The rights of a patient to accept or refuse treatment through the practice of informed consent represents a cornerstone within modern medical ethics (Faden et al, 1986). Meaningful consent may only be given if the individual is deemed competent to understand the information presented before them, analyze this information, and use it in the formation of a rational decision. This competence is defined as having four key decisional abilities. This first is an understanding of the treatment information, the second is appreciation in relation of this information to ones own situation, the third is evaluation of the potential risks, and the final is expression in communication of the patient's decision (Moye et al, 2006). Should patients lack this ability surrogate decision makers must then be sought.
The National Bioethics Advisory Commission defines Alzheimer's disease as a mental disorder that may affect one or more of these key decision making abilities, and thus threatens competence to consent to treatment. It further defines patients with mild to moderate AD as those who experience memory loss and personality changes, but have not progressed to the point of severe cognitive decline, with loss of ability to respond to the environment (Rockville, 1998). Indeed in a study conducted by Karlawish et al (2005), assessing severity of impairment of decision making abilities, only 19 of the 48 patients suffering from mild to moderate Alzheimer's were shown to be competent. Further studies by Moye et al, (2006) who assessed decision-making capacity among older adults with dementia over 9 months, support this conclusion. Indeed results within the dementia strand of their trial identified that 9.4% had impaired capacity at diagnosis, and a further 26.4%, were diagnosed with impairment at the 9-month stage. An indication not only of the fact that patients with AD have an increased chance of impairment, but further that this decline is continual, and its likelihood increases with time.
This ability of physicians to judge capacity to consent is therefore essential; both to preserve the autonomy of the patient, and further protect the physician, given that treatment lacking informed consent may legally be considered equivalent to battery (Dunn et al, 2006). Despite this recent evidence has been toward the contrary, identifying a lack of consistency in this judgment. For instance in a study conducted by Marson et al (1997), the competency judgment of physicians showed high agreement for controls at 98%, but much lower agreement for those patients with Alzheimer's disease, with agreement being only 56%. We therefore aim in this review to assess the content and psychometric properties, as well as more generally strengths and weakness of each instrument available, in the hope of providing physicians with an efficient framework by which judge competence in Alzheimer's disease.
Results -A review of instruments:
Given the time, skills and resources required for clinicians to produce a full formal capacity evaluation, before this, the first and most general tool one must consider, is the Mini Mental State Exam (MMSE). Folstein et al (1973) developed the procedure as a test composed of two sections, taking only 5-10 minutes to complete, and scored out of 30. They identified that the lower the score achieved the greater the decline in ones cognitive function. Thus patients with dementia typically scored in the region of 9.3, thereby separating these individuals from those suffering more general cognitive impairment - 19.1 and depression, 25.1. Research also identified that test is dynamic, in that the score produced is directly related to ones cognitive state, thus suiting patients with Alzheimer's disease, given their cognitive function declines over time.
The research question was presented in a clear and concise manner, identifying the need of a practical method of grading cognitive state, with the protocol methodology and descriptive statistics, assessing this question directly. Further it was conducted in two segments, and consisted of a large sample of 206 subjects. The first segment being a case control study, differentiating each type of test subjects and separating them from controls, whilst the second improving validity through standardization of the test with 137 consecutively examined patients, producing a mean dementia score of 12.2. Validity was then established by correlating the MMS scores in sample B with the Wechsler Adult Intelligence Scale, which was at the time, the current 'gold standard.' This comparison yielded a p value of less than 0.001, indicating the correlation between the two tests did not occur by chance. Similarly Reliability was obtained given that the test was practiced twice in 24 hours by the same tester and the correlation of results produced, showed a Pearson coefficient of 0.887, identifying a strong relationship between the two.
As a consequence the study has conclusively shown that the MMSE is an efficient and practical tool for assessing cognitive impairment. Therefore given Alzheimer's disease is characterized by this dynamic cognitive decline, and further given the test would suit elderly Alzheimer's patients with shorter spans of attention, theoretically it provides an ideal assessment tool by which to judge competence.
