Patient‐centered and experience‐aware mining for effective adverse drug reaction discovery in online health forums

Liu, Yunzhong; Shi, Jinhe; Chen, Yi

doi:10.1002/asi.23929

Cited by 9 publications

(5 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing methods for drug discovery from online content extract drugs at post and statement level. ADR mining systems typically include a named entity recognition (NER) model and a relationship or semantic role labeling model [ 19 , 20 ]. Recent neural approaches address lexical variation in user-generated content – the difficulty faced by traditional keyword matching and rule-based approaches – to improve recognition and labeling components [ 21 , 22 ].…”

Section: Related Workmentioning

confidence: 99%

Neural side effect discovery from user credibility and experience-assessed online health discussions

et al. 2020

View full text Add to dashboard Cite

Background Health 2.0 allows patients and caregivers to conveniently seek medical information and advice via e-portals and online discussion forums, especially regarding potential drug side effects. Although online health communities are helpful platforms for obtaining non-professional opinions, they pose risks in communicating unreliable and insufficient information in terms of quality and quantity. Existing methods in extracting user-reported adverse drug reactions (ADRs) in online health forums are not only insufficiently accurate as they disregard user credibility and drug experience, but are also expensive as they rely on supervised ground truth annotation of individual statement. We propose a NEural ArchiTecture for Drug side effect prediction (NEAT), which is optimized on the task of drug side effect discovery based on a complete discussion while being attentive to user credibility and experience, thus, addressing the mentioned shortcomings. We train our neural model in a self-supervised fashion using ground truth drug side effects from mayoclinic.org. NEAT learns to assign each user a score that is descriptive of their credibility and highlights the critical textual segments of their post. Results Experiments show that NEAT improves drug side effect discovery from online health discussion by 3.04 % from user-credibility agnostic baselines, and by 9.94 % from non-neural baselines in term of F 1 . Additionally, the latent credibility scores learned by the model correlate well with trustworthiness signals, such as the number of “thanks” received by other forum members, and improve credibility heuristics such as number of posts by 0.113 in term of Spearman’s rank correlation coefficient. Experience-based self-supervised attention highlights critical phrases such as mentioned side effects, and enhances fully supervised ADR extraction models based on sequence labelling by 5.502 % in terms of precision. Conclusions NEAT considers both user credibility and experience in online health forums, making feasible a self-supervised approach to side effect prediction for mentioned drugs. The derived user credibility and attention mechanism are transferable and improve downstream ADR extraction models. Our approach enhances automatic drug side effect discovery and fosters research in several domains including pharmacovigilance and clinical studies.

show abstract

Section: Related Workmentioning

confidence: 99%

Neural side effect discovery from user credibility and experience-assessed online health discussions

et al. 2020

View full text Add to dashboard Cite

show abstract

“…In addition, the current technique for future action detection sometimes generates false positives. To address these issues, we will investigate how to leverage part-of-speech and reference resolution techniques [ 47 ] to improve natural language understanding.…”

Section: Discussionmentioning

confidence: 99%

Identifying Influences in Patient Decision-making Processes in Online Health Communities: Data Science Approach

Shi

Chen

2022

J Med Internet Res

Self Cite

View full text Add to dashboard Cite

Background In recent years, an increasing number of users have joined online health communities (OHCs) to obtain information and seek support. Patients often look for information and suggestions to support their health care decision-making. It is important to understand patient decision-making processes and identify the influences that patients receive from OHCs. Objective We aimed to identify the posts in discussion threads that have influence on users who seek help in their decision-making. Methods We proposed a definition of influence relationship of posts in discussion threads. We then developed a framework and a deep learning model for identifying influence relationships. We leveraged the state-of-the-art text relevance measurement methods to generate sparse feature vectors to present text relevance. We modeled the probability of question and action presence in a post as dense features. We then used deep learning techniques to combine the sparse and dense features to learn the influence relationships. Results We evaluated the proposed techniques on discussion threads from a popular cancer survivor OHC. The empirical evaluation demonstrated the effectiveness of our approach. Conclusions It is feasible to identify influence relationships in OHCs. Using the proposed techniques, a significant number of discussions on an OHC were identified to have had influence. Such discussions are more likely to affect user decision-making processes and engage users’ participation in OHCs. Studies on those discussions can help improve information quality, user engagement, and user experience.

show abstract

“…The outcome of previous campaigns on eHealth which focusses on addressing the differences that occur between medical terms in laymen queries and the extracted synonyms terms (expert vocabularies) had clearly shown that the task is challenging with a room for enhancement [10], [8], [16]. [22], [27] used the pre-marketing clinical trials and traditional post-marketing surveillance using voluntary and spontaneous report systems to determine the adverse drug reactions (ADRs) which has now become a major health challenge.…”

Section: Related Workmentioning

confidence: 99%

An Enhanced Concept Based Approach for User Centered Health Information Retrieval to Address Medical and Vocabulary Mismatched Issues

Kontagora¹

2020

IJATCSE

View full text Add to dashboard Cite

With the huge growth in health related information online, user centered health information retrieval has become an obli-gation today. In spite of all its numerous benefits, laymen patients are faced with the challenge of finding relevant and authoritative information online due to vocabulary mismatched issues between laymen queries and expert vocabulary during query expansion. However, this study is focused on designing an enhanced concept based approach for user cen-tered health information retrieval that will address vocabulary mismatched issues. The Proposed Enhanced Concept Based Approach has two special modules fully designed and incorporated in it namely: vocabulary controlled module and the medical terms control module to specifically address vocabulary mismatched issues by ensuring that only the most specific terms in a layman's search query are searched for and their synonyms extracted from the dictionary of the selected dataset. We presented and discussed the performance of the proposed enhanced concept based approach with the existing information retrieval approaches. We concentrated on comparing and evaluating the proposed enhanced concept based approach with three existing approaches used in health information retrieval which are: Concept Based Approach, Query Likelihood Model and the Latent Semantic Indexing. The experimental results obtained using the Met-amap, MeSH, UMLS and Khresmoi Project 6 datasets shows that the Proposed Enhanced Concept Based Approach is the best approach to be applied in addressing vocabulary mismatched issues as it managed to obtain best results in re-spect to maxSim scores and idf weighting values using the Text Similarity Scoring Function with maxSim scored 0.89 (89%) and idf weighting value of 3.61.The Proposed Enhanced Concept Based Approach outshines the existing ap-proaches: Concept Based Approach, Query Likelihood Model and Latent Semantic Indexing with scores ranging be-tween 11% to 15% in respect to maxSim scores and idf weighting values. This better results were obtained by the en-hance concept based approach due to the fact that, it ensures that all the medical terms contained in both the layman's queries and expert vocabularies are vocabulary matched and same. Hence we conclude that the Proposed Enhanced Con-cept Approach is best suited to be deployed in addressing the Vocabulary Mismatched issues encountered during query expansion.

show abstract

Patient‐centered and experience‐aware mining for effective adverse drug reaction discovery in online health forums

Cited by 9 publications

References 30 publications

Neural side effect discovery from user credibility and experience-assessed online health discussions

Neural side effect discovery from user credibility and experience-assessed online health discussions

Identifying Influences in Patient Decision-making Processes in Online Health Communities: Data Science Approach

An Enhanced Concept Based Approach for User Centered Health Information Retrieval to Address Medical and Vocabulary Mismatched Issues

Contact Info

Product

Resources

About