Control status of blood pressure (BP) in hypertensive women as a function of quality of care.
Asch SM, Kerr EA, Lapuerta P, Law A, McGlynn EA. A New Approach for Measuring Quality of Care for Women With Hypertension. Arch Intern Med. 2001;161(10):1329-1335. doi:10.1001/archinte.161.10.1329
Copyright 2001 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2001
Guidelines for care of hypertensive patients have proliferated recently, yet quality assessment remains difficult in the absence of well-defined measurement systems. Existing systems have not always linked process measures to blood pressure outcomes.
A quality measurement system was developed and tested on hypertensive women in a West Coast health plan. An expert panel selected clinically detailed, evidence-explicit indicators using a modified Delphi method. Thirteen indicators (1 screening, 5 diagnostic, 5 treatment, and 2 follow-up indicators) were selected by this process. Trained nurses used a laptop-based tool to abstract data from medical records for the most recent 2 years of care.
Of 15 004 eligible patients with hypertensive and other chronic disease codes, 613 patients were sampled, all eligible for the screening indicator. Of these, 234 women with an average blood pressure of 140/90 mm Hg or more, or a documented diagnosis of hypertension, were studied for the remaining indicators. The average woman received 64% of the recommended care.Most patients did not receive adequate initial history, physical examination, or laboratory tests. Only 37% of hypertensive women with persistent elevations to more than 160/90 mm Hg had changes in therapy or lifestyle recommended. The average adherence proportion to all indicators was lower in patients with uncontrolled blood pressure (>140/90 mm Hg) than in those with controlled blood pressure (54% vs 73%; P<.001).
Quality of hypertensive care falls short of indicators based on randomized controlled trials and national guidelines. Poor performance in essential care processes is associated with poor blood pressure control.
TREATMENT of hypertension has contributed to a reduction in cardiovascular mortality during the past 3 decades, increasing life expectancy in the United States by 5 years.1 Despite overwhelming evidence of the benefits of controlling hypertension and the availability of a plethora of practice guidelines, almost half of the US hypertensive population remains untreated and less than 27% of hypertensive patients achieve the recommended blood pressure (BP) target of 140/90 mm Hg.2 Lack of access and poor patient adherence to prescribed regimens explain part of this deficit, but quality problems in care processes contribute as well.3- 5
Existing systems that assess quality of care have some limitations. They often use a "leading indicator" approach, collecting data for only a few dimensions of quality of care. For example, hypertensive quality monitoring systems often focus exclusively on treatment and ignore the contribution of other clinical functions, such as screening, diagnosis, and follow-up.4 More generally, quality monitoring systems may fail to collect sufficient clinical detail to evaluate provider decision making, relying instead on less precise but easily obtained electronic data. The scientific rationale for indicators and the selection process is either lacking or not specified. Furthermore, few monitoring systems have been applied selectively to women, even though women account for a majority of visits to physicians.6,7
To address these limitations, an alternate approach (the quality assessment [QA] tool) was developed to assess quality of care. More than 1000 clinically detailed indicators for men, women, children, and adolescents in about 50 clinical areas, including hypertension, were developed.7 The project took a comprehensive approach to measuring process quality for hypertensive care, including screening for high BP at physician visits, diagnostic examination of hypertensive patients, appropriate use of both nonpharmacologic and pharmacologic therapies, adequate monitoring of treatment and response over time, control of BP levels, and provision of adequate follow-up.8 The evidence linking each indicator to desired outcomes was assessed by means of proved quantitative models to measure expert opinion.2,9
This article describes performance of the new measurement system when applied to hypertensive female patients in a single health maintenance organization (HMO). It describes the frequency of quality deficits in hypertension care and evaluates the association of these deficits with BP control.
The methods used to construct the QA tool indicators are described elsewhere.8,10 Explicit process indicators for quality of care of hypertensive patients were developed on the basis of a review of the scientific literature covering the continuum of care from initial screening through diagnosis, treatment, and follow-up. Draft indicators were chosen on the basis of the following criteria:
Link to desirable outcomes of reduced BP, stroke, and myocardial infarction
Strength of evidence supporting use of the process
Importance and likelihood of accurate documentation of component variables
Appropriateness of holding providers accountable for indicated care
Generalizability to a variety of clinical settings
A panel of physicians reviewed the indicators and the supporting evidence. The 9 members were recommended by relevant specialty societies and practiced in internal medicine, cardiology, pulmonology, geriatrics, and family medicine in academic, managed care, and other community settings throughout the country. They rated each draft indicator's feasibility and validity on a 9-point Likert scale, both before and during a face-to-face meeting.10 Indicators were included in the final set if their final median validity score was 7 or higher and median feasibility score was 4 or higher.
