Digitizing Swiss Vaccination Cards with Large Language Models
Sintieh Ekongefeyin (Institute for Implementation Science in Health Care),
Phung Lang (Epidemiology, Biostatistics and Prevention Institute)
Rationale
Digital transformation is a global public health priority, enabling more efficient clinical workflows, data-driven policymaking, and real-time public health surveillance. The Swiss Federal Office of Public Health (FOPH) actively promotes the digitization of health records, including the integration of vaccination records into national electronic health systems such as the Electronic Patient Dossier. Vaccination remains one of the most effective public health interventions, preventing an estimated 4–5 million deaths globally each year. However, vaccination records in Switzerland remain largely paper-based, handwritten, and variable in format, making them susceptible to damage,loss or transcription errors. This limits their use in clinical decision-making and public health surveillance.
To address this gap, we propose digitizing vaccination cards of children as initial step and to explore whether and how artificial intelligence (AI) tools, including large language (LLMs) and deep learning techniques, can transform unstructured handwritten health records into structured digital information suitable for research, clinical use and public health surveillance.
Aim
Overall, we aim to build and evaluate a tool that extracts key immunization information from current Swiss vaccination cards and exports the results in interoperable, standards-compliant formats.
Specifically, we aim to:
- Develop and pilot a pipeline that converts images of the current Swiss paediatric vaccination cards (in German) into machine-readable HL7 FHIR format.
- Assess the accuracy, reliability and robustness of the digitization process using a representative set of vaccination cards.
- Create the foundations for a high-quality dataset of digitized vaccination records, annotated for key immunization features that could be used to support surveillance, monitor vaccination coverage trends, and inform evidence-based immunization policies in Switzerland.
Expected outcome
This project will deliver a pilot of an AI pipeline that supports healthcare providers in improving patient care and enables public health authorities to conduct more accurate, real-time vaccination surveillance. It will also provide a reuseable dataset and methodological framework that can serve as an invaluable resource and foundation for interdisciplinary population research and future digital transformation initiatives in health data management.
Team and partnerships
This project is a collaboration between the Epidemiology, Biostatistics and Prevention Institute (EBPI) and the Institute for Implementation Science in Health Care (IfIS). The Immunity & Vaccines Group (EBPI) contributes expertise on vaccination practices and access to representative card formats. The Digital & Mobile Health group (IfIS) leads the development, implementation and secure deployment of the AI model.