The Evolution of Advance Directives

Abstract

THERE ARE 2 WIDELY KNOWN PROBLEMS WITH END-OFlife care in the United States. First, care often does not reflect patients’ values and preferences. Second, end-of-life care constitutes a disproportionate amount of health care expenditures. There is a need to improve the value of health care spending because of changing economic and political climates and because Medicare and Medicaid costs account for nearly a quarter of the federal budget. Despite recognition that advance directives have conceptual limitations, their appeal persists because of the theoretical possibility that advance directives can simultaneously improve the patient centeredness of care and decrease health care expenditures near the end of life. The evolution of advance directives has mirrored that of many new medical technologies: initial unbridled enthusiasm evolved into skepticism as empirical evidence raised questions about the current practice, followed by a wiser, more constrained application. In contrast to early studies suggesting little benefit of advance directives, recent observational and interventional studies in selected populations suggest that advance directives substantially improve the patient centeredness of care near the end of life. However, the evidence about the effect of advance directives on costs is conflicting and incomplete. In this issue of JAMA, Nicholas and colleagues make an innovative contribution by examining whether the association between advance directives and financial expenditures varies according to a region’s average level of Medicare spending. To accomplish this, the authors linked data from the Health and Retirement Study—in which decedents’ surrogates were queried about whether patients had a treatment-limiting directive—with Medicare claims data. Two findings are especially important. First, in regions characterized by high levels of Medicare spending, treatment-limiting directives were associated with modestly lower levels of Medicare spending, lower likelihood of in-hospital death, and higher use of hospice care; however, similar associations were not observed in low-spending regions. Second, advance directives specifying limits on treatment were more common in areas with lower levels of end-of-life spending. The study is limited by the paucity of information regarding the content of patients’ directives and lingering questions about whether the use of surrogates’ reports about the content of advance directives is a reliable measurement strategy. As the authors note, the observational nature of the data precludes causal inferences about whether interventions to increase the use of advance directives will decrease costs. In light of these limitations and the relatively small absolute magnitude of cost savings observed, these findings are insufficient to justify changes in health policy regarding advance directives. Nevertheless, the authors have opened new avenues of inquiry to understand the effects of advance directives and have highlighted the potential importance of local practice norms on the outcomes associated with advance directives. The results raise several important questions about the “mechanisms of action” of advance directives. First, what might explain the finding of lower end-of-life care costs among patients with treatment-limiting directives in high-spending regions? The most obvious explanation that advance directives reduced the application of expensive life-prolonging technologies prior to death does not appear to be the entire explanation because there was not a significant difference in use between patients with and without treatment-limiting directives. An alternative explanation, which was not assessed in the study, is that treatment-limiting directives affected the duration of use of life support rather than the incidence of its application. This distinction is crucial because it raises the possibility that the work advance directives do may be different from the work they were conceived to do. Advance directives were originally designed to allow patients to determine in advance what treatment they would receive in the event of their incapacity. Instead, it may be that treatment-limiting advance directives work not by making the decisions but by giving surrogates and physicians psychological permission to cease life support at some point. This hypothesis is consistent with substantial empirical evidence that surrogates are reluctant to not do everything for patients and experience substantial distress when they believe they do not understand patients’ values. This distinction is important because it may signal the need to refocus advance care planning efforts to help surrogates and clini-