Variability in performance measures for assessment of hypertension control

Hypertension

Variability in performance measures for assessment of hypertension control Ann Marie Navar-Boggan, MD, PhD, Bimal R. Shah, MD, MBA, Joel C. Boggan, MD, MPH, Judith A. Stafford, MS, and Eric D. Peterson, MD, MPH Durham, NC

Background Definitions of multiple performance measures exist for the assessment of blood pressure control; however, limited data on how these technical variations may affect actual measured performance are available. Methods

We evaluated patients with hypertension followed routinely by cardiologists at Duke University Health System from 2009 to 2010. Provider hypertension control was compared based on reading at the last clinic visit vs the average blood pressure across all visits. The impact of home blood pressure measurements and patient exclusions endorsed by the American Heart Association, the American College of Cardiology, and the Physician Consortium for Performance Improvement were evaluated using medical record reviews.

Results Among 5,552 hypertensive patients, the rate of blood pressure control based on last clinic visit was 69.1%; however, significant clinic-to-clinic variability was seen in serial clinic blood pressure measurements in individual patients (average 18 mm Hg). As a result, provider performance ratings varied considerably depending on whether a single reading or average blood pressure reading was used. The inclusion of home blood pressure measurements resulted in modestly higher rates of blood pressure control performance (+6% overall). Similarly, excluding patients who met guideline-recommended exclusion criteria increased blood pressure control rates only slightly (+3% overall). In contrast, excluding patients who were on 2 or more antihypertensive medications would have raised blood pressure control rates to 96% overall. Conclusion Depending on definitions used, overall and provider-specific blood pressure control rates can vary considerably. Technical aspects of blood pressure performance measures may affect perceived quality gaps and comparative provider ratings. (Am Heart J 2013;165:823-7.)

Hypertension is one of the most prevalent, as well as one of the most modifiable, cardiovascular risk factors, affecting nearly 1 in 3 Americans. 1 Despite the availability of numerous effective antihypertensive therapies, challenges remain in achieving optimal blood pressure (BP) control. According to the most recent National Health and Nutrition Examination Survey data, nearly 50% of patients with hypertension have not reached their BP goal. 1 Given that hypertension plays a significant role in incident cardiovascular disease, numerous performance measures have been developed to assess BP control in hopes of assessing quality of care. The Centers for Medicare & Medicaid Services (CMS) includes hyperFrom the Duke Clinical Research Institute, Duke University Medical Center, Durham, NC. This project was supported by Grant No. U19HS021092 from the Agency for Healthcare Research and Quality. Hector O. Ventura, MD, served as guest editor for this article. Submitted November 19, 2012; accepted January 7, 2013. Reprint requests: Ann Marie Navar-Boggan, MD, PhD, 2400 Pratt St, Durham, NC 27705. E-mail: [email protected] 0002-8703/$ - see front matter © 2013, Mosby, Inc. All rights reserved. http://dx.doi.org/10.1016/j.ahj.2013.01.003

tension control as part of the Physician Quality Reporting Initiative on which providers can receive incentive payments. The CMS 2010 Physician Quality Reporting Initiative Narrative Measure Specification for hypertension defined hypertension control as follows: numerator, patients with last systolic BP (SBP) measurement b140 mm Hg and a diastolic BP (DBP) b90 mm Hg, and denominator, all patients 18 years and older with hypertension who had a BP measurement during the last office visit. The only acceptable exclusions are “documentation of medical reason(s) for not recording a blood pressure measurement (diagnosis of ESRD and pregnancy are the only acceptable exclusions).” 2 In 2011, BP control was evaluated for 3 groups of patients: patients with diabetes, patients with ischemic vascular disease, and patients with chronic kidney disease. 3 For patients with diabetes and ischemic vascular disease, the measure specified a numerator of patients whose most recent BP was b140/ 90 mm Hg of a total population of patients with each comorbidity. The measure differed for patients with stage 4 or 5 chronic kidney disease, with a target of b130/80 mm Hg. In addition, the numerator included patients

