Fried TR, Bradley EH, Towle VR. Valuing the Outcomes of TreatmentDo Patients and Their Caregivers Agree?. Arch Intern Med. 2003;163(17):2073-2078. doi:10.1001/archinte.163.17.2073
Copyright 2003 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2003
Treatment outcomes are an important determinant of patients' treatment preferences. Although studies have examined how well surrogates agree with patients' preferences for specific treatment interventions, agreement regarding the valuation of health states as treatment outcomes is unknown.
Cross-sectional cohort study consisting of in-home interviews with 193 persons 60 years or older and seriously ill with cancer, congestive heart failure, or chronic obstructive pulmonary disease and their caregivers. Patients were asked whether, facing an exacerbation of illness, they would find a series of health states acceptable as a result of treatment (a rating of "unacceptable" meant they would prefer to die than to receive treatment). Caregivers were asked whether they would find these states acceptable for the patient.
There was 80% or greater agreement for health states that were overall rated either acceptable (current health, mild memory impairment, mild pain, or other symptoms) or unacceptable (coma). There was 58% to 62% agreement (κ = 0.10-0.25) about states with more severe physical or cognitive impairment. When disagreement occurred, caregivers were more likely to rate the state as acceptable. There was 61% to 65% agreement (κ = 0.20-0.28) about states with severe pain or other symptoms. When disagreement occurred, caregivers and patients were equally likely to rate the state as acceptable.
Patient-caregiver agreement about the acceptability of health states with functional or cognitive impairment, severe pain, or other symptoms was poor. Caregivers making surrogate decisions based on considerations of treatment outcomes may not effectively represent patients' preferences.
DYING PATIENTS are frequently unable to participate in treatment decision making.1 In these cases, physicians typically turn to a surrogate decision maker, who is asked to decide according to the principle of substituted judgment which treatment the patient would have wanted. Many studies have demonstrated that surrogates are poor predictors of patients' preferences for specific interventions.2- 8 Furthermore, even patients and surrogates who have had multiple end-of-life discussions are unlikely to have discussed the use of specific interventions.9 Surrogates' accuracy in predicting patients' valuations of outcomes of treatment has not been similarly examined. Because treatment outcomes are a primary determinant of patients' end-of-life treatment preferences,10- 13 agreement of surrogate and patient valuations of treatment outcomes is an important measure of the quality of surrogate decision making.
Although most studies examining surrogate decision making ask surrogates to decide as they think the patient would, the ability of surrogates to apply the principle of substituted judgment has been called into question. Surrogates' predictions of preferences for specific interventions have been shown to resemble more closely their own preferences than the preferences of the person on whose behalf they would be making decisions.14 When asked on what standard they would make treatment decisions, only 4% of spouses of patients with Alzheimer disease described substituted judgment as the only principle they would use.15
The purpose of the present study was to compare seriously ill patients' and surrogates' attitudes toward treatment outcomes described in terms of different health states resulting from an intervention. To reflect the finding that surrogates may not use substituted judgment when making treatment decision, surrogates were asked to assess these health states according to what they believed was best for the patient. By asking surrogates to think about these health states as they would when actually faced with a treatment decision, this examination provides an evaluation of how closely surrogates' decisions would match patients' preferences.
Patients for this study were drawn from a study of treatment preferences among seriously ill older persons.10 Briefly, we applied objective criteria to identify patients 60 years or older with potentially limited life expectancy secondary to cancer, congestive heart failure (CHF), or chronic obstructive pulmonary disease (COPD) in both the inpatient and outpatient setting. The study protocol was approved by the human investigations committee of each of the participating hospitals (Yale-New Haven Hospital, New Haven, Conn; Hospital of St Raphael's, New Haven; VA Connecticut Healthcare System, West Haven, Conn; Bridgeport Hospital, Bridgeport, Conn; and Middlesex Hospital, Middletown, Conn).
We screened sequential medical records of patients meeting age and diagnosis criteria for criteria defining limited life expectancy. These were criteria of Connecticut Hospice Inc (Branford),16 which were developed for use in clinical settings to select patients appropriate for hospice services because of an approximate life expectancy of 6 months or less, or the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment (SUPPORT) eligibility criteria,17 which define a patient population with an aggregate 6-month mortality rate of 29% to 62% when used to identify patients with CHF, COPD, or certain cancers at hospital admission.18 An additional eligibility criterion that was determined by a telephone screen was the need for assistance with at least 1 instrumental activity of daily living (IADL),19 which was selected both to improve the prognostication of limited life expectancy20 and to ensure that the patient had a caregiver who was actively involved in his or her care.
