Magnitude of Impairment in Decisional Capacity in People With Schizophrenia Compared to Normal Subjects: An Overview

Abstract

The capacity of individuals with schizophrenia to make decisions related to research participation or clinical treatment has received increasing empirical attention. A number of studies have compared patients with schizophrenia to nonpsychiatric comparison subjects (NPCs) on structured measures of decision-making capacity. In this review, we evaluated the magnitude of the difference between schizophrenia and NPC groups reported across these studies, as well as the influence of sample characteristics on observed effect sizes. We also computed the effect sizes of group differences in psychopathology and cognitive deficits. Twelve studies met the search criteria; a majority of them reported data using the MacArthur Competence Assessment Tool for Clinical Research (MacCAT-CR) or for Treatment (MacCAT-T). The mean effect size (evaluated in terms of Cohen's d) for group differences on the Understanding subscale of the MacCAT instruments was 0.88 (SD = 0.40); it was twice as high among inpatient samples as among outpatients. Similar differences were observed in terms of Appreciation and Reasoning subscales, but the effect sizes for Expression of Choice were small (mean d = 0.29, SD = 0.24). Notably, these observed effect sizes were generally smaller than those for differences between schizophrenia and NPC groups in psychopathology (mean d = 2.06, SD = 1.03) and cognition (mean d = 1.01, SD = 0.61). The published studies demonstrate a substantial heterogeneity in decision-making capacity among people with schizophrenia, as well as among NPCs, suggesting that the presence of schizophrenia does not necessarily mean the patient has impairment in capacity.

The capacity to consent to research and even clinical care has become an increasing target of scientific inquiry over the past decade.¹ The ethical conduct of medicine and related research hinges on whether informed, voluntary consent can be provided. Probably the greatest concern over capacity has been raised about the protection of vulnerable populations, particularly those with progressive dementias or with serious mental illnesses, such as schizophrenia, as described in other articles in this issue of Schizophrenia Bulletin. Schizophrenia is associated with poor insight and cognitive dysfunction,² which may influence decisional capacity.³

Decision-making capacity involves several components, which typically include the following: (1) understanding (ie, comprehending the nature of the consent-relevant information), (2) appreciation (ie, understanding how the information applies to one's own condition and situation), (3) reasoning (with the information provided), and (4) evidencing a choice (about participation versus nonparticipation).^4,5 Due to the affective and cognitive complexity of these tasks, some clinicians may make assumptions about consistently impaired decision-making capacity in people with schizophrenia that are not always warranted.^6–8 The report of the National Bioethics Advisory Commission (NBAC)⁹ focused on the decision-making capacity of persons with major mental disorders, and it has been criticized for its rather stereotypical and potentially stigmatizing view of decisional ability among mentally ill individuals.⁷ There is a risk in assuming that decision-making capacity of persons with schizophrenia is always impaired, when they are able to make autonomous decisions, and in considering their decision-making capacity as permanently or globally impaired by virtue of their diagnosis. These concerns must be balanced against a danger of improper enrollment of vulnerable individuals in high-risk clinical trials, when the patients do not fully comprehend the nature of the procedures to which they are consenting.

In a recent survey 395 psychiatrists and psychologists indicated that attributions of decision-making capacity (or lack thereof) were frequently made in a nonspecific manner, particularly when these judgments were made in reference to persons with mental disorders.^6,10 Yet, there is a substantial amount of variability among people with schizophrenia in cognitive and functional characteristics,^2,11–14 as well as empirical data that demonstrates that some patients with schizophrenia or dementia do retain decision-making capacity (see reviews by Dunn and Roberts 2005;¹ Ganzini et al. 2003;¹⁰ Ganzini et al. 2005⁶). A related problem is the assumption that persons without psychiatric or cognitive disorders are intact in their decision-making capacity.^6,10

