The digit spans forwards and backwards were left as raw spans. The full data set was subsequently subjected to the same Rasch analyses as described above. Demographic and clinical characteristics of the patient sample are shown in Table 1. The population we studied was similar to that in recent work on the mild cognitive impairment spectrum in HIV-infected patients from centres in North America. Patients were predominantly men (92%), were relatively well

educated (with 44% having completed at least some university-level education), had a long history of HIV infection (mean disease duration 13.9 ± 6.7 years), and were treated with HAART. The majority had undetectable viral loads at the time of testing. Two patients were coinfected with hepatitis C virus. About a Raf inhibitor third of the sample reported current drug use, most commonly marijuana. About 10% reported consuming more than 7 units of alcohol per week. Over half the sample (55%) was taking one or more psychoactive medications. These were most commonly antidepressants Selleck Navitoclax or sedatives/hypnotics. Patients were tested in either English or French. For most, one of these was their native language, although for a minority (12%) neither was their

mother tongue. This was a clinic-based convenience sample. Cognitive impairment was not an inclusion criterion, but we suspect that both referring clinicians and patients were more likely to consider participation if there were pre-existing concerns about cognition. This sample is thus likely to be enriched with patients representative of those who are presenting with mild cognitive complaints.

Consistent with this supposition, subjective cognitive complaints were present in 47% of the sample. Depressive symptoms ranging from mild to severe were also common, being present in 56% of Urease the sample. Ten patients (13%) were classified as severely depressed (BDI >28). It is worth noting that other recruitment approaches, such as selecting patients with poor viral control, might yield different sample characteristics, but would be unlikely to substantially affect the goal of this study, which was to develop a method of measuring cognitive ability, rather than to categorize patients within the existing diagnostic framework for cognitive impairment. Response categories for serial 7s were rescored when some scores (e.g. 0/5) occurred with insufficient frequency to produce reliable estimates of their thresholds. Four other naming and orientation items were removed because they failed to contribute information to the measurement of cognition (correct in 100% of patients). The resulting set of 24 items showed good fit to a Rasch model of cognitive ability, including absence of an item–trait interaction (χ2=48.92; P=0.44). The items ranged in difficulty, encompassing over 95% of the construct of cognitive ability, from −2.313 logits (easiest) for the clock contour to +2.061 logits for letter-F fluency (Fig. 1).