Assessing the decision making capacity of terminally ill patients with cancer

Abstract

Objective

Despite the clinical, ethical and legal magnitude of end-of-life decision-making, the capacity of terminally ill patients to make the medical decisions they often face is largely unknown. In practice, clinicians are responsible for determining when their patients are no longer competent to make treatment decisions, yet the accuracy of these assessments is unclear. The purpose of this study was to explore decision-making capacity and its assessment in terminally ill cancer patients.

Method

Fifty-five patients with advanced cancer receiving inpatient palliative care and 50 healthy adults were administered the MacArthur Competence Assessment Tool for Treatment (MacCAT-T) to evaluate decision-making capacity with regard to the four most commonly used legal standards: Choice, Understanding, Appreciation and Reasoning. Participants made a hypothetical treatment decision about whether to accept artificial nutrition and hydration for treatment of cachexia. Participants’ physicians independently rated their decision-making capacity.

Results

Terminally ill participants were significantly more impaired than healthy adults on all MacCAT-T subscales. Most terminally ill participants were able to express a treatment choice (85.7%), but impairment was common on the Understanding (44.2%), Appreciation (49.0%) and Reasoning (85.4%) subscales. Agreement between physician-rated capacity and performance on the MacCAT-T subscales was poor.

Conclusions

The use of the MacCAT-T revealed high rates of decisional impairment in terminally ill participants. Participants’ physicians infrequently detected impairment identified by the MacCAT-T. The findings from the present study reinforce the need for establishing comprehensive advance directives for patients with advanced cancer.

Patients with advanced cancer must make many important treatment decisions. Patients are often asked to appoint a health care proxy, consent to a DNR order, initiate palliative care or hospice services, and continue or discontinue life-prolonging treatments (1). In addition, patients in some regions of the U.S., and some countries around the world, may face decisions about whether to seek physician aid in dying, euthanasia and other interventions that hasten death. These decisions have important clinical, ethical, and legal ramifications.

A 2011 systematic review estimated that the prevalence of impaired decision-making capacity to make specific medical decisions (as assessed either by a physician or instrument for assessing capacity) was 26% in the general medical population (2). However, decision-making capacity often declines as patients approach the end of life, with up to 67% of patients exhibiting impairment in the last week of life (3). In the context of cancer, patients may experience impaired decision making capacity as a result of age (4, 5), hospitalization (6), or the cancer and treatment itself (7). However, few studies have attempted to understand decision-making capacity in advanced or terminally ill cancer patients with regard to the legal standards of decisional competence. Those studies that do exist have revealed rates of decisional impairment in patients in advanced cancer that ranged from approximately 33% to 50% (7, 8).

In treatment settings, physicians are largely responsible for determining when a patient suffering from progressive disease no longer has the capacity to make specific decisions, and they have an ethical and legal obligation to ensure that their patients are capable of providing informed consent (9). This entails making sure that each patient understands his or her treatment options, is aware of their condition, and can exercise free will (10). Through the assessment of decision-making capacity, clinicians attempt to resolve the tension that can arise between patient autonomy (i.e., their right to make treatment decisions) and their ethical obligation to protect patients from the harm that might ensue from impaired decision making (10, 11). Most clinicians make these determinations by conducting an interview and asking questions, using an approach that is best described as unstructured or unaided (6). However, research shows poor agreement between clinicians’ evaluations of capacity, particularly when using unaided clinician assessments (12–14). Even when using standardized measures of decision-making capacity, agreement between clinicians in the assessment of decisional capacity is generally poor (15–17), particularly in adults with cognitive impairment (13). In one recent study, physicians identified older cancer patients as lacking capacity to make informed treatment decisions at a rate no greater than chance when compared to a standardized measure (17).

