home monitoring-by-month, behavorial-by-month2, home monitoring-by-month2, and behavioral-by-home monitoring-by-month2.
Patients were analyzed based on initial randomization group (intention to treat); n=634 patients were included in the analyses (see Figure 1). All available data, including data from participants who subsequently discontinued the study, were used for analyses. Our analysis techniques assumed ignorable dropout, meaning that the probability of dropout may depend on covariates in the model or participants’ previous responses, but not on current or future responses (22).
The funding sources did not have any role in the study’s design, data collection, administration of the interventions, analysis, interpretation, data reporting, or the decision to submit these findings for publication.
For the 636 patients in the study, mean age was 61 years, 49% were African-American, 66% were female, and 19% reported having inadequate incomes (see Table 1). Baseline mean systolic BP was 125 mmHg (SD=18) and diastolic BP was 71 mmHg (SD=11). Seventy-three percent of participants had their BP under control at baseline. There were no meaningful differences in baseline characteristics by treatment group.
Non-participants (n=1424) were similar to study participants (n=636) in gender (64% female) and diagnosis of diabetes. However, non-participants were slightly older (mean age 63) and more likely to be African American (57%) than were study participants. There was no cross-over; however, it is possible that individuals not assigned to the home BP intervention groups used their own monitors. There were no study-related adverse events in any intervention group.
Overall, 475 (75%) patients had BPs at the 24 months follow up (see Figure 1). The percent of patients with 24 month follow-up BPs in the usual care, behavioral self-management, home BP self-monitoring, and combined home BP monitoring/behavioral intervention groups was 81%, 78%, 72%, and 69%, respectively (p=0.075). Patients with lower literacy levels at baseline and those enrolled at one of the two clinics also had lower study completion rates. Sixty-three percent of lower literacy patients versus 79% of higher literacy patients (p<0.001), and 57% of patients at one of the clinics versus 82% (p<0.001) at the other clinic, had BP readings at 24 months.
Patients randomized to the combined behavioral and home BP monitor group showed the greatest improvement in the proportion of BP control over the study period (see Figure 2). At 24 months, the adjusted improvement in proportion of BP control for patients in the combined group compared to usual care was 11.0% (95% CI: 1.9%, 19.8%; p=0.012) (see Table 2). As compared to usual care, the relative improvement for the behavioral group was 4.3% (95% CI: −4.5%, 12.9%; p=0.34) and for the home BP monitoring group was 7.6% (95% CI: −1.9%, 17.0%; p=0.096) at 24 months. The 3-way interaction of behavioral-by-home monitoring-by-month was not significant (p=0.99) indicating that the two interventions did not have a synergistic effect upon BP control over time.
The largest sustained improvement for systolic and diastolic BP was observed in the combined home BP monitor and behavioral intervention group (see Figures 2 and and3).3). In these analyses, the 3-way interaction of behavioral-by-home monitoring-by-month2 was significant (p=0.041 for systolic and p=0.004 for diastolic model) indicating that the main effects of home BP monitoring and the behavioral interventions on blood pressure over time are enhanced by the presence of each other (i.e., the combined effect) (see Table 2).
Changes in estimated mean systolic BPs over time are shown by intervention group in Figure 3. At 12 months, the mean systolic BP was 1.6 mmHg (95% CI: −3.9, 0.7; p=0.174) lower in the behavioral group, 3.7 mmHg (95% CI: −6.1, −1.2; p=0.004) lower in the home BP monitor group, and 3.3 mmHg (95% CI: −5.7, −0.8; p=0.009) lower in the combined group than the usual care group. However, by 24 months, the mean systolic BP was statistically significantly lower only in the combined group as compared to the usual care group indicating the synergistic effect of the home monitoring and behavioral interventions on improving BP over time. Relative to usual care, the adjusted 24 month difference was +0.6 mmHg (95% CI: −2.2, 3.4, p=0.67) among patients in the behavioral group, −0.6 mmHg (95% CI: −3.6, 2.3, p=0.69) among patients in the home monitoring group, and −3.9 mmHg (95% CI: −6.9, −0.9, p=0.010) among patients in the combined intervention.
Estimated mean diastolic BPs over time are shown in Figure 4. At 12 months, the mean diastolic BP was 1.4 mmHg (95% CI: −2.6, −0.14; p=0.029) lower in the behavioral group, 3.1 mmHg (95% CI: −4.4., −1.8; p<0.001) lower in the home BP monitor group, and 2.2 mmHg (95% CI: −3.5, −0.8; p=0.001) lower in the combined group relative to the usual care group. Similarly to systolic BP, the synergistic effect of the home monitoring and behavioral interventions is evident by the end of follow-up. At 24 months, relative to usual care, the adjusted difference was 0.4 mmHg (95% CI: −1.1, 1.9, p=0.61) higher among patients in the behavioral intervention group, 1.2 mmHg (95% CI: −2.9, 0.4, p=0.132) lower among patients in the home monitoring group, and 2.2 mmHg (95% CI: −3.82, −0.6, p=0.009) lower among patients in the combined intervention.
160 patients were randomized to the tailored behavioral self-management intervention. The nurse completed 1682 phone calls to these patients during the study. The mean number of completed calls per patients was 11 (SD=2), out of a possible 12, and the mean call length was 16 minutes (SD=7).
158 patients were randomized to the home BP monitoring intervention. Of these, 91% of patients in the first two months of the study and 64% of patients in the last two months turned in BP logs with at least one recorded BP reading. The percentage of patients who turned in their logs was higher among patients who completed the study.