Of the 10 draft hypertension indicators, the panel process modified the text of 7 (including dropping 5 subparts), accepted 3 unchanged, and added 3. Examples of the dropped indicator subparts included certain initial laboratory examinations (eg, calcium, uric acid, and electrocardiography) and a requirement that patients with stage 3 hypertension receive lifestyle modification before pharmacotherapy.2
The final 13 quality-of-care indicators included 1 screening indicator, 5 diagnostic indicators, 5 treatment indicators, and 2 follow-up indicators (Table 1). Six were supported by randomized controlled trials and 7 by expert opinion, such as the Joint National Committee guidelines.2,11 Diagnostic indicators reflect current knowledge and practice with respect to categorizing patients accordin g to the following factors: BP measurement and risk stratification as recommended by the Joint National Committee guidelines (indicators 2, 3, and 6), searching for secondary hypertension (indicators 3 and 5), looking for potential effects of certain drugs on BP (indicator 4), and establishing a metabolic baseline for monitoring adverse effects of antihypertensive therapy (indicator 5). Treatment indicators included patient counseling for nonpharmacologic high BP management (indicators 7 and 8), conditions for prescribing antihypertensive medication (indicators 9 and 10), and recommended medication for specific comorbidities (indicator 11). Finally, follow-up indicators reflected the importance of regular monitoring of hypertensive patients to control the disease (indicators 12 and 13) (Table 1).
The variables needed to calculate these indicators were incorporated into medical record abstraction software by means of a branching logic to improve data collection efficiency. Adherence to the selected quality-of-care indicators was evaluated in a female population as part of a larger study of the quality of women's health care. Medical record data for female patients in a group-model HMO were reviewed during a 2-year period.
The target population included all women continuously enrolled for at least 13 months who visited a provider at a single HMO site at least once during calendar years 1996 and 1997. Annual indicators were calculated with the most recent 13-month period, allowing some room for patient scheduling and other difficulties. With the use of presence or absence of relevant International Classification of Diseases, Ninth Revision, codes for encounter data during this period, the target population was divided into 3 strata according to age and presence or absence of specific chronic conditions. The numbers of eligible patients in each strata were as follows: stratum 1 (<52 years and current diagnosis of hypertension, diabetes, asthma, or breast mass), 1130 patients; stratum 2 (<52 years and none of the chronic conditions listed in stratum 1), 13 194 patients; and stratum 3 (≥52 years and current diagnosis of hypertension), 680 patients.
Patients were then randomly sampled in each stratum. In total, 613 patients (271 in stratum 1, 243 in stratum 2, and 99 in stratum 3) were selected for this study.
Nurses hired specifically for this project were trained to use the data collection software. Data were abstracted simultaneously for all providers from centrally located medical records. Data included all BP measurements performed during the study period, all medications prescribed, number and dates of physician visits, and other variables needed to calculate the quality-of-care indicators for preventive care (not reported herein) and hypertension. Average abstraction time was approximately 42 minutes per patient, about 30% of which was relevant to hypertension alone.
Person-weights were calculated for each sample stratum and assigned to individuals selected within that stratum to extrapolate results so that they would represent the entire eligible plan population. Sample person-weights were calculated as the inverse of the ratio of the number sampled to the number in the stratum. Both weighted and unweighted patient numbers are presented, although means, proportions, and SEs are adjusted for sample weights.