The purpose of this study was to examine decision-making capacity and its assessment in terminally ill cancer patients using a structured measure that targets the cognitive abilities thought to comprise decisional competence (18). Specifically, decision-making capacity was evaluated with regard to the four legal standards most commonly used in the U.S., that have been articulated in the legal doctrine of informed consent: the patient’s ability to express a choice (Choice), understand information relevant to treatment decisions (Understanding), appreciate the significance of his or her condition, situation and the treatment decision (Appreciation) and rationally manipulate information in order to make comparisons and weigh treatment options (Reasoning) (19). The application of these widely accepted legal standards increases the external validity and policy implications of the study results.

A standardized measure of decision-making capacity, the MacArthur Competence Assessment Tool for Treatment (MacCAT-T) (20), was used to identify the nature and extent of decisional impairments in a sample of terminally ill cancer patients compared to a community sample of demographically similar participants who were not acutely ill. This was part of a larger investigation into the neuropsychological correlates of decision-making capacity and thus we focused on individual, rather than shared medical decision-making. The study also examined concordance between unaided, more global, physician assessments of their patient’s decisional capacity and the patient’s performance on the MacCAT-T. We hypothesized that the sample of terminally ill patients with cancer would demonstrate significantly more impairment than the comparison group in their ability to understand information related to a hypothetical treatment decision (Understanding), appreciate the nature of a hypothetical medical condition and its treatment (Appreciation) and identify risks and benefits of the treatment decision (Reasoning). We also hypothesized that agreement between clinician ratings of capacity and performance on the MacCAT-T would be poor.

Method

Participants

Participants were 105 English-speaking adults free of significant visual or auditory impairment. Individuals with psychotic symptoms, delirium, aphasia, or who were in “critical condition” as determined by medical record review, or examiner assessment, were ineligible for participation.

Terminally ill cancer patients were recruited from a 200-bed palliative care hospital in an urban, ethnically diverse neighborhood. Eligible patients in this facility had a life expectancy of less than six months and a diagnosis of advanced cancer. The majority of these patients admitted to the study site during the recruitment period were ineligible for study participation (See Figure 1). Ultimately, 55 participants provided study data.

Figure 1.

Recruitment of terminally ill patients with cancer receiving inpatient palliative care

Fifty comparison participants were recruited from two community-based organizations that serve an equally demographically and socioeconomically diverse, urban region. Prospective comparison participants were excluded from participation if, by self-report on the Charlson Comorbidity Index (21), they were diagnosed with a life-limiting medical illness (e.g., cancer, heart disease, dementia). Because comparison participants were recruited via flyer and presentation (that indicated the eligibility and exclusion criteria), all participants who volunteered were eligible for study participation.

Most participants were female (60%; n = 63), and age ranged from 50 to 88 (M = 67.5, SD = 10.9). Participants were diverse with regard to race and ethnicity, cancer diagnosis, and socioeconomic status (22). Demographic data are presented in Table 1. The terminally ill sample was significantly older than the comparison sample (M=69.60; versus M=65.26), t(103)=2.07, p=.04. In addition, the terminally ill group had a significantly higher percentage of White participants (74.5%) than the comparison sample (46.0%), whereas the comparison sample had a higher percentage of Black participants (44.0%) than the terminally ill sample (23.6%), χ²(df=4)=13.16, p=.01.

Table 1.

Descriptive Characteristics of Participants

	Sample	Terminally Ill (n = 55)		Comparison (n = 50)
		n	%	n	%
Gender	Male	23	41.8	19	38.0
	Female	32	58.2	31	62.0
Race	White	41	74.5	23	46.0
	Black	13	23.6	22	44.0
	Asian or Pacific Islander	0	0.0	3	6.0
	American or Alaska Native	0	0.0	2	4.0
	Other	1	1.8	0	0.0
Ethnicity	Hispanic	5	9.1	5	10.0
	Not Hispanic	50	90.9	45	90.0
Marital Status	Single	14	25.9	18	36.0
	Married/Cohabitating	14	26.0	11	22.0
	Divorced/Separated	13	24.1	18	36.0
	Widowed	13	24.1	3	6.0
Socio-economic status	Unemployed	6	11.3	8	16.0
	Manual laborer	2	3.8	8	18.0
	Skilled laborer	7	13.2	6	12.0
	Clerical/Sales worker	15	28.3	10	20.0
	Technician/semi-professional	5	9.4	6	12.0
	Manager/mid-level professional	9	17.0	9	18.0
	Administrator	3	5.7	2	4.0
	Higher executive/professional	3	5.7	0	0.0
Cancer diagnosis	Lung	10	18.2
	Pancreas	8	14.5
	Breast	8	14.5
	Ovary/Uterine	8	14.5
	Skin	4	7.3
	Esophagus	3	5.5
	Colon	3	5.5
	Stomach	3	5.5
	Lymphoma	3	5.5
	Other	5	9.0
		M	SD	M	SD
Age.		69.6	10.1	65.3	11.3
Years of Education		13.8	3.7	13.9	3.0

