Physician Characteristics as Predictors of Blood Pressure Control in Patients Enrolled in the Hypertension Improvement Project (HIP)

Abstract

The authors sought to examine the relationship between physician characteristics and patient blood pressure (BP) in participants enrolled in the Hypertension Improvement Project (HIP). In this cross‐sectional study using baseline data of HIP participants, the authors used multiple linear regression to examine how patient BP was related to physician characteristics, including experience, practice patterns, and clinic load. Patients had significantly lower systolic BP (SBP) (−0.2 mm Hg for every 1% increase, P=.008) and diastolic BP (DBP) (−0.1 mm Hg for every 1% increase, P=.0007) when seen by physicians with a higher percentage of patients with hypertension. Patients had significantly higher SBP (0.8 mm Hg for every 1% increase, P=.002) when seen by physicians with a higher number of total clinic visits per day. Patients had significantly lower DBP (−4.4 mm Hg decrease, P=.0002) when seen by physicians with inpatient duties. Physician’s volume of patients with hypertension was related to better BP control. However, two indicators of a busy practice had conflicting relationships with BP control. Given the increasing time demands on physicians, future research should examine how physicians with a busy practice are able to successfully address BP in their patients. J Clin Hypertens (Greenwich). 2011;13:106–111. © 2010 Wiley Periodicals, Inc.

It is estimated that hypertension affects more than 65 million Americans. ¹ , ² , ³ Despite numerous effective treatments and widely disseminated management guidelines, ⁴ the percentage of individuals whose hypertension is controlled remains suboptimal. ¹ One contributing factor to the low rate of blood pressure (BP) control involves physician adherence to guidelines. There is evidence of wide variation in physician management of hypertension even though guidelines for the treatment of hypertension (the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure [JNC 7]) are readily available. ⁵ Physician factors such as lack of motivation and awareness have been attributed to reduced adherence to current guidelines. ⁶ However, other factors such as time constraints and resources might influence physician adherence to recommendations. ⁶ Certain physician characteristics or practice patterns may also be associated with BP management. ⁷ , ⁸ Understanding these relationships can improve the design of interventions to improve BP control.

The purpose of this study was to identify the relationship of physician characteristics such as demographics, training, experience, practice patterns, and clinic load, on BP control in patients participating in the Hypertension Improvement Project (HIP) trial.

Methods

Overview

Our study was a cross‐sectional analysis of baseline data from the HIP trial, which was a 2×2 nested, randomized controlled trial of a physician intervention, patient intervention, and both interventions combined, compared with neither intervention. The methods for this trial are reported in more detail elsewhere. ³ , ⁹ In summary, the HIP trial included community‐based primary care practices in central North Carolina. A total of 4 matched pairs of practices were randomized between 2005 and 2007 to either the physician intervention or control (usual care). Practices were matched with regard to physician specialty and patient socioeconomic mix. Eligible patients were randomized to the patient intervention or control (usual care) independent of physician randomization assignment. All study procedures were approved by the Duke institutional review board.

The physician intervention lasted 18 months and involved 3 main elements: 2 online continuing medical education (CME) modules completed at baseline, an evaluation, and a treatment algorithm pocket card summarizing JNC 7 recommendations, behavioral change techniques, and continuous quality improvement–type procedures. The first CME module addressed JNC 7 guidelines and the second addressed lifestyle modifications for BP control. Each module required approximately 45 minutes to complete with a quiz to provide immediate feedback. The quality improvement procedure audited clinical performance and provided quarterly feedback to physicians. Physicians and/or their staff completed a brief, 1‐page form at each clinical encounter for each enrolled patient. The form collected pertinent clinical information (BP at current visit and prior visit, comorbidities, height, and weight) followed by clinical decision‐making information (current BP medications, whether BP was at goal, actions to manage BP, and follow‐up interval). These forms were also anonymously completed one clinic day per month for patients who were not enrolled in the study to enrich feedback to each physician. Data were summarized and presented quarterly in a letter to each physician. The reports displayed the proportion of hypertensive patients with adequately controlled BP for that quarter and previously, the proportion not at goal who had medication adjustments and/or received lifestyle modification counseling, the proportion taking JNC 7–recommended medications (ie, thiazide diuretics in patients without diabetes or chronic kidney disease [CKD], and angiotensin‐converting enzyme inhibitors or angiotensin receptor blockers in patients with diabetes or CKD), and comparisons with peer physicians for these categories.