Abstracted data were then analyzed to determine the eligibility of each patient for every indicator. Eligibility was determined from a set of specific predetermined criteria for each indicator. The entire study population satisfied criteria for indicator 1, the screening indicator. For the other 12 indicators, patients must have had a recorded diagnosis of hypertension or an average BP greater than 140/90 mm Hg during the study period, depending on the indicator. For example, to be eligible for indicator 3a (Table 1), there had to be a progress note indicating that hypertension was a new diagnosis. In many cases, stage of hypertension from the sixth report of the Joint National Committee2 had to be calculated to determine eligibility. Stage 1 or mild hypertension includes patients with average systolic BP (SBP) of 140 to 159 mm Hg or average diastolic BP (DBP) of 90 to 99 mm Hg; stage 2 (moderate), SBP of 160 to 179 mm Hg or DBP of 100 to 109 mm Hg; and stage 3 (severe), SBP of 180 to 209 mm Hg or DBP of 110 to 119 mm Hg. If systolic and diastolic stages conflicted, the higher of the 2 was chosen. If stage was not calculable from the record (BP measurement missing), the patient was not eligible for the indicator.
Adherence to indicators was similarly determined. For example, for indicator 3a, the patient had to have the presence or absence of any of the following noted sometime during the study period: family or personal history of coronary artery disease, cerebrovascular accident, diabetes, or hyperlipidemia. Eligible women who received the care defined in each indicator were counted and total adherence (percentage of eligible patients who received the care) to each indicator was calculated. Standard errors were calculated by means of binomial assumptions, correcting for design effects.
The screening indicator was excluded from summary scores. Summary scores were calculated as the proportion of indicators for which each patient was eligible where the patient received the recommended care. For example, if a woman was eligible for 3 indicators but received the indicated care for only 2, her summary score would be 67%. Alternative calculations of the summary score using the indicator rather than the person as the unit of analysis were also performed. In these calculations, the total number of indications in the population formed the denominator, while the total number of instances in which patients received the indicated care formed the numerator.
The average adherence scores were divided into tertiles, and the association with the proportion of patients with adequate control (average BP, ≤140/90 mm Hg) was calculated for each tertile. Similarly, the average adherence was calculated for patients with controlled (average, ≤140/90 mm Hg) and uncontrolled (>140/90 mm Hg) BP. A χ2 statistic was used to determine whether average adherences for the comparisons were statistically different. Analyses were performed with SAS software package (SAS Institute Inc, Cary, NC).
Sensitivity analyses were performed by varying eligibility criteria for 3 indicators. For treatment indicators 7 and 8, only new hypertensive patients rather than all hypertensive patients were included. For follow-up indicator 13, the definition of persistent elevation was changed from an SBP of greater than 160 mm Hg to an SBP of greater than 140 mm Hg.
With the screening indicator, 467 eligible women were included in the study. The hypertensive population, ie, patients eligible for the nonscreening indicators (indicators 2-13), included 234 subjects. All means, proportions, and SEs are adjusted for design effects. This hypertensive population had been enrolled in the plan for an average of 8.6 years and for 22 months of the 24-month study period. They had an average of 4.1 other acute or chronic medical diagnoses each (eg, upper respiratory tract infection and diabetes) during the 2-year study period. The average age was 49 years, and 73% had an average SBP of greater than 140 mm Hg or an average DBP of greater than 90 mm Hg.
Eligibility and adherence data for each indicator are reported in Table 1. Eligibility varied greatly between indicators. All 467 patients were eligible for the screening criteria. For the diagnosis indicators, eligibility ranged between 1 and 160 of 234 hypertensive patients. The number of women eligible for treatment indicators varied between 5 and 119, and follow-up indicators had the highest number of patients eligible (234 and 151 for indicators 12 and 13, respectively).
Adherence proportions also varied greatly between indicators, from 0% to 100%. Although some data on initial history and physical examination of patients were documented, no patient had a comprehensive set of information recorded (indicator 3). Medical counseling for nonpharmacologic management of hypertension was recorded for only 3.7% of patients with mild to moderate (stage 1-2) hypertension (indicator 7). At least some laboratory tests were performed for a large proportion of patients, but they were comprehensive for only 4.1% of the eligible population (indicator 5). Many patients with persistent uncontrolled mild to moderate hypertension after 6 months of lifestyle modifications did not receive any pharmacotherapy (indicator 9). About half of hypertensive patients with diabetes received appropriate antihypertensive treatment (indicator 11).