Informed consent

Because this study sought to examine impaired decisional capacity, and involved very little risk to study participants, the Institutional Review Boards of Fordham University and Calvary Hospital approved this study and deemed an “assent” standard appropriate. An assent standard is consistent with both case law and the research literature on informed consent (19), as the acceptable threshold for participation in research can be quite low when the risks are also very low. All participants had sufficient cognitive ability to provide informed consent after receiving a complete description of the study.

Procedure

After providing informed consent, participants were administered the MacCAT-T (20) through a semi-structured interview. Although the test is intended to be individualized based on the patient’s medical condition, we utilized a standardized version of this instrument in order to permit comparisons between patients and across groups. Because decision-making capacity is decision-specific, rather than global, participants were presented with a vignette that described a hypothetical, but very common condition, along with the recommended treatments and associated risks and benefits, and potential alternative treatments. Specifically, participants were asked to decide whether they would choose to accept artificial nutrition and hydration (ANH) for the treatment of cachexia in the context of advanced illness. The measure (and vignette) was found to be feasible and acceptable for use in the study population (23). The MacCAT-T assesses ability to make a choice by whether or not the patient articulates a preference based on the material presented. Understanding is assessed by the patient’s ability to paraphrase the information presented. Appreciation is assessed with questions about the application of the information to the patient’s life and current situation. Reasoning is assessed through questions about the consequences of treatments and alternatives.

Following study completion, the attending physician for each terminally ill participant responded to two questions regarding their patient’s decisional capacity based on their previous interactions with the patient. Specifically, physicians were asked whether they discussed any treatment decisions with the identified patient since his or her admission (e.g., do-not-resuscitate orders, use of analgesic medications). If so, the physician provided an assessment of the patient’s overall decision-making capacity using a 3-point scale (i.e., unimpaired, moderate impairment, or severe impairment). The physician responded to these questions within 48 hours of the patient’s study participation to minimize discrepancies due to changes in cognitive function over time.

Data analysis

To categorize level of impairment in decisional capacity based on MacCAT-T scores (i.e., unimpaired, moderately impaired or severely impaired), we used cutoff scores derived from comparison group. This method of identifying levels of decisional capacity has been used in previous studies of the decision-making capacity (7, 24, 25), where cut-scores for identifying impairment have not been established. Moderate impairment was defined as a score between 1.5 and 2.5 standard deviations below the mean of the comparison group and severe impairment was defined as a score more than 2.5 standard deviations below the comparison group mean. However, due to a lack of variance on the Choice subscale in the comparison group, a score of 1 was defined as moderate impairment and a score of 0 was defined as severe impairment. Analysis of covariance, with age as a covariate, was used to compare levels of decisional impairment between the terminally ill and comparison participants. Data analysis was conducted using SPSS statistics package, version 23.

Kappa coefficients were used to quantify the level of agreement between physicians and MacCAT-T performance. However, the initial analyses of agreement between MacCAT-T and physician ratings of capacity revealed an unequal distribution of data across categories, which can introduce bias into the Kappa rating (26). Thus, Kappa statistics were adjusted for prevalence and bias were calculated using the PABAK-OS (Prevalence and Bias Adjusted Kappa-Ordinal Scale) Calculator (27). Interpretation of the agreement between physician rating of capacity and the MacCAT-T subscales was based on guidelines identified by Landis and Koch (28); Values less than .20 were characterized as poor, .21to .40 as fair, .41 to .60 as moderate, .61 to .80 as substantial and .81 to 1.0 as almost perfect agreement.

