Lessons learned from screening literature on person-centred care: Exploration of text-mining functions for semi-automatization of study identification.
Notice bibliographique
Résumé
Short introduction: To develop evidence-based person-centred care practice, informed by synthesised international knowledge, there is a need to comprehensively identify relevant citations from systematic data-base searches. However, systematic searches on ‘thick’ concepts with diffuse conceptual boundaries like person-centred care will generate to many citations possible to screen and assess manually. Background and purpose: The implementation of person-centred care is hindered by the fact that research results are not easily accessible. Searches of databases for original studies generate at least 90,000 unique citations. In addition to this large number of publications, there is also a diversity in terminology and conceptualisations used. When decision-makers, practitioners and researchers are unable to review all existing knowledge, the obvious risk is confusion in implementation. As part of an ongoing project, which aims to map available international literature of centredness in healthcare, our specific purpose with this presentation is to share some lessons learned from literature screening supported by text-mining functions. This was performed by a team of researchers specialised in interprofessional person-centred care who in the larger team includes patient partners, students and project assistants. Our experience is of importance not only for researchers, but also for policy-makers, decision-makers and healthcare professionals. Method: The use of text-mining functions to semi-automate the process of citation screening is a way to tackle the great amount of research citations available today. Database searches for literature on person-centred care resulted in the retrieval of 94 236 unique citations. A random sample of 5455 records was screened manually by two reviewers independently against inclusion- and exclusion criteria. Results from that screening was used to build two project tailored text-mining classifier models, one built manually by a language technologist based on word frequencies, and one machine learning classifier constructed in the software EPPI-reviewer using the sci-kit-learn library. The 1000 highest-ranking records were retrieved and manually screened for both models. Result: In the exploration, manual screening of the first random sample of 5455 citations resulted in 3,7% of the sample to be included in the specific mapping study. When screening the 1000 highest ranked citations using the manually built classifier model 23,5% of the sample was included. . The EPPI reviewer classifier model resulted in 83,4% to be included, while applying the same criteria for inclusion. Discussion: For our purposes the classifier model built in EPPI-reviewer showed promise in identifying relevant citations earlier in the process as compared to a manually built classifier. Both models performed better than random manual screening. Making use of classifying model software is merited to facilitate processes of screening and sifting citations in knowledge fields that are hard to conceptually delimit in larger data bases, such as the field of person-centred care research.
Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.
Comment cette classification a été obtenuedéplier
Prédiction machine sur la base complète
Imitation des enseignantsNi prévalence calibrée, ni vérité terrain. Validation humaine à venir. Le volet Gemma est une étiquette directe du modèle pour chaque travail de la base, lue sur la notice réduite au titre. Le volet Codex est un classifieur appris des 10 348 étiquettes directes de Codex et calibré sur les taux pondérés de l'échantillon; les champs sans appui suffisant ne portent aucun appel Codex. Le mode candidate est l'union des deux volets; le consensus est leur intersection. Ces sorties portent le statut machine_predicted_unvalidated et ne sont pas des étiquettes humaines.
Scores du classifieur distillé par catégorie (deux têtes)
| Catégorie | Codex | Gemma |
|---|---|---|
| Métarecherche | 0,352 | 0,680 |
| Méta-épidémiologie (sens strict) | 0,003 | 0,002 |
| Méta-épidémiologie (sens large) | 0,005 | 0,005 |
| Bibliométrie | 0,037 | 0,033 |
| Études des sciences et des technologies | 0,004 | 0,005 |
| Communication savante | 0,020 | 0,017 |
| Science ouverte | 0,007 | 0,012 |
| Intégrité de la recherche | 0,003 | 0,005 |
| Charge utile insuffisante (le modèle a refusé de juger) | 0,008 | 0,005 |
Scores machine (provisoires)
Les deux têtes enseignantes du modèle étudiant, lues sur ce travail. Un score ordonne la base pour la relecture; il n'affirme jamais une catégorie, et le statut de validation accompagne chaque rangée tel quel.
Scores de référence d'un modèle non mature (critères de maturité non atteints, 7 itérations). Un score ordonne; il n'affirme jamais une catégorie.
score_only:v0-immature-baseline · tel quel depuis la passe de notation : score_only signifie que le nombre peut ordonner les travaux, et qu'aucune étiquette de catégorie n'en découleClassification
machine, non validéePrédiction automatique; l’étiquette directe de Gemma et le classifieur distillé Codex s’accordent sur ce qui est montré ici.
Le détail, modèle par modèle et score par score, se trouve en fin de page sous « Comment cette classification a été obtenue ».