Impact of family structure on HIV infection in 20 countries in Sub Sahara Africa: a multivariate analysis approach

Jeroen van Ginneken, Netherlands Interdisciplinary Demographic Institute (NIDI)
George Groenewold, Netherlands Interdisciplinary Demographic Institute (NIDI)

This proposal focuses on the role of family structure as one group of causes of the HIV/AIDS epidemic in Sub-Sahara Africa. This will be done by making comparisons among countries with varying HIV infection rates. Data are used from Demographic and Health Surveys conducted in 20 countries in Africa among men and women 15-49(54) years old between 2002 and 2008. Both bivariate and multivariate techniques of analysis will be used. Attention is also paid to the interrelationships between family structure and other demographic, social and economic factors. This is facilitated by using a model in which the various factors are classified into three groups: proximate, intermediate and underlying. In our project we will focus initially on the impact of variables at the proximate and intermediate levels. There is evidence of substantial differences, at the level of proximate factors, in high-risk sexual and other behaviour between countries with high and low HIV prevalence. This difference is - at the intermediate level - partly due to the characteristics and functioning of the family system (married couples and partners living together) and informal sexual unions (premarital and extramarital sex). Settled family situations are less prevalent in countries with high than in countries with low HIV prevalence. Informal unions are, on the other hand, more common in countries with high than in countries with low HIV prevalence. Conclusions are drawn at the end concerning the importance of the variables to be included in the multivariate analysis with emphasis on the role of variables dealing with family structure. We also mention the implications of the provisional findings for development of HIV/AIDS strategies and policies in countries with high and low HIV prevalence.