Using Group-Based Trajectory and Prediction Modeling to Understand Under-Five Mortality

The graph shows the under-five mortality of LMICs from 2000 to 2019. The four darker lines reflect the four groups identified.

An estimated 5.2 million children died in 2019 due to causes that were often preventable and treatable. Though this is a significant decrease from the 12.6 million deaths in 1990, there is still substantial progress to be made in protecting children from preventable deaths (WHO, 2020). Identifying the factors that decrease the risk of under-five mortality can help inform decision making to improve child health.

An ongoing study by Data for Impact (D4I), funded by the United States Agency for International Development (USAID), examined the factors that correlate to lower and higher mortality rates among children under age five in 78 low- and middle-income countries (LMICs), including more than 40 countries in Africa, between 2000 to 2019. The study utilized group-based trajectory modeling, a statistical technique that identifies groups of individuals (in this case, countries) that have similar trajectories for a particular outcome (for this study, under-five mortality rates) to identify four mortality groups and show how the groups change over time. The researchers identified 18 different health and non-health covariates (Table 1) that were averaged over the 20-year study period to create time-stable predictors of mortality rates. The predictors with higher average scores were found to be less likely to be associated with high under-five mortality rates.

Four predictors were found to be highly likely to predict a country’s membership in one of the higher mortality groups: fertility rate, female secondary school enrollment rate, receiving the DTP 1 vaccine, and receiving the DTP 3 vaccine.

In addition to identifying important predictors of under-five mortality, this study yielded important findings pertaining to grouping the countries by mortality rate. All 21 LMICs in the group with the lowest mortality rates have already achieved Sustainable Development Goal Three, which aims to provide adequate healthcare and end preventable child deaths. Encouragingly, the 18 countries in the group with the highest mortality rates showed the highest declines in under-five mortality rates over the study period.

Viewing trends over time and identifying factors that can lead to declines in under-five mortality helps provide donors and countries with a better understanding of child mortality in LMICs and can offer useful insight into areas for improvement. Currently, researchers at D4I are looking into the possibility of exploring time-varying predictors and studying multivariable models.

Table 1: