In:
Circulation, Ovid Technologies (Wolters Kluwer Health), Vol. 146, No. Suppl_1 ( 2022-11-08)
Abstract:
Background: Correct identification of futile prehospital resuscitation in out-of-hospital cardiac arrest (OHCA) may reduce unnecessary transports. Reliable prediction variables are needed for 'termination of resuscitation' (TOR) rules in OHCA. However, making such prediction models requires a large number of samples and time-consuming statistical analysis. Deep learning (DL), a form of artificial intelligence (AI), is attractive for building such prediction models. We made a TOR prediction model using a DL software, Prediction One (Sony Network Communications Inc., Tokyo, Japan). The purpose of this study was to evaluate the efficacy of this predictive model using general purpose AI compared to the Universal TOR rule for out - of hospital cardiac arrest. Methods: A retrospective, population-based review of OHCA victims without prehospital ROSC from January 1, 2010 to December 31, 2017 in the All-Japan Utstein Registry. We divided half (2010-2014) as a training dataset and half (2015-2017) as an external validation dataset. Prediction one made the prediction model using the training dataset with internal cross-validation. We used both the created model and the Universal TOR rule to predict outcomes using the external validation set. Results: 989,929 OHCA cases, 18 years of age or older, were registered in the All-Japan Utstein Registry and 575,346 cases were of presumed cardiac origin. Of these, 354,356 cases were used as the training dataset and 220,990 cases were used as the external validation dataset. The model made by Prediction One using 11 variables had AUC of 0.969 and F value of 0.969, and its AUC for the validation cohort was 0.964 (95%CI 0.963-0.966) with 99.3% accuracy. The model made by Prediction One using three variables of universal TOR rule had AUC of 0.948 and F value 0.967, and its AUC for the validation cohort was 0.939 (95%CI 0.936-0.942) with 99.1% accuracy. Conclusions: The accuracy of prediction models using Prediction One was good. More research into the utility of using AI for TOR prediction is required.
Type of Medium:
Online Resource
ISSN:
0009-7322
,
1524-4539
DOI:
10.1161/circ.146.suppl_1.206
Language:
English
Publisher:
Ovid Technologies (Wolters Kluwer Health)
Publication Date:
2022
detail.hit.zdb_id:
1466401-X
Permalink