The patient intervention was based on key theoretic constructs for guiding healthy behavioral changes. The patient intervention involved 20 weekly sessions provided during a period of 6 months conducted by a trained interventionist who was assisted by a community health advisor. The community health advisor was identified by physicians in the practice as a respected and locally involved community member capable of learning how to assist patients with behavior change. During the intervention sessions, participants met in small groups (10–15 patients per group) and received a comprehensive educational and behavioral program regarding diet, physical activity, and adherence to antihypertensive medications. After the initial 6‐month group counseling intervention, the community health advisors contacted participants by telephone each month for 1 year to offer brief lifestyle counseling. Patients randomized to the control program received a brief visit with the interventionist, during which they received advice and materials on lifestyle modifications for BP control based on JNC 7. At the end of the trial, these participants were offered an abbreviated version of the group counseling intervention that consisted of 6 weekly sessions.

The main results showed that the combination of medical doctor and patient interventions resulted in lower systolic BP (SBP) with a decrease of 9.7±12.7 mm Hg (P=.0072). In addition, there was a significant interaction between medical doctor and patient interventions (P=.03). Similar results were seen for participants’ diastolic BP (DBP), but these results did not reach statistical significance. ⁹

Study Population

Participating physicians were recruited from practices in the Duke Primary Care Research Consortium. In each practice, all physicians were invited to participate, with the goal of enrolling 4 physicians at each clinic. The number of physicians recruited for the study included 32 primary care practitioners from 4 internal medicine and 4 family medicine practices in central North Carolina serving a diverse patient mix in terms of age, sex, race, and health care coverage. Patients enrolled in HIP included adults (25 years and older) with a diagnosis of hypertension and currently under the care of one of the enrolled physicians. Patients were excluded if they had self‐reported CKD, a cardiovascular disease event within the past 6 months, or were pregnant, breastfeeding, or planning a pregnancy. All study participants (physicians and patients) provided written informed consent.

Data Collection

The following data were collected at baseline from each randomized physician using a self‐administered questionnaire: demographic data (age, sex, ethnicity/race), years since completed medical school, years since completed residency, percentage of patients seen with hypertension, percentage of patients seen with diabetes, average amount of time spent addressing hypertension at a visit, average amount of time spent counseling patients on lifestyle changes at a visit, and whether the physician had inpatient duties. Data collected from patients at baseline included height, weight, and BP, with BP measured twice at each of two separate visits (within 1 week) and averaged. All study measurements were collected by trained and certified study personnel who were blinded to treatment assignment. For the purposes of this analysis, only the measurements at baseline were used.

Analysis

This was a cross‐sectional analysis using baseline data of physician and patient participants. For this study we conducted the analysis in two steps. First, we conducted the analyses using the patient as the unit of analysis in order to maximize power. Second, we conducted the analyses using physician as the unit of analysis to confirm the patient level analyses. The latter sensitivity analyses did not yield substantially different results from those using the patient as the unit of analysis, therefore only the former analytic results are reported. In univariable regression models, we examined the association between BP (systolic and diastolic separately) and each of the physician characteristics. In multivariable regression models, we included as independent variables those variables found to be statistically significant (defined as P≤.10) in the univariable regression models. For the multivariable regression model, a statistical significance was defined as P<.05. The statistical analysis was conducted with SAS Enterprise guide 9.1.3 (SAS Institute, Cary, NC). All data are presented as mean ± standard deviation unless otherwise specified.

Results

Sample Characteristics

Characteristics of participating physicians (N=32) are presented in Table I. Mean age was 47.9 years (range, 35–78 years). Twenty‐five were white, 5 were African American, 1 was Hispanic, and 1 was Asian; 11 were women. All physicians were full‐time clinicians who saw a mean of 21.2 patients per day (range, 15–30). By self‐report, the mean percentage of patients seen with hypertension in a single day was 28.9%, whereas those with diabetes constituted 20.9%. Physicians reported spending a mean of 9.5 minutes per encounter addressing hypertension and 4.5 minutes counseling patients on lifestyle changes.

Table I.

 Physician Characteristics

Variable	Value
Study physicians, No.	32
Age, mean, (SD), y	47.9 (9.9)
Women, No.	11
Ethnicity, No.
White	25
Black/African American	5
Hispanic	1
Asian	1
Family medicine specialty, No.	17
Internal medicine specialty, No.	15
Years since completed medical school, mean (SD)	20.7 (10)
Years since completed residency, mean (SD)	17.8 (9.6)
Number of clinic visits per day, mean (SD)	21.2 (3.2)
Percentage of patients with hypertension, mean (SD)	28.9 (16)
Percentage of patients with diabetes, mean (SD)	20.9 (10.3)
Average amount of time spent addressing hypertension during a clinic visit, mean (SD), min	9.5 (4.6)
Average amount of time spent counseling patients on lifestyle changes, mean (SD), mean	4.5 (2.4)
Physicians with inpatient duties, No.	9
Very familiar with JNC 7 guidelines, No.	10
Percentage of adult patients, % (SD)	90 (10.6)

Abbreviations: SD, standard deviation; JNC 7, Seventh Report of the Joint National Committee on Prevention, Evaluation, Detention, and Treatment of High Blood Pressure.

