One-month stability of scores on the understanding, appreciation,reasoning, and expression of a choice subscales of the MacArthur CompetenceAssessment Tool for Treatment (MacCAT-T) among patients with schizophreniaor schizoaffective disorder (n = 53). To ease visual comparison across theMacCAT-T subscales, each patient's score was transformed to a percentage ofthe maximum possible score on the respective subscale. The height of eachbar represents the mean of these transformed scores; the error bars representthe standard deviations of these transformed scores.
Palmer BW, Dunn LB, Appelbaum PS, Jeste DV. Correlates of Treatment-Related Decision-Making Capacity Among Middle-Agedand Older Patients With Schizophrenia. Arch Gen Psychiatry. 2004;61(3):230-236. doi:10.1001/archpsyc.61.3.230
Antipsychotic medications constitute the backbone of treatment for schizophrenia.
Current guidelines require clinicians to obtain patients' informed consent
for treatment, but few empirical studies of the capacity of patients with
schizophrenia for meaningful consent in this context exist. This issue may
be particularly relevant for middle-aged and older patients, as the cognitive
changes associated with normal aging may have an adverse impact on decision-making
processes. We examined the range, stability, and correlates of treatment-related
decisional capacity in this patient population.
Participants included 59 middle-aged and older patients with schizophrenia
or schizoaffective disorder and 38 normal comparison subjects. Baseline measures
included the MacArthur Competence Assessment Tool for Treatment (MacCAT-T),
psychopathology rating scales, and the Mattis Dementia Rating Scale. Patients
also completed a neuropsychological test battery. The MacCAT-T was readministered
to patients at a 1-month follow-up.
Relative to the comparison subjects, the patients had worse understanding
of disclosed material; however, a wide range of performance was observed among
patients. Variability in MacCAT-T performance was not predicted by demographic
characteristics; there were no significant correlations between psychopathology
ratings and MacCAT-T scores. Cognitive test scores were often significant
correlates of capacity, particularly in terms of understanding and reasoning.
The MacCAT-T scores were stable during the 1-month follow-up.
Overall, middle-aged and older outpatients with schizophrenia had worse
understanding of disclosed information than did normal comparison subjects,
but such group comparisons obscure remarkable heterogeneity among patients.
Differences in capacity appeared more related to cognitive functions than
to severity of psychopathology. Such information about barriers to capacity
may help in developing more effective methods of providing informed consent.
Antipsychotic medications form the backbone of effective treatment ofchronic psychotic disorders such as schizophrenia.1- 3 Currentguidelines require clinicians to obtain patients' informed consent for treatmentwith these medications, but reports have suggested that some psychiatric patientsmay not sufficiently understand treatment-relevant information as commonlydisclosed.4- 8 Suchfindings raise the question of decisional capacity, which is widely consideredto require more than just understanding of disclosed information. It alsoinvolves the capacity to appreciate the significance of the information forone's own condition and situation, to reason with the information, and toexpress a choice about whether to accept or reject the proposed treatment.9
Decisional capacity may be of particular relevance as patients age,because the risk-benefit ratios of treatments tend to become more complex.At the same time, cognitive changes associated with normal aging may makeit more difficult for some patients to understand, appreciate, and/or reasonabout the relative risks and benefits of a particular treatment being proposed.Little is known about the effects of aging on the capacity of patients withschizophrenia to consent to treatment.
Although there has been recent debate and some empirical research regardingthe capacity of patients with schizophrenia to make informed decisions whenconsenting to research participation,10- 14 littleempirical attention has focused on factors associated with better or worsetreatment-related decisional capacity among patients with schizophrenia, witha few notable exceptions.15,16
One of the most comprehensive studies of treatment-related decisionalcapacity was the MacArthur Treatment Competence Study conducted by Appelbaumand Grisso and colleagues.9,15,17 Theseinvestigators examined treatment-related decisional capacity among predominantlyyounger inpatients with schizophrenia or depression compared with medicalinpatients who were hospitalized for coronary heart disease and community-dwellingnormal comparison (NC) subjects. As a group, the patients with schizophreniashowed worse understanding, appreciation, and reasoning relative to the othergroups. Severity of psychopathology was significantly correlated with worseunderstanding of disclosed material.