Due to disease severity and fatigue, 87.3% (n=48) of the terminally ill sample completed all items on the MacCAT-T; 100.0% of the comparison participants completed the entire assessment. Hence, sample sizes vary by analysis. Physician ratings of capacity were obtained for 53 of the 55 terminally ill participants (96.3%).

Results

Levels of decisional impairment were generally low in the comparison sample across subscales. Conversely, participants in the terminally ill sample had much higher levels of impairment (see Table 2). Most terminally ill participants (n = 42, 85.7%) were able to express a treatment choice without difficulty. However, nearly half of participants (n = 23, 44.2%) evidenced some level of impairment on the Understanding and Appreciation subscales, and most were impaired on the Reasoning subscale (n = 41, 85.4%). Nearly ninety percent of terminally ill participants (n = 44, 89.8%) were impaired on at least one MacCAT-T subscale, whereas three participants (5.5%) were impaired on all four MacCAT-T subscales.

Table 2.

MacCAT-T Impairment

Terminally Ill Participants										Comparison Participants
MacCAT-T Subscale	Range	Mean	S.D.	Unimpaired		Moderate Impairment		Severe Impairment		Range	Mean	S.D.	Unimpaired		Moderate Impairment		Severe Impairment
				N	%	N	%	N	%				N	%	N	%	N	%
Choice	0 – 2	1.76	0.59	42	85.7	3	6.1	4	8.1	2	2.00	0.00	49	100.0	0	0.0	0	0.0
Understanding	0 – 6	3.49	1.40	26	53.1	11	22.4	12	21.8	2.1 – 6	4.89	0.88	45	90.0	2	4.0	2	4.0
Appreciation	0 – 4	3.08	1.18	25	51.0	0	0.0	24	49.0	3 – 4	3.91	0.27	45	90.0	0	0.0	4	8.0
Reasoning	0 – 8	4.50	2.07	7	14.6	13	27.1	28	58.3	5 – 8	7.46	0.79	44	88.0	3	6.0	2	4.0

Note. SD = Standard deviation

The terminally ill sample evidenced significantly lower scores on each MacCAT-T subscale than the comparison group after controlling for age (Choice: F (2, 98) = 3.86, p = .02; Understanding: F (2, 98) = 17.35, p < .001; Appreciation: F (2, 98) = 11.79, p < .001; and Reasoning F (2, 98) = 44.45, p < .001). The effect sizes for these comparisons ranged from medium (Choice, η² = .08) to large (Understanding, η² = .27; Appreciation, η² = .20; and Reasoning, η² = .49). In the terminally ill sample, Understanding and Appreciation were significantly correlated, r=.61, p < .001, and Reasoning was significantly correlated with all of the other subscales (r’s = .42, .44, and .60 for Choice, Understanding and Appreciation respectively).

Physician ratings of capacity

The majority of terminally ill participants (64.1%; n=34) were rated as unimpaired with regard to decision-making capacity; one third of participants were rated as moderately impaired (n=18, 33.9%) and one participant (1.8%) was rated as severely impaired. Interestingly, the one patient rated (by the treating physician) to be severely impaired was rated as unimpaired on three of the four MacCAT-T subcales (Choice, Understanding, and Appreciation) and as moderately impaired on the Reasoning subscale. Conversely, of the 34 individuals rated by their physician as unimpaired, rates of impairment on the MacCAT ranged from 4 (for Choice) to 17 (for Reasoning).

Prevalence and Bias Adjusted Kappa–Ordinal Scale indicated poor agreement between physician ratings and each of the MacCAT-T subscales (see Table 3). The highest (though still very poor) level of agreement was observed for the Choice subscale, PABAK-OS = −0.06, 95% CI [0.09, 0.36], with even poorer agreement for the subscales measuring Understanding, PABAK-OS = −0.03, 95% CI [0.09, 0.37], Appreciation, PABAK-OS = −0.05, 95% CI [−0.18, 0.08], and Reasoning, κ=−0.03, 95% CI [−0.40, −0.13].

