We established a retrospective cohort of 697 HIE patients from March 2010 to March 2020 and a prospective cohort of 107 from June 2021 to June 2023. The inclusion criteria were: 1) age ≥18 years, 2) EEG and neuron specific enolase (NSE) levels, and 3) clear medical record, including prognosis at discharge. The primary outcome was mortality at discharge. The clinical explanatory variables included suppression ratio (SR), NSE, age and highly malignant EEG patterns. SR was calculated with pre-defined algorithm installed in an EEG machine. A multivariate logistic regression with a nomogram was analyzed, and a C5.0 machine learning algorithm was used to verify the explanatory variables involved in the model.
A total of 257 patients satisfied the inclusion criteria, with 106 of these enrolled from a prospective cohort. Age, NSE and the SR were significant explanatory variables for mortality prediction. A nomogram was constructed using logistic regression to aid in decision-making regarding the continuation of life-supporting therapy in real-world situations.
The research found that the nomogram, employing a logistic regression model that includes SR, age, and NSE, can effectively predict mortality. The findings indicate that it is prudent to avoid discontinuing life-sustaining treatment in patients with a lower probability of mortality, especially when mortality predictions may lead to a self-fulfilling prophecy. Our study lacks imaging data, necessitating future research that includes quantitative or semi-quantitative analysis using imaging data.