Sudden acceleration of diabetes quality measures.

THE MODERN ERA IN DIABETES PERFORMANCE MEAsurement began in 1997, when the Diabetes Quality Improvement Project (DQIP), the first national disease-specific measure, set recommended thresholds for intermediate outcomes (glycated hemoglobin [HbA1c] levels, blood pressure, and low-density lipoprotein cholesterol [LDL-C] levels). These thresholds, if not achieved, would clearly result in greater population morbidity and mortality. The DQIP explicitly stated that such threshold measures were neither guidelines nor standards and that lower intermediate outcome values should not be used for public reporting owing to lack of evidence for efficacy and the absence of validated risk-adjustment models that would allow for fair comparisons. Also in the 1990s, major randomized trials such as the Diabetes Control and Complications Trial, Scandinavian Simvastatin Survival Study, United Kingdom Prospective Diabetes Study, West of Scotland Coronary Prevention Study, and Hypertension Optimal Treatment Study established stronger evidence for treatment of intermediate-outcome risk factors. These pivotal studies provided the impetus for studies conducted from 2000 onward to determine optimal values for these risk factors to reduce cardiovascular risk in patients with longer duration of disease. The successor to the DQIP, the National Quality Improvement Alliance (Alliance), made few major changes to the measure set from 2001 through 2005 other than to decrease threshold levels for HbA1c from 9.5% to 9.0% and for blood pressure from 140/90 mm Hg to 140/80 mm Hg and to add a threshold level for LDL-C of less than 100 mg/dL. The Alliance members, including liaisons from federal agencies and major professional societies, were aware that planned major trials would inform the choice of threshold values for optimal measures by the end of the decade. However, beginning in 2004, the National Committee for Quality Assurance (NCQA) added optimal measures for HbA1c, blood pressure, and cholesterol to the Diabetes Physician Recognition Program—HbA1c values less than 7%, blood pressure less than 130/80 mm Hg, and LDL-C values less than 100 mg/dL—which were adopted by the Bridges to Excellence pay-for-performance program. In 2005 the Minnesota Community Measurement Program developed a composite measure, termed the Diabetes 5 (D5), that was based on these optimal measures for HbA1c, blood pressure, and LDL-C; administration of aspirin; and nonsmoking status. Despite unanimous opposition from the Technical Advisory Panel of the Alliance, the NCQA approved the optimal measure set in May 2006. Even though inconsistent with guidelines from nonspecialty societies, these measures were widely viewed as evidence-based standards of care because they were based on the guidelines of specialty societies that lent their leadership to the industry-sponsored public service campaign. Coincidentally, 1 week after approval of the optimal measure set, scientific concerns were again raised at an Agency for Healthcare Research and Quality–sponsored conference on diabetes performance measures attended by representatives from key federal agencies, professional societies, the NCQA, and members of academic medicine. In addition to methodological concerns related to selection bias, lack of case-mix adjustment, patient preferences, and measurement variability, the possibility was raised of unintended—and unmeasured—consequences such as harms, costs, and polypharmacy. Indeed, only 2 months later the National Quality Forum did not endorse the “optimal measures.” Much has since changed. The glycemic treatment group in the ACCORD (Action to Control Cardiovascular Risk in Diabetes) study was prematurely terminated because of increased deaths in the intensive treatment group, and the NCQA measure was then modified. Then, the hypertension treatment approach in ACCORD failed to demonstrate benefit of tighter blood pressure control, resulting in modification of NCQA’s optimal measure of 130/80 mm Hg. Although the lipid treatment intervention was designed to evaluate the efficacy of fibrates, the subanalysis did not demonstrate benefit of lower lipid levels; however, this finding needs further evaluation. In the absence of new evidence from 1998 to 2008 and despite ongoing major trials