Development and validation of a predictive model for delayed neurological sequelae in acute carbon mon-oxide poisoning
10.3969/j.issn.1006-5725.2025.10.015
- VernacularTitle:急性一氧化碳中毒迟发性神经后遗症预测模型构建与效能验证
- Author:
Shaolin LI
1
;
Xiaohong MA
;
Dehe ZHANG
;
Peng SONG
Author Information
1. 新乡市中心医院(新乡医学院第四临床学院)康复医学科(河南 新乡 453000)
- Publication Type:Journal Article
- Keywords:
acute carbon monoxide poisoning;
delayed neurological sequelae;
predictive model;
no-mogram
- From:
The Journal of Practical Medicine
2025;41(10):1533-1539
- CountryChina
- Language:Chinese
-
Abstract:
Objective To construct a predictive model for delayed neurological sequelae(DNS)following acute carbon monoxide poisoning(ACMP)and to verify its efficacy.Methods A retrospective analysis of the gen-eral data of 183 patients with ACMP was conducted.The factors influencing the occurrence of DNS were analyzed using a multivariate Logistic regression model.A corresponding predictive model was then established and its efficacy was verified.Results The multivariate logistic regression model showed that age,smoking history,severe poisoning,blood lactate,time from poisoning to hyperbaric oxygen therapy,and pulmonary infection were independent risk factors for DNS following ACMP(P<0.05).The area under the curve(AUC)of the model for predicting DNS in the development set was 0.933,with a sensitivity of 94.12%and specificity of 89.77%.In the validation set,the AUC was 0.906,with a sensitivity of 90.00%and specificity of 92.68%.The Hosmer-Lemeshow test showed that the predicted probabilities of DNS in both the development and validation sets were not significantly different from the actual probabilities(P>0.05).The predictive model achieved clinical net benefit within the risk threshold ranges of 0.11～0.98 for the development set and 0.12～0.92 for the validation set.Conclusions Age,smoking history,severe poisoning,blood lactate,time from poisoning to hyperbaric oxygen therapy,and pulmonary infec-tion are independent risk factors for DNS following ACMP.The corresponding predictive model has been verified to have good clinical efficacy.
