Assessing Quality of Care for Diabetes: Conference Final Report
Diabetes is one of the leading causes of death in the United States. The number of adults with diabetes has increased from less than 8-million in 1990 to over 18-million today, and that number is expected to reach nearly 40 million by 2050.1 Diabetes is associated with long-term complications and often leads to blindness, stroke, kidney failure, and amputations. Heart disease and stroke cause about 65 percent of deaths among people with diabetes.
In 1997, the Diabetes Quality Improvement Program (DQIP), with support from over 25 key organizations, including the Centers for Medicare & Medicaid Services (CMS), the National Committee for Quality Assurance (NCQA), the American Diabetes Association (ADA), and the Department of Veterans Affairs, Veterans Health Administration (VHA), developed a performance measurement set for diabetes.2 The measures focused on technical quality, such as frequency of blood tests and specified exams, and proportion with poor glycemic control. These measures have been incorporated into the NCQA Health Plan Employer Data and Information Set (HEDIS), the NCQA-American Diabetes Association Provider Recognition Program, the American Medical Association Diabetes Measures Group, and the VHA External Peer Review Program.
Over the past 5-10 years, there have been substantial improvements in many processes of medical care (e.g., checking laboratory tests at indicated intervals) but less dramatic improvements in intermediate outcomes (e.g., value of blood pressure or glycemic control). Indeed, a recent NCQA report (State of Health Care 2006) shows that while appropriate monitoring rates for glycemic and lipid control are both close to 90%, nearly 30% of commercial managed care patients have poor glycemic control, and over 30% have LDL values above 130 mg/dl.3 Additionally, recent studies have shown that at least half of patients with diabetes may have suboptimal blood pressure control.4 These substantial quality gaps in glycemic, blood pressure and lipid control are of considerable concern.
The ultimate goal of quality measurement in diabetes is to motivate quality improvement and decrease long-term diabetes complications. There is evidence that some improvements in processes of care were motivated by the quality monitoring process itself, especially when incentives for better performance were incorporated.3,5-6
Although the DQIP measurement set has been used to profile performance at the facility, health plan, and even individual physician level, we do not have a clear understanding of how to further motivate quality improvements and what or who can activate change. Moreover, new measures may be needed (or existing measures modified) to stimulate further improvements in diabetes care quality. These types of measures should encourage appropriate treatment of diabetes to prevent complications and promote self-management, without creating an incentive for undesirable (and unintended) consequences such as inefficient use of resources, polypharmacy, or risk to patient safety.
The National Diabetes Quality Improvement Alliance,® which represents a collaboration between 13 private and public national organizations, is the successor to DQIP. Its mission has been to maintain and update the measure set for accountability and quality improvement and to recommend areas of investigation for a next-generation measurement set for diabetes quality to improve patient outcomes. While quality measurement in diabetes has advanced beyond that of many other conditions, several challenges remain in developing these next-generation measures. These challenges include promoting good risk-factor (blood pressure, glycemic and lipid) control, considering the special needs of the elderly and those with low socioeconomic position (SEP), and incorporating patient assessments of quality.
A key challenge is constructing measures that are designed to improve metabolic, lipid, and blood pressure control and also minimize unintended consequences. The DQIP recognized that achieving optimal levels of glycemic control depends not only on health care quality, but also on patient disease severity and regimen adherence. Without case-mix adjustment, these measures could unfairly penalize some providers over others. The DQIP panel also understood that even in published randomized controlled trials (which only include patient volunteers), most intensive control studies (such as the Diabetes Complications and Control Trial [DCCT] and the UK Prospective Diabetes Study [UKPDS]) have failed to achieved an A1c <7% in half of their intensive control patient volunteers.7 Therefore, the original measurement set included only a measure of inadequate care—poor glycemic control (A1c>9.5%, later changed to 9%)—intended to motivate providers to minimize the percentage of patients with poor control who have the greatest risk of poor downstream outcomes from hyperglycemia.
