ABSTRACT
Objective:To establish a risk prediction model for the occurrence of low 1 min Apgar scores in extremely premature infants (EPIs).Methods:From January 2017 to December 2021, EPIs delivered at our hospital were retrospectively analyzed and randomly assigned into training set group and validation set group in a 7∶3 ratio. 17 clinical indicators were selected as predictive variables and low Apgar scores after birth as outcome variables. Lasso regression and multi-factor logistic regression were used within the training set group to select the final predictors for the final model, and the calibration, distinguishability and clinical decision making curves of the final model were evaluated in the validation set group.Results:A total of 169 EPIs were enrolled, including 117 in the training set group and 52 in the validation set group. 4 indicators including gender, fetal distress, assisted conception and delivery time were selected as the final predictors in the final model. Both the training set group and the validation set group had good calibration curves. The area under the receiver operating characteristic curve (AUC) of the prediction model was 0.731, the sensitivity was 72.2%, the specificity was 60.5% and the AUC of the external validation curve was 0.704. The clinical decision making curve showed that the model had a greater benefit in predicting the occurrence of low Apgar score in EPIs within the threshold of 2% to 75%.Conclusions:The clinical prediction model established in this study has good distinguishability, calibration and clinical accessibility and can be used as a reference tool to predict low Apgar scores in EPIs.