Dynamics of Neuropsychological Testing

Dynamics of Neuropsychological Testing Michael H. Coen 1,2 , Timothy S. Chang 3 , Bruce Hermann 4 , Asenath La Rue 5 , Mark Sager 5 Department of Biostatistics and Medical Informatics 1 Department of Computer Sciences 2 Institute for Clinical and Translational Research 3 Department of Neurology 4 Wisconsin Alzheimer’s Institute 5 University of Wisconsin, Madison, WI 53706 mhcoen@biostat.wisc.edu, {tschang3, larue, masager}@wisc.edu, hermann@neurology.wisc.edu enormously from trial to trial. This may occur even in simple tests such as the Rey Auditory-Verbal Learning Test (AVLT), which is repeated only a few times. In this paper, we provide evidence this is not simply due to naturally occurring variance. Rather, individual trials can be far more informative than aggregate summary scores. We believe that during tests involving a relatively small number of repetitions, subjects are simultaneously acclimatizing and responding to the testing procedure itself, which may conflate interpretations of their responses when viewed via aggregate summary measures. In other words, they are still ―learning‖ the test while they are ―taking‖ the test; we believe the shift between these two processes accounts for much variance across trials. We have previously demonstrated (Coen et al., 2009) that even in non-memory based experiments, such as the Conceptual Set Shifting Task (Milner, 1964) conducted on human and macaque subjects, performance varies substantially but predictably over the course of the trials. Specifically, subjects’ performance on the first few rounds is both slow and inaccurate. However, by the third round, it Abstract How should we analyze repeated trials in neuropsychological testing? It has long been known that experimental subjects display distinct stages of acclimatization and subsequent saturation during cognitive testing (Thurstone, 1927). For example, in list learning tests examining memory, it has been demonstrated that repeated exposure to a fixed enumeration of items can improve recall. However, we think it is equally important to examine acclimatization of the subjects to the test taking procedure itself. In other words, subjects must grow comfortable with the paradigm of the test before we can assume the results correspond with our interpretations of them. In this paper, we examine results of the Rey Auditory- Verbal Learning Test administered to the largest Alzheimer’s disease family history cohort. We demonstrate the most informative signal in a neuropsychological test may contradict a priori assumptions about the test’s interpretation. Keywords: Neuropsychological testing; statistical analysis; Alzheimer’s disease; memory Introduction Psychological tests often employ repeated trials of similar or identical tasks. Sometimes, these repetitions are intended to allow subjects to acclimatize to the stimulus and/or decision making paradigms, e.g., as in forced choice experiments (Mitchell & Jolley, 2009). Indeed, in psychoacoustic experiments, subjects may be unable to even distinguish phenomena of interest without substantial prior exposure and early practice rounds are commonly discarded as uninformative. In neuropsychological tests, it is commonplace to conduct multiple trials of a test, from which summary scores may be derived (Lezak et. al, 2004). Multiple trials can also reveal perseverative errors, which are characteristic of a number of cognitive pathologies. However, in tests focused on evaluating memory, it has been demonstrated that enhanced performance may occur after repeated examinations (Benedict & Zgaljardic, 1998). Thus, subjects are often tested on a smaller number of rounds than might otherwise be desired for acclimatization. Although examining summary scores averaged over trials is commonplace, the individual trial scores can vary Figure 1 – Viewing the Conceptual Set Shifting Task (CSST) by individual trial. In the CSST, each trial consists of a lengthy procedure of trying to guess a hidden concept correctly 10 times in a row, after which a new secret concept is selected for the next trial. By examining both the times taken per decision and the subjects’ error rates, it appears clear that the results become meaningful according to the test’s desiderata by the third trial, whereas the first two trials reflect acclimatization. We see this in the dramatic decrease in the average time taken between decisions and the precipitous drop in error rate. This illustrates both that aggregate summary scores combining all five trials are conflating (at least) two different phenomena, and it additionally provides a signal that the subject’s results after trial 3 more meaningfully reflect performance. (Figure adapted from Coen et al. (2009)).