The utilization of this export process has been successful with cohorts exceeding 15000 patients, integrating data from over 10 distinct sources. This implementation facilitated validation within a targeted cohort and enabled the successful dissemination of patient-level results to stakeholders, while ensuring compliance with privacy and integrity requirements.

Integration of NLP-derived results into the OMOP CDM led to significant enhancements in data richness. Firstly, we identified 224% more patients across four different hospitals in Spain who met the inclusion criteria thanks to NLP-derived data. Moreover, the dataset incorporating NLP demonstrated a substantial increment in the proportion of records across different OMOP domains compared to the dataset without NLP. The structured inclusion of NLP-derived results facilitated more comprehensive analyses, enabling deeper insights into treatment patterns and patient outcomes.

A total of 5137 medical records of patients diagnosed with thyroid cancer between 2015 and 2022 were included. The median follow-up (interquartile range) was 29.7 months (8.8-55.8). The mean age at the time of diagnosis was 55 years (SD 18), and 67% were women. The stage could be classified in a subgroup of 520 patients, of which 60% (n=313) had advanced stages. Metastasis was observed in 2177 patients (42%) during the follow-up, mainly in lymph nodes (44%). It was also identified that the majority of patients (71%; n=3629) had some comorbidity.

This work shows that it is possible to build competitive automatic systems when both data and the right evaluation metrics are available. The annotated data, the implemented evaluation scripts, and the section identification Language Model are open-sourced hoping that this contribution will foster the building of more and better systems.

Results from the DERMACLEAR study will increase the real-world evidence of clinical practice, obtaining a large amount of information on patients with the studied diseases. The NLP system used is precise in identifying patients diagnosed with HS, PsO, CU and/or AD, and other medical variables from EHRs, highlighting that it is a valid system to use in the DERMACLEAR study.

Our automated concept mapping system provides an efficient way of mapping source codes to OMOP concepts. By utilizing text-based vector representations and knowledge transfer, our system can find equivalent mappings from other hospitals, thereby reducing the time and effort required for manual mapping.

Finding the false negatives of a NER/NEL system is fundamental to improve it, and is usually done by manual annotation of texts. However, in an environment with a huge volume of unannotated texts (e.g. a hospital) and a low frequency of positives (e.g. a mention of a particular disease in the clinical notes) the task becomes very inefficient.

Although OMOP CDM provides a NOTE_NLP table to store the outputs of NLP algorithms, queries to this table can become clumsy and slow, so we designed and extended the OMOP CDM with our own NLP schema to store the results generated in the annotation process of NLP. We designed an extension of the OMOP CDM able to store the output of NLP solutions while integrating with the vocabulary normalization process of the OMOP CDM.

Taking advantage of electronic health records in clinical research requires the development of natural language processing tools to extract data from unstructured text in dif ferent languages. A key task is the detection of contextual modifiers, such as understanding whether a concept is negated or if it belongs to the past. We present ContextMEL, a method to build classifiers for contextual modifiers that is independent of the specific task and the language, allowing for a fast model development cycle.