Survey validity prior to HIV voluntary counseling and testing (VCT): significant pre-intervention changes

Citation: 
E. Reddy , B. Njau, M. Shorter, N. Thielman, D. Itemba, B. Lema, J. Ostermann (2008) (J Acquir Immune Defic Syndr 2008;49:585–591)
Publication year: 
2008

Background: Studies of voluntary counseling and testing (VCT) have revealed disparate results regarding its effectiveness as an education and prevention tool. Methods: As part of a study of rural mobile VCT (MVCT) and HIV knowledge, attitudes and risks in four villages in Northern Tanzania, 640 randomly selected community respondents were surveyed in their homes prior to free village-based MVCT. 168 respondents were re-surveyed at home prior to MVCT (same venue); 165 were re-surveyed during MVCT pre-test counseling (changed venue). Agreement between responses was analyzed for 25 common survey questions in 4 domains using kappa coefficients for nominal responses and Wilcoxon Sign-Rank tests for mean domain scores. Results: Mean (median) time between surveys was 22 (17) days. Across 7 internalized stigma questions, k ranged from 0.33-0.56 for same venue and 0.25-0.46 for changed venue. Mean internalized stigma scores (range 1-16) were 6.4 (first survey) vs. 5.5 (second survey) in the same venue and 6.5 vs. 5.5 for changed venue (p < 0.01). Five witnessed stigma questions yielded k of 0.12-0.46 for same venue and 0.08-0.27 for changed venue; aggregate scores (range 1-5) were 2.1 vs. 2.2 (p=0.52) and 2.2 vs. 1.9 (p=0.02) for same and changed venue, respectively. HIV knowledge (k 0.12-0.46 same venue, k 0.04-0.37 changed venue) aggregate scores (range 1-9) changed from 5.2 to 6.0 (p< 0.01) in same venue and from 5.2 to 6.3 (p< 0.01) in changed venue. Reports of risk behaviors (k 0.27-0.66 same venue, 0.22-0.66 changed venue) did not change on aggregate. Conclusions: Resurveying a community cohort prior to MVCT revealed variable reliability of questions in all domains. On aggregate respondents reported decreased stigma and increased knowledge; reports of risk behaviors did not change. Venue change resulted in greater differentials. Data analysis without survey validation could lead to false interpretation of the impact of behavioral interventions.