|Home | About | Journals | Submit | Contact Us | Français|
In general, adherence to blood pressure guidelines is low. We assessed whether hypertension recognition and control in North Carolina was consistent with the seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) in primary care practices participating in a quality improvement study regarding the implementation of the ATP3 cholesterol management guideline in primary care in North Carolina (GLAD Heart).
Demographic and clinical data were abstracted from 5,073 charts (patients aged 21–84 years, seen from June 1, 2001 to May 31, 2003) at 60 practices. Sites were non-university based primary care practices from 22 North Carolina counties. A dyslipidemia screening was defined as a lipid profile performed when not on lipid-lowering therapy. Among patients receiving a lipid profile, the proportion with diagnosed, undiagnosed, and controlled hypertension, was calculated according to JNC 7 guidelines. Practice level hypertension control was examined using the median and interquartile range across practices.
Among 1,763 patients screened for dyslipidemia, 49.4% had diagnosed hypertension. Only 67 individuals (3.8%) had undiagnosed hypertension. Although 85.8% of hypertensive patients were treated, the median proportions of patients with blood pressure below goal (<140/90, <130/80 with diabetes) was 33.3% (21.8% – 43.7%), with women more likely to be controlled and individuals treated by a solo provider less likely to be controlled.
These data were abstracted from the charts of patients who received a lipid profile; therefore, they are only generalizable to individuals who are screened for hyperlipidemia.
There remains a need to improve hypertension management in North Carolina primary care among patients screened for hyperlipidemia.
Hypertension is a strong and prevalent risk factor for cardiovascular disease (CVD). From 1999–2002, 30% of American adults had hypertension,1 which represents a significant public health burden, as CVD is the leading cause of death in the United States. Approximately 37% of individuals with hypertension are unaware of their condition.2 Elevated blood pressure is the leading diagnosis patients receive from their primary care provider (PCP) and accounts for 12 million office visits per year.3 Current blood pressure guidelines were released in 2003 in the seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7).4 The opportunity to measure blood pressure is present at every office visit, but according to the 2003 National Ambulatory Medical Care Survey, blood pressure is only measured at 52.4% of all physician office visits.3 Multiple measurements afford PCPs the opportunity to make frequent decisions regarding management of elevated blood pressure measurements. However studies suggest over two-thirds of individuals with hypertension have not achieved good blood pressure control.2
Despite widespread dissemination of JNC 7 guidelines, there is evidence that physicians are not following (and may not agree with) the recommendations. In a study by Huse and colleagues, only 60% of internists and 58% of general/family practitioners agreed with 140/90 as a threshold for hypertension.5 Oliveria and colleagues surveyed PCPs treating patients with hypertension and found that only 38% of visits resulted in initiation of or changes in medications, despite documented hypertension (>140/90) for at least six months.6
The Guideline Adherence for Heart Health (GLAD Heart) Trial is a practice-based, randomized controlled trial designed to test technology-based interventions on adherence to two cardiovascular disease prevention guidelines: the Third Report of the National Cholesterol Education Program’s Adult Treatment Panel (ATP3) and the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7). Providers in practices randomized to the ATP3 arm received a personal digital assistant with an ATP3 guideline-based cholesterol management software program. Practices in the JNC 7 arm received automated blood pressure devices. The goal of the current study is to report the levels of recognition and control of hypertension among individuals screened for dyslipidemia in practices recruited for the GLAD Heart Trial.
The current study consists of patient-level baseline blood pressure data from the GLAD Heart Trial.7 Details regarding the recruitment of our sample have been published elsewhere.8 Briefly, we recruited 61 primary care practices that were within an approximate three-hour driving radius of Winston-Salem, North Carolina; by design none was located at an academic medical center. Physician and nonphysician providers consented to participate and to have chart abstraction performed at baseline and at follow-up. Patient-level consent for chart review was not deemed necessary by the Wake Forest University Institutional Review Board. We complied with the Health Insurance Portability and Accountability Act (HIPAA) privacy directives.
Practice characteristics were obtained by way of a standard survey administered to participating providers at the initial orientation session and included provider training (internal or family medicine), education (physician, physician assistant, or nurse practitioner), practice size, and provider gender and ethnicity. Practices were considered to consist of predominantly female or minority providers if greater than 50% of providers (including non-physician providers) were women or non-white, respectively. Practices were considered to be urban if they were located in a city or town designated by the United States Census as an urbanized area; otherwise they were categorized as rural.
