Backgroundː
Post-induction hypotension (PIH) refers to arterial hypotension occurring within the first 20 minutes after anesthesia induction or from anesthesia induction to the beginning of surgery. Identifying high-risk patients with PIH is of great significance for medical staff to take corresponding preventive measures and formulating intervention plans. Therefore, this study aims to construct a PIH prediction model for patients undergoing general anesthesia (GA) and varify the performance of the model. It was hypothesized that we could create a prediction model with a sensitivity/specificity > 85%.
Methodsː
This is a cross-sectional, observational study performed in a tertiary hospital in southwest China, among 290 patients who underwent elective non-cardiac surgery under GA from March 2023 to May 2023. The data came from medical records and anesthesia information collection system. Variables included patient age, gender, heart rate (HR), body mass index (BMI), disease diagnosis, complications, drug use, Charlson comorbidity index (CCI), American society of anesthesiologists physical status classification (ASA), the last measured blood pressure (BP) in the ward, the BP before anesthesia induction, and the lowest BP during anesthesia induction. the lowest BP during anesthesia induction was measured by invasive measurement method. PIH was defined as a decrease of mean arterial blood pressure (MAP) during induction of more than 30% compared with the MAP measured before anesthesia induction. The data was divided into trainning set and validation set according to the ratio of 7:3. The least absolute shrinkage and selection operator (LASSO) binary logistic regression was used for feature selection and model training. The area under the receiver operating characteristic curve (AUROC) was used to test these hypotheses. A calibration curve and the Hosmer-Lemeshow (H-L) chi-square test were used to evaluate the calibration degree of the model. Decision curve analysis (DCA) was used to evaluate the performance of the modeling in supporting clinical decision-making. The model was then visualized using a nomogram.
Results
PIH was presented in 8% patients in the training set and 10% in the test set. The predictors of this model included BMI, changes in MAP, pre-operative HR, and pre-operative use of angiotensin-converting enzyme inhibitors (ACEIs)/angiotensin receptor blockers (ARBs). For the training and test sets, the AUROC using LASSO regression was 0.894 [95% CI, (0.78, 1.00)] and 0.883 [95% CI, (0.718, 1.00)], with respective sensitivity (0.880 and 0.901) and specificity ( 0.875 and 0.889). The H-L test of calibration curve was 3.42 and 11.265, with respective p value 0.905 and 0.187. The DCA demonstrated that using the model obtained higher net benefit (NB) than not using it. This model composed of these four independent variables showed good calibration, and clinical efficiency, which is helpful for medical staff to identify patients with high risk of PIH and formulate corresponding prevention and intervention strategies
Conclusions
BMI, MAP change, HR, and ACEIs/ARBs were predictive of PIH by LASSO regression. This model composed of these four independent variables showed good discrimination, calibration, and clinical efficiency, which is helpful for medical staff to identify patients with high risk of PIH and formulate corresponding prevention and intervention strategies. The prediction and validation model with a sensitivity/specificity > 85% means the model was “successful”.