Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Arch Intern Med. Author manuscript; available in PMC 2007 September 18.
Published in final edited form as:
PMCID: PMC1978221




Instructional forms of advance care planning depend upon patients' ability to predict their future treatment preferences. However, preferences may change with changes in patients' health states.


We performed in-home interviews with 226 older community-dwelling persons with advanced cancer, congestive heart failure, or chronic obstructive pulmonary disease at least every four months for up to 2 years. Patients were asked to rate whether treatment for their illness would be acceptable if it resulted in one of four health states.


The likelihood of rating as acceptable treatment resulting in mild (OR [95% confidence interval]) (1.11 [1.06, 1.16]) and severe (OR 1.06 [1.03, 1.09]) functional disability increased with each month of participation. Patients who had a decline in their ability to perform instrumental activities of daily living were more likely to rate as acceptable treatment resulting in the states of mild (OR 1.23 [1.08, 1.40]) and severe (OR 1.23 [1.11, 1.37]) disability. Although the overall likelihood of rating treatment resulting in a state of pain as acceptable did not change over time (OR 0.98 [0.96, 1.01]), patients who experienced moderate to severe pain were more likely to rate this treatment as acceptable (OR 2.55 [1.56, 4.19]) than those who did not.


For some patients, the acceptability of treatment resulting in certain diminished states of health increases over time, and increased acceptability is more likely among patients experiencing a decline in that same domain. These changes pose a challenge to advance care planning, which asks patients to predict their future treatment preferences.


Patients with diminished states of health rate these states more highly than does the general public.1-3 Further, cancer patients are more willing to undergo intensive therapy with a small likelihood of benefit than are physicians or the general public.4, 5 These findings suggest that changes in patients' health status may affect their treatment preferences as patients become more willing to tolerate a diminished state of health.6 Patients' valuations of health states are likely to be associated with treatment preferences because these preferences have been shown to be based largely on the outcome achieved by any given intervention.7-9 Changes in treatment preferences as a result of changing health have profound implications for instructional forms of advance care planning, which ask patients their preferences for end-of-life care.

However, there has been little longitudinal examination of the association between patients' health states and their end-of-life treatment preference among persons with advanced illness, for whom advance care planning is most pertinent. The majority of longitudinal studies of patients' treatment preferences examined preferences for specific interventions without specifying the health states resulting from those interventions.10-18 These studies could therefore not assess whether changes in preferences resulted from changes in patients' valuation of health states. Two studies specifying the health states resulting from intervention, one performed among a general population of older persons,19 and the other among persons with AIDS,20 found preferences to be moderately unstable, with no clear associations between preferences and changes in participants' own health.20

The purpose of this study was to examine changes over time in end-of-life treatment preferences, measured in terms of willingness to undergo treatment based on the health state that would result from the treatment, among a cohort of older persons with advanced illness.



Participants for this study were 226 community-dwelling older persons with advanced chronic illness. The Human Investigations Committee of each of the hospitals participating in the study approved the study protocol, and each participant provided written informed consent. We screened sequential charts of persons age 60 years or older with a primary diagnosis of cancer, congestive heart failure (CHF), or chronic obstructive pulmonary disease (COPD) for the primary eligibility requirement: advanced illness, as defined by Connecticut Hospice criteria21 or SUPPORT criteria.22 Charts were identified according to the patient's age and primary diagnosis in the subspecialty outpatient practices in the greater New Haven area and in three hospitals: a university teaching hospital, a community hospital, and a VA hospital. Of the 26 practices approached for participation, 3 (12%) did not permit screening of their charts. An additional eligibility criterion was the need for assistance with at least one instrumental activity of daily living (IADL),23 determined during a telephone screen, and selected to improve the identification of patients who had advanced disease.24 Because of our interest in the relationship between disease diagnosis and preferences, screening and enrollment occurred in a stratified manner to enroll approximately equal numbers of patients with cancer, CHF, and COPD.