American Heart Journal May 2013

824 Navar-Boggan et al

with BP ≥130/80 mm Hg but who had a documented plan of care for BP management. The plan of care could include “recheck blood pressure at a specified future date, initiate or alter pharmacologic or non-pharmacologic therapy, or review of home blood pressure log.” All of these measures used the BP from the most recent clinic visit. Similarly, the National Quality Foundation endorsed using a single outpatient clinic BP measurement of b140/ 90 mm Hg for all patients with hypertension, with a denominator of all patients with hypertension aged 18 to 85 years, excluding patients with end-stage renal disease and pregnancy. 4 Currently, hypertension performance metrics endorsed by the National Quality Foundation in use by CMS reference BP measurements from a single clinic visit. BP at a single clinic visit does not accurately reflect a patient's true BP control; as a result, this isolated measurement may overestimate BP status. 5 This potential error raises methodological issues with the metric because variability in BP measurements across clinic visits may affect assessment of BP control by either patients or providers. The American Heart Association (AHA), American College of Cardiology (ACC), and the Physician Consortium for Performance Improvement (PCPI) have endorsed a new hypertension control performance metric for BP control to account for challenges in current metrics. The numerator for this measure is as follows: BP b140/90 mm Hg OR “patients with a BP ≥140/90 mm Hg and prescribed 2 or more antihypertensive medications during the most recent office visit.” The denominator is all patients 18 years and older with hypertension. This metric includes several key differences from currently used CMS performance measures. First, the AHA/ACC/ PCPI guidelines allow home BP measurements to be used for the assessment of overall control. 6 In addition, unlike other BP control metrics, patients prescribed 2 or more antihypertensive medications are considered controlled, regardless of their most recent home or clinic BP measurement. Also, the AHA/ACC/PCPI recommends patient exclusion from the metric if there is at least 1 documented reason for continually elevated BP (ie, no insurance, noncompliance, and/or medical indication for higher BPs). These departures from currently used performance metrics may be less easily captured, and the impact of these exclusions on overall rates of BP control remains uncertain. Using outpatient clinic data from a large tertiary academic cardiology practice, we evaluated the applicability and performance of these new guidelines in clinical practice by assessing the following: (1) the variability in provider BP control metrics for 1 year using an average clinic BP rather than a single BP measurement, (2) the impact of including home BP measurements, (3) how other patient exclusion criteria (endorsed by the AHA/ ACC/PCPI) affect provider BP control rates, and (4) the

proportion of patients with uncontrolled hypertension who are taking 2 or more medications.

Methods We identified all patients with hypertension who had at least 2 visits with the same cardiologist at Duke University Health System between June 1, 2009, and June 1, 2010. Patients with hypertension were identified using the International Classification of Diseases, Ninth Revision codes of hypertension (401-405) at any time before or within the first 6 months of the 12-month evaluation period. Blood pressure measurements were taken from electronic records of triage vital signs. In cases where BP was repeated and updated electronically or patients had multiple visits on the same day, the lowest SBP and DBP measurements were used. Cardiologists were included if they cared for at least 10 patients with hypertension during the study period. Patient-level demographic and clinical data were abstracted electronically and/or via medical record review. Using the final BP in the yearlong evaluation period, BP control for each patient was initially defined as SBP b140 mm Hg and DBP b90 mm Hg. Blood pressure control was then reevaluated based on the mean BP across all visits (average BP control) with similar cut points. Within-patient variability in BP measurements was assessed by calculating the difference between each individual reading and the patient's mean BP for the entire year. For patients with at least 2 visits ≥30 days apart, BP readings from the first and last visit in the study period were compared to assess changes in BP. Provider-specific statistics were calculated by evaluating the rate of BP control among patients seen by each provider. Patients were assigned to providers based on the last provider seen in the 12-month period. The proportion of patients with BP at goal was reported for each provider, and provider rankings were calculated using both last BP and mean BP. Correlations between provider ranking using last and mean BPs were assessed using unadjusted univariable linear regression. To assess the impact of recorded home BP measurements on rates of BP control, we performed manual medical record reviews for a random sample of 300 patients with uncontrolled BP based on last visit. This number was selected to obtain a precise estimate of the proportion of patients with explanations for uncontrolled BP, as the width of 95% CIs for point estimates using this sample size would be approximately 10% or less. The proportion of patients with elevated clinic BP but normal home BP was used to estimate the impact of including home BP measurements on overall hypertension control rates. Home BP control was determined based on qualitative provider documentation that home BP measurements were acceptable or within goal. Similarly, these medical record reviews were also used to determine if the patient met other AHA/ACC/PCPIacceptable exclusion criteria based on an explanation for uncontrolled BP. Acceptable exclusion criteria included medical indication for higher goal, patient noncompliance, or controlled BP at other clinic visits. Finally, medication records were evaluated with these medical record reviews. The proportion of patients with uncontrolled BP who met AHA/ACC/PCPI exclusion criteria and the percent of patients taking 2 or more medications were used to estimate the impact of these exclusion criteria on control rates. The Duke University institutional

American Heart Journal Volume 165, Number 5

Navar-Boggan et al 825

Table. Characteristics of the patient population (n = 5552)

Mean age (y) Female Race, white Race, black Private insurance Medicare/Medicaid No insurance Mean BMI (kg/m 2) Chronic kidney disease (GFR b60) Type 2 diabetes mellitus Congestive heart failure No. of visits with the same cardiologist

Frequency 66.4 41.2% 72.6% 24.7% 30.4% 68.0% 1.3% 30.6 33.2% 41.6% 51.1% 2.8 (range 2-20)

BMI, Body mass index; GFR, glomerular filtration rate.

