AI Model to Predict Latent Shock in Emergency Department
Author Information
Author(s): Wu Mingzheng, Li Shaoping, Yu Haibo, Jiang Cheng, Dai Shuai, Jiang Shan, Zhao Yan
Primary Institution: Zhongnan Hospital of Wuhan University
Hypothesis
Can an ICU-based noninvasive model effectively predict latent shock risk in the emergency department?
Conclusion
The adult noninvasive model can effectively predict latent shock occurrence in EDs, which is better than using shock index and nSBP.
Supporting Evidence
- More than 80% of latent shock patients were identified more than 70 minutes earlier using the noninvasive model.
- The area under the receiver operating characteristic curve (AUROC) of the noninvasive model was 0.90.
- The model outperformed traditional methods like shock index and nSBP.
- Statistically significant differences were observed in the AUC per 10 minutes of external validation.
Takeaway
Doctors can use a special computer model to find out if patients might have a serious condition called latent shock, helping them treat patients faster.
Methodology
Multiple regression analysis was used to compare datasets and develop a noninvasive model based on MIMIC-IV-ICU data, validated in ED populations.
Potential Biases
Imbalance in sample sizes between stable and unstable groups may lead to prediction biases.
Limitations
The model may not include all relevant features and requires further optimization and validation across different patient subgroups.
Participant Demographics
Patients aged 18 and older, with a total of 50,636 from MIMIC-IV-ICU and 2,142 from ICCA-ED.
Statistical Information
P-Value
p<0.05
Confidence Interval
95% CI: 0.84–0.96
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website