Capturing Social Determinants of Health using Machine Learning for Integrated Care Program Refinement and Spread
Notice bibliographique
Résumé
Background: Social determinants of health (SDOH) such as language preference, health literacy, housing access, food insecurity, social isolation and supports, transportation, depression and addiction, can significantly impact health outcomes and exacerbate disparities surrounding care transitions in and out of hospital. There is growing momentum among healthcare systems to capture SDOH through point-of-care surveys for implementation efforts. However, the sporadic, unstructured nature to these surveys when patients are acutely unwell can lead to low response rates. To be more effective, these efforts can benefit from a systematic approach to capturing SDOH in electronic health records (EHRs). Approach: A comprehensive list of social determinants of health relevant to care transitions was informed based on literature review across different countries then narrowed down to balance feasibility of capture among health records at point of care in two provinces (Ontario and Alberta). A random sample of 3075 charts were then manually reviewed among admitted patients enrolled in integrated care program supporting patients around an acute care admission for these SDOH. Additionally, A specialized keyword list was made to narrow down the search within EHRs. The keywords were selected based on their relevance to the SDOH as highlighted in the literature and based on similar social determinants of health research conducted across other countries. Result: The hospital-level survey (N=3075 of ICP participants) demonstrated poor response and capture of several SDOH, particularly for income-related SDOH. However, the use of chart records (including admission, consultant and other point of care notes) demonstrated feasible and usable capture of key search terms for SDOH. Of the 50 patient charts reviewed so far, 45 individuals had SDOH captured in chart-level records, with the majority of these being related to language barriers and a minority being related to transportation access. Similarly, other SDOH also showed equally low response rates in the survey. These results imply that although it is possible to capture SDOH from electronic health records, existing approaches need significant improvement to become more efficient and scalable. Implication: The use of health care records from clinicians documenting at point-of-care presents a unique opportunity for capturing SDOH in electronic health record systems. Shared learnings from this project will greatly widen institutions with EHRs feasibility and success of capturing SDOH for the evaluation, refinement and spread of integrated care models surrounding acute care admissions. Next steps include collaboration with decision support and analytic teams across Ontario and Calgary to develop machine learning algorithms with the use of natural language processing to perform more extensive data pulls and analytics. This process will also involve implementing the list of keywords that the algorithms will use to improve the accuracy and range of data extraction. These steps would capture a larger and more accurate number of social needs.
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 | 0,011 | 0,040 |
| Méta-épidémiologie (sens strict) | 0,001 | 0,000 |
| Méta-épidémiologie (sens large) | 0,002 | 0,003 |
| Bibliométrie | 0,006 | 0,005 |
| Études des sciences et des technologies | 0,001 | 0,000 |
| Communication savante | 0,002 | 0,001 |
| Science ouverte | 0,001 | 0,002 |
| Intégrité de la recherche | 0,001 | 0,002 |
| Charge utile insuffisante (le modèle a refusé de juger) | 0,002 | 0,000 |
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; un appel candidat d’une seule source (Gemma direct ou Codex distillé), pas un consensus.
Le détail, modèle par modèle et score par score, se trouve en fin de page sous « Comment cette classification a été obtenue ».