60

Provider Rank Using Average BP

Characteristic

Figure

50

40 R2 = 0.71

30

20

10

0 0

review board approved this study. All analyses were conducted using STATA version 9.0 (College Station, TX). The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the manuscript, and its final contents.

Results Study population and BP variability We identified 5,552 unique patients with a diagnosis of hypertension who were seen at least twice by the same cardiology provider within a 12-month period within the Duke University Health System. Patients in the analysis cohort were seen by 51 different cardiologists. Table presents the characteristics of the study population. The average age of patients was 66 years. The majority (68%) were covered by Medicare or Medicaid. On average, each cardiologist had 469 (range 11-1,684) visits with eligible patients who had hypertension. Of the 5,552 patients included, 4,816 (87%) had at least 2 visits ≥30 days apart with BP available. These patients had a mean of 3 (range 2-16) visits with a cardiologist in which BP was documented. Forty-six percent (n = 2,231) of these patients had consistently controlled BP measurements (b140/90 mm Hg) across all visits, whereas 15% (n = 709) of patients had consistently elevated BP measurements across all visits. The remaining 1,876 patients (39%) had at least 1 clinic visit with an elevated BP (≥140/90 mm Hg) and 1 clinic visit with a controlled BP (b140/90 mm Hg) in the 12-month assessment period. The average within-patient variability was 18 mm Hg. Therefore, an individual clinic BP reading was, on average, 9 mm Hg above or below that patient's mean BP over the entire year. Of 4,816 patients with 2 or more visits, 1,653 (34.3%) had uncontrolled BP at the last clinic visit in the study period. Overall, 256 (15.4%) of these

10

20

30

40

50

60

Provider Rank Using Last Clinic BP

Provider performance ranking for clinic population BP control as defined by BP at last clinic visit vs average BP. Scatterplot of provider rankings for hypertension control performance as defined by BP at the last clinic visit and by average BP; correlation evaluated using linear regression.

1,653 patients had a ≥10-mm Hg decrease in SBP from the first clinic visit, whereas 196 patients (11.9%) had a ≥10-mm Hg increase in SBP from the first clinic visit.

Final vs mean BP We found that 1,715 (30.9%) of 5,552 patients had uncontrolled BP at their last clinic visit in the study period, and 423 (7.6%) had BP ≥160/100 mm Hg. Using average BP measurements, 1,554 (28.0%) had BP ≥140/ 90 mm Hg and 307 (5.5%) had BP ≥160/100 mm Hg. For 763 (13.7%) patients, classification of control using BP at the last visit differed from the classification derived from the average BP. Rates of uncontrolled BP within provider panels varied from 11.1% to 46.6% using average BP and from 15.6% to 47.3% using last clinic BP. Provider performance varied depending on whether average or last BP was used. There was moderate correlation between a provider's ranking generated by using average BP and ranking generated using last BP (r 2 = 0.71; Figure). Inclusion of home BP records Overall, 98 (32.7%) of 300 patients had documentation of home BP recordings. In manual medical record reviews of a random sample of patients with uncontrolled BP at the last clinic visit (n = 300), 14% (n = 43) of patients had documentation of home BP recordings at goal without other explanation for elevated BP. An additional 16 patients (n = 5.3%) had BP control documented at home but met ACC/AHA/PCPI exclusion criteria. Thirty-

826 Navar-Boggan et al

nine patients (13%) had documentation of uncontrolled home BP readings. Extrapolating to the study population, if 19.3% of all patients with uncontrolled hypertension in the study population had home BP recordings at goal and were considered “controlled,” the overall rate of hypertension control would increase by 6% (from 69% to 75%).

Inclusion of AHA/ACC/PCP patient exclusion criteria In manual medical record reviews of a random sample of patients with uncontrolled BP at the last clinic visit (n = 300), 13% of patients met AHA/ACC/PCPI-acceptable exclusion criteria. These reasons included patient noncompliance (7%), medical reasons for higher BP targets (5%), and goal BPs documented at other clinic visits (1%). Applying this to the entire study population by excluding these patients from the denominator, the rate of controlled hypertension would increase by 3% (from 69% to 72%). Patients on 2 or more antihypertensive medications Finally, of the 300 patients selected for medical record reviews, 86% of patients with uncontrolled BP were taking 2 or more antihypertensive medications. Across the study population, reclassifying 86% of patients with uncontrolled BP as controlled results in an increase in the rate of controlled hypertension from 69% to 96%.