A second study by Grisso and Appelbaum16,18 wasconducted as part of the validation for a standard instrument for assessingdecisional capacity, the MacArthur Competence Assessment Tool for Treatment(MacCAT-T). In that study, they compared MacCAT-T performance among 40 predominantlyyounger patients hospitalized for schizophrenia with that among 40 NC subjects.Relative to the NC subjects, the patients with schizophrenia had significantlyworse scores on the MacCAT-T understanding and reasoning subscales, but evenamong patients there was a wide range of performance observed on each MacCAT-Tsubscale.
A series of studies by Marson and colleagues19- 21 andDymek et al22 of patients with Alzheimer andParkinson diseases suggest that specific cognitive deficits may differentiallyrelate to specific aspects of decision-making capacity, particularly executivefunctions and verbal memory. Such findings may be relevant to schizophreniaand aging, as these are among the most frequently impaired ability areas inindividuals with schizophrenia and in adults undergoing normal aging. However,only a few studies have specifically examined the relationship of cognitivetest performance to decisional capacity. The only cognitive testing in theMacArthur Treatment Competence Study9,15,17 wasa composite score consisting of the vocabulary, similarities, and digit symbolsubtests from the Wechsler Adult Intelligence Scale–Revised. This compositescore was significantly correlated with understanding of treatment-relevantinformation. In an investigation of capacity to consent to research participationinvolving a sample of predominantly younger patients with schizophrenia, Carpenteret al12 found that understanding and reasoningwere each strongly correlated with overall performance on a brief neurocognitivebattery.
Another important aspect of decisional capacity is the degree to whichit may be stable over time. There is general consensus that consent is notsynonymous with having a patient sign a form, but rather is a process resultingfrom an ongoing dialogue between the provider and patient. Appelbaum et al23 recently examined the stability of capacity to consentto research participation among women with depression, and found stable appreciationand reasoning from baseline to an 8- to 10-week follow-up, with a slight improvementin understanding scores. We have found no published studies of the stabilityof decisional capacity among patients with schizophrenia.
The aims of the present study were to examine the level, range, correlates,and stability of treatment-related decisional capacity among middle-aged andolder patients (age, ≥40 years) with schizophrenia or schizoaffective disorder.We hypothesized that patients would have lower levels of decisional capacitythan NC subjects, and that the patients' level of capacity would be stableduring the 1-month follow-up. However, we also expected substantial variabilitywithin the patient group and that this variability would be strongly associatedwith psychopathology and cognitive deficits, particularly in the cognitivedomains of learning/memory and abstraction/cognitive flexibility.
The following unique features of the present study may be worth highlighting.(1) This study focused on 4 dimensions of treatment-related decisional capacitymeasured with the MacCAT-T, whereas much of the recent literature has focusedon capacity to consent to research, and/or has been primarily focused on theunderstanding dimension of capacity. (2) This study focused on middle-agedand older patients with schizophrenia or schizoaffective disorder and NC subjects,rather than younger adults who were the focus of most earlier studies. (3)The patients in our study underwent evaluation with a comprehensive neuropsychologicaltest battery. (4) We examined the stability of the patients' decisional capacityduring a 1-month interval.
Participants included 59 middle-aged and older community-dwelling outpatientswith schizophrenia (n = 49) or schizoaffective disorder (n = 10). Fourteenof the patients provided limited baseline data for a preliminary report.24 The study was conducted through the Advanced Centerfor Interventions and Services Research (ACISR) at the University of California–SanDiego (UCSD), which focuses on studies of psychosis in older persons. TheACISR recruits patients through a variety of outpatient settings, includingthe UCSD Psychiatry Services, the Veterans Affairs San Diego Healthcare SystemPsychiatry Service, referrals from individual psychiatrists and physicians,and direct recruitment at San Diego–area board-and-care facilities.Most of the patients are residents of privately run, community-based assistedliving facilities (board-and-care homes). These facilities provide patientswith an intermediate level of care (including lodging, meals, and medicationmanagement) between independent living and institutionalized care. The focusof our study was on the correlates and stability of decision-making capacityamong patients with schizophrenia or schizoaffective disorder. However, wealso collected limited data from 38 middle-aged and older NC subjects. TheNC subjects were individuals without major DSM-IV psychiatricdisorders who were recruited from the general San Diego area community.
Because we were interested in decision-making capacity among patientslikely to be independently responsible for providing consent, we did not enrollpatients with designated conservators or those whose physicians had diagnosedthem as having Alzheimer disease or other dementias. Other exclusion criteriaincluded physical or medical problems interfering with the patient's abilityto complete the assessments or a lack of fluency in English. All participantsreviewed and signed an informed consent form approved by the institutionalreview board before participation. Less than 10% of those who reviewed theconsent form declined enrollment.