However, a key flaw within the MMSE is that of competence, and the fact that this exam, though successful in identifying cognitive decline, relies on practitioners forming the basic assumption that the score produced by this exam, may also be directly applicable to competence. Indeed given this research article was published in 1973, though at the time intuitively this may be assumed to be correct, more recent literature has shown this not be the case, as is evident in studies conducted by Kim et Caine (2002). These studies examined the MMSE as screening tool for the evaluation of competence in patients with AD. The utility was determined by comparison with the MacArthur Competence Assessment Tool, Clinical Research version (MacCAT-CR).
The study used statistical analysis to compare the results of the two, which identified that only an MMSE of 26 or greater produces sensitivity high enough - 90% + to clearly identify patients who are capable of providing informed consent. However given most patients with mild to moderate AD score in the range of 21-25, at this range the MMSE is at best a modest discriminator of decisional capacity, producing sensitivities between 50-70%, and as noted by the authors, this 'power may only reflect the instruments insensitivity in detecting executive dysfunction.'
Further given the research question posed by this study relates directly to this literature review, and was conducted using outpatient participants with a probable diagnosis of Alzheimer's and an MMSE of 18 or higher, the results of this study are directly applicable to our question of competence. These results can be assumed valid, given their comparison to the MacCAT-CR; a validated capacity instrument, and the current gold standard. Whilst the use of a randomized, double blinded, placebo controlled trial has prevented bias and the receiver operating characteristic curve, plotting sensitivity against 1-specificty allows accurate evaluation of the MMSE .
The only limitations one must consider of this study are the small sample size used - 37 people, and thus the need for inferential statistics in relating the results to the wider population, potential issues of reliability given that each test was carried out only once, and finally the fact that our review concerns capacity to consent to treatment, whereas the research of Kim et Caine (2002) is directed toward consent to research. Despite this the evidence is still applicable, particularly given its consistency, and the fact that the MacCAT-CR measures research competence along the same guidelines used in competence to consent to treatment (understanding, appreciation, evaluation and communication). As a result one may therefore conclude that although the MMSE may be useful for predictions in cognitive decline, and further may provide absolute values at which we can be sure an individual is competent, for those with mild to moderate AD, its use is very much limited.
Given the limitations identified above, the federal practice guidelines for psychologists recommend the use of specific instruments, which are able to assess capacity. These instruments, unlike the MMSE, are able to consider more specific context dependent abilities. Indeed in a review conducted by Dunn et al (2006), assessing clinical capacity for research or treatment, 23 existing instruments for assessing capacity were analyzed, and results specifically identified the use of the Macarthur Competence Assessment Tool (MacCAT-T), the Hopemont Capacity Assessment Interview (HCAI), and Competence to Consent to Treatment Interview (CCTI), as tools for the assessment of competence in Alzheimer's disease.
The research was produced in a systematic manner, considering a wide range of literature published between 1980-2004, concerning the ability of current instruments to assess capacity. Articles were selected with specific inclusion and exclusion criteria, and further for each instrument factors such as the domain assessed (whether each instruments construct was consistent with the current theory of competence) and formalities of admission, were considered. Interrater reliability was examined ensuring results were not observer defendant and test-retest reliability ensured the scores were consistent over brief follow-up.
Of the 23 instruments considered, 15 focused on consent to treatment, nine of these assessed all four decisional abilities. Of the remaining 9, results concluded the use of the MacCAT-T, CCTI and HCAI in the treatment of Alzheimer's disease, given that in addition to considering all four decisional abilities, they provided detailed manuals on their use, and further were reliable, with an interrater reliability score of greater than or equal to 0.80. The only limitations one must account for in this review are the fact that the identification of instruments was based on the current view of decisional capacity, which is open to interpretation, and thus threatens the validity of results, and further that the results are not specifically applicable to patients with AD.