Of the 548 patients identified by chart review, 470 persons received the telephone screen (physician refused permission to call (n=30), patient died prior to call (n=24), patient refused screening (n=18), patient could not be reached (n=6)). Those who were not screened were significantly older (75 vs. 72 years, p<.001) but did not differ according to gender or diagnosis. Of those screened, 362 required IADL assistance. Exclusion criteria included cognitive impairment (n=77) and part-time Connecticut residence (n=6). Of the 279 eligible patients, 2 died prior to participation and 51 refused participation, resulting in 226 participants. Non-participants did not differ from participants according to age or gender. Among eligible patients with CHF, 8% refused participation, compared to 19% among patients with cancer and 25% among patients with COPD (p=.02). Of the 226 participants, 8 withdrew after the initial interview (4%), 26 died before completing a follow-up interview (12%), and 3 were unable to participate in follow-up interviews (1%). Of the surviving 124 participants at the end of the first year of the study, 98 (79%) consented to a second year of participation.

Data collection

We interviewed patients in their homes, obtaining all variables by self-report. We subsequently interviewed patients at least every four months for up to two years. If the patient had a decline in health status, determined by a monthly telephone call, the next interview was scheduled immediately. We then conducted subsequent interviews every four months, unless the patient had another decline. We employed this strategy to balance the burden imposed by frequent interviews with the desire to interview patients as their illness worsened but before they died. We defined decline in health status as: 1) a new disability in a basic activity of daily living (ADL),25 2) a prolonged hospitalization (≥ 7 days) or a hospitalization resulting in a discharge to nursing home or rehabilitation facility, or 3) introduction of hospice services.

Descriptive and analytic variables included sociodemographic, health, and psychosocial status measures. Ordinal variables were dichotomized such that at least 10% of respondents were in the category of most severe status. Sociodemographic variables included age and education (continuous variables), gender, ethnicity, sufficiency of monthly income,26 marital status, and living arrangement. Health status variables included self-rated health, with response categories of “excellent, very good, good”, versus “fair, poor;” number of IADL disabilities (range 0-14), self-rated life expectancy, for which patients were asked, “If you had to take a guess, how long do you think you have to live?” and level of pain, for which patients were asked, “How would you describe your worst pain during the last 24 hours?” with response categories of “no pain, mild pain” versus “moderate pain, severe pain.” Psychosocial variables included quality of life, with response categories of “best possible, good” versus “fair, poor, worst possible;” depression, measured using the 2-item PRIME-MD instrument;27 and whether the patient had a living will. Health and psychosocial variables were obtained at each interview.

The outcome variables, assessed at each interview, were obtained by asking patients to think about whether 4 health states that could result from treatment represented either an acceptable or unacceptable quality of life. Similar to the concept of “states worse than death,”28 they were told that rating the health state as acceptable meant that they would want to undergo treatment and that rating the health state as unacceptable meant that they would prefer to die rather than to undergo treatment. The states included: 1) mild physical disability, described as being unable to get out of the house to visit family, attend religious services, go to work, do volunteer work, or do hobbies, 2) severe physical disability, described as only being able to get from bed to chair; requiring help with bathing, dressing, and grooming; 3) cognitive impairment, described as having severe problems with memory such that you cannot recognize family members; 4) pain, described as being in moderately severe pain daily, like having a broken bone or appendicitis. Test-retest reliability was established by re-interviewing 20 participants one week after their initial interview. This time period was chosen to be long enough to decrease the likelihood that participants would remember their initial responses but short enough that participants would not undergo a change in health status. Raw agreement for the four states was 89%, 80%, 95%, and 87%, with corresponding kappa coefficients of .75, .58, .64, and .72. The percentage of respondents rating the second and third states as unacceptable at baseline was high (70%, 95%), resulting in lower kappa coefficients.29

At the baseline interview, the correlation in ratings among the 4 health states was ≤.43, except for the two disability states, for which the correlation was .59.

Statistical analysis

To describe the population, we used mean and standard deviation for continuous variables and frequency and percentage for categorical variables. We defined five trajectories for health state ratings over time: 1) acceptable at all times, 2) unacceptable at all times, 3) change from acceptable to unacceptable without further changes, 4) change from unacceptable to acceptable without further changes, 5) variable (i.e. multiple changes in rating). We examined the frequency of these trajectories according to four trajectories for functional status over time, defined by the number of IADL disabilities: 1) unchanged, 2) improved, 3) declined, 4) other (both improvement and decline over time).