We assessed 4 dimensions of treatment-related decision-making capacityusing the corresponding subscale scores of the MacArthur Competence AssessmentTool for Treatment (MacCAT-T).16,18 Thesesubscales included (1) understanding (the ability to comprehend disclosedinformation regarding the condition [schizophrenia] and proposed treatment[rated from 0-6]); (2) appreciation (the ability to appreciate the significanceof the disclosed information for one's own condition and situation [ratedfrom 0-4]); (3) reasoning (the ability to manipulate the relevant informationrationally, eg, in comparing the risks and benefits of treatment options andthe likely consequences of one's choices [rated from 0-8]); and (4) expressionof a choice (the ability to arrive at and communicate a choice regarding theproposed treatment [rated from 0-2]). In the present study, the MacCAT-T disclosuresand items referred to treatment of schizophrenia or schizoaffective disorderwith atypical antipsychotic medications. Consistent with the way the MacCAT-Twas developed and used by its authors,16,18 ourfocus was on the subscale scores. A MacCAT-T total score is not calculatedbecause deficits in one dimension might translate into incompetence, evenwhen the other dimensions are intact. Also, as with the NC subjects in theoriginal MacCAT-T validation study,16,18 weomitted the appreciation subscale for NC subjects, as the concept of appreciatingone's own condition and need for treatment does not apply to people who donot have that condition. As in the MacCAT-T validation study, the NC subjectsin the present study were told to imagine that they had the condition beingdescribed in the MacCAT-T disclosure/interview.
The MacCAT-T was administered and scored as described in the MacCAT-Tmanual16 and the training videotape.25 Administration and scoring were conducted by 1 of3 trained research assistants who were kept unaware of the patient's psychopathologyratings and neuropsychological scores. We initially had 2 research assistantscollecting MacCAT-T data, and as reported previously,24 theirinterscorer reliability was established by having each independently score13 of the initial protocols. The interclass correlations were 0.85 for understanding,0.87 for appreciation, and 0.75 for reasoning. When we subsequently hireda third research assistant, her training included sitting in with the otherresearch assistants during MacCAT-T interviews, independently scoring eachresponse to the MacCAT-T item, and discussing the scoring after each interview.This process continued until she achieved general consistency with each ofthe first 2 research assistants.
Severity of psychiatric symptoms was assessed with the Positive andNegative Syndrome Scale,26 which includes positiveand negative subscale scores and incorporates the 18-item version of the BriefPsychiatric Rating Scale27 to measure overallseverity of psychiatric symptoms.
Cognitive deficits were evaluated with the Mattis Dementia Rating Scale(DRS).28 The DRS provides a total score (range,0-144) and the following 5 subscale scores: attention (range, 0-37), initiation/perseveration(range, 0-37), construction (range, 0-6), conceptualization (range, 0-39),and memory (range, 0-25). Lower scores represent worse cognitive functioning.
Cognitive functioning was further evaluated in the patient sample witha comprehensive neuropsychological test battery, including those subtestsfrom the Wechsler Adult Intelligence Scale (WAIS)–Third Edition (WAIS-III)29 that are included in the WAIS-III index scores andselected measures of abstraction/cognitive flexibility and learning/memory.The individual test scores were grouped into the following cognitive abilityareas: (1) Verbal Comprehension (WAIS-III vocabulary, similarities, and informationsubtests), (2) Perceptual Organization (WAIS-III picture completion, blockdesign, and matrix reasoning subtests), (3) Attention/Working Memory (WAIS-IIIarithmetic, digit span, and letter-number sequencing subtests), (4) ProcessingSpeed (WAIS-III digit symbol and symbol search subtests), (5) Abstraction/CognitiveFlexibility (the Wisconsin Card Sorting Test–64 Card Version,30 Booklet Category Test,31 andTrail-Making Test Part B32), and (6) Learning(the Hopkins Verbal Learning Test–Revised [total recall trials 1-3],33 Story Memory Test [learning score],34 BriefVisual-Spatial Memory Test–Revised [total recall trials 1-3],35 and family pictures subtest from the Wechsler MemoryScale–Third Edition [immediate recall score]36).To place scores on a common metric so that we could create ability compositescores, raw test scores were converted to scaled scores using previously establishednorms,29- 36 whereinthe mean scaled score among neurologically and psychiatrically healthy adultsis 10 and the SD is 3. The scaled scores were coded such that higher scoresrepresent better performance. We then calculated the mean scaled score withineach ability area. The focus of our analyses was on these mean ability area–scaledscores, but to describe cognitive functioning in the patient sample we alsocalculated patients' WAIS-III index scores (normative mean of 100 and an SDof 15).
