Estimating Numbers of Orphans and Vulnerable Children – A Test of Regression Modeling

Author(s): Brodish P, Charyeva Z, Foreit K

Year: 2017

Abstract:

An earlier study reported a tight linear fit between national adult HIV prevalence and the percentage of children living in a household with at least one HIV-positive adult. MEASURE Evaluation extended this analysis to all existing DHS data sets with HIV testing, to determine the feasibility of using regression modeling to estimate the size of two priority groups: (1) children living with at least one adult who is HIV-positive, and (2) orphans and coresident children living with at least one adult who is HIV-positive.

At the national level, we found reasonably tight linear relationships between HIV prevalence among adults and the proportion of children living with at least one HIV-positive adult and between adult HIV prevalence and the proportion of orphans and coresident children living with at least one HIV-positive adult. However, at the subnational level, owing to small sample sizes, we found greater variation at the same level of HIV prevalence in the proportion of children living with an HIV-positive adult. Although the slopes and intercepts of the national and subnational regression lines are almost the same, the confidence intervals for the subnational level estimates are very wide and many data points fall outside the prediction.

PEPFAR has given priority for OVC programs to 21 countries. Of these, 17 have a recent biomarker survey: Burundi, Cameroon, Cote d'Ivoire, Democratic Republic of Congo, Ethiopia, Haiti, Kenya, Lesotho, Malawi, Mozambique, Namibia, Rwanda, Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe. For these countries, we recommend using the results of our secondary DHS analyses for programmatic estimates. Our Excel workbook includes point estimates and 95-percent confidence intervals at both the national and subnational levels of the proportions of children who live in a household with an HIV-positive adult and the proportion of children who are orphans or coresident children living in a household with an HIV-positive adult, and extrapolates the observed proportions to numbers of children.

At present, four countries in the PEPFAR OVC portfolio do not have a recent DHS survey with HIV testing: Botswana, Nigeria, South Africa, and South Sudan. For these countries, the linear regression is a reasonable way to estimate at the national level the proportion of children who live in a household with an HIV-positive adult and the proportion of children who are orphans or coresident children living in a household with an HIV-positive adult. We have created an Excel workbook to predict these two indicators at the national level and to extrapolate the predicted proportions to numbers of children.

Access the Population Size Estimation Tool for Programs Serving Orphans and Vulnerable Children tool at www.measureevaluation.org/resources/publications/tl-18-07/.

Access the DHS estimates for the size of OVC groups at the national and sub-national levels tool at www.measureevaluation.org/resources/publications/tl-18-08/.