Research on daily medication and drug efficacy evaluation for chronic diseases based on natural language processing (NLP)

Objectives: This study aims to harness Natural Language Processing (NLP) to improve chronic disease medication management, focusing on the antihypertensive drug Lisinopril. The objectives are: (1) to develop an intelligent drug safety monitoring and personalized intervention platform integrating patient feedback; (2) to accurately identify drug side effects and analyze their associations with patient demographics; and (3) to build a personalized medication recommendation system for enhanced clinical decision-making. Methods: We used user reviews from authoritative medical databases (WebMD and Kaggle) spanning 2008-2020, forming a high-quality dataset of 345,845 samples related to Lisinopril. Data preprocessing involved redundancy removal, missing value imputation (e.g., using the KNN algorithm), and text standardization (e.g., through UMLS terminology mapping). NLP techniques included sentiment analysis and entity recognition with BERT models, topic modeling using LDA algorithms to extract key side-effect patterns (e.g., dry cough, dizziness), as well as the integration of knowledge graphs and decision trees for personalized recommendations, and the development of an interactive risk heatmap dashboard for dynamic monitoring. Results: The analysis revealed that common side effects of Lisinopril included dry cough (38%), dizziness (28%), and fatigue (22%), with nearly half of users experiencing these issues-particularly among older adults and long-term users. Sentiment analysis showed that 58% of reviews were negative, primarily due to side effects. The personalized recommendation system, tested in simulated scenarios, significantly improved medication adherence by suggesting alternatives (e.g., ARBs for cough-sensitive patients). The NLP-driven framework achieved 89.7% accuracy in sentiment classification and identified side-effect patterns 15 times faster than manual annotation. Conclusion: NLP effectively extracts insights from patient feedback, enabling rapid identification of drug efficacy and side effects. The study demonstrates the potential of NLP to support chronic disease management and precision medicine by facilitating early risk detection and personalized interventions. Future work should address subjective bias in reviews and integrate multi-source data for comprehensive assessment models

Health care · Knowledge base · MEDLINE · Preprocessor · Sentiment analysis · SNOMED CT · Terminology · Unified Medical Language System · Machine Learning in Healthcare · Medication Adherence and Compliance · Pharmacovigilance and Adverse Drug Reactions

• Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (Dalys), and healthy life expectancy (Hale) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990–2021

  Open Access•Alize J Ferrari, Damian Francesco Santomauro et al.•The Lancet•2024

  Add to list

• High-performance medicine

  Open Access•Eric J Topol•Nature Medicine•2019

  Add to list

• The global landscape of AI ethics guidelines

  Open Access•Anna Jobin, Marcello Ienca et al.•Nature Machine Intelligence•2019

  Add to list