The Carolinas Center for Medical Excellence abstracted data via chart reviews at 60 of 61 practices (one practice was not open during the baseline data collection period). Chart abstraction methods are described in detail elsewhere.7 In short, a random sample of charts was pulled at each practice and those with a recent lipid profile were abstracted fully. Eligible patients included adults aged 21 through 84 years who had been seen in participating primary care practices from June 2001 to June 2003. Full medical record abstractions were completed on patients who were not on lipid-lowering therapy prior to the data abstraction window and had a lipid profile during the window, as the primary foci of the project were cholesterol screening and management. Data elements collected included demographics (age, gender, race/ethnicity), major comorbidities (coronary heart disease and diabetes), lipid profile values, additional cardiovascular disease diagnoses (stroke, peripheral vascular disease), CVD risk factors (smoking, diagnosed hypertension, antihypertensive medicine prescription, family history of heart disease), lipid-lowering medication, and blood pressures measured before and after the lipid profile.
To assess the intra- and interobserver reliability of the data being abstracted, 14 records per practice were reabstracted. A total of 858 records were abstracted in duplicate including, at baseline, 491 for intra- and 367 for interobserver reliability. Intra- and interobserver agreement was 95.2% and 89.9% respectively.
A person was considered to have diagnosed hypertension if (1) there was a note in the medical chart regarding a diagnosis of hypertension (HTN, or high blood pressure) or (2) they were prescribed antihypertension medications. We considered a person to have undiagnosed hypertension if (1) the two documented blood pressure values were at least 140/90 mmHG and (2) there was no documentation in the medical chart of a hypertension diagnosis or hypertension medications.
Blood pressure readings used in analysis consisted of clinic blood pressure readings taken at the last visit before the lipid profile and the first visit after the lipid profile. Persons with only one blood pressure value during the chart abstraction window could not meet criteria for undiagnosed hypertension; however 74 patients had a documented diagnosis and only one blood pressure value. Blood pressures were considered to be controlled if (1) below 140/90 mmHG in the absence of diabetes or (2) below 130/80 mmHG with diabetes.4 JNC 7 guidelines were used to define control, in order to be consistent with baseline feedback given to the providers and the intervention itself.
Frequencies, proportions, means, and standard deviations were calculated to describe the sample. Sampling weights based on approximations of the total number of patients in each practice were used to provide estimates applicable to the study base of patients receiving care in these practices. Weighted estimates of blood pressure control were calculated for each practice. The median and interquartile range for blood pressure control rates across all practices were computed. Multivariate logistic regression modeling was used to evaluate the relationship between blood pressure control and patient-level and practice-level characteristics. Sampling weights and a specification for clustering by practice were taken into account in all models. We considered the following characteristics: age (in categories of 21–44, 45–64, and 65–84 years), gender, race (non-white/white), diabetes (yes/no), cardiovascular disease (yes/no), solo/group practice, family/internal medicine, majority female/male practice, majority non-white/white practice, and urban/rural practice. Because race was missing from many records, we utilized multiple imputation to assign race for 198 patients. Multiple imputation is a technique whereby multiple simulated analyses are performed; in each analysis the missing data is replaced by plausible values and the results are combined to produce estimates and confidence intervals that account for the missing-data uncertainty.9 The patient-level characteristics and minority and solo practice variables were included as covariates in the model after displaying p-values less than 0.05 in unadjusted analyses. Significance was determined using two-tailed tests and an alpha=0.05. All analyses were performed using SAS version 9.1.
Assumptions that are made in the analysis are that the unweighted and weighted estimates are unbiased; the imputation methods preserve an unbiased complete structure of the data; the models are free of colinearity; and our weights and cluster structure give appropriate variance estimates.9 In our analyses, sampling weights were constructed to account for the unequal probability of selection of charts due to different screening rates across different practices. The application of sampling weights allows us to make unbiased inference about the target population. We also used a parametric (regression) method to build a multiple imputation model to impute missing data for race. The major underlying statistical assumptions are (1) the missingness of race can depend on the race observed in the dataset, but does not depend on the unobserved values and (2) the parameters of the data model and the parameters of the model for the missing data indicators are distinct.
We obtained chart review data from 60 primary care practices. Most were group practices (mean number of providers=3.6, range 1–14) and were staffed by family medicine providers. Our practices were diverse with respect to provider gender and ethnicity and included both rural and urban practices (see Table 1). A total of 5,742 charts were examined; of these, 5,073 patients were eligible to be screened for lipids. Figure 1 describes the exclusions which led to our final analytic sample of 1,763 patients. Patient demographics and comorbidities are presented in Table 1. Diabetes (prevalence=15.8%, range 0–42%) and heart disease (prevalence=8.3%, range 0–46%) were common comorbidities among patients in these practices.
Among these patients, 786 (44.6%) were determined to have a diagnosis of hypertension by a notation in their chart and an additiona l85(4.8%) were classified as having hypertension because they were on hypertension medications, yielding a total of 871 (49.4%) meeting our criteria for diagnosed hypertension. Only 67 individuals (3.8%) had undiagnosed hypertension. Table 2 describes the treatment status of individuals with hypertension. A minority, 14.8%, were not currently prescribed hypertension medications. However, those with a diagnosis who were not treated had a mean blood pressure that was below treatment goal, suggesting nonpharmacologic management may have been prescribed.
