Hans A. Eguia (Denmark)
Lægehuset Rudkøbing Primary CareAuthor Of 1 Presentation
ARTIFICIAL INTELLIGENCE IN PRIMARY CARE - EBM RECOMMENDER SYSTEM
Abstract
Abstract Body
PURPOSE:
Evaluate the accuracy of the “Health Operation for Personalized Evidence” (HOPE) project with recommendations based on actual clinical cases from Primary Care patients, evaluating the results by a physician.
METHODS:
HOPE using ontologies and Natural Language Processing (NLP) was used in real medical records (EMR) from patients. Accuracy was reviewed by two GPs, the Kappa-Cohen Coefficient measured the level of agreement between them. The precision was also measured, to model the satisfaction of a user who is presented with different documents.
RESULTS:
HOPE showed results almost immediately, with a mean time of 17,4 seconds. Regarding the medical information found in PubMed, the mean for specificity was 69% with a mean for the sensibility of 49%. The Kappa-Cohen coefficient for the 2 raters in the PubMed recommendation, with 150 subjects analysed was 0,735 (z=9.02 and p-value=0) which represents an “important association” between the raters.
CONCLUSIONS:
Our research seems to be the first to use ontologies and NLP in a non-endogamic way to find reliable information in primary care. HOPE seems to find adequate information in a short time which could potentially let to better diagnosis and treatment, as well as eventually reduce consultation time. This tool could be used in many other languages only using the right thesaurus.
Seems plausible the possibility of a system like HOPE that can recommend reliable health content to the However, it is still necessary to develop the system more so that it understands the EMR better.
Presenter of 1 Presentation
ARTIFICIAL INTELLIGENCE IN PRIMARY CARE - EBM RECOMMENDER SYSTEM
Abstract
Abstract Body
PURPOSE:
Evaluate the accuracy of the “Health Operation for Personalized Evidence” (HOPE) project with recommendations based on actual clinical cases from Primary Care patients, evaluating the results by a physician.
METHODS:
HOPE using ontologies and Natural Language Processing (NLP) was used in real medical records (EMR) from patients. Accuracy was reviewed by two GPs, the Kappa-Cohen Coefficient measured the level of agreement between them. The precision was also measured, to model the satisfaction of a user who is presented with different documents.
RESULTS:
HOPE showed results almost immediately, with a mean time of 17,4 seconds. Regarding the medical information found in PubMed, the mean for specificity was 69% with a mean for the sensibility of 49%. The Kappa-Cohen coefficient for the 2 raters in the PubMed recommendation, with 150 subjects analysed was 0,735 (z=9.02 and p-value=0) which represents an “important association” between the raters.
CONCLUSIONS:
Our research seems to be the first to use ontologies and NLP in a non-endogamic way to find reliable information in primary care. HOPE seems to find adequate information in a short time which could potentially let to better diagnosis and treatment, as well as eventually reduce consultation time. This tool could be used in many other languages only using the right thesaurus.
Seems plausible the possibility of a system like HOPE that can recommend reliable health content to the However, it is still necessary to develop the system more so that it understands the EMR better.