Variables with significantly skewed distributions were transformed usingsquare root, logarithmic, or inverse functions when necessary to meet theassumptions for parametric analyses. We evaluated differences between theNC and patient groups in terms of demographic characteristics, psychopathologyratings, cognitive deficits, and decisional capacity with independent samples t tests (2-tailed) for continuous variables and Pearson χ2 for categorical variables. We used the Pearson correlation to evaluateassociations between the patient characteristics and each of the MacCAT-Tsubscales. A relatively conservative α level of P<.01 (2-tailed) was chosen to define significance to minimize therisk for type I error. Stepwise regression analyses were used to determinethe relative contribution of the various patient characteristics as predictorsof each of the decisional capacity subscales. For each MacCAT-T subscale model,we used only variables that were significant bivariate correlates (at the P<.01 level) of the respective MacCAT-T subscale score.Stability of MacCAT-T scores among the patients during the 1-month follow-upwas evaluated in terms of the test-retest correlations (Pearson r) and paired-samples t tests. As with theother analyses, significance was defined as P<.01(2-tailed).
Baseline demographic characteristics, mean scores on the MacCAT-T subscales,psychopathology ratings, and overall cognitive deficits (DRS total score)among the NC subjects and patients are described in Table 1. Although recruited in the same age range (age, ≥40 years),the mean age of patients was younger than that of the NC subjects; the patientsalso had completed fewer years of education and were more likely to be male(as well as having the expected differences in terms of psychopathology andcognitive deficits). The mean MacCAT-T understanding score among the patientswas significantly worse than that among the NC subjects (t90.5 = 2.91; P = .004). The latterdifference remained significant when age was added as a covariate (F1,94 = 7.420; P = .008). The differences inreasoning and expression of a choice remained nonsignificant, even when agewas added as a covariate. Gender generally has little demonstrable influenceon cognitive tasks,34,37 and wedid not include education as a covariate because truncated education is oftenpresent in samples of patients with schizophrenia and may reflect an aspectof the disorder itself.38,39
The mean DRS total score among patients was 128.6 (of a total possible144), which is near the boundary of traditional cutoff scores for definingimpaired vs unimpaired performance. There was substantial heterogeneity amongpatients, however, as the patients' scores ranged from 97 to 142. Patients'mean WAIS-III scores were also generally in the low-average range of performance,but again with substantial heterogeneity within the sample. For example, themean (age-corrected) WAIS-III Verbal Comprehension index score (a measureof crystallized verbal intelligence, similar to the traditional Verbal IQ)was 89.1 (SD = 17.7), but ranged from 50 (extremely low range) to 138 (verysuperior range).
Among the NC subjects, the only significant correlations with MacCAT-Tsubscale scores were the positive correlations between the DRS total scoreand the MacCAT-T reasoning subscale score (r = 0.43; P = .008) and between the DRS memory score and the MacCAT-Tunderstanding subscale (r = 0.46; P = .004). The correlation between education (in years) and the MacCAT-Tunderstanding subscale was 0.34 (P = .03), but thiswas not significant at the P<.01 level. Cognitiveassessment of the NC subjects was limited to the DRS; they did not undergoevaluation with the comprehensive cognitive test battery.
Among patients, there were no significant correlations between patients'performance on any of the MacCAT-T subscales and their demographic characteristicsor psychopathology ratings (absolute value, r = 0.01to r = 0.26; P>.05). However,as shown in Table 2, there wereseveral significant bivariate correlations between patients' cognitive anddecisional capacity scores. Those that were significant bivariate correlatesat the preselected criterion of P<.01 were includedas potential independent variables in the multiple regression analyses.
We used stepwise multiple regression analyses to evaluate the relativecontribution of each of the significant (P<.01)bivariate correlates of patients' decisional capacity to the overall predictionof understanding, reasoning, and/or expression of a choice. For each model,the criterion for entry was set at P<.05. We didnot use appreciation as a dependent variable in these models, because theonly significant bivariate correlate of appreciation was the mean Attention/WorkingMemory scaled score (r = 0.37; P = .009). We also excluded the DRS total score from the list of potentialindependent variables (IVs), because the variance overlaps with the DRS subscales.
