L’evidenza che generiamo deve diventare anche conoscenza condivisa.
Esplora pubblicazioni, poster e comunicazioni scientifiche sviluppate con Latrikos. Ogni lavoro mostra come una metodologia clinica strutturata, tracciabile e riproducibile possa trasformare la documentazione assistenziale in evidenza utile per ricercatori, ospedali, società scientifiche e promotori.
Automated generation of real-world clinical databases in Allergy: a data-driven research framework
1Ramón y Cajal Hospital, Madrid, Spain
EAACI Congress · 2026
A reproducible, data-driven framework, implemented on the Latrikos platform, that turns routine clinical narratives into longitudinal, high-granularity and interoperable Allergy research registries — with local de-identification, expert-validated extraction (target accuracy ≥ 99.5%) and automated analytics — to enable real-world evidence generation and data reuse.
- Allergy
- Real-world evidence
- Clinical data structuring
- Research registries
- Longitudinal data
- Natural language processing
Vedi l’abstract
- Introduzione
- Most clinically relevant information in Allergy remains embedded in unstructured clinical narratives, making large-scale data collection labor-intensive, poorly reproducible and difficult to reuse for research purposes.
- Obiettivo
- To develop a reproducible methodology for generating longitudinal, high-granularity and interoperable Allergy research registries from routine clinical narratives.
- Materiali e metodi
- Reproducible pipeline structured in six stages: (1) clinical ontology definition — variable generation and definition, manual and expert review, and search/mapping in the hospital information system (HIS); (2) processing-model optimization on EHR/clinical text via the Latrikos platform, tuning preprocessing to a target accuracy ≥ 99.5%; (3) processing with local de-identification (pseudonymization) to produce longitudinal records; (4) random-sample validation by manual expert review and accuracy assessment, iterating on variables, definitions or engineering level until accuracy is satisfactory; (5) automated statistics (descriptive, inferential, data-driven and clustering); and (6) a structured database with analytics, live visualization and PDF export.
- Risultati
- The framework yields structured, longitudinal, high-granularity and interoperable Allergy registries directly from routine clinical narratives, with a preprocessing target accuracy of ≥ 99.5% and expert-in-the-loop validation. On top of the resulting database it provides automated descriptive, inferential, data-driven and clustering analyses, with live visualization and PDF export.
- Conclusioni
- This methodology enables the generation of reproducible, longitudinal and high-granularity Allergy research registries directly from routine clinical narratives, facilitating real-world evidence generation and future reuse across evolving clinical paradigms.
Automated generation of structured clinical databases in drug hypersensitivity: real-world validation in patients with proton pump inhibitor reactions
1Ramón y Cajal Hospital, Madrid, Spain
EAACI Congress · 2026
Real-world validation of an automated methodology, implemented on the Latrikos platform, for building structured clinical databases from routine medical reports. In 172 patients with suspected proton pump inhibitor (PPI) hypersensitivity, the approach extracted 13,244 data points across a 77-variable ontology with 98.96% accuracy in 2 hours.
- Drug hypersensitivity
- Proton pump inhibitors
- Real-world evidence
- Clinical data structuring
- Drug allergy
Vedi l’abstract
- Introduzione
- Most clinically relevant information in drug hypersensitivity remains embedded in routine clinical narratives. The unstructured nature of these reports limits the generation of reproducible and scalable clinical databases for research.
- Obiettivo
- To validate a reproducible methodology for generating structured clinical databases from routine clinical narratives in patients with suspected PPI hypersensitivity.
- Materiali e metodi
- Retrospective, single-centre study on routine clinical reports. The methodology comprised: (1) definition of a domain ontology; (2) ingestion of EHR clinical text; (3) local de-identification and pseudonymization; (4) NLP-based entity extraction; (5) generation of structured longitudinal registries; (6) an anonymized, interoperable real-world dataset; and (7) automated descriptive and inferential analysis. The pipeline was implemented on the Latrikos platform (hosted on OVHcloud).
- Risultati
- 357 patients with a PPI listed in the diagnosis field were identified; 185 were excluded (eosinophilic esophagitis on PPI therapy or patients not evaluated), leaving n = 172 (2018–2025). Using a 77-variable ontology (demographics, symptoms, treatments, skin tests, BAT/LTT, DPT and cross-reactivity), 13,244 data points were extracted with an overall extraction accuracy of 98.96% (138 errors / 13,244 variables) and a processing time of 2 hours.
- Conclusioni
- Structured clinical databases can be generated accurately and reproducibly from routine medical reports in patients with suspected PPI hypersensitivity. These findings support the use of the proposed methodology, implemented through the Latrikos platform, as a scalable framework for real-world research in drug allergy.