Discussion Performance measures for a number of chronic conditions have been developed, and some are currently used for physician incentive payments. The CMS has included hypertension (along with diabetes mellitus, heart failure, and coronary artery disease) among 4 disease conditions on which data would be gathered in making determinations about provision of payment incentives for clinical services. 7 Multiple performance measures exist for hypertension, with some variation across the technical aspects of defining uncontrolled hypertension; however, limited data are available on how these variations affect measured performance. In this analysis of a large, tertiary academic cardiology practice, we found significant variability in individual BP measurements for patients followed over a 12-month period. More than 1 in 3 patients was found to have at least 1 BP measurement above and 1 BP measurement below the cutoff for control of b140/90 mm Hg. Mean BP is less impacted by visit-to-visit variability and, therefore, may be preferable to most recent clinic BP when evaluating provider performance. Multiple other exceptions to a straightforward definition of BP control using most recent clinic BP have been proposed. Current performance measures from the CMS and the National Committee for Quality Assurance evaluate BP control using the most recent clinic BP and do not consider home BP recordings in the assessment of

American Heart Journal May 2013

BP control. Failure to include home readings results in a relatively small underestimate of BP control. We found that inclusion of home BP measurements would result in an absolute improvement in BP control rates of 6%. As home BP monitoring use becomes more prevalent, the importance of including home BP in performance measure assessment increases. 8 However, most electronic health records including those at our institution do not allow for easy documentation, or even easy uploading, of home BP measurements. In addition, lack of home BP cuff standardization and selective patient reporting may limit the accuracy of home measurements. Finally, the recommended cutoff for home BP goals is lower than for clinic visits (b135/85 mm Hg) and depended on patient report of home BPs. Patients were considered “in goal” if providers stated so qualitatively in the note. However, if providers use higher cutoffs than recommended, or if patients selectively report desirable BPs, the impact of home BP monitoring would have been overestimated by this study. The AHA/ACC/PCPI guideline-endorsed patient exclusion criteria allow for patient exclusion from performance assessment for a number of medical or social reasons. Ultimately, we found that excluding patients with AHA/ACC/PCPI endorsed reasons for elevated BP led to improvement in BP control rates by only 3%. One concern regarding performance assessment for hypertension control is how to assess patients with resistant hypertension. The AHA/ACC/PCPI performance measure considers patients on 2 or more antihypertensive agents as equivalent to meeting the requirement for BP control. However, in our cohort of patients followed by cardiologists, nearly 9 of 10 patients with elevated BP were taking 2 or more antihypertensive medications. This finding is not surprising given that AHA/ACC guidelines for coronary artery disease and heart failure include 2 medications that are classified as BP medications: ace inhibitors and beta blockers. If these patients were classified as appropriately controlled, then the results would reflect near-perfect rates of BP control, thereby limiting the use of the performance measure. This specification in the AHA/ACC/PCPI should be modified to parallel current definitions of resistant hypertension, requiring patients to be adherent to full doses of at least 3 medications including a diuretic. One additional option for evaluating BP control in patients with resistant hypertension is to consider whether they have improvements in BP over time, yet variability in individual clinic readings limits the ability to identify patients with clinically significant improvements in BP measurements over a year. In this cohort, nearly as many patients had improvements of 10 mm Hg in SBP as had increases in SBP of 10 mm Hg. Although performance measures provide important frameworks to guide quality assessment, the limitations of these measures need to be carefully considered. Unfortunately, the more nuanced the exclusion criteria