Excluding indicator 1, hypertensive patients were eligible for an average of 3 hypertension indicators. The average woman received care for 64% of the indicators for which she was eligible. When the summary score was calculated at the indicator rather than the person level, 60% of all indications for care were satisfied.
None of the women in the lowest tertile of quality summary scores had a controlled average BP (≤140/90 mm Hg) (Figure 1). Slightly more than a quarter (26%) had controlled BP in the middle tertiles of quality, and 38% had controlled BP in the top tertile of quality. Similarly, providers achieved a 73% score on indicators for which they were eligible for women with average SBP of less than 140 mm Hg and DBP of less than 90 mm Hg, compared with 54% for those with BP above 140/90 mm Hg (P<.001). Changing the systolic threshold to a less stringent threshold of 160 mm Hg did not affect the direction or statistical significance of these results. The direction and statistical significance persisted if only indicators based on randomized controlled trials were incorporated into the summary score and if only nontreatment indicators were incorporated.
Because infrequently prescribed treatments in patients with long-standing hypertension might have been overlooked, sensitivity of the results for 2 treatment indicators (indicators 7 and 8) to include only patients with newly diagnosed hypertension was tested. For treatment indicator 7, the number of eligible patients decreased (from 119 to 32) and the proportion receiving indicated care increased (from 3.7% to 32.5%). The same sensitivity analysis performed on treatment indicator 8 did not demonstrate any major difference between new hypertensive patients and all hypertensive patients.
Since the development of the QA tool, changes in optimal care for hypertensive patients with diabetes have occurred and calcium channel blockers may not be a suitable treatment for this category of patients.12,13 The impact of dropping calcium channel blockers as a recommended treatment for hypertensive patients with diabetes was assessed (indicator 11). Of the 7 patients who were eligible for this indicator, all 3 patients who received care received only an angiotensin-converting enzyme inhibitor as therapy. Thus, dropping the patients treated with calcium channel blockers had no impact on the results.
The results for follow-up indicator 13 were insensitive to changing the definition of persistent SBP elevation from 160 to 140 mm Hg; the adherence proportion for both thresholds was very similar (35.1% vs 36.6%).
This study described the development of a new measurement system to assess the quality of care for hypertension. Its component indicators were clinically detailed, were subjected to rigorous evidence-explicit review, and covered a continuum of care from screening through diagnosis, treatment, and follow-up. Testing this part of the QA tool on a cohort of women enrolled in a single HMO showed that care fell far short of the national guidelines. Better performance in essential care process as assessed by the quality indicators was associated with better BP control.
We found that the average hypertensive woman received 64% of recommended care. This is consistent with a recent review of quality-of-care studies in the United States, which reported that about 60% of patients received the care recommended for their chronic diseases.14 Our results are also consistent with the average adherence rate for selected clinical measures from the Health Plan Employer Data Information Set (HEDIS) reported by the National Committee for Quality Assurance, which fell in the 60% to 70% range.15
Our study was designed to measure overall quality of hypertensive care rather than to generate estimates at the individual indicator level. Nonetheless, the pattern of deficits across the hypertensive care spectrum deserves mention. Screening for elevated BP was performed on 84.2% of patients, a better result than that reported by Ornstein and Jenkins.16 Almost all patients received annual follow-up. However, diagnostic procedures at initial history, physical examination, and laboratory testing were documented for less than a quarter of sampled patients; the number of new hypertensive patients in the study was small, though, decreasing our precision. Many hypertensive patients who should have received lifestyle interventions or pharmacotherapy did not, even when analysis was restricted to new hypertensive patients to assess ongoing infrequent prescriptions. Perhaps of even greater concern is that only a third of those whose BP remained elevated (SBP of ≥140 or 160 mm Hg) after 6 months of therapy had any change in their therapy. Since a long-term reduction of 5 to 6 mm Hg in DBP is associated with 35% to 40% lower risk of stroke and 20% to 25% lower risk of coronary heart disease,17 a dramatic need for more aggressive treatment is indicated.