Of the 548 patients identified by medical record review, 469 received the telephone screen, and of these patients, 108 were excluded because they required no IADL assistance. Additional exclusion criteria included cognitive impairment, as measured by the Short Portable Mental Status Questionnaire21 and the Executive Interview (EXIT), a test of executive functioning22 (n = 76), and part-time Connecticut residence (n = 6). Of the 279 patients eligible for participation, 2 died prior to participation and 51 refused participation, resulting in 226 participants. Nonparticipants did not differ from participants according to age or sex. Among eligible patients with CHF, 8% refused participation compared with 19% among patients with cancer and 25% among patients with COPD (P = .02).
Of the caregivers identified by the patient as providing the greatest assistance with IADLs during the telephone screen, 193 participated. Of the remaining 33, the patient refused the request to approach a caregiver for participation (n = 9), the patient had only a paid, formal caregiver (n = 9), or the caregiver refused participation (n = 15). Patients with participating caregivers did not differ from patients without participating caregivers according to age, education, ethnicity, income, activities of daily living (ADL) status,23 or self-rated health. Compared with patients with participating caregivers, patients without participating caregivers were more likely to be female (58% vs 40%; P = .07), to be unmarried (70% vs 38%; P<.001), and to live alone (58% vs 18%; P<.001). Only the 193 patients who had participating caregivers are included in the analyses for this study.
Patients and caregivers underwent independent interviews in their homes. They were asked to think about the patient having an exacerbation of illness requiring treatment and to rate 11 health states resulting from treatment as representing either an acceptable or unacceptable quality of life. Using the concept of "states worse than death,"24 they were told that rating the health state unacceptable meant that the death of the patient would be preferable to undergoing treatment. The states included the following:
The patient's current health
Three states of physical disability, described as (1) not being able to get out of the house to visit family, attend church, go to work, do volunteer work, or do hobbies; (2) only being able to get from bed to chair and requiring help with bathing, dressing, and grooming; and (3) being bed bound and requiring help going to the bathroom
Two states representing mild and severe cognitive impairment, described as (1) having some problems with memory such that you forget things that people have told you and cannot remember things like the month or year and (2) having severe problems with memory such that you cannot recognize family members
Four states representing the presence and severity of different symptoms, described as (1) being in mild pain daily; (2) being in more severe pain daily; (3) having other mild symptoms daily, such as fatigue, nausea, shortness of breath; and (4) having other more severe symptoms daily
Patients were asked to rate the state according to whether they thought it would be acceptable or unacceptable; caregivers were asked to rate the state according to whether they thought it would be acceptable or unacceptable for the patient.
We assessed both patients' and caregivers' sociodemographic, health, and psychosocial status. Patients' sociodemographic variables included age, sex, race, education, marital status, and income. Health-related variables included primary diagnosis, self-rated health (in which the participant was asked to rate his or her health as excellent, good, very good, fair, or poor), and number of dependencies in ADLs. Psychosocial status included depression, as measured by the 2-item Primary Care Evaluation of Mental Disorders (PRIME-MD) instrument,25,26 and global quality of life, in which the participant was asked to rate his or her quality of life as best possible, good, fair, poor, or worst possible.
Caregivers' sociodemographic variables included age, sex, race, education, marital status, income, and relationship to the patient. Health status was measured by self-rated health. Psychosocial status included caregiver stress, which was measured using a subsection of the Caregiver Difficulties Questionnaire.27
Patients and caregivers were also asked to provide an estimate of what they believed the patient's life expectancy to be. Patients were asked if they had a durable power of attorney for health care or a health care proxy.
We used simple frequencies and proportions to describe the patient and caregiver populations. We examined patient and caregiver agreement regarding the acceptability of the 11 health states in terms of (1) the percentage of patient/caregiver pairs in which both agreed that the state was acceptable or unacceptable and (2) the κ coefficient to correct for agreement due to chance alone. We used the McNemar test to determine whether there were systematic differences in the direction of the disagreement within the patient/caregiver pairs (ie, whether there were significant differences in the proportion of patient/caregiver pairs in which the patient rated the health state as acceptable when the caregiver did not vs the proportion of pairs in which the caregiver rated the health state as acceptable when the patient did not). We examined patient and caregiver factors associated with patient-caregiver agreement vs disagreement about the acceptability of individual health states using the χ2 test for categorical variables.
The 193 patients had a mean ± SD age of 73.1 ± 7.2 years, 40% were female, and 91% were white. Patients with CHF were older than patients with cancer or COPD, and a larger proportion were nonwhite and male. A larger percentage of patients with COPD rated their health as poor and had at least 1 impairment in ADLs compared with patients with cancer or CHF. A total of 41% of patients reported that they had a durable power of attorney for health care or health care proxy (Table 1). The 193 caregivers had a mean ± SD age of 61.6 ± 14.3 years, 78% were female, and 92% were white. Spouses accounted for 57% of caregivers, children for 27%, other relatives for 9%, and others for 7% (Table 2).
