Problems in Measuring and Interpreting Cognitive Decline
Problems in Measuring and Interpreting Cognitive Decline
复制标题
测量和解释认知衰退的问题
DOI：
10.1111/j.1532-5415.1998.tb01547.x
复制
发表时间：
1998
影响因子：
6.3
通讯作者：
W. Kukull
中科院分区：
文献类型：
--
作者：
W. Kukull
在大约20至30岁时，缓慢下降到大约60岁，并且在该点之后可能随着年龄的增长而下降得更快。对这一现象的早期估计主要是基于规范性的横截面样本。在不同年龄段测试的横截面的代表性往往受到质疑，特别是对于最老年龄组的人。这种智力下降的模式通常被解释为与衰老相关的“正常”下降，而不是由疾病或疾病过程引起的下降。除了全球智力能力的一些普遍下降外，涉及痴呆症定义的认知领域（例如，记忆、执行功能）在过去十年中引起了越来越多的兴趣。通过评估认知功能下降早期识别痴呆过程可能表明早期治疗，并可能延迟残疾的有效治疗方法被发现。I-I，在这个问题上thelournu/，文章由O 'Hara et al.提请注意测量和解释认知下降的困难。评估老年人的认知能力需要进行与功能谱相适应的测试。如果“正常”的人用一种旨在区分坦率的痴呆程度的仪器进行测试，“正常”的受试者可能会得到最高分或接近满分的分数。可变性将丢失或受到限制，并且在上限处的那些得分之间的区分将是不可能的。同样，一些智力测试可能不适合区分处于不同痴呆阶段的个体。精神错乱的受试者可能会在测试楼层得分，并排除额外的描述或歧视。因此，下限和上限效应是工具选择中的重要因素;必须考虑到研究人群中认知功能的分布。某一特定仪器的下限和上限的一个难以捉摸的伙伴可能被称为临床阈值。将受试者称为受损而不是正常的临界点或测试评分的可用性具有明显的临床实用性。在确定阈值后，它们往往受到所研究样本中其他因素的影响，如年龄、教育和种族，从而使解释变得复杂。当没有可用或可接受的测试或电池的临床阈值时，研究人员将注意力集中在连续分数以及它们如何随时间变化。测试分数被认为是真实分数成分和错误成分（或噪音）的组合;在两个或两个以上的场合测量一个人会导致观察到的分数（真实恐怖成分），而不管涉及的工具。在线性情况下，还将估计由恶化或改善引起的点之间的斜率。货车贝儿和31.“讨论评估阿尔茨海默病患者变化分数的可靠性的问题（A”测试中观察到的分数差异，传统上被描述为连续达到峰值）。他们指出，可靠性取决于（1）患者之间真实变化的变异性，（2）患者内部真实变化的剩余变异性，以及（3）观察的数量和间隔？（1）与（2）的比值表征为仪器的“信噪比”，而（3）是研究者确定的研究设计特征。研究人员如何确定观察到的分数差异是可重复的差异？个体被测试的次数或个体被测试的时间段的长度是否会对变化的可靠性产生更大的影响？货车贝儿等人的结果表明，仪器的选择，更好的信噪比，是很重要的;和更长的时间长度的病人进行测试有一个更大的影响比观察的数量的可靠性。
at around ages 20 to 30, declining slowly to about age 60, and possibly declining more rapidly with age after that point. Early estimates of this phenomenon were based primarily on normative, cross-sectional samples. The representativeness of the cross-sections tested at the different ages was often questioned, especially for persons comprising the oldest age groups. This pattern of intellectual decline is generally interpreted as “normal” decline associated with aging rather than decline resulting from a disease or disease process. In addition to some general decline in global intellectual capacity, areas of cognition involved in the definition of dementia (e.g., memory, executive functioning) have generated increasing interest during the last decade. Early recognition of a dementing process through evaluation of cognitive decline could indicate early treatment and, possibly, delay of disability as effective treatments are discovered. I-Iowever, in this issue of thelournu/, the article by O’Hara et al. draws attention to the difficulties in measuring and interpreting cognitive decline.’ Evaluating cognition in older individuals requires tests appropriate to the spectrum of function. If “normal” individuals are tested with an instrument designed to distinguish levels of frank dementia, the “normal” subjects will likely score at the ceiling or near perfect score. Variability will be lost or limited, and discrimination among those scoring at the ceiling will not be possible. Similarly, some tests of intellectual capacity may be inappropriate for distinguishing between individuals who are at different stages of dementia. The demented subjects would likely score at the test floor and preclude additional description or discrimination. Floor and ceiling effects are, therefore, important factors in instrument choice; they must be considered in regard to the distribution of cognitive function in the population under study. An elusive partner of the floor and ceiling for a particular instrument might be called the clinical threshold. Availability of a cutpoint or test score at which subjects are called impaired rather than normal has obvious clinical utility. When thresholds have been defined, they are often influenced by other factors, such as age, education, and ethnicity, in the sample being studied, thus making interpretation complicated. When no clinical threshold for a test or battery is available or accepted, researchers focus their attention on continuous scores and how they change with time. Test scores are considered to be a combination of a true score component and an error component (or noise); measuring an individual on two or more occasions leads to observed scores (true -terror components) regardless of the instrument involved. In the linear situation, slope(s) between points caused by deterioration or improvement would also be estimated. van Belle et 31.’ discuss the problem of assessing the reliability of change scores (difference in observed scores on A” tests, has traditionally been described as reaching a peak successive occasions) in Alzheimer’s disease patients. They note that reliability depends on (1) the variability of the true changes among patients, (2) the residual variability about the true change within a patient and (3) the number and spacing of observations? The ratio of (1) to (2) is characterized as a “signal to noise ratio” for the instrument, whereas ( 3 ) is a study design characteristic established by the investigator. How sure can researchers be that an observed difference in scores is a reproducible difference? Can the reliability of the change be influenced more by how many times the individual is tested or the length of the time period over which the individual is tested? van Belle et al.’s results indicate that instrument choice, better signal to noise ratio, is important; and the longer the length of time over which the patient is tested has a greater impact on reliability than does the number of observations.