Of the 574 participating patients, the mean age was 60.5 years (range, 28–94 years), 61% were women, and 37% were African American. At baseline, mean systolic BP (SBP) was 133.1±16.1 mm Hg and diastolic BP (DBP) was 74.1±11.3 mm Hg; 30% had diabetes, 3% had CKD, and 16% had a previous cardiovascular disease event (Table II).

Table II.

 Patient Characteristics

Variable	Value
Study patients, No.	574
Age, mean (SD), y	60.5 (11.4)
Women, No. (%)	350 (61)
Race/Ethnicity, No. (%)
White	349 (60.8)
Black/African American	214 (37)
Other (includes Asian, American  Indians, Hispanic)	11 (1.9)
Systolic BP, mean (SD), mm Hg	133.1 (16.0)
Diastolic BP, mean (SD), mm Hg	74.1 (11.3)
Systolic BP in patients with diabetes, mean (SD), mm Hg	134.5 (17.3)
Diastolic BP in patients with diabetes, mean (SD), mm Hg	71.6 (11.1)
Systolic BP in patients with CKD, mean (SD), mm Hg	142.5 (16.3)
Diastolic BP in patients with CKD, mean (SD), mm Hg	78.0 (10.2)
Systolic BP in patients with past cardiovascular disease event, mean (SD), mm Hg	131.9 (15.2)
Diastolic BP in patients with past cardiovascular disease event, mean (SD), mm Hg	70.8 (10.9)
Diabetes, No. (%)	171 (30)
CKD, No. (%)	17 (3)
Past cardiovascular disease event, No. (%)	93 (16)

Abbreviations: BP, blood pressure; CKD, chronic kidney disease; SD, standard deviation.

Univariable and Multivariable Regression

Unadjusted analyses revealed a significant association (P≤.10) between BP (systolic and/or diastolic) and the following: average number of clinic visits per day, percentage of patients seen who were adults, percentage of patients seen with diabetes, percentage of patients seen with hypertension, physician familiarity with JNC 7 guidelines, average amount of time spent addressing hypertension during a clinic visit, physician race, and inpatient duties. In the adjusted analyses, systolic BP was positively associated with average number of clinic visits per day and negatively associated with percentage of patients seen with hypertension. For each additional clinic visit per day, there was a 0.8 mm Hg increase in SBP (P=.002). For each 1% increase in percentage of patients seen with hypertension, there was a 0.2 mm Hg decrease in patient SBP (P=.008) (Table III). DBP was negatively associated with percentage of patients seen who were adults, percentage of patients seen with hypertension, and inpatient duties. For each 1% increase in percentage of patients seen who were adults, there was a 0.1 mm Hg decrease in DBP (P=.020). Similarly, for each 1% increase in percentage of patients seen with hypertension, there was a 0.1 mm Hg decrease in DBP (P=.0007). Lastly, having inpatient duties was associated with a 4.4 mm Hg decrease in DBP (P=.0002) (Table IV).

Table III.

 Multivariable Model Results: Predictors of Systolic Blood Pressure at Baseline

Variable	Parameter Estimate	P Value
Physician race, white	−2.5	.175
Percentage of adult patients	−0.1	.174
Average number of clinic visits per day	0.8	.002
Percentage of patients seen with hypertension	−0.2	.008
Percentage of patients seen with diabetes	−0.1	.408
Average amount of time spent addressing hypertension during a clinic visit, min	−0.3	.132
Familiarity with JNC 7 guidelines	0.9	.557

Abbreviation: JNC 7, Seventh Report of the Joint National Committee on Prevention, Evaluation, Detention, and Treatment of High Blood Pressure.

Table IV.

 Multivariable Model Results: Predictors of Diastolic Blood Pressure at Baseline

Variable	Parameter Estimate	P Value
Percentage of adult patients	−0.1	.020
Average number of clinic visits per day	0.2	.298
Percentage of patients seen with hypertension	−0.1	.0007
Percentage of patients seen with diabetes	0.1	.233
Physicians with inpatient duties	−4.4	.0002
Familiarity with JNC 7 guidelines	−1.5	.115

Abbreviation: JNC 7, Seventh Report of the Joint National Committee on Prevention, Evaluation, Detention, and Treatment of High Blood Pressure.