For 5 of the 11 health states, patients and caregivers had 80% or higher agreement about the acceptability of the state (Table 3). These states included the patient's current health, mild memory impairment, mild pain, mild other symptoms, and being comatose. Despite the high level of agreement, the κ coefficients were low, ranging from 0.14 to 0.32. However, because of the distribution of responses (ie, most patient/caregiver pairs rated the state either as acceptable or as unacceptable), it may not be appropriate to use the κ coefficient as a measure of chance-adjusted agreement.28 Of the patient/caregiver pairs, 78% of pairs agreed about whether being confined to the house was acceptable. Substantially smaller percentages of pairs (58%-62%) agreed about the more physically and cognitively impaired states (κ = 0.10-0.25). For each of the states of physical impairment and severe cognitive impairment, when disagreement occurred, caregivers were significantly more likely to rate the state as acceptable (McNemar P<.01 for each state). Agreement was also low (61%-65%) for the states of severe pain and severe other symptoms (κ = 0.20-0.28). When disagreement occurred in these states, neither patients nor caregivers were more likely to rate the states as acceptable (McNemar P>.05 for each state).
Because caregivers were more likely to rate states of physical and cognitive impairment as acceptable than were patients, we explored whether there was a consistent group of caregivers who were simply unwilling to rate any health state as unacceptable. It appears instead that caregivers' ratings of acceptability were scenario specific. Among 72 caregivers who rated being bed bound as acceptable when the patient did not, 33 (46%) agreed with the patient that severe memory problems were unacceptable. Among 61 caregivers who rated severe memory problems as acceptable when the patient did not, 21 (34%) agreed with the patient that being bed bound was unacceptable and 1 (2%) rated bed bound as unacceptable when the patient did not.
To examine factors associated with patient-caregiver disagreement, we selected the health states in which the levels of disagreement were highest. These were being bed bound, having severe cognitive impairment, having severe pain daily, and having other severe symptoms daily (Table 4). No patient or caregiver characteristics were associated with disagreement about the state of being bed bound. For the state of severe cognitive impairment, female patients were more likely than male patients to disagree with their caregivers (48% vs 34%; P = .05). Patients who were not white were also more likely to disagree with their caregivers (65%) than were patients who were white (38%; P = .03). Similar results were obtained when the race of the caregiver was examined. Patients without a durable power of attorney for health care were more likely to disagree with their caregivers (50%) than were patients with a durable power of attorney (25%; P<.01). For both the states of severe pain and severe other symptoms, patients' quality of life was associated with disagreement about the acceptability of the state. Patients who rated their quality of life as the best possible or good were more likely to disagree with their caregivers about the acceptability of severe pain (43%) and severe other symptoms (44%) than were patients who rated their quality of life as fair, poor, or worst possible (18% [P<.01] and 29% [P = .04]). In contrast to the finding for the state of severe cognitive impairment, patients who had a durable power of attorney were more likely to disagree with their caregivers about the state of severe other symptoms (48%) than were patients without a durable power of attorney (32%; P = .03). Patients' and caregivers' education and perception of the patients' life expectancy, patients' depression, and caregivers' age, sex, relationship to the patient, and level of caregiving stress were not associated with disagreement about any of the 4 health states.
In this study of seriously ill older persons and their caregivers, each was asked to decide whether certain health states resulting from treatment would be acceptable for the patient, with unacceptable ratings meaning that they believed it would be preferable for the patient to die of his or her underlying disease. There was substantial agreement between patients and caregivers only for states that were rated as acceptable by most patients and caregivers, such as the patient's current health, mild memory impairment, mild pain, mild other symptoms, or rated as unacceptable by most patients and caregivers, such as coma. Other health states with more variability in their acceptability ratings, including moderate and severe functional impairment, severe cognitive impairment, severe pain, and severe other symptoms, had lower levels of agreement.
These results are consistent with a body of literature in which patient-surrogate agreement regarding preferences for specific interventions on behalf of the patient has been examined and found to be low.2- 8 These studies have led to concern that when patients are incapacitated and surrogates are called on to make treatment decisions, these decisions are not concordant with how the patient would have decided if he or she were able. Specific interventions, however, are fundamentally different depending on clinical circumstances, which determine the outcome the treatment is likely to provide.29 Because treatment outcomes have been shown to be a strong determinant of patients' treatment preferences,10- 13 it has been suggested that the elicitation of preferences center on the valuation of treatment outcomes.30,31 Although a consideration of treatment outcomes to guide treatment decision making might better capture patients' preferences, the results of the present study suggest that it may not help surrogates to make decisions consistent with the preferences of patients. In addition, the results demonstrate that caregivers' agreement with patients' valuations of outcomes are associated with few caregiver or patient characteristics. Furthermore, the magnitude of these associations was modest, the characteristics differed according to the health state being assessed, and, given the large number of variables examined and found not to be associated with agreement, may represent chance findings rather than true associations. The findings of our study suggest that caregivers who are more or less likely to agree with patients cannot be easily identified.
