Background An index of wealth is widely used in national surveys to create economic profiles. We constructed such an index within a population-based cohort study of 1,788 preschool children and their primary caregivers living in five isiZulu tribal areas of KwaZulu-Natal, South Africa, an extremely poor area which remains at the epi-center of the South African HIV/AIDS epidemic.

Methods Study Design: Information on household assets, employment, household structure was obtained by door-to-door survey alongside a screen for child disability in developing countries. All children were invited to a structured medical and psychological assessment for disability and HIV status of child and primary caregiver. Wealth Index: A household asset index developed within our study population used many items identical to those in WHO DHS surveys, employing a principal components approach. Asset indicators were grouped into 3 categories: land ownership, ownership of consumer goods, and characteristics of household dwelling (building material, water sources, toilet facility, energy source, etc.). Factor analysis was performed: variables with zero variances and all variables with prevalence less than 2% were removed. The first component explained 16% of the variance and a KMO of 0.532. We ranked the factor scores on the first component in ascending order, standardized to a range of 0 to 4, and grouped into tertiles (1=lowest third, 2=middle third, 3=top third).

Results The wealth index gave information about study participation: Of 1788 children screened, children from households in the poorest third were significantly more likely to be brought to the assessment: 91% compared to 83 and 89 % in the middle and top third respectively. Significant differences in the wealth index were found between the five areas in the sample. With regard to caregiver responses about child disability, those in the poorest third were most likely to report that their child had a disability (43% compared to 46 and 49% - significant on test for trend). Though doctors found a lower rate of disability than reported by parents, there was no difference in disability or false positives rate by wealth index tertile. HIV seropositivity of caregiver varied by tertile (30.3%, 28.5% and 21.8 % in poorest, middle and least poor) but not that of children.

Conclusion Even within extremely deprived areas in a low or middle income country, a wealth index can assess comparative risk among groups suffering more or less material disadvantage and also provide important information in assessing possibility of selection bias in the findings.