In order to determine factors associated with ratings, we utilized generalized linear mixed effects models30, 31 by implementing repeated measures logistic regression with inclusion of a patient-level random effect. We chose the mixed effects model approach over a competing one, e.g., a marginal model approach using generalized estimating equations,32 because we wanted to account for irregular interview times and draw inferences at the subject-specific level. We developed four multivariable models, using a forward selection approach, with no correction for multiple comparisons. For each model, the dependent variable was the rating of a given health state as acceptable or unacceptable as a result of treatment at each time point. Independent variables were eligible for inclusion in each of the multivariable models if they were associated with the health state rating in a bivariate model with P < 0.20. To be included in the final model, the variable needed to maintain P < 0.10. In each model we included age, gender, race, marital status and time, regardless of their bivariate associations. Time was measured in months since the start of the study for each individual. All analyses were carried out using SAS software (version 8.2), using Proc NLMIXED to fit the generalized linear mixed effect models.33


Patient population

Table 1 provides a description of the patient population. During the two years of follow-up, 77% of patients with cancer, 43% of patients with COPD, and 46% of patients with CHF died. Of the cohort, 68% had at least 3 interviews and 36% had 5 or more. The median number of interviews was 2 for patients with cancer, 4 for patients with CHF, and 5 for patients with COPD. Among non-dropouts, ascertainment of outcome data was 90% complete, and, among the 10% of missing data, 89% was due to the participant being too cognitively impaired or too ill to participate in the interview.

Description of 226 participants at baseline:

Description of trajectories of health state ratings

For each of the health states, at least 49% of participants provided the same rating (acceptable or unacceptable) throughout the study (Table 2). For the two states representing mild and severe functional impairment, larger proportions rated the states as acceptable at all interviews (56% and 32%) than as unacceptable (8% and 19%). In addition, a larger proportion of participants changed their ratings from unacceptable to acceptable (19% and 20%) than from acceptable to unacceptable (6% and 6%). In contrast, for the states of cognitive impairment and pain, larger proportions rated the state as unacceptable at all interviews (75% and 37%) than as acceptable (2% and 12%). A larger proportion of participants changed their ratings from acceptable to unacceptable (8% and 17%) than from unacceptable to acceptable (3% and 6%).

Number (percent) of participants with different trajectories of health state ratings:*

For each functional status trajectory, at least 44% of respondents provided the same rating over time for the state of severe functional impairment (Table 3). Among those who had a decline in functional status, 27% changed their rating from unacceptable to acceptable while 9% changed from acceptable to unacceptable.

Trajectories of health state rating for the state of being confined from bed-to-chair according to trajectories of functional status:*

Multivariable correlates of changes in health state ratings

The likelihood of patients' rating the states of mild and severe functional status impairment as acceptable increased significantly over time (Table 4). In contrast, the likelihood of rating the state of cognitive impairment as acceptable decreased significantly over time. The likelihood of rating the state of severe pain as acceptable did not change significantly over time.

Factors associated with a health state rating of acceptable among all 226 participants:

Patients with greater IADL disability were more likely to rate the states of functional impairment as acceptable, with a 23% [8%, 40%] increase in the odds of rating the state of mild impairment and a 23% [11%, 37%] increase in the odds of rating the state of severe impairment as acceptable for each additional IADL disability (Table 4). However, an increase in IADL disabilities was not associated with the ratings for cognitive impairment or pain. Patients who experienced moderate or severe pain were more likely to rate the state of severe pain as acceptable (OR 2.6 [1.6, 4.2]) than those who did not. However, experiencing pain was not associated with the ratings for functional or cognitive impairment.


The results of this study illustrate that for some older, seriously ill persons, changes in their health are associated with changes in their valuations of the outcomes that may result from treatment for their illness. Those patients who experienced a decline in instrumental activities of daily living were more likely to rate more severe functional disability as an acceptable outcome of therapy than those who do not experience such a decline. Similarly, those who experienced moderate or severe pain were more likely to rate severe pain as an acceptable outcome of therapy than those who experienced no or mild pain. Changes in the valuations of health states resulting from treatment are specific to the state itself. Even after accounting for changes in functional status as well as sociodemographic and other health and psychosocial characteristics, the acceptability of functional disability increased over time, whereas the acceptability of severe cognitive impairment decreased. However, in absolute terms, the majority of participants did not change their ratings of the acceptability of these health states.