Based on the NBAC report and surveys of clinicians,^6,15 there appears to be an existing bias that assumes almost everyone with schizophrenia has impaired decisional capacity, whereas nonpsychiatric comparison subjects (NPCs) are not impaired. As a result, some institutional review boards now require investigators to evaluate decisional capacity in persons with schizophrenia but not in NPCs. In recent years some data have begun to accumulate comparing patients and various comparison groups using structured measures of decisional capacity (eg, the MacArthur Competence Assessment Tool for Clinical Research, or MacCAT-CR).¹⁶ Although there is no “gold standard” method for assessing capacity, these instruments allow for a scientific study of decision-making capacity. Previous studies have identified decisional capacity impairment in persons with schizophrenia, yet it is unclear to what degree this impairment is evident relative to NPCs and what circumstances moderate the impairment. To our knowledge, this is the first quantitative review of capacity instruments in schizophrenia. We computed effect sizes as benchmarks for assessing the magnitude of differences between schizophrenia patients and NPCs. Furthermore, we wanted to compare these effect sizes to those for more established differences between people with schizophrenia and NPCs on 2 of the hallmarks of schizophrenia—psychopathology and cognition. To enhance the clinical relevance of these findings, we additionally described the effect sizes in terms of the percent overlap between schizophrenia and NPC groups.¹⁷ A large effect size (and small overlap between these 2 groups) would indicate a need for evaluating capacity in practically all the patients with schizophrenia but in very few of the NPCs. On the other hand, a small effect size (and large overlap) would argue against basing the requirement for capacity assessment on a diagnosis of schizophrenia only. The effect size estimations would also indicate varied needs for developing interventions to enhance decisional capacity in people with schizophrenia and in NPCs. Finally, we examined sources of variability in decisional capacity among people with schizophrenia (eg, sample characteristics).

METHOD

Articles were identified with the use of PubMed, PsychINFO, and www.scholar.google.com searches (using the terms “consent,” “capacity,” “schizophrenia,” “decision making”), and the compilation of reference lists of all articles found from those sources. We selected only English-language articles published in peer-reviewed journals that reported empirical comparisons of decisional capacity between patients with schizophrenia and NPCs. Articles were excluded if they did not use structured measures of capacity to compare a group with a diagnosis of schizophrenia to a group of NPCs. No exclusions were made for year of publication. Where possible, we computed effect sizes (Cohen's d¹⁷) for differences reported between groups with schizophrenia and NPCs on measures of capacity, as well as for measures of psychopathology and cognition.

Results

The search procedures revealed 12 articles that met our selection criteria (Table 1). The mean sample size of groups with schizophrenia was 38.4 (range 6–80, SD = 22.3), whereas that of NPCs was 30.7 (range 15–82, SD = 18.0). The studies varied in terms of sample characteristics—for example, whether schizophrenia groups were strictly inpatients,^16,18–22 outpatients,^13,23 or mixed samples of both.²⁴ Some samples were restricted to middle-aged and elderly patients.¹³ In general, few exclusion criteria were listed, although most excluded persons with dementia.^23,25 The composition of NPC groups also varied from samples of healthy community volunteers,^13,23,26 to inpatients with non-neuropsychiatric medical diagnoses (eg, angina, 16 HIV-infected individuals²⁷), to outpatients with diabetes.²⁵ Some studies selected demographically matched NPCs (eg, Grisso and Appelbaum 1995³), while others statistically controlled for differences (eg, Palmer et al. 2005²⁵). Also, as noted above, decisional capacity is generally assessed in reference to a specific research protocol or treatment; all but 2 studies involved consent to some form of medication clinical research trial or treatment. (The 2 exceptions were Grisso and Appelbaum 1995;¹⁹ Grisso et al. 1997²⁰). Seven of the 9 studies of capacity related to a specific research project or treatment involved hypothetical protocols (Dunn et al. 2002²³ and Palmer et al. 2004¹³ did not). The reports differed in terms of whether the protocol was procedurally simple or complex, and whether the risks were minimal or greater.

Table 1.