Table 3.

Agreement between MacCAT-T Subscales and Physician Ratings of Impairment

MacCAT –T
	Unimpaired	Moderate impairment	Severe impairment	Total
Choice
Unimpaired	23	3	4	30
Moderate impairment	16	0	0	16
Severe impairment	1	0	0	1
Total	40	3	4	47
Understanding
Unimpaired	16	6	8	30
Moderate impairment	7	5	4	16
Severe impairment	1	0	0	1
Total	24	11	12	47
Appreciation
Unimpaired	14	0	16	30
Moderate impairment	9	0	7	16
Severe impairment	1	0	0	1
Total	24	0	23	47
Reasoning
Unimpaired	4	8	17	29
Moderate impairment	3	3	10	16
Severe impairment	0	1	0	1
Total	7	12	27	46

Note. Highlighted boxes indicate agreement between physician ratings of capacity and participant performance on individual MacCAT-T subscales.

Discussion

Given the magnitude of many end-of-life medical decisions, an understanding of terminally ill patients’ ability to make these decisions has become an important aspect of clinical care. Previous studies found substantial rates of impaired decisional capacity in patients at the end of life (8, 29). However, these studies typically assessed cognitive abilities more generally, not the ability to make specific treatment decisions. This study is the first to examine decision-making capacity in terminally ill cancer patients relative to the four most commonly used legal standards of competency in light of a hypothetical, yet very common treatment decision. The use of the MacCAT-T provides more comprehensive information about decisional capacity than instruments that assess capacity as a single ability or cognitive functioning more generally. Analysis of the MacCAT-T revealed strikingly high rates of decisional impairment when compared to a comparison sample of non-medically ill elderly adults, but also highlighted the heterogeneity of impairment across individuals and MacCAT-T subscales.

One of the most important findings from this study was the high prevalence of impaired decisional capacity in the terminally ill sample. The terminally ill sample evidenced significantly greater impairment than the comparison sample on each MacCAT-T subscale, even after controlling for the modest age difference between the two samples. The Choice subscale is generally considered the “easiest” standard of decision-making competence. To express a choice, the participant only needs to indicate a preference for a treatment option. However, 15% of the terminally ill participants evidenced impairment on this seemingly easy standard. Past studies of decisional capacity with seriously ill individuals found little impairment on the ability to make a choice (7, 25, 30–32). Moreover, nearly half of the participants were impaired on the Understanding and Appreciation subscales and 85% were impaired on the Reasoning subscale. In fact, all but five of the terminally ill participants (i.e., 90%) were impaired on at least one of the MacCAT-T subscales. These findings highlight the need for clinicians to be aware of the high likelihood that patients at the end of life will be impaired in their ability to make treatment decisions, and the need for a careful assessment of decisional capacity.

Impairment on multiple standards was common, but was not the rule, as rates of impairment increased as the rigorousness of the legal standard increased. This finding indicates that decisional impairment is not a global phenomenon, but differs depending on how capacity is operationalized. Hence, it may be important for evaluators to assess across a range of domains to identify impairments that may be present. The MacCAT-T findings provide important information about decisional capacity in terminally ill cancer patients, but final determinations about capacity status require an expert clinical evaluation.

The findings of high rates of decisional impairment in terminally ill patients reinforce the need for participation in advance care planning when patients are more likely to have capacity to make important medical decisions. Advance care planning is a process that aims to “help ensure that people receive medical care that is consistent with their values, goals and preferences during serious and chronic illness.(33)” This approach helps patients, surrogates and providers to prepare for end-of-life decision-making and improves both clinicians’ and surrogate decision-markers ‘understanding of patient preferences for end of life care (34). Providers can encourage frequent participation in this process to gain an understanding of patient goals and needs over the course of their illness and updating advance directives. Within a palliative care setting, advance care planning offers a solution to clinicians’ quest to balance respect for patient wishes with the responsibility to protect patients from harm resulting from impaired decision-making.

