Ai Okajima

Ai Okajima

Auto-generated profile, based on publicly available information

Ai Okajima

Yuimedi Inc., Tokyo, Japan

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
31 October 2024 in Healthcare Informatics Research
Verified
Journal Article
Open access
PDF file
EKEizen Kimura
Eizen Kimura
YKYukinobu Kawakami
Yukinobu Kawakami
SIShingo Inoue
Shingo Inoue
AOAi Okajima
Ai Okajima
Objectives: This study evaluated the efficacy of integrating a retrieval-augmented generation (RAG) model and a large language model (LLM) to improve the accuracy of drug name mapping across international vocabularies.Methods: Drug ingredient names were translated into English using the Japanese Accepted Names for Pharmaceuticals. Drug concepts were extracted from the standard vocabulary of OHDSI, and the accuracy of mappings between translated terms and RxNorm was assessed by vector similarity, using the BioBERT-generated embedded vectors as the baseline. Subsequently, we developed LLMs with RAG that distinguished the final candidates from the baseline. We assessed the efficacy of the LLM with RAG in candidate selection by comparing it with conventional methods based on vector similarity.Results: The evaluation metrics demonstrated the superior performance of the combined LLM + RAG over traditional vector similarity methods. Notably, the hit rates of the Mixtral 8x7b and GPT-3.5 models exceeded 90%, significantly outperforming the baseline rate of 64% across stratified groups of PO drugs, injections, and all interventions. Furthermore, the r-precision metric, which measures the alignment between model judgment and human evaluation, revealed a notable improvement in LLM performance, ranging from 41% to 50% compared to the baseline of 23%.Conclusions: Integrating an RAG and an LLM outperformed conventional string comparison and embedding vector similarity techniques, offering a more refined approach to global drug information mapping.

21 February 2025 in Data in Brief
Verified
Journal Article
Open access
EKEizen Kimura
Eizen Kimura
YKYukinobu Kawakami
Yukinobu Kawakami
SIShingo Inoue
Shingo Inoue
AOAi Okajima
Ai Okajima
Observational Health Data Sciences and Informatics (OHDSI) is an international research community dedicated to large-scale observational studies using real-world healthcare data. Participation in OHDSI requires mapping local terminology systems to the OHDSI Standard Vocabulary (OSV) and transforming healthcare data into the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM), a standardized database schema. Adherence to the OSV and CDM enables the integration of datasets from different countries and regions, facilitating international cross-sectional analyses and supporting the discovery of large-scale evidence and new medical knowledge. Despite the globally recognized healthcare technology and systems excellence in Japan, Japanese real-world data (RWD) remain underutilized in international research. This is primarily due to reliance on domestically managed controlled terminologies in Japan that are not aligned with international controlled terminologies such as SNOMED CT, making mapping Japanese RWD to the CDM challenging. In addition, the wide variety of pharmaceutical products in Japan has hindered the establishment of mappings to RxNorm, the standardized drug terminology used in OHDSI. Previously, we used a Large Language Model (LLM) to map Japanese pharmaceutical data to RxNorm. A sampling-based evaluation confirmed that the LLM could accurately identify mapping candidates. Pharmacists and a medical informatics researcher validated these mappings, resulting in an ingredient-based mapping of Japanese pharmaceutical terms to RxNorm. Researchers interested in pharmacoepidemiology, pharmacoeconomics, and drug-related clinical decision support systems integrated with Japanese RWD can benefit significantly from this dataset. It also contains information about the target drugs, their translated names, LLM-generated suggestions, and reference data, making it suitable for developing and validating natural language processing and machine learning techniques for terminology mapping.