Cognitive models of physicians' legal standard and personal judgments of competency in patients with Alzheimer's disease

OBJECTIVES: To investigate measures of patient cognitive abilities as predi ctors of physician judgments of medical treatment consent capacity (compete ncy) in patients with Alzheimer's disease (AD). DESIGN: Predictor models of legal standards (LS) and personal competency ju dgments were developed for each study physician using independent neuropsyc hological test measures and logistic regression analyses. SETTING: A university medical center. PARTICIPANTS: Five physicians with experience assessing the competency of A D patients were recruited to make competency judgments of videotaped vignet tes from 10 older controls and 21 patients with AD (10 with mild and 11 wit h moderate dementia). MEASUREMENTS: The 31 patient and control videotapes of performance on a mea sure of treatment consent capacity (Capacity to Consent to Treatment Instru ment) (CCTI) were rated by the five physicians. The CCTI consists of two cl inical vignettes (A-neoplasm and B-cardiac) that test competency under five LS. Each study physician viewed each vignette videotape individually, made judgments of competent or incompetent under each of the LS, and then made his/her own personal competency judgment. Physicians were blinded to partic ipant diagnosis and neuropsychological test performance. Stepwise logistic regression was conducted to identify cognitive predictors of each physician 's LS and personal competency judgments for Vignette A using the full sampl e (n = 31). Classification logistic regression analysis was used to determi ne how well these cognitive predictor models classified each physician's co mpetency judgments for Vignette A. These classification models were then cr oss-validated using physician's Vignette B judgments. RESULTS: Cognitive predictor models for Vignette A competency judgments dif fered across individual physicians, and were related to difficulty of LS an d to incompetency outcome rates across LS for AD patients. Measures of sema ntic knowledge and receptive language predicted judgments under less diffic ult LS of evidencing a treatment choice (LS1) and making the reasonable tre atment choice (LS2). Measures of semantic knowledge, short-term verbal reca ll, and simple reasoning ability predicted judgments under more difficult a nd clinically relevant LS of appreciating consequences of a treatment choic e (LS3), providing rational reasons for a treatment choice (LS4), and under standing the treatment situation and choices (LS5). Cognitive models for ph ysicians' personal competency judgments were virtually identical to their r espective models for LS5 judgments. For AD patients, shortterm memory predi ctors were associated with high incompetency outcome rates (over 70%), a si mple reasoning measure was associated with moderately high incompetency out come rates (60-70%), and a semantic knowledge measure was associated with l ower incompetency outcome rates (30-60%). Overall, single predictor models were relatively robust, correctly classifying an average of 83% of physicia n judgments for Vignette A and 80% of judgments for Vignette B. CONCLUSIONS: Multiple cognitive functions predicted physicians' LS and pers onal competency judgments. Declines in semantic knowledge, short-term verba l recall, and simple reasoning ability predicted physicians' judgments on t he three most difficult and clinically most relevant LS (LS3-LS5), as well as their personal competency judgments. Our findings suggest that clinical assessment of competency should include evaluation of semantic knowledge, v erbal recall, and simple reasoning abilities.