University of Washington AI helps ambulances and ICUs weigh emergency risks and cost considerations
por John R. Fischer, Senior Reporter | June 02, 2022
Artificial Intelligence
Emergency Medicine
Risk Management
A new AI tool at the University of Washington can determine which factors to use to predict risk in emergency situations.
To save on time and treat patients faster in emergency situations, researchers at the University of Washington have developed a new AI tool to identify the crucial variables for accurately predicting risks, even within a tight budget.
The solution is called CoAI (Cost-Aware Artificial Intelligence) and predicts outcomes by calculating the importance of variables and then recommending which ones to use. This not only saves time but determines risks using less factors but without sacrificing accuracy.
Using a massive data set, the scientists trained it to combine patient features, prediction labels, expert annotations of feature cost and a budget of total acceptable financial and non-financial costs. “We believe that in the near future, the number of clinical variables in the EHR and the number of clinical risk scores will continue to grow, making cost-aware risk scores like CoAI increasingly important,” Gabriel Erion, who is pursuing his M.D. as part of the UW's Medical Scientist Training Program, told HCB News.
CoAI takes into account cost shifts, when features become more or less expensive after the model is trained. It can adapt any predictive AI system to be cost-aware and make predictions at lower costs in various care settings. It also is generalizable in different medical specialties, such as cancer screening, where different factors have to be considered, including financial expenses.
With it, the scientists predicted increased bleeding risk in trauma patients heading to the hospital in ambulances and in-hospital mortality risk of critical care ICU patients. Data acquisition cost decreased by around 90% in trauma responses, and similar results were found among ICU patients.
“The mathematical framework applies to any situation where some predictive variables are more burdensome to gather than others,” said Erion. “One example that we discussed in the paper is outpatient clinics in which the goal may be to reduce the financial cost of the diagnosis (lab tests) rather than the time cost in minutes. We found that CoAI could substantially reduce the total cost of lab tests with little to no reduction in accuracy.”
He adds that CoAI could have a role in medical imaging “by selecting which images are the most informative and easiest to collect.”
Research was conducted by the Medical Scientist Training Program, the UW School of Medicine, and the UW’s Paul G. Allen School’s AIMS Lab.
The team discussed the solution in a paper published in Natural Biomedical Engineering.
Back to HCB News
The solution is called CoAI (Cost-Aware Artificial Intelligence) and predicts outcomes by calculating the importance of variables and then recommending which ones to use. This not only saves time but determines risks using less factors but without sacrificing accuracy.
Using a massive data set, the scientists trained it to combine patient features, prediction labels, expert annotations of feature cost and a budget of total acceptable financial and non-financial costs. “We believe that in the near future, the number of clinical variables in the EHR and the number of clinical risk scores will continue to grow, making cost-aware risk scores like CoAI increasingly important,” Gabriel Erion, who is pursuing his M.D. as part of the UW's Medical Scientist Training Program, told HCB News.
CoAI takes into account cost shifts, when features become more or less expensive after the model is trained. It can adapt any predictive AI system to be cost-aware and make predictions at lower costs in various care settings. It also is generalizable in different medical specialties, such as cancer screening, where different factors have to be considered, including financial expenses.
With it, the scientists predicted increased bleeding risk in trauma patients heading to the hospital in ambulances and in-hospital mortality risk of critical care ICU patients. Data acquisition cost decreased by around 90% in trauma responses, and similar results were found among ICU patients.
“The mathematical framework applies to any situation where some predictive variables are more burdensome to gather than others,” said Erion. “One example that we discussed in the paper is outpatient clinics in which the goal may be to reduce the financial cost of the diagnosis (lab tests) rather than the time cost in minutes. We found that CoAI could substantially reduce the total cost of lab tests with little to no reduction in accuracy.”
He adds that CoAI could have a role in medical imaging “by selecting which images are the most informative and easiest to collect.”
Research was conducted by the Medical Scientist Training Program, the UW School of Medicine, and the UW’s Paul G. Allen School’s AIMS Lab.
The team discussed the solution in a paper published in Natural Biomedical Engineering.
Back to HCB News
1(current)
You Must Be Logged In To Post A CommentRegistroRegistrarse es Gratis y Fácil. Disfruta de los beneficios del Mercado de Equipos Médicos Nuevos y Usados líder en el mundo. ¡Regístrate ahora! |
|