Studies of Decisional Capacity Comparing Persons with Schizophrenia to Nonpsychiatric Comparison Subjects

Study	Sample Characteristics	Capacity Measure and Interscorer Reliability	Proposed Study or Intervention	Effect Sizes/Findings	Comments
Grisso et al. (1995)38 Grisso and Appelbaum (1995)3	SC Group: N = 75; Inpatients; Mean (SD) age = 34.8 (7.5) years	MacCAT original version: Understanding treatment disclosures; Perceptions of disorder; thinking rationally about treatment; Expressing a choice	N/A	Understanding (Uninterrupted): Hospitalized SC vs. Hospitalized Angina = 0.71	In a community NPC group, 4.0% were impaired on compound measure of capacity
	NPC group: N = 82 with angina; Inpatients; Mean age = 55.3 (10.4)	Reliability (ICC):		Appreciation (No SDs provided): Hospitalized SC = 13–35% with score < 4 vs. Hospitalized Angina = 0–9.8% with score <4	In the Angina group, 12.2% were impaired
	Demographically matched (age, gender, raced, education, occupation)	Understanding Treatment Disclosures = 0.91		Reasoning: Hospitalized SC vs. Hospitalized Angina = 1.11–1.17 (2 subtests)	In the SC group, 52.0% were impaired
		Thinking Rationally about Treatment = 0.91
Grisso et al. (1997)20	SC Group: N = 40; Inpatients; Mean age = 39	MacCAT-T	N/A	Understanding = 1.19	Subjects included if between age 18 and 65; English speaking, and excluded if too agitated or disoriented to participate
			Reliability (ICC):			Reasoning = 0.45
	NPC Group: N=40; Community-dwelling volunteers	Understanding = 0.99			0% of NPCs and 18% of SC group had Understanding scores below 3
			Appreciation = 0.87
	Demographically matched (age, gender, raced, education, occupation)	Reasoning = 0.91			6% of NPC and 20% of the SC group had Reasoning scores below 4
			Expressing a Choice = 0.97
Carpenter et al. (2000)24	SC Group: N = 30; Mostly inpatient; Age = 40.2 (8.8), Education = 12.2 (2.4)	MacCAT-CR	RCT of an atypical antipsychotic; Hypothetical	Understanding = 1.1	54% of the NPCs had scores below 20 on the Understanding subscale
			Reliability (ICC):			Reasoning = 0.90
			Understanding = 0.98			Appreciation = 1.19
	NPC group: N = 24; Age = 39.7 (10.2); Education 11.5 (2.1)	Reasoning = 0.84		Choice = 0.56 (NS)	No differences between SC and NPC groups in Understanding scores after an educational intervention
			Appreciation = 0.84
Wong et al., (2000)26	SC Group: N = 20; Inpatient and Outpatient; Mean age = 40.1	Authors' version of the MacCAT relevant to treatment	Consent to a blood test	No effect sizes calculable	Participants excluded if met criteria for learning disability or dementia; had no verbal expressive communication; required a complicated discussion of implications of blood testing (eg, HIV)
	NPC Group: N = 20; Mean age = 53.4	Reliability (Kappa): 0.87		90% of SC group and 100% of NPC group had adequate level of decisional capacity
Dunn et al. (2002)23	SC Group: N = 80; Outpatient; Age > 40; Mean age = 51; Mean education = 12.4	Comprehension assessment; 20 yes/no and short answer questions after the consent procedure	Participation in the Intervention Research Center; Real protocol	Trial 1 Comprehension difference: 1.12 Percentage scoring 100%	SC group had fewer years of education than the NPC group
	NPC Group: N = 19; Community volunteer; Age > 40; Mean age = 54; Mean education = 14	Reliability:	Consent presented in a routine or enhanced (PowerPoint) format	SC Group: Trial 1: 31% enhanced; 15% routine Trial 2: 85% enhanced; 63% routine
			N/A		NPC Group: Trial 1: 78% enhanced; 40% routine Trial 2: 100% enhanced; 100% routine
Moser et al. (2002)27	SC Group: N = 25; Mostly outpatient; Mean age = 31.6 (SD = 9.8); Mean education = 13.2 (SD = 2.6)	MacCAT-CR	RCT of a cognition-enhancing agent; Hypothetical	Understanding = 0.53	SC group was younger than NPC group
			Evaluation to sign consent		Appreciation = 0.59	Restricted to ages 18–55
