Development and validation of a prediction model for in-hospital mortality among patients with acute diquat poisoning: A retrospective cohort study

建立和验证急性敌草快中毒患者院内死亡率预测模型:一项回顾性队列研究

阅读:2

Abstract

Diquat has become the leading bipyridine herbicide, surpassing paraquat and causing more poisoning cases, yet knowledge about it is limited. Our study sought to create and validate a nomogram to predict in-hospital mortality in patients with acute diquat poisoning. This retrospective cohort study, conducted from January 2016 to April 2024 in a tertiary hospital, identified prognostic factors using best subsets regression with the lowest Bayesian information criterion. A nomogram based on 5 prognostic factors was validated through bootstrap sampling validation and leave-one-out cross-validation. The model's effectiveness was assessed using a decision analysis curve. Of the 98 acute acute diquat poisoning cases included, 58.2% were male, with a 34.7% in-hospital mortality rate. Five prognostic factors (post-ingestion time, plasma diquat concentration, plasma lactate concentration, occurrence of respiratory failure and need for blood transfusion) were used to create a predictive nomogram with area under the curve was 0.874 (95% confidence interval: 0.806-0.941) and concordance index of 0.874. For the bootstrap sampling validation, the corrected concordance index was 0.846. In the leave-one-out cross-validation, the concordance index was 0.849. The Hosmer-Lemeshow test (χ2 = 8.125, P = .421) and brier score (original 0.142, mean 0.167 in cross-validation) supported this. Clinical decision analysis indicated patient benefit at threshold probabilities of 1% to 76%. We developed a predictive model and nomogram to forecast in-hospital mortality in acute diquat poisoning patients using baseline characteristics. This model performs well and can help clinicians identify patients at high risk of in-hospital mortality.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。