Among a cohort of older persons with advanced chronic illness interviewed at multiple time points, trajectories of preferences for potentially life-sustaining treatment, assessed in terms of participants' willingness to endure high-burden treatment for a given chance to avoid death or risk disability in order to avoid death, were frequently inconsistent. Many participants became more and then less willing (or vice-versa) over time to undergo future high-burden therapy or to risk severe disability. The likelihood of a consistent versus an inconsistent trajectory differed according to the trade-off under consideration, and according to the participant's primary diagnosis, nature of initial preference, and health state trajectory.
The inconsistency in preferences over time poses an extreme challenge to advance care planning. Advance directives, documents in which patients could detail their treatment preferences for times of incapacity, were proposed in order to allow patients to express what were assumed to be deeply held and therefore presumably stable values regarding medical treatment.36
Early reviews of the stability of end-of-life treatment preferences concluded either that preferences were indeed stable37
or that the rate of change was similar to other major life decisions, such as estate wills, and that stability could be improved by ensuring that decisions were well informed and carefully considered.38
More recent studies have demonstrated that preferences are only moderately stable.7, 9, 10
However, because changes in preferences have been associated with changes in the patient's health state,7, 9
the interpretation of this change is that it occurs in a consistent and predictable direction. The notion is that, because patients adapt to diminishing health,12
as their own health declines and states that they imagined as intolerable are actually experienced, older persons may become increasingly willing to undergo life-sustaining treatment.39
In contrast to the notion that preferences change in a consistent manner, the findings of this study demonstrate that many persons had an inconsistent pattern of preferences over time. The study supports several explanations for these findings. The first is that it is true that, as previously shown, changes in health status do affect preferences in a predictable way. However, in contrast to the assumption that the health state of persons with advanced chronic illness changes in a predictable way; ie: their health state steadily declines over time, in this study we found that health state may be variable over time. This explanation is supported by the finding that participants with inconsistent health state trajectories were more likely to have inconsistent preference trajectories than were participants with consistent health state trajectories. The frequency of inconsistent health state trajectories confirms prior work demonstrating variable functional status trajectories both among older persons in general40
and in the last year of life among persons dying with organ failure.41
However, the high frequency of inconsistent preference trajectories even among participants with consistent health state trajectories suggests that there are additional influences on preferences. These inconsistent trajectories supports prior work demonstrating that preferences are influenced by transient immediate circumstances and affective states that may change according to these circumstances.42, 43
One recent study, which examined preferences for life-sustaining treatment among a cohort of older persons before, shortly after, and then several months after being in the hospital demonstrated that hospitalization transiently changed preferences for a number of specific interventions.16
This notion of the potential importance of transient influences on preferences is further supported by the observation that the will of patients terminally ill with cancer to live fluctuates greatly over short time periods and is associated both with affective states of anxiety and depression and with the physical symptom of shortness of breath.44
The frequencies of inconsistent preference trajectories in the current study were greater than in a prior analysis performed in this cohort, which utilized a measure of treatment preferences based solely on a consideration of the health state that would result from treatment.45
This comparison suggests that the increased complexity of the choices participants were asked to consider in the present study also contributed to inconsistent trajectories. Although it is not surprising that a more complex choice would lead to less consistency in response, it is both critically important and highly challenging to determine why. It is possible that the complexity of the assessment may, by overwhelming the cognitive capacity and understanding of respondents, fail to reflect participants' core values and beliefs. The items used in this study demonstrated only moderate test-retest reliability, performed over a one-week period, with intraclass correlation coefficients of .49 -.77. However, if preferences are inherently unstable, this could also account for changes in response seen after one week, suggesting that test-retest reliability may not be a measure of the quality of assessment tools, and that alternative methods need to be used to determine how well these tools are understood.
Can the process of advance care planning overcome these challenges? It has been eloquently argued that, for some patients, the ability to indicate their preferred care in advance of decisional incapacity, ensures that they will end their lives according to the deeply held values with which they lived them.46
For patients with unchanging preferences, advance care planning may be a reliable process that reveals their deeply held and stable core values. Patients whose preferences are changing because of adaptation to declining health may be able to incorporate knowledge of how their preferences are changing into their valuation of future states of health and disability.14, 15
Whether patients whose preferences are influenced by variable trajectories of health and/or transient affective states can do the same is unknown and needs to be determined. These considerations suggest that the process of advance care planning would be greatly strengthened by asking patients to re-evaluate their preferences over time, with an explicit consideration of the reasoning underlying these preferences. This would serve to identify that subgroup of patients whose preferences reflect stable core values. For patients with more variable preferences, it would help them to recognize the factors influencing their preferences, which holds the potential to improve the advance care planning process.
The study was limited by the small numbers in the analyses examining health state trajectories and by the lack of racial and ethnic diversity in the study population, which decreases its generalizability.
Even with its potential limitations, our measure assessed preferences in a more systematic way than is typically done in clinical practice. That such an approach is characterized by a high degree of variability in response over time highlights the caution that must be applied to the interpretation of the treatment preferences that are elicited in clinical contexts. This variability also highlights the need to understand not only what patients' preferences may be but also how they have formulated these preferences and to recognize that inconsistency in the trajectory of treatment preferences is common.