Scaling up prevention of mother-to-child HIV transmission programs in sub-Saharan African countries: a multilevel assessment of site-, program- and country-level determinants of performance
1 Biostatistics and Epidemiology Unit, Hôtel Dieu, Assistance Publique Hôpitaux de Paris, Université Paris Descartes, Sorbonne Paris Cité, Paris, France
2 Philip R. Lee Institute for Health Policy Studies, University of California, San Francisco, USA
3 Axios International, Paris, France
4 Epidemiology and Public Health Department, Faculty of Medicine, Rouen University Hospital, Rouen, France
BMC Public Health 2013, 13:286 doi:10.1186/1471-2458-13-286Published: 1 April 2013
Uptake of prevention of mother-to-child HIV transmission (PMTCT) programs remains challenging in sub-Saharan Africa because of multiple barriers operating at the individual or health facility levels. Less is known regarding the influence of program-level and contextual determinants. In this study, we explored the multilevel factors associated with coverage in single-dose nevirapine PMTCT programs.
We analyzed aggregate routine data collected within the framework of the Viramune® Donation Programme (VDP) from 269 sites in 20 PMTCT programs and 15 sub-Saharan countries from 2002 to 2005. Site performance was measured using a nevirapine coverage ratio (NCR), defined as the reported number of women receiving nevirapine divided by the number of women who should have received nevirapine (observed HIV prevalence x number of women in antenatal care [ANC]). Data on program-level determinants were drawn from the initial application forms, and country-level determinants from the Demographic and Health Surveys (DHS) and the World Bank (World Development Indicators). Multilevel linear mixed models were used to identify independent factors associated with NCR at the site-, program- and country-level.
Of 283,410 pregnant women attending ANC in the included sites, 174,312 women (61.5%) underwent HIV testing after receiving pre-test counselling, of whom 26,700 tested HIV positive (15.3%), and 22,591 were dispensed NVP (84.6%). Site performance was highly heterogeneous between and within programs. Mean NCR by site was 43.8% (interquartile range: 19.1-63.9). Multilevel analysis identified higher HIV prevalence (Beta coefficient: 25.1, 95% confidence interval [CI] 18.7 to 31.6), higher proportion of persons with knowledge of PMTCT (8.3; CI 0.5 to 16.0), higher health expenditure as a proportion of Gross Domestic Product (3.9 per %; CI 2.0 to 5.8) and lower percentage of rural population (-0.7 per %; CI -1.0 to -0.5) as significant country-level predictors of higher NCR at the p<0.05 level. A medium ANC monthly activity (30-100/month) was the only site-level predictor found (-7.6; CI -15.1 to -0.1).
Heterogeneity of nevirapine coverage between sites and programs was high. Multilevel analysis identified several significant contextual determinants, which may warrant additional research to further define important multi-level and potentially modifiable determinants of performance of PMTCT programs.