Although this study focused on eliciting patients' valuations of health states, it assessed these valuations in the context of whether the patient would choose to have therapy resulting in the given health state. The majority of previous studies have examined the relationship between clinical status and preferences for specific treatment interventions. Several of these studies demonstrate a relationship between declining health and a greater likelihood of a preference for high-burden life-sustaining treatment.12,18 These studies suggest that declining health status is associated with a greater willingness to bear the burdens of therapy. By systematically evaluating the attitudes of older persons facing a decline in their health toward a variety of health states, this study suggests that an increased desire to undergo therapy results, at least in part, from the willingness to tolerate diminished states of health, a finding that confirms previous studies.34, 35 These findings are consistent with the theory of “response shift,” whereby patients change their self-evaluation of quality of life with changes in their health status.36 This concept is further supported by the finding that higher self-rated quality of life, independent of pain, function, or other markers of health, was associated with a greater likelihood of rating several of the health states as acceptable.

The results of this study pose a serious challenge to the instructional form of advance care planning. The main purpose of these instructions is to allow patients to express their preferences for care in the case of future circumstances in which they are unable to speak for themselves. This presupposes that patients' projections of the treatment they want for themselves in some future state accurately reflect how they would actually feel in that state. However, the findings of this study illustrate that some patients cannot accurately predict their future valuations. This limitation is well recognized in certain fields, such as psychology37 and decision-making,38 but it has not been a prominent part of considerations regarding advance care planning.

The problems with predicting future hypothetical states of health have led some to conclude that instructional advance directives are a misguided means of care planning.39 However, the methods we used to demonstrate the problem of predicting future health states provide a partial solution to the problem Many patients were still able to think about and express their preferences regarding these states after experiencing a change in their health status. This finding implies that, if advance care planning is conducted as a process over time, in which patients are asked to reflect on their preferences after experiencing a change in their health, they will have the opportunity to reflect upon how their preferences may be changing. In addition, the appointment of a health care proxy, another form of advance care planning, is not dependent upon patients' ability to predict future states of health. The finding of changes in preferences supports the notion of allowing surrogates leeway in the interpretation of advance directives, an approach that is endorsed by a large majority of patients.40, 41

Because the study examined the preferences of patients with advanced illness, missing data are unavoidable. The largest cause of missing data in the study was mortality. It is unclear if these data are “missing” in the sense that this term is traditionally used, since these data, along with the data from participants who became cognitively impaired or more acutely ill, are not recoverable. However, there were also missing data from participants who dropped out of the study for other reasons or who failed to consent to a second year of participation. The issue of missing data in longitudinal research is one of current investigation, and there are no easy or straightforward methods for assessing the impact of missing data. Therefore, we cannot know whether these missing data introduce some bias into the results. Nonetheless, the high overall rates of participation and completeness of data collection among patients who remained in the study suggest that data collection was as complete as possible in this challenging population.

The study population included only a small number of non-white participants. Because there was a trend for non-white race to be associated with ratings of the acceptability of health states, the findings of this study regarding the rates of different trajectories of health state ratings may not be applicable to all populations of older persons with advanced chronic illness. Further, this study leaves unanswered the important question of the reason(s) why valuations of health states change over time.42 Nonetheless, the changes in older persons' treatment preferences over time and the association between changes in older persons' health and treatment preferences highlight the need for repeated assessments of these preferences.


The authors thank Carm Joncas, R.N. and Barbara Mendes, R.N. for their interviewing skills. Supported by grant PCC-98-070-1 from VA HSR&D, R01 AG19769 from the National Institute on Aging, P30 AG21342 from the Claude D. Pepper Older Americans Independence Center at Yale, and a Paul Beeson Physician Faculty Scholars Award. Dr. Fried is supported by K02 AG20113 from the National Institute on Aging.


The funders had no role in the design and conduct of the study; the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.

Dr. Fried had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.


