Census-derived migration data as a tool for informing malaria elimination policy

Nick Ruktanonchai, Darlene Bhavnani, Alessandro Sorichetta, Linus Bengtsson, Keith H. Carter, Roberto Córdoba, Arnaud Le Menach, Xin Lü, Erik Wetter, Elisabeth zu Erbach-Schoenberg, Andrew J. TatemMalaria 2016 · DOI

Abstract

These results show long-term migration data can effectively predict the relative flows of infected people to direct malaria elimination policy, a particularly relevant result because migration data are generally easier to obtain than short-term movement data such as mobile phone records. Further, predicted relative flows highlight policy-relevant population dynamics, such as major exporters across the region, and Nicaragua and Costa Rica's strong connection by movement of infected people, suggesting close coordination of their elimination efforts. Country-specific applications are discussed as well, such as predicting areas at relatively high risk of importation, which could inform surveillance and treatment strategies.

Introduction

As global efforts push toward malaria elimination, the strategic focus shifts from broad control to managing the risk of reintroduction. A primary challenge in this phase is 'importation'—the movement of infected individuals across borders, which can spark new outbreaks in areas where local transmission has ceased. To effectively manage this risk, health officials require high-resolution data on human movement. While mobile phone Call Detail Records (CDRs) are considered the gold standard for tracking short-term movement, they are often inaccessible due to cost, privacy concerns, and lack of cross-border coverage. This study investigates whether widely available census-based migration data can serve as a reliable proxy for these critical movement patterns.

Validation: Census Data as a Proxy for Movement

The researchers first validated the use of census data by comparing 2003 census records from Haiti with 2010 mobile phone data. Using logistic regression and gravity models, they analyzed how both datasets responded to variables such as population size, distance, and urbanization.

Comparison of ranked movement flows between CDR and census data.

A key finding was that while census data significantly under-predicts the absolute volume of movement, the relative connectivity—identifying which paths are the most heavily traveled—remained remarkably consistent, with an $R^2$ of 0.69. For public health policy, identifying the primary 'parasite highways' is often more critical than knowing the exact number of travelers, making census data a powerful tool for countries lacking CDR access.

Mapping Parasite Connectivity in Mesoamerica

Applying this methodology to Mesoamerica, the study utilized census microdata from El Salvador, Costa Rica, and Nicaragua to predict population flows across the region. By combining these movement predictions with 2013 malaria incidence data from the Pan American Health Organization (PAHO), the researchers estimated the flow of infected individuals.

Predicted population flows across Mesoamerica based on census migration data.

The resulting maps identified 'major exporters'—regions that act as hubs for spreading the parasite to neighboring areas. For instance, northern Nicaragua was identified as a critical source of malaria cases for both Costa Rica and El Salvador. Targeting these source regions with interventions like bed nets or active case detection offers a more efficient use of resources for the entire region.

Strategic Coordination and Policy Communities

Using a walktrap community detection algorithm, the study identified 'movement communities'—groups of regions so tightly linked that they should coordinate elimination efforts as a single unit. A prominent example is the strong connection between Nicaragua and Costa Rica, suggesting that Costa Rica's elimination success is inextricably linked to the timing and strategy of its northern neighbor.

Furthermore, the model allows for subnational targeting. In Costa Rica, the province of Limon showed high incidence but low importation risk, suggesting transmission is primarily local. In contrast, Alajuela showed low incidence but high importation risk, indicating that cases there are likely 'visitors' requiring different surveillance and treatment strategies.

Conclusion and Future Directions

This research demonstrates that census-derived migration data is a viable and accessible tool for informing malaria elimination policy. While the model has limitations—such as assuming population homogeneity and relying on potentially dated census records—it provides a scalable framework for identifying high-risk corridors. Future work could refine these models by integrating travel history surveys from clinics to better capture the movement patterns of high-risk subgroups, such as seasonal workers or miners.

Summary updated

Abstract

These results show long-term migration data can effectively predict the relative flows of infected people to direct malaria elimination policy, a particularly relevant result because migration data are generally easier to obtain than short-term movement data such as mobile phone records. Further, predicted relative flows highlight policy-relevant population dynamics, such as major exporters across the region, and Nicaragua and Costa Rica's strong connection by movement of infected people, suggesting close coordination of their elimination efforts. Country-specific applications are discussed as well, such as predicting areas at relatively high risk of importation, which could inform surveillance and treatment strategies.
BibTeX
@inproceedings{ruktanonchai2016census,
  author = {Nick Ruktanonchai and Darlene Bhavnani and Alessandro Sorichetta and Linus Bengtsson and Keith H. Carter and Roberto Córdoba and Arnaud Le Menach and Xin Lü and Erik Wetter and Elisabeth zu Erbach-Schoenberg and Andrew J. Tatem},
  title = {{Census-derived migration data as a tool for informing malaria elimination policy}},
  booktitle = {Malaria},
  year = {2016},
  doi = {10.1186/s12936-016-1315-5},
}
Figure