			Reliability:		Reasoning = 0.19 (NS)	No matching
		N/A		Expression of Choice = 0.40 (NS)
	NPC Group: N = 25; Group HIV-positive; 15 with AIDS; Mean age = 37.4 (SD = 9.77); Education = 13.0 (SD = 2.1); 11 had major depression or anxiety disorder				On the evaluation to sign consent; 80% of SC group and 96% of HIV group scored in the adequate range
Saks et al. (2002)28	SC Group: N = 39 Inpatients and outpatients over age 50; Mean age = 51.0; Mean education = 12.3	California Scale of Appreciation	N/A	Across 3 raters, mean effect size = 0.84	NPC group was significantly older and more likely to be female than the SC group
			Reliability (percent agreement):			64.1% to 76.9% of patients were rated as “capable”
	NPC Group: N = 15; Mean age = 65.4; Mean education = 13.1	Schizophrenia group = 71.8–97.4 %		100% of NPCs were rated as “capable”
Kovnick et al. (2003)21	SC Group: N = 27 Long-stay inpatients; Mean age = 39.1; Mean length of stay = 7 years	MacCAT-CR	RCT of a new medication; Hypothetical	Understanding = 1.54	No significant differences on demographic variables between groups
			Reliability (Kappa):		Reasoning = 0.88	Ages 18–65
			Understanding = 0.69		Appreciation = 1.25
	NPC Group: N = 24; Mean age = 39.7; same NPC as in Carpenter et al. (2000)24		Reasoning = 0.53			Those with mental retardation or traumatic brain injury excluded
			Appreciation = 0.79
Cohen et al., (2004)22	SC Group: N = 6; Inpatients; Mean age = 34.1; Mean education = 14.0	MacCAT-CR	2 Hypothetical Studies:	None calculable; Not enough N for statistical comparisons	All NPCs scored above 20 on the Understanding scale, as did 5 of 6 SC patients, in both High- and Low-risk conditions
			Reliability:	Low-risk drug study
	NPC Group: N = 20; Mean age = 41.1; Mean education = 15.9	N/A (2 raters independently scored and resolved discrepancies)	High-risk Ketamine challenge study with little benefit
Palmer et al. (2004)13	SC Group: N = 59; Outpatient; Age = 50.2 (6.8); Education = 11.9 (2.6)	MacCAT-T	Treatment with atypical antipsychotic;	Understanding = 0.51	SC and NPC groups differed in age, education, percent male (higher values in NPC group)
			Reliability (ICC):			Reasoning = 0.53 (NS)
	NPC Group: N=38; age =56.8 (9.2); education = 14.3 (2.1)	Understanding = 0.85	Real for SC group (all already receiving antipsychotics); Hypothetical for NPC group	Expression of Choice = 0.20 (NS)	Those with dementia and/or physical/medical problems interfering with assessment were excluded
				Reasoning = 0.75
				Appreciation = 0.87
Combs et al. (2005)18	SC group: N = 25; Inpatient; Age = 37.9 (SD= 7.4); Education = 12.0 (1.8)	8 questions about the study; uncued recall and cued recognition, each yielding an understanding score	Study on the “assessment of psychosis”; Hypothetical	SC group Recall = 61%	NPC group had greater education and was younger
							SC group Recognition = 88%
							NPC group Recall = 82%
	NPC Group: N = 25; Community group matched on ethnicity and gender; Age = 22.4 (SD = 7.4); Education = 14.5 (1.0)	Reliability:		NPC group Recognition = 93%	Age, education, reading level, and attention span were controlled for in Capacity comparisons
			N/A			SC and NPC groups differed on Recall only
Palmer et al. (2005)25	SC group: N = 35; Older, clinically-stable outpatients; Mean age = 65.7; Mean education = 12.0	MacCAT-CR	RCT of a cognition enhancing agent; Hypothetical	Understanding = 0.67	Subjects included if over age 60; fluent in English; without comorbid diabetes in the SC group; with MMSE score greater than 18
				alpha > 0.80			Appreciation = 0.69
				3-item questionnaire (Purpose, Risks, and Benefits of the study)			Reasoning = 0.49 (NS);
	NPC Group: N = 36; With Diabetes Mellitus; Mean age = 70.9; Mean education = 14.5	Reliability (ICC):		Expression of a Choice = 0.0 (NS)
			0.90		3-item questionnaire = 0.84

