An important implication of good fit to a Rasch model is the potential for developing adaptive tests. Subjects who pass a given item would not need to be tested on those items shown to measure lesser degrees of cognitive ability. Depending

on the accuracy required and the ability of the subject, only a few items might need to be administered to measure cognitive ability. This item-bank approach reduces test burden without loss of information, even across a wider range of cognitive deficits. It also allows clinicians to continuously monitor the impact of therapies without the artificial interruption in scores introduced when having to switch from a ‘hard’ test to an ‘easy’ test if cognitive KU-60019 learn more impairment worsens. The adaptive approach to cognitive measurement was recently validated for geriatric mild cognitive impairment in a study that combined test items from the MoCA and the MMSE (S. Konsztowicz et al., unpublished observations). The data we present here provide a basis for an adaptive approach to measuring cognition, but further

work will be needed to implement and fully validate such a method. Some limitations to this study must be considered. Firstly, the use of computerized measures adds inconvenience when compared with a brief pencil-and-paper test, although web-based testing software could be developed to minimize that inconvenience. A computerized approach has the additional advantage of greatly simplifying the

process of administering a test in an adaptive format, automatically selecting the next items to be administered based on the pattern of previous responses and stopping once a criterion is reached for confidence in the accuracy of the resulting score. This approach has been used successfully to evaluate cognition in patients with cerebrovascular disease [41] and in a rehabilitation clinic population [42]. Secondly, the particular computer tests we used are drawn from the experimental cognitive neuroscience literature, Immune system and so have not undergone the extensive normative testing of more conventional measures. However, they are in the public domain and thus readily available for evaluation and development by others. At the very least, the present work illustrates a methodological path that could be profitably pursued as we seek to improve on current tools for the assessment of cognitive ability in people with HIV infection. This work was supported by operating grants from CIHR and CECR to LKF, by salary support from the MUHC Research Institute (LK) and from CIHR and FRSQ (LKF), by a Canada Graduate Studentship (AT), and by a McGill Faculty of Medicine Research Bursary (EW). We thank the patients and family members who volunteered for this study, and the clinicians who provided referrals.