Generating fair, reliable, and accurate neuropsychological test norms for people with HIV in a low- or middle-income country.

Effective neuropsychological assessment of people with HIV (PWH) in low- and middle-income countries (LMICs) is hampered by the unavailability of adequate test norms. We aimed to: (1) develop demographically-corrected (regression-based) South African (SA) normative data for an HIV appropriate neuropsychological test battery for Xhosa home-language speakers; (2) compare the utility of those norms to that of (i) internal standardization norms and (ii) US test publisher norms; and (3) determine the criterion validity of the newly-developed norms. 114 controls and 102 demographically comparable Xhosa home-language people living with HIV completed a well-establised, standard HIV neuropsychological test battery assessing seven cognitive domains. Using a common performance metric (z-score), we compared control and PWH test performance and examined the extent to which the three different normative datasets embedded demographic effects e.g., education. Using internal standardization norms, analyses detected medium-sized correlations of overall test performance with age and education. Correlations were fully corrected for by the newly-developed demographically-corrected norms. Using demographically-corrected norms, PWH performed significantly more poorly than controls in five cognitive domains, whereas using internal standardization norms and test-publisher norms, PWH performed significantly more poorly than controls in one and two domains, respectively. Demographically-corrected norms estimated 43.1% of PWH were cognitively impaired; these estimates were 22.5% using test-publisher norms and 19.6% using internal standardization norms. Demographically-corrected SA norms were more sensitive to cognitive impairment in PWH than the other sets of norms. Expansion of this regression-based method to create population-appropriate norms will benefit research and clinical practice in LMICs.