Harnessing AI to Transform Global Immunization Funding

23 September, 2024 By Fabrizio Celli (Frontier Data Network) and Nikhil Mandalia (UNICEF Immunization Programme)

The Global Immunization Budget Database (G-IBD) represents a major leap forward in supporting countries on their path to achieving sustainable financing for immunization programmes. By leveraging advanced AI, the G-IBD enhances the efficiency of tracking government immunization budgets, ensuring that financial resources are allocated effectively to close coverage gaps and sustain immunization services. In the future, the data technologies developed for this project can be applied to assess government budget data for other programs, such as education and nutrition.

Tackling a major global health challenge

Ensuring the continuity and effectiveness of immunization services globally is an enormous task. Accurate and transparent tracking of funding for immunization within national budgets is essential for ensuring government accountability in the delivery of these services. Public Finance Management (PFM) systems are the backbone of effective health financing, but without robust data, countries risk misallocating resources, disrupting service delivery, and failing to meet immunization targets. Unfortunately, current practices for tracking funding for immunization are inadequate due to data inconsistencies, time-consuming processes, and policymaking challenges.

Budget and expenditure data collection mechanisms are hampered by inaccuracies and inconsistencies due to varying interpretations of reporting requirements and challenges in retrieving information from government financial systems. Obtaining relevant information from government budgets is also time intensive, requiring manual searches through lengthy budget documents presented in different formats across countries. The unreliability of immunization budget data limits its use by countries, hampering effective decision-making and policy formulation, leading to inefficient resource allocation, ultimately risking service disruptions.

Solution: Leveraging Technology to Improve Data Quality

To tackle these issues, UNICEF’s Immunization Team contacted the Frontier Data Network, which developed an innovative approach to data collection, processing and classification. The resulting product, the G-IBD, harnesses advanced AI to not only streamline data extraction but also improve the comparability of immunization budget data across countries.

Key features of the Global Immunization Budget Database:

  • AI-Powered Data Extraction: The application uses artificial intelligence (AI) to extract data from unstructured documents, including image-based PDFs, and classifies budget data according to standardized classification frameworks.

  • Comprehensive Data Manipulation: Through various stages of data manipulation, including machine learning-driven classification, G-IBD ensures that relevant data are retained and organized effectively for analysis.

  • Web-Based Accessibility: G-IBD is a web-based application with a robust backend that supports data uploads, processing, and retrieval through APIs. This facilitates the creation of interactive dashboards for data visualization.

By utilizing publicly accessible government budget documents, G-IBD minimizes the reporting burden placed on country governments and enhances the organization and efficiency of budget data management. G-IBD relies on the Azure Document Intelligence Service for data extraction, Data Bricks for training machine learning models, and GitHub and Azure Functions for hosting and querying the AI models.

Navigating technical and operational challenges

The G-IBD project team had to overcome several project-specific challenges to develop a working product – from managing document format variability to multiple languages. Budget documents vary widely in presentation and format across different countries and even within the same country over time. This required the development of flexible data extraction and classification methods. The initial stages of data processing involved manual interventions to ensure accuracy, guiding the AI in distinguishing relevant data from irrelevant content. Supporting multiple languages presented a complex challenge, requiring the team to utilize advanced tools like the Azure OpenAI API to effectively manage multilingual document processing.

Future directions: Expanding and enhancing the database

The minimum viable product (MVP) was launched on July 31, 2024 for UNICEF staff. The project team aims to further enhance the G-IBD with several key developments:

  • Expanding Data Collection: The team plans to gather budget documents from additional countries, providing a richer dataset for deeper analysis.
  • Developing Dashboards: Dynamic data visualizations, created using PowerBI, will facilitate routine analysis and provide insights into immunization budgets.
  • Expanding User Access: The platform will be made available to users from country governments, development partners, NGOs, and CSOs, ensuring that a broad range of stakeholders can benefit from the enhanced data and insights provided by G-IBD.
  • Implementing Multilingual Support: Enhancing the system to process documents in various languages will extend G-IBD’s reach and utility.
  • Broadening Programmatic Scope: The data processes developed for G-IBD will be adapted to support other UNICEF program areas, such as health, education, and social protection, leveraging the comprehensive dataset for wider applications.

A promising step towards better immunization coverage

The G-IBD project represents a significant advancement in PFM and immunization budget data management. By addressing the critical challenges related to data inconsistency, and time and labor-intensive processes, G-IBD is set to strengthen data driven decision-making around sustainable immunization financing globally. As the project evolves, its impact will extend beyond immunization, supporting various UNICEF programme areas and contributing to a more data-driven approach to global health initiatives.

Acknowledgements

The Frontier Data Network (FDN) at the UNICEF Chief Statistician Office is dedicated to accelerating adoption of innovative data technologies and practices to enhance delivery of impactful, reproducible, data-driven solutions for the world’s most vulnerable children. We achieve this by cultivating community and partnerships to scale talent, data infrastructure, and new analytics initiatives.  

Project team includes Ulla Griffith (Program Group – Immunization), business owner and project director; Nikhil Mandalia (PG-I), senior user; Yves Jaques (FDN tech unit), senior supplier; Fabrizio Celli (FDN tech unit), lead engineer; Perfect Igbadumhe (FDN tech unit), .NET developer; Piotr Krosniak (Program Group – Immunization), data viz and data scientist.

This project was made possible in part through the Spanish Agency for International Development Cooperation’s (AECID) generous support of the Frontier Data Network, along with the Government of Canada and U.S. Centers for Disease Control and Prevention’s (CDC) invaluable support to UNICEF in strengthening immunization and primary health care.