Patients with uncontrolled hypertension received less of the indicated care than did patients with controlled BP, suggesting a relationship between process (quality of care measured by indicators) and outcome (BP levels) of care. Linking process and outcome is an important goal of quality-of-care assessment.4 However, previous attempts have not always established a link between quality of hypertensive care and control of hypertension. Haynes and colleagues18 reported a positive link between 3 items of antihypertensive treatment and reducing BP levels: decision to treat, vigor of prescribed medication, and compliance. Using chart review analyses, Berlowitz and colleagues4 examined the care of hypertensive men in Veterans Affairs settings and made the link between more intensive medical therapy and better BP control. Our study is the first, to our knowledge, to link a broad measure of quality of care to outcome. This relationship persisted if indicators measuring the quality of therapy were excluded, leaving only diagnostic, screening, and follow-up measures, although inferences about causation are premature in the absence of longitudinal data.
The broader nature of these measures points the way to improving quality. Leading quality standards such as HEDIS focus on the intermediate outcome of BP,19 but in the absence of data on care processes, providers may not be aware of what behavioral changes are likely to improve outcome. Systems like the one presented herein allow for focused evaluation of care for hypertensive patients and can be used longitudinally to evaluate targeted quality improvement interventions.7
The quality measurement instrument reported herein has some limitations. First, some care may have been delivered but poorly documented and thus not detected by our instrument, biasing quality estimates downward. Nevertheless, chronic disease care is increasingly delivered by teams of providers rather than by a single individual, especially in managed care settings. This makes clear documentation essential and the failure to document potentially detrimental to patient care. The component indicators were selected explicitly so that the absence of documentation was in and of itself evidence of poor quality. Second, although some indicators are clearly more closely linked to outcomes than are others, each indicator was weighted equally in its contribution to the overall quality score in this first test of the QA tool. We elected to use equal weights because we did not have a compelling method for assigning differential weights. Third, we tested the instrument in only 613 patients in a single health plan, and more extensive testing is under way. Fourth, there are no randomized controlled trials supporting or refuting some of the indicators. However, as much as 70% to 80% of all care delivered lacks such support,20 and so rigorously conducted expert panel methodology is a viable tool to extend the reach of quality measurement instruments. Finally, optimal care of patients is evolving as research data are being published. Since the development of the QA tool, recommendations on management of hypertensive patients have changed (eg, calcium channel blockers may not be a suitable treatment for hypertensive patients with diabetes12,13), and indicators need to be updated regularly to capture the most recent advances in medical care.
Finally, the clinically detailed indicators presented herein require chart review, a more costly process than pure electronic data analysis. However, as part of a global measure designed to analyze charts of patients unselected for condition, the hypertensive portion of the QA tool will benefit from maximizing the amount of relevant data analyzed per abstracted case. We have estimated that the QA tool will require half the sample size of the average HEDIS report, substantially reducing costs.7 Systems with similar sampling schemes cost $15 000 to $25 000 per measurement cycle for 300 to 500 records, an amount within the resources of most plan quality assessment budgets.21
Future work will develop methods of weighting indicators according to their relative importance to patient outcomes. Further research should also explore the validity and generalizability of the results by expanding evaluation of the tool to other populations and larger sample sizes. Examining the correlation of quality of care for hypertension with that for other conditions also may provide valuable information about which factors most significantly affect the quality of care within a health plan.
Using a new approach of detailed evaluation of quality of care in an HMO indicates that physicians may not be aggressive enough in managing hypertension. Data obtained with a performance evaluation system such as the QA tool can pinpoint deficiencies in health care and provide critical insight to guide quality improvement efforts.
Accepted for publication October 3, 2000.
This study was supported by grants from Bristol-Myers Squibb, Princeton, NJ; the Agency for Health Care Research and Quality, Washington, DC; the California HealthCare Foundation, Oakland; and the Health Care Financing Administration, Baltimore, Md. Drs Asch's and Kerr's time is supported by career development awards from the Veterans Affairs Health Services Research and Development Service, Washington.
Presented in part at the meeting of the American College of Cardiology, Anaheim, Calif, March 14, 2000.
We thank Peggy Wallace, Alison DeCristofaro, Jennifer Hicks, and Karen Connor for invaluable research assistance and Micki Fujisaki for her skilled programming efforts.
This article was prepared with the assistance of BioMedCom Consultants Inc, Montreal, Quebec.
Corresponding author and reprints: Steven M. Asch, MD, MPH, RAND, 1700 Main St, Santa Monica, CA 90407 (e-mail: email@example.com).