Note: SC = Schizophrenia; NPC = Nonpsychiatric Comparison Subjects; NS = Non-significant; MacCAT-CR = MacArthur Competence Assessment Tool for Clinical Research; MacCAT-T = MacArthur Competence Assessment Tool for Treatment; RCT= Randomized Controlled Trial; ICC = Interclass Correlation Coefficient

Data were based on different instruments to assess capacity. Two utilized the original instruments by Grisso and Appelbaum from the MacArthur Competence Assessment studies—the Understanding of Treatment Disclosures, Perception of Disorder, and Thinking Rationally About Treatment scales.¹⁶ The MacCAT-T (MacArthur Competence Assessment Tool for Treatment) was an abbreviated form of those original scales that was created by Grisso and Appelbaum to be more readily usable in clinical settings.¹⁶ Two of the studies of decisional capacity in schizophrenia employed the MacCAT-T.²⁰ The MacCAT-CR was adapted by Grisso and Appelbaum from the MacCAT-T, as a measure of capacity to consent to research; 5 of the 12 studies used the MacCAT-CR.²⁹ Thus, although most of the studies used measures having common ancestry, the component items and structures of these scales varied in some key aspects, so comparison of results across studies should be done with appropriate caution.³⁰ Nine of 12 studies reported some form of interscorer reliability, and the majority reported high levels of agreement (eg, kappa or interclass correlation coefficient > 0.80).

From 10%²⁶ to 52%¹⁹ of the people with schizophrenia and from 0%²⁶ to 18%¹⁸ of NPCs were classified as being impaired in capacity. However, across studies, criteria (and measures) for defining impairment in capacity varied, and not all of the studies reported the proportion classified as impaired.

On the various MacArthur instruments, comparisons between groups with schizophrenia and NPCs showed large effect sizes on the Understanding subscale (mean d = 0.88, SD = 0.40, 7 studies), as well as on the Appreciation subscale (mean d = 0.93, SD = 0.34, 4 studies) (Figure 1). On the Reasoning (mean d = 0.65, SD = 0.34, 7 studies) and Expression of Choice (mean d = 0.29, SD = 0.24, 4 studies) subscales, effect sizes were small to medium. Based on effect sizes ranging from 0.29 to 0.93, we found that schizophrenia and NPC groups overlapped by 48% to 79% on the MacCAT measures,¹⁷ suggesting that decisional impairment was not a distinguishing feature of schizophrenia in a majority of cases. In general, effect sizes comparing different MacArthur scales were larger (range 0.45–1.54; median = 1.17) for studies comparing mostly inpatients with schizophrenia to NPCs^19,21,24,28 and smaller (range 0.0–0.84, median = 0.53) for reports that included clinically stable, community-dwelling outpatients with schizophrenia.^13,25,27 Furthermore, these effect sizes were generally lower than those for comparisons between schizophrenia patients and NPCs on measures of psychopathology (either the Brief Psychiatric Rating Scale³¹ or Positive and Negative Syndrome Scale Total³²) (mean effect size = 2.06, SD = 1.03; 4 studies) or neurocognition (either the Mini-Mental State Examination,³³ Mattis Dementia Rating Scale,³⁴ or Repeatable Battery for the Assessment of Neuropsychological Status³⁵ (mean effect size = 1.01, SD = 0.61; 6 studies).

Fig. 1.

Effect Sizes for Differences on Subscales of MacArthur Competence Assessment Tools (Mac-CAT) Comparing Schizophrenia and Nonpsychiatric Comparison Groups. Bars represent means with standard deviation. Understanding Scale = 7 studies; Appreciation Scale = 4 studies; Reasoning Scale = 7 studies; Expression of Choice Scale = 4 studies; Psychopathology (Brief Psychiatric Rating Scale or Positive and Negative Syndrome Scale) = 4 studies; Neurocognition (Mini-Mental State Examination, Mattis Dementia Rating Scale, or Repeatable Battery for the Assessment of Neuropsychological Status) = 6 studies.

