55EXTERNAL VALIDATION OF A SHORT SCREENING TOOL FOR DEMENTIA IN A RESOURCE POOR, LOW LITERACY SETTING

Abstract

Introduction: Cognitive screening tools for dementia have been repeatedly shown to be biased in populations with low levels of education and literacy. The IDEA-IADL study (Collingwood, Global Health Action, 2014; 7) validated a tool in the Hai District of Tanzania which was free from educational bias, incorporating a cognitive screen and an informant interview assessing function. This comprised instrumental activities of daily living (IADLs), developed to be culturally appropriate for the region. Internal validation of the data found that reducing the number of IADL items from 11 to 3 minimally decreased the Area Under the Receiver Operating Curve (AUROC) from 0.939 to 0.923. The study aims to externally validate this shortened tool, with increased efficiency for use in a resource poor environment. Methods: 329 participants were screened using the shortened IDEA-IADL tool in four villages in the Hai District. This was in conjunction with the local enumerator for each village (responsible for demographic surveillance). A stratified sample was subsequently assessed according to 'gold standard' DSM-IV dementia criteria by a doctor, blinded to screening results. AUROC values were calculated to determine the screen performance. Results: AUROC was lower than expected at 0.878. Assessment of the sub-sections of the screen found the IADLs were the primary contributor to this reduced score, varying from 0.687 to 0.939, depending on assessor. On regression modelling, whilst controlling for age, gender and presence of dementia, there was no association between a positive screen and the absence of formal education (Odds Ratio 2.776, p=0.106). Conclusions: The shortened screening tool has been externally validated. Whilst it remains free from educational bias, the research exposes the unreliable nature of more subjective measures of function when carried out by unskilled assessors. Screen performance appears to be assessor dependent and the effect of training and experience in healthcare needs further analysis.