Abstract

Reliable estimates for district-level adolescent health indicators are required to inform policy and program planning. The decentralised health systems in Malawi require accurate statistics at district-level, where programs are implemented, to achieve effective program implementation and monitoring. Empirical evidence supporting policies on adolescent health indicators at disaggregated levels relies on population surveys, but smaller sample sizes of adolescents from the survey result in district-level estimates with higher sampling variability. Small area estimation (SAE) are commonly used statistical methods that estimate a domain-specific parameter of interest from a survey for disaggregated units with small or no samples in which survey estimates are unreliable. The univariate Fay-Herriot SAE methods have improved small area estimates by linking the target indicator variable with auxiliary information. However, adolescent health indicators, like most public health measurements, are often correlated, and modeling them jointly improves estimation through borrowing strength between outcomes. Spatial dependence may also exist in adolescent health indicators, as neighboring districts often share similar characteristics and access to services. Incorporating both multivariate and spatial relationships within the SAE methodology offers an added advantage in improving estimation for small areas through multivariate spatially dependent error structures. This study develops a multivariate spatial Fay-Herriot model to jointly model three district-level adolescent sexual and reproductive health indicators on adolescent contraceptive use, condom use, and family planning demand in Malawi. Direct survey estimates from the 2024 Malawi Demographic and Health Survey will be linked with auxiliary information from the 2018 Malawi Population and Housing Census. Correlated district random effects will account for dependence among indicators and enable borrowing of strength across outcomes, and spatial dependency will allow for correlated random district effects between adjacent areas through a simultaneous autoregressive structure process. By borrowing strength through correlated indicators while accounting for spatial dependency, the study seeks to produce reliable district-level adolescent health statistics that support policy priorities and provide
evidence for district-level planning of adolescent sexual and reproductive health services in Malawi.