Like the present study, previous research has demonstrated poor agreement between clinician assessments of decisional capacity and performance on measures of decision-making (15, 16). Clinicians are more likely to find that their patients have intact decisional capacity compared to patient performance on standardized capacity instruments (14). Likewise, in this study treating physicians infrequently detected impairment identified by the MacCAT-T. Because the gold standard for capacity assessment is assessment by a physician who has been trained to conduct, and has completed an extensive number of capacity evaluations (2), it is not possible to comment on whether the MacCAT-T or unaided physician assessment determinations of capacity were correct in any particular instance.

There are several possible explanations for the poor agreement between assessment methods found in this study. The lack of agreement between physician ratings of capacity and MacCAT-T performance may have been due to the study methodology. First, we asked physicians to make a non-specific, global decision (capable/moderately impaired/severely impaired) with regard to patient capacity rather than evaluating the patients on each of the four standards with regard to a specific treatment decision as in the MacCAT-T. Physicians made this global rating based on their previous interactions with a patient, often shortly after admission to the hospital. Additionally, the possibility of errors in physician recall of patient interactions exists, or there may have been changes in decisional capacity between prior conversations and study completion. Ideally, physicians would have assessed each participant based on ability to make the same specific medical decision with regard to each legal standard at approximately the same time as study participation, but this “ideal” methodology would have substantially increased physician burden (and likely detracted from time available to care for the patients). Further, the physicians may have been unlikely to rate patients as “severely impaired” because they were able to assent to participate in the study, and therefore represented some of the most cognitively intact patients at the study site. Because of these methodological limitations, the poor agreement between MacCAT-T subscale scores and physician ratings of capacity should be interpreted with caution. A future study with a more detailed and specific physician rating of capacity is necessary to describe this relationship. Finally, physicians are frequently unsure about how to assess decision-making capacity. Many physicians receive no formal training in capacity assessment and may hold erroneous beliefs about decisional capacity (35). Indeed, in the study facility there is no formal training in the evaluation of decision-making, although a consultation service is available for cases in which capacity is unclear.

Overall, the study results provide information about patients with cancer at the very end of life. The sample included in this study may have limited the generalizability to the broader population of palliative care patients. Additionally, we only looked at individuals’ ability to make medical decisions in isolation. This is a study limitation as shared decision-making is prevalent and is considered standard of practice in healthcare delivery (36). Future study into decision-making capacity in the context of shared decision-making would provide information that is more generalizable. Our comparison group was free of major medical illness and thus may represent a healthier group than the general population of older adults. Cross-sectional findings (29) indicate that patients receiving inpatient palliative care had significantly higher rates of cognitive impairment than those patients who were receiving outpatient hospice care. The inclusion of a terminally ill outpatient sample or patients with advanced cancer but with a longer life expectancy would have strengthened the study. Further, it is worth considering the value of a comparison group of hospitalized, but not terminally ill participants to better assess if impaired capacity is more likely the result of hospitalization or the disease process.

These limitations notwithstanding, this study is the first to examine decision-making capacity in terminally ill cancer patients relative to the four most commonly used legal standards of competency. The use of the MacCAT-T provides more comprehensive information about decisional capacity than instruments that assess capacity as a single ability or cognitive functioning more generally. The MacCAT-T also revealed strikingly high rates of decisional and cognitive impairment in this sample, with nearly 90% demonstrating impairment on at least one legal standard. These results underscore the necessity for conducting research to inform best practices in palliative care. The designation of a systematic method for capacity assessment will allow clinicians to better preserve the autonomy of their patients and protect them from harm while providing minimally invasive assessments.

Highlights.

• Terminally ill patients with cancer were significantly more impaired than healthy adults on a measure of decisional capacity aligned to legal standards of decisional competence (Choice, Understanding, Appreciation, and Reasoning).

• Impairment on multiple standards was common.

• There was poor agreement between physician assessment of capacity and performance on the standardized measure of decisional capacity.