Discussion

As with other ethical issues in obtaining informed consent (eg, voluntariness, disclosure of information in language understandable by the recipient), empirical data are only just beginning to accumulate concerning decisional capacity. Our review of 12 published empirical studies that compared structured measures of decision-making capacity between persons with schizophrenia and NPCs revealed considerable variability in study design in terms of sample sizes (mean sample sizes were 35 with schizophrenia and 30 NPCs), setting (inpatient versus outpatient), age range of the participants, measures used to assess capacity, and the nature of the proposed study or intervention. Decisional capacity is a context-specific construct, and it is difficult to compare the absolute magnitude of decisional capacity scores across studies. Therefore, definitive statements about the magnitude of differences between patients with schizophrenia and NPCs are difficult to make.

Despite these caveats, this quantitative review supports the following conclusions. First, impairment in capacity is not a distinguishing feature of schizophrenia. Among the 4 studies that reported a proportion of participants with impaired capacity, a majority of people with schizophrenia was deemed to have adequate decision-making capacity. The magnitude of the difference (ie, effect sizes) comparing MacCAT subscales between groups with schizophrenia and NPCs ranged from 0.29 to 0.93. These effect sizes, ranging from small to large, can be interpreted as showing that schizophrenia and NPC groups overlapped by 48% to 79% on the MacCAT measures.¹⁷ Moreover, these mean effect sizes were only about 50% of those associated with differences in measures of psychopathology in these same samples, and slightly lower than those in measures of cognition. Therefore, measures of decisional capacity do not seem to reliably distinguish persons with schizophrenia from NPCs, as measures of psychopathology do; this illustrates the flaw in equating diagnostic status with risk for impaired decisional capacity. The reviewed studies suggest that a small proportion (up to 18%) of NPCs may have significant impairment in decision-making capacity. Much of the research on decisional capacity in mentally ill persons has focused on patients with schizophrenia. The degree to which deficits in decisional capacity in these patients reflect the burdens of a chronic condition, rather than unique aspects of schizophrenia per se, requires further empirical inquiry.

Second, there also exists considerable within-group heterogeneity in decisional capacity among patients with schizophrenia, with standard deviations on capacity measures commonly twice those reported in NPC groups. One of the identified sources of variability in capacity among patients with schizophrenia was whether the samples were composed of inpatients or community-dwelling and/or clinically stable outpatients, as outpatients were much closer to NPCs in performance on capacity measures. This last point is particularly important, as a great majority of persons with schizophrenia reside in the community.³⁶ Moreover, psychotic inpatients have several characteristics (distinguishing them from outpatients) that might temporarily limit their decisional capacity, such as greater severity of positive and negative symptoms, experiencing a stressful life event (hospitalization), and often receiving higher doses of medications that may adversely impact cognition.³⁷ Given the heterogeneity among persons with schizophrenia, further research identifying the variables that moderate decisional capacity (eg, cognition or place of residence) will help recognize subgroups at greater risk for impaired decisional capacity.

The lack of a specific association between decisional impairment and psychiatric diagnosis, as well as the variability in decisional impairment among persons with schizophrenia, have several potential implications for practice and research. Persons with schizophrenia may be generally at risk for impaired decision-making capacity, yet such impairment is not invariable. Inappropriately limiting autonomy is as much an ethical breach as is failing to obtain a fully informed consent. Similarly, NPCs are not universally intact in decision-making capacity. Therefore, the need for capacity assessment should not be restricted to persons with certain psychiatric diagnoses. However, lengthy assessments of capacity are burdensome and unlikely to be routinely implemented. Future research in this arena should include the development and validation of brief and easily interpretable screening methods for identifying individuals who might require more in-depth capacity assessment, such as the 3-item questionnaire we have described elsewhere.²⁵ In addition, investigations of methods to improve the comprehension of consent forms—for example, through the use of interactive and/or technological formats (such as PowerPoint)—may prevent or compensate for some of the manifestations of decisional capacity impairment. Finally, for people deemed to have impaired capacity, interventions designed to remediate deficits in specific components of capacity would be useful. Initial intervention studies in the enhancement of understanding in persons with schizophrenia show promise,^23,24 and it will be important to determine whether other domains of capacity (eg, reasoning) can also be enhanced in this manner.