Application of Causal Inference Methods to Evaluate the Impact of Malaria Control Interventions on Pregnancy Outcomes
- Author(s): Roh, Michelle
- Advisor(s): Gosling, Roly
- et al.
Sub-Saharan Africa is faced with a challenging road ahead if it is to meet its United Nations Sustainable Development Goals of reducing rates of neonatal mortality to 12 per 1,000 live births by 2030. Currently, the region contributes to 90% of the global burden of newborn deaths and progress toward reducing these rates has been slow . Although the causes of newborn death are multifactorial, in malaria-endemic areas, infection with the Plasmodium falciparum parasite during pregnancy is a leading cause of low birthweight (LBW) and stillbirth, which are both risk factors for neonatal mortality [2, 3]. Each year, malaria in pregnancy is estimated to cause nearly one million LBW deliveries, 220,000 stillbirths, and 110,000 neonatal deaths [3-6].
To prevent the adverse consequences of malaria in pregnancy, the World Health Organization (WHO) recommends all pregnant women living in areas of moderate and high malaria transmission receive an effective method of vector control, namely long-lasting insecticidal nets (LLINs), and malaria chemoprevention known as intermittent preventive treatment (IPTp) with sulfadoxine-pyrimethamine (SP) . Though LLINs and IPTp-SP have been shown to be highly effective in trial settings [8, 9], their current efficacy is threatened by the emergence of mosquito resistance to pyrethroids, the insecticide most commonly used in LLINs, and parasite resistance to the SP antimalarial in Eastern and Southern Africa. This has led researchers to evaluate alternative strategies for malaria control, including new types of LLINs that can overcome pyrethroid resistance, more efficacious antimalarials, and non-pyrethroid-based vector control tools. However, little is known of how these alternative strategies will affect pregnant women and their birth outcomes.
The overall goal of this dissertation is to apply advanced causal inference methods to determine the effectiveness of these alternative strategies for preventing adverse pregnancy outcomes among women at-risk for malaria. The dissertation is organized into three chapters, each describing one of these alternative strategies. The first chapter pools data from three randomized controlled trials that compared IPTp efficacy of SP to dihydroartemisinin-piperaquine (DP), a highly efficacious antimalarial. Data from the pooled analysis found that while DP is a more efficacious antimalarial, it did not confer greater benefits on pregnancy outcomes compared to SP. Through applying causal mediation analyses, we found that this was due to the greater, ‘non-malarial’ and possible antimicrobial effects of SP which were counteracting DP’s greater antimalarial effects. Findings from this study suggest future IPTp regimens should consider adding an antimicrobial to the IPTp-DP regimen to achieve a greater impact on improving birth outcomes. The second chapter compares the effectiveness of conventional, pyrethroid-based LLINs to a new type of LLIN additionally treated with piperonyl butoxide (PBO), a chemical shown to restore pyrethroid sensitivity. Using quasi-experimental analyses such as interrupted time series analyses and difference-in-differences models, the study found PBO LLINs conferred a 22% greater reduction in LBW and a 33% greater reduction in stillbirth compared to conventional (non-PBO) LLINs, supporting the latest WHO recommendation for deployment of PBO LLINs. The third chapter evaluates the impact of indoor residual spraying (IRS), an existing but highly underutilized malaria vector control intervention, on preventing adverse birth outcomes. We used a novel, machine learning method to relax some of the strict assumptions of traditional quasi-experimental analyses to find that high-coverage IRS can reduce LBW incidence up to 17%, a finding consistent with conferring full protection against malaria via LLINs, IPTp, or both. Given the scale-up of LLINs and IPTp has been traditionally low across sub-Saharan Africa, our results suggest IRS can play a complementary role in preventing malaria-associated adverse pregnancy outcomes and thus, efforts should be made in expanding its use.
Together these findings provide important policy implications for identifying alternative and potentially more effective malaria interventions for pregnant women, particularly in light of growing pyrethroid and SP antimalarial resistance.