Nevertheless, the study results provide distinct patterns of disagreement according to the type of health state. In the case of functional or severe cognitive impairment, caregivers who disagreed with patients were significantly more likely to rate the states as acceptable when the patient did not. One reason for patients' frequent rating of dependent states as unacceptable may be their concerns about being a burden to family members because of their illness.13,32,33 Caregivers, however, appear more willing than patients to tolerate states of patient dependence, despite the potential for increasing caregiving responsibilities and independent of their perception of caregiver burden. In the case of pain and other symptoms, patients and caregivers who disagreed were equally likely to rate the states as acceptable. Caregivers' ratings of symptomatic states as unacceptable suggest that there are circumstances in which they would be less willing than the patient to have the patient undergo therapy. This is similar to the finding that more patients than their surrogates wish that they live as long as possible despite suffering9 and contrasts with previous studies demonstrating that surrogates tend to overestimate patients' preferences for invasive treatment interventions in hypothetical scenarios.5,11 Literature discussing the "demanding" surrogate who requests futile care or care that is not in the best interest of the patient34- 36 further focuses attention on surrogates who, as medical decision makers, are reluctant to forgo therapy, leading to potential overtreatment of the patient. Relatively little attention has been paid to surrogates whose decisions may potentially lead to the undertreatment of patients.37
In our study, caregivers were asked to rate the acceptability of health states based on what they thought was best for the patient. This raises the question of whether agreement would have been higher had caregivers explicitly been asked to use substituted judgment and to rate the health states based on how they believed patients would answer. However, although substituted judgment is the standard by which decisions to uphold patient autonomy would ideally be made,38 substituted judgment is a difficult standard to use.39 Moreover, studies of surrogate decision making suggest that surrogates do not use this standard.12,40- 42 Therefore, asking the question in terms of what the surrogate thinks is best for the patient without specifying the standard(s) the surrogate should use may be a more accurate reflection of how decisions are actually made.
A limitation of the study is that we do not know if the participating caregiver is the person the patient would specifically designate as his or her health care proxy. However, given that the caregiver was chosen as the person with the greatest involvement in providing functional assistance, it is likely that this person would also be involved in medical decision making. The major strength of the study is that health state ratings were examined among a group of patients who, because of a serious underlying illness, have had experience with the health care system and with medical interventions, such that consideration of treatment outcomes is highly relevant.
The health states examined in this study are a central component of treatment preferences, and it is important for surrogates to know the states in which patients do not want to be kept alive.9 Nonetheless, agreement between surrogate and patient valuations of these states is poor and is no better than agreement between valuations of specific treatment interventions. However, several findings from this and other studies suggest that agreement between patients and surrogates can be improved with increased communication. First, agreement about specific interventions has been shown to be better when patients and surrogates have discussed end-of-life care in detail than when such communication has not occurred.8 Second, a pilot randomized controlled trial demonstrated that patient-caregiver agreement regarding preferences for specific interventions can be improved with the use of a trained nurse facilitator leading discussions with patients and caregivers.43 Finally, for one of the health states in this study, severe cognitive impairment, agreement between patients and caregivers about the acceptability of the state was significantly better when the patient had formally appointed a surrogate decision maker. Therefore, seriously ill patients and their surrogates need to be encouraged to discuss their attitudes toward states of functional and cognitive impairment and states with pain and other symptoms, for which little patient-caregiver agreement currently exists. Although many patients want their surrogates to have leeway when making decisions and not be bound by patients' previously expressed wishes,44 increased discussion about patients' preferences for treatment outcomes can only improve surrogates' understanding of patients' hopes and fears as they face progressive serious illness.
Corresponding author and reprints: Terri R. Fried, MD, Geriatrics & Extended Care, 240, VA Connecticut Healthcare System, 950 Campbell Ave, West Haven, CT 06516 (e-mail: email@example.com).
Accepted for publication November 21, 2002.
The study was supported by grant PCC-98-070-1 from VA Health Services Research & Development (HSR&D), Washington, DC; grant K02 AG20113 From the National Institute on Aging, Bethesda, Md; and a VA HSR&D Career Development Award and Paul Beeson Physician Faculty Scholars Award (Dr Fried).
This study was presented in part at the annual meeting of the American Geriatrics Society; May 10, 2002; Washington, DC.
We thank Carm Joncas, RN, and Barbara Mendes, RN, for their interviewing skills; John O'Leary, MA, for his careful data management; and the patients and their caregivers for their willingness to grapple with difficult issues.