1. Sackett DL, Torrance GW. The utility of different health states as perceived by the general public. J Chronic Dis. 1978;31:697–704. [PubMed]
2. Boyd NF, Sutherland HJ, Heasman KZ, Tritchler DL, Cummings BJ. Whose utilities for decision analysis? Med Decis Making. 1990;10:58–67. [PubMed]
3. Dolan P. The effect of experience of illness on health state valuations. J Clin Epidemiol. 1996;49:551–64. [PubMed]
4. Slevin ML, Stubbs L, Plant HJ, Wilson P, Gregory WM, Ames PJ, et al. Attitudes to chemotherapy: comparing views of patients with cancer with those of doctors, nurses, and general public. BMJ. 1990;300:1458–60. [PMC free article] [PubMed]
5. Donovan KA, Greene PG, Shuster JL, Partridge EE, Tucker DC. Treatment preferences in recurrent ovarian cancer. Gynecol Oncol. 2002;86:200–11. [PubMed]
6. Emanuel LL. Advance directives and advancing age. J Am Geriatr Soc. 2004;52:641–2. [PubMed]
7. Rosenfeld KE, Wenger NS, Kagawa-Singer M. End-of-life decision making: a qualitative study of elderly individuals. J Gen Intern Med. 2000;15:620–5. [PMC free article] [PubMed]
8. Fried TR, Bradley EH. What matters to older seriously ill persons making treatment decisions? A qualitative study. J Pall Med. 2003;6:237–244. [PubMed]
9. Fried TR, Bradley EH, Towle VR, Allore H. Understanding the treatment preferences of seriously ill patients. N Engl J Med. 2002;346:1061–6. [PubMed]
10. Everhart MA, Pearlman RA. Stability of patient preferences regarding life-sustaining treatments. Chest. 1990;97:159–64. [PubMed]
11. Silverstein MD, Stocking CB, Antel JP, Beckwith J, Roos RP, Siegler M. Amyotrophic lateral sclerosis and life-sustaining therapy: patients' desires for information, participation in decision making, and life-sustaining therapy. Mayo Clin Proc. 1991;66:906–13. [PubMed]
12. Danis M, Garrett J, Harris R, Patrick DL. Stability of choices about life-sustaining treatments. Ann Intern Med. 1994;120:567–73. [PubMed]
13. Emanuel LL, Emanuel EJ, Stoeckle JD, Hummel LR, Barry MJ. Advance directives. Stability of patients' treatment choices. Arch Intern Med. 1994;154:209–17. [PubMed]
14. Carmel S, Mutran EJ. Stability of elderly persons' expressed preferences regarding the use of life-sustaining treatments. Soc Sci Med. 1999;49:303–11. [PubMed]
15. Rosenfeld KE, Wenger NS, Phillips RS, Connors AF, Dawson NV, Layde P, et al. Factors associated with change in resuscitation preference of seriously ill patients. The SUPPORT Investigators. Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments. Arch Intern Med. 1996;156:1558–64. [PubMed]
16. Ditto PH, Smucker WD, Danks JH, Jacobson JA, Houts RM, Fagerlin A, et al. Stability of older adults' preferences for life-sustaining medical treatment. Health Psychol. 2003;22:605–15. [PubMed]
17. Kohut N, Sam M, O'Rourke K, MacFadden DK, Salit I, Singer PA. Stability of treatment preferences: although most preferences do not change, most people change some of their preferences. J Clin Ethics. 1997;8:124–35. [PubMed]
18. Straton JB, Wang NY, Meoni LA, Ford DE, Klag MJ, Casarett D, et al. Physical functioning, depression, and preferences for treatment at the end of life: the Johns Hopkins Precursors Study. J Am Geriatr Soc. 2004;52:577–82. [PubMed]
19. Lockhart LK, Ditto PH, Danks JH, Coppola KM, Smucker WD. The stability of older adults' judgments of fates better and worse than death. Death Stud. 2001;25:299–317. [PubMed]
20. Weissman JS, Haas JS, Fowler FJ, Jr., Gatsonis C, Massagli MP, Seage GR, 3rd, et al. The stability of preferences for life-sustaining care among persons with AIDS in the Boston Health Study. Med Decis Making. 1999;19:16–26. [PubMed]
21. The Connecticut Hospice I . Summary Guidelines for Initiation of Advanced Care. John Thompson Institute; Branford, CT: 1996.
22. Murphy DJ, Knaus WA, Lynn J. Study population in SUPPORT: patients (as defined by disease categories and mortality projections), surrogates, and physicians. J Clin Epidemiol. 1990;43:11S–28S. [PubMed]
23. Lawton MP, Brody EM. Assessment of older people: self-maintaining and instrumental activities of daily living. Gerontologist. 1969;9:179–86. [PubMed]
24. Inouye SK, Peduzzi PN, Robison JT, Hughes JS, Horwitz RI, Concato J. Importance of functional measures in predicting mortality among older hospitalized patients. JAMA. 1998;279:1187–93. [PubMed]
25. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in the aged: the index of ADL: a standardized measure of biological and psychosocial function. JAMA. 1963;185:914–919. [PubMed]
26. Pearlin LI, Lieberman MA, Menaghan EG, Mullan JT. The stress process. J Health Soc Behav. 1981;22:337–56. [PubMed]
27. Whooley MA, Avins AL, Miranda J, Browner WS. Case-finding instruments for depression. Two questions are as good as many. J Gen Intern Med. 1997;12:439–45. [PMC free article] [PubMed]
28. Pearlman RA, Cain KC, Patrick DL, et al. Insights pertaining to patient assessments of states worse than death. J Clin Ethics. 1993;4:33–41. [PubMed]
29. Feinstein AR, Cicchetti DV. High agreement but low kappa: I. The problems of two paradoxes. J Clin Epidemiol. 1990;43:543–9. [PubMed]
30. Breslow NE, Clayton DG. Approximate inference in generalized linear mixed models. J Am Stat Assoc. 1993;88:125–134.
31. McCulloch CE, Searle SR. Generalized, Linear, and Mixed Models. John Wiley & Sons; New York: 2001.
32. Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:13–22.
33. SAS Institute Inc. The NLMIXED Procedure. SAS/STAT User's Guide. SAS Institute Inc.; Cary, NC: 1999. Version 8.
34. Patrick DL, Pearlman RA, Starks HE, Cain KC, Cole WG, Uhlmann RF. Validation of preferences for life-sustaining treatment: implications for advance care planning. Ann Intern Med. 1997;127(7):509–517. [PubMed]
35. Pearlman RA, Starks HE, Cain KC, Cole WG, Patrick DL, Uhlmann RF. Integrating preferences for life-sustaining treatments and health state ratings into meaningful advance care discussions. In: van der Maas PJ, editor. Proceedings from the Royal Netherlands Academy of Arts and Sciences Colloquium on Epidemiological and Clinical Aspects of End-of-life Decision-Making; 2001.
36. Sprangers MA, Schwartz CE. Integrating response shift into health-related quality of life research: a theoretical model. Soc Sci Med. 1999;48:1507–15. [PubMed]
37. Loewenstein G, Schkade D. Wouldn't it be nice? Predicting future feelings. In: Schwarz N, editor. Well-Being: The Foundations of Hedonic Psychology. Russell Sage Foundation; New York: 1999. pp. 85–108.
38. Hibbard JH, Slovic P, Jewett JJ. Informing consumer decisions in health care: implications from decision-making research. Milbank Q. 1997;75:395–414. [PubMed]
39. Dresser R. Precommitment: a misguided strategy for securing death with dignity. Tex Law Rev. 2003;81:1823–47. [PubMed]
40. Sehgal A, Galbraith A, Chesney M, Schoenfeld P, Charles G, Lo B. How strictly do dialysis patients want their advance directives followed? JAMA. 1992;267(1):59–63. [PubMed]
41. Hawkins NA, Ditto PH, Danks JH, Smucker WD. Micromanaging death: process preferences, values, and goals in end-of-life medical decision making. Gerontologist. 2005;45(1):107–117. [PubMed]
42. Ubel PA, Loewenstein G, Jepson C. Whose quality of life? A commentary exploring discrepancies between health state evaluations of patients and the general public. Qual Life Res. 2003;12:599–607. [PubMed]