Data science

Using geocoded big data to identify causal links between infectious diseases and child developmental outcomes

Getulio Vargas Foundation (FGV)

Project Presentation

  • What it is?

    Infectious diseases may have only transitory impacts on pregnant mothers, but they can have lasting impacts on children. Can public interventions mitigate these impacts? This project aims to identify how exposure to localized epidemiological risk factors in the fetal period influences developmental outcomes for children through the early years of life. It will evaluate the extent in which the access to primary health care and social welfare programs mitigate negative impacts in child development.

  • How was the experiment?

    A comprehensive data analysis was performed on the entire population of Rio de Janeiro from 2000 to 2017, focusing on recorded births, deaths, and notifications of infectious diseases—specifically dengue, syphilis, and tuberculosis. The goal was to examine maternal exposure and its effects on birth outcomes and infant mortality. To do this, the research team geocoded mothers’ addresses, linking them to census tracts and healthcare coverage areas, enabling the replication of these methodologies nationwide. Using survival curves and regression models, the study revealed that the spread of infections is influenced by local factors such as entomological conditions, population density, and community networks. Additionally, a literature review indicated mixed evidence regarding the impact of infections during pregnancy on babies, highlighting significant gaps in the assessment of the role of primary care in this area.

  • Main Results

    The project provided detailed descriptive information about the incidence of diseases and adverse outcomes, mapping the city for each disease, the proportion of women infected during pregnancy, and the 1-year infant mortality rate (IMR) for births from infected mothers. A high heterogeneity was observed among the diseases—both the incidence and the infection-conditioned IMR vary across regions, with syphilis indicators notably higher in all cases. This finding suggests that the incidence of diseases among mothers and their effects are highly localized within the territory. Additionally, notable differences were found between infection rates and the infection-conditioned IMR for each disease. Importantly, maps illustrating the infection-conditioned IMR reveal that adverse effects are often even more localized within the city and do not always coincide with the incidence map. This is particularly evident with tuberculosis, where impacts on IMR are concentrated in specific areas. Furthermore, regression analyses indicated that infections during pregnancy generally result in negative consequences for birth outcomes and child health in the first years of life. However, these adverse effects are specific to combinations of diseases and outcomes and tend to be more localized within the city than the disease incidence itself. The results suggest that the role of primary health care in mitigating these effects varies among different disease and outcome combinations, which can lead to significant compositional effects as well as both positive and negative unintended consequences. On a positive note, there is suggestive evidence that the onset of infection may promote greater engagement of mothers with health services, potentially increasing child survival in the years following birth (e.g., in the case of tuberculosis). Conversely, there is also suggestive evidence that access to primary care is associated with a reduction in fetal death, which may eventually lead to an increase in infant mortality after birth (e.g., in the case of syphilis).

  • What problems it seeks to solve?

    Infections are one of the main causes of maternal mortality and morbidity worldwide, with the greatest impact estimated in low- and middle-income countries. According to recent global estimates on disease incidence, maternal sepsis and other pregnancy-related infections totaled about 12 million cases in 2017, corresponding to a ratio of 1 case for every 11 live births worldwide. However, maternal infections can leave even deeper scars on global health indicators and human capital development if they also affect their children. Despite extensive clinical research documenting if and how infections can impact maternal and fetal health, and the potentially disruptive impacts that maternal infections can have on child health and long-term human development, there is still very little causal evidence directly linking in utero infection exposure to child health. The infection status of women is often unobserved individually and rarely associated with children’s outcomes after birth. This is especially common in low- and middle-income countries, where vital statistics are often scarce and infectious diseases are widespread. There is even less evidence on the extent to which socioeconomic status and access to healthcare, especially primary care services, can mitigate potentially harmful impacts.

  • Brazilian Health implications

    The project results are relevant as they contribute to product innovation and policy refinement. Three main contributions stand out: 1) Development of linkage and georeferencing algorithms for the Datasus/MS information systems (SIM, SINASC, SINAN), which have been made publicly available and can be replicated by the academic community. 2) Using the final database, it was possible to map in an extremely refined way (i) the incidence of infectious diseases among pregnant women and (ii) outcomes for child health (up to 5 years of age), conditional on maternal infection. The maps are available. It is important to highlight that both incidence and mortality conditioned on incidence are highly localized within the territory, enabling targeted interventions. 3) Finally, a set of econometric results was made available, allowing the identification of disease-outcome combinations for which adverse effects are more severe; and for which access to primary care has been more/less protective.