159 patients were randomized to the combined intervention group. Of these, 89% and 59% turned in BP logs with at least one recorded BP reading for the first and last 2 months of the study, respectively. Among participants with a BP measurement at 24 months (n=110), the first and last BP log “turn-in” percents were: 99% and 81%, respectively. Nurses completed 1589 phone calls to 156 of the 159 patients in the combined intervention group during the study. The mean number of completed calls per patient was 10 (SD=3), out of a possible 12. The mean call length was 16 minutes (SD=7).
The numbers of outpatient encounters over the 24 months were similar across the four groups; median counts ranged from 13–15 (p=0.73). There were no differences in the proportion of individuals hospitalized (range from 19.5% to 22.6%, p=0.91). Two-year medical costs across the four groups showed a mean of $15,641 (SD=$25,769, median= $6698).
Intervention costs over two years were estimated at $345 for the behavioral intervention, $90 for home BP monitoring, and $416 for the combined intervention (not including patient time costs). In sensitivity analyses, mean costs to implement the combined behavioral and home BP intervention ranged from $208 to $811 per patient depending on assumptions regarding the mean time between successful encounters and various direct costs of implementing the intervention.
Self-reported medication adherence and exercise at 6 and 24 months improved slightly over time though comparisons of changes in these variables between interventions arms and the usual care group were not statistically significant (data not shown).
We examined the effects of a patient behavioral intervention delivered by telephone, home BP monitoring, and a combination of both in an attempt to improve BP among hypertensive adults treated in primary care. Neither intervention alone resulted in improvements in BP control at 24 months; however, the combination of home BP monitoring and behavioral phone intervention resulted in a clinically significant improvement in blood pressure control, resulting in an 11% improvement at 24 months compared to usual care. Patients in the combined group also showed a clinically meaningful decrease in systolic BP of 3.9 mm Hg compared to usual care. These effects were observed with a brief telephone intervention (mean phone call length=16 minutes; mean number of calls=10 over 24 months) implemented bi-monthly and the use of home BP monitors which required minimum patient training (<5 minutes every 6 months) over 24 months.
Home BP monitoring alone has been well studied as a method to improve BP control. Its main effect is thought to be on BP recognition, which may lead to improved adherence and better control. Our results regarding home BP monitoring are consistent with a meta-analysis of 18 randomized controlled trials that compared home BP monitoring with usual care and found that home BP monitoring resulted in small improvements (2.2/1.9 mm Hg reduction) in BP (27). Positive feedback (i.e., seeing BP decrease) encourages the patient to continue treatment. Continued high BP readings may encourage appropriate alterations in lifestyle or therapy (28) and medical regimens may be changed sooner (29).
A literature search of pertinent work using the following terms, ‘self-management’, ‘behavior intervention’ through June 2009 yielded seven randomized controlled trials. These trials reported mean differences in systolic BP ranging from −15.7 to 0.6 mmHg (30) with an average follow-up length of less than 12 months. However, the trials involved in-person contact, which potentially reduces scalability and increases intervention costs. In a separate review of 10 studies testing mostly in-person behavioral interventions, results favored counseling over usual care with improvements in systolic BP of 11.1 mmHg and diastolic BP of 3.5 mmHg (31). Few studies have implemented a multidimensional intervention that is tailored to patients’ needs and delivered by telephone, yet based in primary care practices. Furthermore, few studies have examined changes in BP for longer than 12 months, which was particularly relevant in the current study; the effects of using home BP monitors at 12 months indicated an almost 4 mmHg decrease in systolic BP, but these findings dissipated by 24 months for all but the combined intervention group. We also demonstrated that the intervention did not increase health care use and the cost was approximately $400 for two years to implement the combined home BP monitor and behavioral intervention.
The patient self-management intervention was only effective when combined with home BP monitoring. One proposed explanation is that self-management may be most effective when it includes on-going disease monitoring by patients that creates the opportunity to respond to new information.(32)
One limitation of our study was the high rate of BP control rate in the study population at baseline. The relationship between blood pressure and cardiovascular risk is continuous and blood pressure reductions translate to reductions in clinical events, even with observed relatively low baseline BP.(21) Changes in process measures such as medication use and salt intake during the trial were not measured in all groups. Finally, although 25% of the sample was not available at 24 months, we used mixed-effects models as our primary analysis tool. This analysis includes all patients with any BP measurements and leads to valid inferences.
Despite the availability of evidence-based hypertension guidelines, only a third of all U.S. hypertensive patients have their BP under effective control (21). A combined home BP monitoring and a tailored brief behavioral intervention resulted in a statistically significant improvement in BP control and decreases in systolic and diastolic BPs at 24 months with minimal costs. The combination of home monitoring with patient self-management interventions may be a valuable tool for improving blood pressure control rates.
The views expressed in this manuscript are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Grant support: Primary Funding Source: This research is supported by a NHLBI grant R01 HL070713, a Pfizer Foundation Health Communication Initiative award, and Established Investigator Award from the American Heart Association to the first author.
This is the prepublication, author-produced version of a manuscript accepted for publication in Annals of Internal Medicine. This version does not include post-acceptance editing and formatting. The American College of Physicians, the publisher of Annals of Internal Medicine, is not responsible for the content or presentation of the author-produced accepted version of the manuscript or any version that a third party derives from it. Readers who wish to access the definitive published version of this manuscript and any ancillary material related to this manuscript (e.g., correspondence, corrections, editorials, linked articles) should go to www.annals.org or to the print issue in which the article appears. Those who cite this manuscript should cite the published version, as it is the official version of record.
No authors have conflicts of interest.
The study protocol and analytic dataset may be requested from the first author.
Trial Registration: ClinicalTrials.gov