However despite the strength of this review in advocating the use of the HCAI, CCIT and MacCAT-T, research conducted by Guerra et al (2007) has identified that though individually these tools may be useful in assessing competence, their agreement with one another is limited. This therefore leads one to question, which if any is correct. The paper presented by Guerra et al (2007) assesses the agreement between the three instruments for rating treatment decisional capacity using a sample consisting of 38 men and 41 women with mild to moderate AD, its participants and their findings are thus directly relevant to this review. The participants were selected using telephone screening, which, though not ideal is effective, having used the diagnostic and statistical manual of mental disorders, 4th edition, as a guideline. They were then compared to a control group, using the MacCAT-T, CCTI and HCAI. The results were compared using statistical analysis, measuring both agreements between each instrument and the 4 key decisional abilities individually; this ensured the results were valid.
Principal Component Analysis was then used to accurately assess each decisional ability under the three instruments, with results supporting the findings of Dunn et al (2006). All three instruments measured the 4 abilities, with understanding showing the largest variance of 79.9%, and expression of choice with the narrowest, producing a variance of 44.6%. However when a Kappa coefficient was then used to measure agreement between the three, the value produced was 0.451. This value indicates only fair agreement for overall decisional capacity, whilst the kappa for communication, appreciation and evaluation was 0.158, -0.039 and 0.047 respectively, providing an indication of poor agreement. This therefore leads one to question the validity of each instrument.
This paper was however limited given that the kappa statistic used to measure agreement is designed in the use of categorical variables, and decisional capacity is continuous, thus resulting in a lower than realistic agreement. Further validity is questionable given each instrument has a different definition of decisional capacity, thereby limiting their comparison. Indeed agreement is high with respect to understanding given a kappa of 0.618, whilst similarly there is strong agreement between the MacCAT-T and HCAI, with a kappa 0.802 - an indication this difference in agreement may lie only with the CCTI. Finally, the paper more generally may be questioned on its reliability, given the sample size was small -79, inferential statistics relating to the general population are not provided, and further the procedure was not repeated. Despite these limitations, the results still however highlight the lack of overall consensus amongst these tools, indicating that though individually their use is justified, potential contradictions may occur when used together.
Conclusion:
In the process of this review we have clearly identified the need for efficient and effective tools in the judgment of competence in patients with mild to moderate Alzheimer's disease. Current literature has identified that MMSE, though useful in the more general assessment of cognitive decline, lacks the ability to consider specific contextual dimensions required in ones capacity to consent to treatment (Kim et Caine, 2002). We therefore considered more specific, exhaustive tools for the judgment of capacity. Dunn et al (2006) identified 23 such tools with the use of the CCTI, HCAI and Mac advocated, given their consideration of the four key decisional abilities, further confirmed by their PCA scores, and the extensive material available in their use. However Guerra et al (2007) identified that these same tools were shown to have only fair agreement beyond chance, producing a kappa statistic of 0.451 for overall decisional capacity, with low kappa values for communication, appreciation and evaluation.
This therefore leaves one in a position of difficulty over which tool they must choose. Dunn et al (2006) argued the potential singular use of the MacCAT-T tool, given its validity in a number of populations, as well as the extensive material available in its use. However, even if one were to solely implement this tool, substantial limitations are present. This includes the requirement for significant training, the lack of a specific cutoff by which to distinguish competency and further refinement in the use of this instrument to assess appreciation, evaluation and communication. Other alternatives may include the combined use of the HCAI and MacCAT-T or the combination of all three tools solely to assess the ability of understanding; both these methods produced strong kappa statistics. Further given the kappa agreement of all three tools is similar to that of pairwise comparison with expert physicians. Though agreement is low, the combined use of the three may prove useful for those younger lesser-experienced clinicians.
However, overall no 'gold standard' currently exists. Given the development of modern medicine with a multitude of treatment options available, with substantial variance in their risk reward ratio; whilst further considering changes in population demographic, with an increase in elderly suffering from Alzheimer's disease, the need for effective, valid and reliable tools in ones judgment of competence is undeniable. Further research must therefore be conducted in order that one-day, such a gold standard, for the assessment of competence in mild to moderate AD, may exist.
