Literaturnachweis - Detailanzeige
| Autor/inn/en | Balyan, Renu; McNamara, Danielle S.; Crossley, Scott A.; Brown, William, III; Karter, Andrew J.; Schillinger, Dean |
|---|---|
| Titel | Employing Computational Linguistics to Improve Patient-Provider Secure Email Exchange: The ECLIPPSE Study |
| Quelle | (2022), (32 Seiten)
PDF als Volltext |
| Zusatzinformation | ORCID (Balyan, Renu) ORCID (Karter, Andrew J.) ORCID (Schillinger, Dean) Weitere Informationen |
| Sprache | englisch |
| Dokumenttyp | gedruckt; online; Monographie |
| Schlagwörter | Computational Linguistics; Patients; Physicians; Information Security; Electronic Mail; Natural Language Processing; Readability; Health Promotion; Literacy; Artificial Intelligence; Profiles |
| Abstract | Online patient portals that facilitate communication between patient and provider can improve patients' medication adherence and health outcomes. The effectiveness of such web-based communication measures can be influenced by the health literacy (HL) of a patient. In the context of diabetes, low HL is associated with severe hypoglycemia and high rates of complications. Health outcomes could be improved for these patients by increasing quantity and quality of health-related communications. However, current measurement tools for HL are time-consuming and require in-person administration. Similarly, not much is known about the readability of physicians' secure messages (SMs). Natural language processing (NLP) tools have been previously used to measure medical text readability, but not to examine physician-patient communications. The ECLIPPSE project sought to develop and validate automated patient literacy profiles (LPs) and physician language complexity profiles (CPs) by assessing the linguistic features of SMs generated by diabetes patients and primary care physicians, respectively. The study data was obtained from the KPNC Diabetes Registry, including only SMs that were between patients and primary care providers. The final ECLIPPSE dataset for the patient HL measure constitutes 283,216 SMs written by 6,941 patients to their primary care providers, a validated HL scale containing items measuring HL self-efficacy, and expert ratings of HL based on the quality of patients' SMs. A Suite of Automatic Linguistic Analysis Tools (SALAT) was used to extract linguistic features from patient SMs. Using machine learning (ML) techniques, five separate LP prototypes were developed. Three were baseline profiles that used increasingly sophisticated NLP indices, one selected 185 linguistic features to best predict self-reported HL, and one used 8 linguistic features to predict expert human ratings of HL. The authors explored associations between each LP's classification of HL and patients' sociodemographic features, reported provider communications, diabetes-related outcomes, and healthcare utilization. The final ECLIPPSE dataset for the CP measure contained 724 unique SMs written by 592 primary care providers to 486 patients. Readability of physicians' SMs were rated by two experts and categorized into CPs of "low" or "high". The Model of Text Readability in Physicians (MoTeR-P) used 24 linguistic features to predict expert ratings of readability, achieving an accuracy of 0.749. Several challenges were overcome during the creation of LPs and CPs, including maintaining both patient security and data accessibility, identifying messages written by patient proxies, selecting the best set of linguistic indices for training machine learning models, and working across multiple scientific disciplines and geographical locations. The chapter concludes that applying innovative NLP and ML approaches to generate a patient-physician LP and CP from their SMs is a feasible strategy for identifying patients with limited HL, and to identify those physicians who write complex messages to their patients. Future work will focus on the development and evaluation of an online, automated feedback prototype embedded in the patient portal. [This chapter was published in: "Natural Language Processing in Healthcare," CRC Press, 2022, pp. 211-241.] (As Provided). |
| Erfasst von | ERIC (Education Resources Information Center), Washington, DC |
| Update | 2024/1/01 |
Literaturbeschaffung und Bestandsnachweise in Bibliotheken prüfen
Permalink als QR-Code
Inhalt auf sozialen Plattformen teilen (nur vorhanden, wenn Javascript eingeschaltet ist)
Teile diese Seite:
