Although individual clinicians and performance measures both focus on improving the quality of care, they do so from different perspectives on the healthcare system. Performance measures evaluate health systems by measuring processes and outcomes of care and aim to drive improvement by decreasing undesirable variations in care, informing consumers and influencing payment.[1, 2] By reporting the proportion of patients receiving a given evidence-based intervention divided by the number of patients eligible for that intervention, performance measures provide an explicitly population-level view of healthcare. Paradoxically, to improve care, performance measures must influence individual-level decisions by changing the practice of individual clinicians who must maintain a patient-level perspective. Often these two perspectives align, with the performance measure encouraging an intervention that the clinician recognizes is likely to help an individual, but because no intervention is beneficial for all individuals with a specific disease, there are many instances in which performance measures encourage an intervention that the clinician believes is unlikely to help an individual.[4-7] In today's healthcare environment, with financial incentives tied to performance measures, it is critical to understand how clinicians reconcile this tension between performance measures and their clinical judgment.
In this issue of the Journal of the American Geriatrics Society, Woodard and colleagues highlight a common situation in which this tension occurs by focusing on individuals with diabetes mellitus (DM) and limited life expectancy in the Veterans Affairs (VA) health system. Because medically complex patients with limited life expectancy have greater risks and less benefit with intensive glycemic control,[9, 10] guidelines recommend less-intensive control for them,[9, 11, 12] but VA performance measures (like nearly all performance measures) do not exclude these patients. Thus, for individuals with limited life expectancy, performance measures may encourage more aggressive glycemic treatments which are more likely to harm than help. Focusing on individuals with limited life expectancy, Woodard and colleagues describe how clinicians in the VA health system are navigating this situation.
In a national sample of veterans receiving care in the VA health system, Woodard and colleagues found that individuals with DM who had a limited life expectancy were less likely to meet the glycemic control performance measure at baseline (77% vs 78%, P < .001) and were less likely to receive treatment intensification if they were not at goal (20% vs 28%, P < .001). Some may view these results as showing that, despite the incentives tied to performance measures, clinicians continue to individualize care and appropriately treat individuals with DM with limited life expectancy differently from those with a longer life expectancy. Others may view the small differences between the individuals with and without limited life expectancy and conclude that performance measures are decreasing appropriate variations in care and encouraging clinicians to provide “one size fits all” care. Without more-detailed clinical data, it is impossible to know which of these two interpretations of the results is more accurate. Additional research with detailed patient-level clinical data is needed to determine whether the small differences seen in this study are appropriate.
The article highlights three important areas for future research.
First, this study highlights the need for further research to determine the appropriate level of glycemic control for individuals with DM with limited life expectancy. At least 8 years of intensive glycemic control (glycosylated hemoglobin (HbA1c) < 7%) is required before decreases in microvascular outcomes are seen.[12, 13] Thus, individuals with limited life expectancy are unlikely to benefit from fewer microvascular complications. However, individuals with limited life expectancy may benefit from avoiding poor glycemic control (HbA1c > 9%), because poor control can lead to immediate symptoms. Because of glycosuria, most experts recommend avoiding poor glycemic control in individuals with urinary incontinence.[14, 15] Studies in younger individuals suggest that hyperglycemia may lead to fatigue. Thus, there are reasons to believe that HbA1c greater than 9% may lead to worse outcomes (such as incontinence or fatigue) for individuals with limited life expectancy. Studies in these individuals are urgently needed to determine whether poor glycemic control (HbA1c > 9%) is associated with worse outcomes. Only then will the evidence base be available to guide ideal clinical care and determine whether a performance measure of HbA1c of less than 9% is appropriate for individuals with limited life expectancy.
Second, this study highlights the need to design, test, and implement performance measures for individuals with limited life expectancy rather than excluding them from performance measurement altogether. Although individuals with DM and limited life expectancy may not benefit from intensive glycemic treatment, they are likely to benefit more from other interventions such as advance care planning. Thus, although individuals with limited life expectancy may need to be excluded from some quality measures, they need their own set of quality measures that encourage the care that is most likely to help them.
Third, a major methodological achievement of this work is the development of a method for identifying individuals with limited life expectancy using administrative data. The authors convened an expert panel to develop administrative data algorithms for five common life-limiting conditions to identify patients with a limited life expectancy. The algorithm successfully identified patients who had a 55% risk of 5-year mortality (compared with 15% risk for patients not identified through the algorithm). Because guidelines recommend targeting intensive glycemic control (and cancer screening) to individuals with a long life expectancy,[9, 11, 12] this algorithm could be implemented within the VA to identify individuals who are unlikely be benefit from intensive glycemic control or cancer screening. Not incorporating clinical data makes this algorithm less useful as a prognostic index to guide an individual's decision-making, although relying solely on administrative data simplifies the implementation of this algorithm to identify individuals at high risk for limited life expectancy at a population level. Future research should focus on the effect of implementing this algorithm to improve the targeting of intensive glycemic control of older adults with DM with limited life expectancy.