American Heart Journal Volume 165, Number 5

become for performance measures, the more difficult they become for health systems to electronically track using automated data. We found that failure to include home BP or other AHA/ACC/PCPI-endorsed exclusion criteria had only a small impact on the overall rates of hypertension control. Nevertheless, the rate of patients meeting these exclusion criteria is likely to vary by provider. Therefore, inclusion of these exclusion criteria may have a greater impact on relative provider rankings than the overall rate of hypertension control, and failure to use these exclusion criteria may “penalize” providers who care for patients with more difficult social situations or complex medical problems. This study had several limitations. First, this study is representative of a single center's experience with a mostly insured patient population. We recognize that other health systems may have higher rates of home BP monitoring, more accurate capture of reasons for elevated BP or patient compliance, and more consistency across providers. Second, our reasons for elevated BP were based on retrospective medical record review, which is dependent on comprehensive provider documentation. Third, variability in BP measurements may be related to variation in how BP measurements are taken at individual clinics or by individual nurses. We did not explicitly examine BP control by other physicians or providers, such as nurse practitioners or physician assistants, who may have different rates of BP control. Finally, this study depended on vital signs usually taken by nurses at triage who are trained on proper assessment. The quality of these vital signs decreases if proper technique is not followed universally. Although this affects quality of these measurements, this also reflects the real-world nature of this study because current quality measures assess BP using these same types of readings. This study was unable to assess the proportion of patients with “reverse white-coat hypertension” because medical record reviews were only performed for patients with uncontrolled BP in the clinic. This would lead to an underestimation of the rate of hypertension control because home readings may be higher than clinic readings for up to one-third of patients with hypertension. 9 By using only clinic BP measurements, current guidelines do not address this potentially important subgroup of patients with hypertension. Performance measures will be increasingly used for evaluating provider and system performance. Current performance metrics for hypertension control vary in both inclusion and exclusion criteria, and this variability affects estimates of hypertension control. Variability in the technical aspects of current BP performance measures may affect perceived quality gaps and comparative provider rankings. These technical aspects should be evaluated and tested before widespread implementation.

Navar-Boggan et al 827

Acknowledgements The authors thank Erin LoFrese, MS, for her editorial contributions to this manuscript. Ms LoFrese did not receive compensation for her contributions, apart from her employment at the institution where this study was conducted.

Disclosures Dr Navar-Boggan, Dr Boggan, and Ms Stafford have no relevant disclosures to report. Dr Shah reports being a consultant for Castlight Health, LLC. Dr Peterson has received research support from Ortho-McNeil-Jannsen and has served as a consultant for Merck & Co, OrthoMcNeil-Janssen, Pfizer, and Sanofi-Aventis.

References 1. Guo F, He D, Zhang W, et al. Trends in prevalence, awareness, management, and control of hypertension among United States adults, 1999 to 2010. J Am Coll Cardiol 2012;60:599-606. 2. 2010 Physician Quality Reporting Initiative (PQRI) Group Practice Reporting Option (GPRO) Narrative Measure Specifications. Centers for Medicare & Medicaid Services. Accessed September 10, 2011 from http://www.cms.gov/Medicare/Quality-Initiatives-PatientAssessment-Instruments/PQRS/Downloads/ 2010_GPRO_NarrativeSpecifications_111009.pdf. 2011. 3. 2011 Physician Quality Reporting System (Physician Quality Reporting) Measure Specifications Manual for Claims and Registry Reporting of Individual Measures, V5.3. Centers for Medicare & Medicaid Web site. http://www.cms.gov/Medicare/Quality-Initiatives-PatientAssessment-Instruments/PQRS/downloads/ 2011_PhysQualRptg_MeasureSpecificationsManual_033111.pdf. Accessed September 14, 2012. 4. Controlling high blood pressure: percentage of members 18 to 85 years of age who had a diagnosis of hypertension (HTN) and whose BP was adequately controlled (less than 140/90) during the measurement year. National Committee for Quality Assurance (NCQA). HEDIS® 2012: Healthcare Effectiveness Data and Information Set. Vol. 1, narrative. Washington (DC): National Committee for Quality Assurance (NCQA); 2011. various p. Agency for Healthcare Research and Quality Web site. http://www.qualitymeasures.ahrq.gov/content. aspx?id=34655. Accessed September 18, 2012. 5. Bosworth HB, Powers BJ, Olsen MK, et al. Home blood pressure management and improved blood pressure control: results from a randomized controlled trial. Arch Intern Med 2011;171:1173-80. 6. American College of Cardiology Foundation (ACCF), American Heart Association (AHA), Physician Consortium for Performance Improvement® (PCPI™). Hypertension performance measure set. American Medical Association Web site http://www.ama-assn.org/ama1/ pub/upload/mm/pcpi/hypertension-8-05.pdf2011 Accessed September 14, 2012. 7. 2010 Reporting Experience, Including Trends (2007-2011): Physician Quality Reporting System and Electronic Prescribing (eRx) Incentive Program, February 22, 2012. WebMD Interactive Web site. http://web. mdinteractive.com/files/uploaded/2010%20PQRS%20 and%20eRx%20Experience%20Report_03162012.pdf. Accessed September 14, 2012. 8. Ayala C, Tong X, Keenan NL. Regular use of a home blood pressure monitor by hypertensive adults—HealthStyles, 2005 and 2008. J Clin Hypertens (Greenwich) 2012;14:172-7. 9. Wing L, Brown M, Beilin L, et al. “Reverse white-coat hypertension” in older hypertensives. J Hypertens 2002;20(4):639-44.