For models involving the MacCAT-T understanding score as the dependentvariable, the IVs included the DRS initiation/perseveration and memory subscales,and the mean scaled scores for the Attention/Working Memory, Abstraction/CognitiveFlexibility, and Learning ability areas. Only the DRS memory subscale scoreentered the model. The model was significant (R2 = 0.27; F1,42 = 15.32; P<.001);the standardized β for DRS memory was .517 (t =3.91; P<.001). None of the other variables accountedfor significant additional variance in understanding.
For models in which the MacCAT-T reasoning score was the dependent variable,the IVs included the DRS conceptualization and memory subscale scores andthe mean scaled scores for the Verbal Comprehension, Perceptual Organization,Attention/Working Memory, Processing Speed, Abstraction/Cognitive Flexibility,and Learning ability areas. Only the DRS conceptualization subscale enteredthe model (R2 = 0.22; F1,42 =11.94; P = .001); the standardized β for DRSconceptualization was .470 (t = 3.46; P = .001). None of the other variables accounted for significant additionalvariance.
For models in which expression of a choice was the dependent variable,the IVs included the DRS attention and conceptualization subscale scores andthe mean scaled score from the Abstraction/Cognitive Flexibility ability area.The first IV to enter the model was the Abstraction/Cognitive Flexibilityscore. At that step, the overall model was significant (R2 = 0.14; F1,47 = 7.79; P = .008); the standardized β for abstraction/flexibility was .377(t = 2.79; P = .008). Theother 2 IVs did not account for significant additional variance.
Among the 53 patients with follow-up MacCAT-T data, the test-retestcorrelations (Pearson r) during the 1-month follow-upwere highly significant (r = 0.72 for understanding, r = 0.73 for appreciation, r =0.61 for reasoning, and r = 0.45 for expression ofa choice; P<.001), suggesting that the level ofeach patient's performance relative to the patient group as a whole remainedfairly stable. There were no significant differences between baseline andfollow-up scores for appreciation (t52 =0.91; P = .37), reasoning (t52 = 0.42; P = .68), or expression of a choice(t52 = 0.11; P = .91). The mean understandingscore changed from 5.1 (SD = 1.0) to 5.2 (SD = 1.1) (t52 = 2.02; P = .048). The stability in MacCAT-Tperformance during the 1-month retest interval is also illustrated in Figure 1. For the sake of visual comparisonin Figure 1, each patient's scoreon each MacCAT-T subscale was transformed to a percentage of the maximum possiblescore on that respective subscale.
To our knowledge, ours is the first study to examine multiple dimensionsof treatment-related decisional capacity (understanding, appreciation, reasoning,and expression of a choice) among middle-aged and older patients with schizophrenia.We found that understanding of treatment-related disclosures among middle-agedand older patients with schizophrenia or schizoaffective disorder, as a group,was lower than that seen among the NC subjects, but such group comparisonshide the considerable heterogeneity among older patients with schizophrenia.The patients' level of capacity was not associated with age, or (contraryto our hypothesis) with severity of psychopathology. However, consistent withour hypotheses, level of decisional capacity was strongly associated withcognitive test performance. In our previous preliminary report,24 wefound no significant correlation between the DRS total and MacCAT-T performance,but the sample size in that report was only 16. The patients' level of decisionalcapacity was stable during the 1-month follow-up.
In stepwise regression analyses, once any single cognitive variableentered the prediction of a MacCAT-T subscale, the other cognitive variablesgenerally did not predict significant additional variance. For example, DRSmemory was the only IV to enter the model of understanding, although in bivariateanalyses, DRS initiation/perseveration and the Attention/Working Memory, Abstraction/CognitiveFlexibility, and Learning ability areas were also significantly correlatedwith understanding. There was a large number of cognitive correlates of reasoning,although in stepwise regression analysis, DRS conceptualization entered themodel first. Abstraction/Cognitive Flexibility was the only IV to enter forthe model of expression of a choice, but the DRS attention and conceptualizationsubscales were also significant bivariate correlates. Overall, while showingthe general importance of cognitive functions to decisional capacity, thesefindings do not suggest the presence of strong differential relationshipsamong specific cognitive abilities to specific aspects of decisional capacity.
