Although clinical practice guidelines generally acknowledge the importance of juxtaposing life expectancy and the lagtime-to-benefit for preventive interventions, quality indicators encourage preventive interventions for all patients within a certain age range. For example, the US Preventative Services Task Force (USPSTF) colorectal cancer (CRC) screening guideline recommends discontinuing screening for patients with a limited life expectancy. However, the HEDIS quality indicator encourages screening in all adults aged 50–75 years. Thus, the current HEDIS CRC quality indicator would encourage screening for a 70-year-old patient whose oxygen-dependent lung disease limits his life expectancy and makes it unlikely that he would benefit. Paradoxically, the current HEDIS CRC quality indicator would ignore screening in an 80-year-old woman with only mild hypertension, even though she has an extended life expectancy and is more likely to benefit from screening. Because the heterogeneity of life expectancy increases with increasing age,(7
) the problem of age-based indicators is greatest among older adults.
The ACOVE project addressed this issue by identifying quality indicators that would be inappropriate for patients with limited life expectancy.(8
) We propose a more nuanced approach with prevention quality indicators explicitly accounting for life expectancy and encouraging prevention only in those patients whose predicted life expectancy exceeds the intervention’s lagtime-to-benefit. For example, since the lagtime-to-benefit for colorectal cancer screening is estimated to be 7 years,(9
) a life expectancy-based CRC screening quality indicator would only consider those patients who have a life expectancy >7 years to be appropriate candidates for screening. Although estimates of life expectancy are imperfect, combining readily available data such as gender and comorbidities with age would allow more accurate predictions of life expectancy. Similar to how estimated glomerular filtration rate is now routinely calculated and presented to clinicians, age, gender and comorbidity data could be used to calculate an estimated life expectancy.
Depending on the electronic medical record (EMR) system that is available in a given clinical setting, life expectancy-based quality indicators would require differing implementation strategies. In settings with robust EMR systems such as the VA, life expectancy could be calculated for every patient using the rich clinical data within the EMR (i.e. comorbidities and laboratory values). This life expectancy calculation could be used to determine whether a patient is appropriate for a preventive intervention such as CRC screening, leading to the display or suppression of a clinical reminder. In clinical settings without robust EMR, providers could estimate life expectancy explicitly using a variety of published calculators or implicitly using clinical intuition. At the administrative level where quality indicators are calculated, available data including age, gender and comorbidities could be used to determine life expectancy. This calculated life expectancy can then be used to determine whether each patient is an appropriate candidate for the quality indicator.
For many patients, the estimated life expectancy and the lagtime-to-benefit for the intervention may be similar, suggesting that the net benefits are small or uncertain. In these situations, the intervention is discretionary and highly dependent on factors such as patient preference.(10
) The 2008 USPSTF CRC guidelines foreshadows this concept of discretionary care by suggesting that screening for patients age 75-85 years is optional, while recommending screening for patients who are younger (age 50-75) and discouraging screening for patients who are older (age ≥85).(9
) We propose that quality indicators omit patients where the benefits are small or uncertain, so that neither intervention nor the lack of intervention can be viewed as poor quality care.