Discussion

Results indicate that certain physician characteristics are related to better BP control in the physicians’ patients. Lower SBP and DBP levels occurred in patients whose physicians saw a higher percentage of hypertensive patients. Lower levels of DBP also occurred in patients whose physicians saw a higher percentage of adult patients. Conversely, higher SBP occurred in patients whose physicians had a higher number of total clinic visits per day. These results are somewhat intuitive because physicians who see more patients with hypertension (or who see more adult patients) presumably become better at treating hypertension, whereas physicians with busier practices may have less time to spend with patients who have hypertension. Another finding, however, was less intuitive: patients had lower DBP if their physicians had inpatient duties in addition to their outpatient duties.

Results from this study did not confirm the previously reported relationship between physician gender and BP control (data not shown). In one study, patients with diabetes followed by female physicians were more likely to reach BP control than those followed by male physicians. ⁷ Similar sex differences have been reported regarding lipid management and hemoglobin A_1C measurement. ¹⁰

Our results did confirm results from multiple studies of other conditions, however, and of invasive procedures indicating that increased exposure (volume) to the condition or procedure translates into better outcomes. Examples include the medical treatment of conditions such as the acquired immunodeficiency syndrome and lupus erythematosus or invasive procedures such as coronary angioplasty. ¹¹ , ¹² , ¹³ The higher the disease‐specific or procedure‐specific volume, the better the outcome, including survival. This indicates that higher volume might increase the ability to more effectively treat the disease. ¹¹ , ¹² Interestingly, similar findings were not corroborated in conditions such as diabetes and pneumonia. ¹⁴ , ¹⁵ In a meta‐analysis published in 2002 looking at volume‐related outcomes in health care, the authors concluded that higher hospital and physician volumes are associated with better outcomes, but the magnitude of these relationships vary among conditions. ¹⁶ This study, however, did not assess the relationship between volume and hypertension control.

Explanations for some of our findings could include that physicians with a larger proportion of adult or hypertensive patients might be more comfortable treating these patients, more familiar with the intricacies of hypertension management, or more motivated to focus on hypertension control. Physicians who treat a large number of adults with hypertension may also be more likely to have systems in place in the clinic for carefully monitoring BP, intensifying treatment regimens when indicated, providing lifestyle counseling, and following up with patients after treatment recommendations are made. Such systems might not require additional time spent with the physician on hypertension management (which did not predict BP control) but still communicate the importance of BP control to the patient and overcome barriers to BP control.

The explanation for the relationship between inpatient duties and better BP control is not immediately clear; confirmation of this finding should be sought in future studies. One could speculate that physicians with inpatient duties may build systems into their clinic structure to efficiently address issues such as hypertension and that physicians with inpatient duties may also see fewer patients per day; however, it would seem that physicians with larger volume in general (ie, number of clinic visits per day) would also incorporate these efficiencies.

Limitations and Strengths

Our study results are based on physician self‐reported survey data; therefore, physician responses may be subject to recall and social desirability bias. Further, it is possible that participating physicians were more invested in providing better hypertension care and therefore more likely to participate in this research (volunteer bias), limiting the spectrum of our results and generalizability. In addition, relying on only the reported data precludes us from assessing unmeasured variables that might be predictors of BP control. Although we may have missed other potential predictors of BP control that were not measured as part of this investigation, an important strength of this study is that the data were collected from community‐based primary care clinics. Therefore, the results of this study are more likely generalizable to actual primary care practices.

Conclusions

In this study, looking at physician characteristics as predictors of BP control in participants enrolled in a randomized control trial, volume of patients with hypertension or who were adults, and inpatient responsibilities had beneficial relationships with BP control. Some of these findings appear to be consistent with the literature showing that patient outcomes improve as the physician’s exposure to a particular disease or procedure increases. Future work in this area is needed to determine ways that might simulate or substitute the beneficial effect that volume of patients with hypertension has on BP control. For clinicians who have lower volumes of patients with hypertension, establishing clinical systems for the management of hypertension is an intuitive strategy for improving BP outcomes.

Acknowledgments and disclosures:  The investigators grate‐fully acknowledge the valuable contributions of the study participants, research staff members, data and safety monitoring board members, community health advisors, and physicians and staff at the participating clinics (Durham Medical Center, Durham, NC; Harps Mill Internal Medicine, Raleigh, NC; Henderson Family Medicine Clinic, Henderson, NC; Hillsborough Family Practice, Hillsborough, NC; Metropolitan Durham Medical Group, Durham, NC; Oxford Family Physicians, Oxford, NC; Roxboro Medical Associates, Roxboro, NC; Triangle Family Practice, Durham, NC.). Leonor Corsino, MD, MHS, was supported by a T32 DK007012‐30S1 National Institutes of Health training grant at Duke University Medical Center. The Hypertension Improvement Project was supported by R01HL75373. The authors of this manuscript report that there are no conflicts of interest to disclose.