Although cognitive test scores were the best predictors of treatment-relateddecisional capacity, even the strongest correlations explained about 25% ofvariance. This highlights the likelihood that decisional capacity is a multiplydetermined construct, reflecting the interaction of a number of patient characteristicsand contextual or environmental factors.
Previous studies have reported significant correlations between psychopathologyand some aspects of decisional capacity.7,12,15,40 Althoughthe present study focused on treatment-related decisional capacity among middle-agedand older outpatients, most previous studies have involved predominantly youngerpatient samples and many have focused on hospitalized patients and/or assessedcapacity in terms of consenting to research participation. The reasons fordifferent findings may in part rest in one of these factors. However, thepattern we have observed is consistent with the overall findings in the functionaloutcome literature in schizophrenia. Severity of psychopathology tends tohave minimal demonstrable impact on everyday functioning, whereas performanceon neuropsychological tests tends to be among the best predictors of everydayfunctioning.41- 46
Although patients' mean levels of decisional capacity impairment wererelatively mild, and we observed remarkable stability in decisional capacityduring the 1-month follow-up, it should be noted that capacity was evaluatedin a specific context (treatment with atypical antipsychotic medications),and that participants were clinically stable outpatients (although many wereliving in board-and-care homes, where at least some of their daily functionalneeds were met by the board-and-care staff). It is likely that worse impairmentsin decisional capacity would be evident in settings with higher proportionsof patients with worse cognitive impairment (eg, chronic institutionalizedsettings). Also, patients in acute psychotic phases might show more difficultymaking meaningful choices about their own treatment. The observed stabilitymay provide clinicians with some assurance that if a patient initially understooda treatment disclosure, and there has been no significant change in mentalstatus, he or she may retain that intact capacity. The context or contentrelevant to the decision may also be important in that patients may have difficultyunderstanding some types of treatments and their risks and benefits, whileshowing better understanding (or appreciation or reasoning) regarding othertreatment decisions.
As clinicians discuss treatment options with their patients, it is helpfulto attend to possible difficulties that patients may have in understanding,appreciating, and reasoning with the information being disclosed (regardlessof the patient's age or diagnosis). We are not suggesting that the MacCAT-Tmust be routinely used to screen capacity in all middle-aged and older patientswith schizophrenia. It is important that patients make meaningful informedchoices when consenting to or refusing treatment, but when defining the borderof what constitutes "capable," it is also important to avoid arbitrarily holdingthose with psychiatric disorders to a higher standard than that which canbe expected from the general population. Also, capacity should not be viewedas an unmodifiable trait. Patients having difficulty with initial understandingof disclosed material can often benefit from educational efforts designedspecifically to teach them the relevant information.12- 14
Our effort to clarify factors associated with worse or better capacityis in part motivated by a desire to facilitate more effective informed consentprocedures. Such facilitation may come by improving the way treatment disclosuresare delivered.12- 14 Anadditional avenue, however, is the possibility of directly ameliorating theunderlying cognitive deficits that have an impact on decisional capacity.Controversy remains regarding whether the second-generation antipsychoticmedications actually improve cognitive functioning in patients with schizophrenia.However, there is recent interest in developing interventions that directlytarget the cognitive deficits of schizophrenia, rather than medications thatmay or may not aid cognition as a by-product of their antipsychotic effects.The relative importance of cognition to decisional capacity and, in turn,the importance of decisional capacity to fully effective and ethical treatment,should provide further impetus for efforts to develop interventions that mayenhance cognitive functioning.
Corresponding author: Barton W. Palmer, PhD, Geriatric PsychiatryResearch Center 116A-1, Veterans Affairs San Diego Healthcare System, 3350La Jolla Village Dr, San Diego, CA 92161 (e-mail: firstname.lastname@example.org).
Submitted for publication May 15, 2003; final revision received August7, 2003; accepted August 15, 2003.
This study was supported by grants MH66248, MH62341, MH64722, and MH43693from the National Institute of Mental Health, Rockville, Md, and by the NationalAlliance for Research on Schizophrenia and Depression, Great Neck, NY.
A preliminary version of this study was presented at the annual meetingof the American Association for Geriatric Psychiatry; March 31, 2003; Honolulu,Hawaii.
We thank Shahrokh Golshan, PhD, for his statistical guidance for thisstudy, and our research staff for their assistance in data collection.