Since the development of the DQIP measures, there have been considerable advances in our understanding of measurement science and the treatment of diabetes, but these advances have not always been well synthesized and translated for measures of diabetes quality. For example, evidence now suggests that creating "tightly-linked" clinical action measures can reliably assess whether appropriate action was taken in response to substandard risk factor control.8-9 These types of measures not only identify patients with inadequate risk factor control, but they also motivate an appropriate response or action to poor control (increasing dose of a medication or adding another medication, for example).
The appropriate response may differ depending on how far control is from goal and current treatment intensity. For example, if a patient is already on three or four antihypertensive medications, one could consider that performance is adequate on the measure (i.e., that no further intensification is necessary despite suboptimal blood pressure levels), especially since neither the benefits nor the safety of using additional medications has been established. In addition, it currently is unclear the degree to which achieving the above blood pressure (BP) thresholds is important independent of being on at least moderate doses of three to four medications. Further, there are issues in defining thresholds of control in performance and quality improvement measures and in reconciling clinical guidelines and results from randomized controlled trials with goals of performance measurement.
Recently, a weighted continuous measure for A1c has been proposed. This measure gives "partial credit" to achieved A1c levels with differential credit based on potential for Quality Adjusted Life Years (QALYs) saved in different age groups of moving from 7.9-7.0 percent. In this proposed measure, individuals with A1c values of 7-7.9 percent were assigned credit based on their age group as a percentage of maximal QALYs saved in that age group, while A1c values ≥7.9 percent (or not performed) received zero credit, and A1c values <7 percent received full credit. This approach assesses progress toward achieving thresholds, rather than whether the targets were completely met, and motivates movement toward target even if target cannot be fully achieved.10 (Note: One of the goals of the conference described in this report was for experts to seriously examine whether approaches such as tightly-linked measures and weighted continuous measures could be more effective in promoting further improvements in patient metabolic and blood pressure control.)
In addition to scientific issues on how to formulate the next generation of diabetes quality measures, many experts suggest we should develop a better understanding of the issues faced by special populations (especially the elderly, racial and ethnic minorities, and those of low socioeconomic position (SEP) and low health literacy) in attaining common diabetes quality standards, as well as the implications of these challenges to quality measurement. In particular, nearly 50 percent of patients with diabetes are aged 65 or older,11 and those individuals often have different benefits and risks from clinical interventions to improve risk factor control.
Recently, the ADA guidelines and others have acknowledged this heterogeneity,12-14 but these factors have not been translated to quality measures. In addition, experts have suggested the possibility of stratifying quality results by race or SEP in order to better understand the presence of disparities.15 As recently summarized in a review by Brown and colleagues, research suggests that patients with diabetes of low SEP have less access to care, higher diabetes-related complications, and inferior processes of care.16 African-Americans with diabetes have been shown to have worse metabolic control, higher rates of complications, and inferior processes of care compared to white patients.17-20 While these issues have been raised as areas of concern, whether and how to incorporate issues of SEP in performance measurement have been unresolved.
Finally, the DQIP measurement set, while a great advance in the area of standardizing technical quality assessment, did not agree upon measures for the assessment of interpersonal quality (although the initial report did include several test measures). Diabetes is a complicated chronic condition that requires patients to take charge of many aspects of their care. Good self-management has been shown to improve glycemic control and decrease the rates of diabetes complications. Providing adequate information, good patient-physician communication, and the use of participatory decisionmaking are critical to enhancing patient self-management. Measures that assess patients' satisfaction with the information they are provided and the adequacy of self-management support should be considered as additions to technical measures for the evaluation of quality; some measures are included as part of the NCQA/ADA Diabetes Physician Recognition Program and as part of the research efforts of the Improving Chronic Illness Care Program (ICIC).21 Additionally, collecting data from patients would allow us to understand differences in measurement results by race and SEP. While there are several survey instruments that evaluate dimensions of interpersonal care, the scientific community has not assessed which of these may be appropriately adapted for diabetes quality assessment, and what other factors (such as patient comorbidity, disease severity and SES) would need to be considered before using these measures to profile providers or evaluate quality deficits.