The percentage of patients with blood pressure below a common goal (<140/90mmHG for all patients) at practices performing blood pressure management was 41.2% (23.8–54.7%, median and interquartile range (IQR). The proportion of patients under control changes when taking into account the appropriate goal for diabetics. Median (and IQR) percentage of patients with blood pressure below goal (140/90mmHG without diabetes, 130/80mmHG with diabetes) across practices was 33.3% (21.8–43.7%). The substantial variability in hypertension control by practice is demonstrated in Figure 2 based on this more stringent criterion.
Despite an overall mean blood pressure below 140/90mmHG (see Table 1), two-thirds of patients with diagnosed hypertension had uncontrolled blood pressure within the data abstraction window. The mean blood pressure for patients with hypertension under control (by JNC 7 guidelines) was 122.8/74.8 mmHG, whereas the mean blood pressure for those patients with uncontrolled hypertension was 150.5/88.1 mmHG. Among the 484 patients with uncontrolled hypertension, 72.5% had systolic blood pressures between 140 and 159 mmHG, and only 27.5% had systolic blood pressures greater than 160 mmHG. Correlates of blood pressure control are shown in Table 3. Men were less likely to be controlled compared to women (OR = 0.64), however this difference did not reach statistical significance. Patients treated by a provider practicing at a solo practice were less likely to be controlled than those treated at a multiprovider practice. No other patient or practice characteristic was associated with blood pressure control.
In this sample of patients who were screened for dyslipidemia in primary care practices in North Carolina, we found evidence for suboptimal hypertension control. These data suggest there are opportunities for improvement in the diagnosis of hypertension, even among patients who are accessing primary care. Furthermore, despite high treatment rates there is ample opportunity for improving blood pressure control. Finally, we demonstrate substantial variability in the control of hypertension across practices.
These data are a snapshot of blood pressure control around the time of lipid management, and several limitations should be noted. We did not record the number of antihypertensive medications, nor the doses, and thus cannot comment on the intensity of treatment. We also did not obtain multiple blood pressure readings over time, which may have revealed lowered blood pressure with further follow-up. On the other hand, our data come from practices that agreed to participate in a quality improvement program; these practices may have had better care, on average, than general medical care in these communities. Patient level data was not linked to specific providers within practices. While not ideal, randomization was at the practice level and the number of patients from each practice was too low to differentiate by provider. Despite these limitations, our findings are consistent with other reports.
In studies assessing management of hypertension in primary care, reported rates of blood pressure control across the United States range from 23% to 54%.10–12 Control rates from the most recent National Health and Nutrition Examination Survey (NHANES) were 36.8% nationwide.13 We found, however, few data regarding hypertension treatment and control specifically in North Carolina. According to the Behavioral Risk Factor Surveillance System (BRFSS) survey, in 2005, 29.2% of North Carolinians said they had been told by a health professional that they had high blood pressure, and 77.4% of those with hypertension were currently taking medication for it.14 This survey only reports prevalence of self-reported hypertension, however, and does not include data regarding control.
Most of the literature assessing control of blood pressure has not considered the impact of the lower blood pressure goals for persons with diabetes or renal disease. We did not have measures of renal function or abstract renal insufficiency as a comorbidity. However, as the prevalence of diabetes was substantial in this population, the proportion controlled are significantly lower when taking into account diabetes. The poor control of blood pressure in persons with diabetes in this study is consistent with other reports.15
Given the significance of hypertension as a public health problem in the United States and the many clinical trials that have demonstrated the efficacy of blood pressure control in reducing CVD events,16 efforts to improve the diagnosis and control of hypertension should be a health care quality improvement priority. Furthermore, improved hypertension control in a state with a high stroke and heart disease incidence would likely yield meaningful reductions in the burden of CVD in the population.
Erica L. Rosenberger, Department of Epidemiology and Prevention in the Division of Public Health Sciences at the Wake Forest University School of Medicine. She can be reached at erosenbe (at) Email: ude.cmbufw.
David C. Goff, Jr, Department of Epidemiology and Prevention in the Division of Public Health Sciences and a professor in the Department of General Internal Medicine at the Wake Forest University School of Medicine.
Cralen C. Davis, Department of Biostatistical Sciences in the Division of Public Health Sciences at the Wake Forest University School of Medicine.
Caroline S. Blackwell, Department of Epidemiology and Prevention in the Division of Public Health Sciences at the Wake Forest University School of Medicine.
Alain G. Bertoni, Departments of Epidemiology and Prevention and General Medicine in the Division of Public Health Sciences at the Wake Forest University School of Medicine.