Objective: To develop predictive models that can distinguish acute strangulation from simple intestinal obstruction by multivariate analysis. Methods: All the 571 patients involved in the study were the cases admitted...Objective: To develop predictive models that can distinguish acute strangulation from simple intestinal obstruction by multivariate analysis. Methods: All the 571 patients involved in the study were the cases admitted with acute mechanical intestinal obstruction in the departments of general surgery of three hospitals affiliated to Beijing Medical University in the past 7 to 10 years. Logistic regression model was developed and an algorithm for the classification of intestinal obstruction was constructed by classification and regression trees (CART) analysis. These two models were tested by the method of cross validation and compared with the surgeons’ experience. Results: The pattern of abdominal pain, tender mass, ascites, abdominal rigidity, history of abdominal operation and history of similar attack were important variables to develop the models; Sensitivity(SEN) of logistic model was 71.8%, specificity(SPE) was 73.1%, and SEN of CART model was 73.1% , SPE was 69.8% . SEN of the surgeons judgement was only 45.2%. SEN of both models were much higher than the one of surgeons experience and intuition. Conclusion: Logistic model and CART model are the useful methods to differentiate strangulated intestinal obstruction(SIO) from the simple obstruction and can be applied in the prediction for SIO. (J Beijing Med Univ, 2000,32:183 186)展开更多
文摘Objective: To develop predictive models that can distinguish acute strangulation from simple intestinal obstruction by multivariate analysis. Methods: All the 571 patients involved in the study were the cases admitted with acute mechanical intestinal obstruction in the departments of general surgery of three hospitals affiliated to Beijing Medical University in the past 7 to 10 years. Logistic regression model was developed and an algorithm for the classification of intestinal obstruction was constructed by classification and regression trees (CART) analysis. These two models were tested by the method of cross validation and compared with the surgeons’ experience. Results: The pattern of abdominal pain, tender mass, ascites, abdominal rigidity, history of abdominal operation and history of similar attack were important variables to develop the models; Sensitivity(SEN) of logistic model was 71.8%, specificity(SPE) was 73.1%, and SEN of CART model was 73.1% , SPE was 69.8% . SEN of the surgeons judgement was only 45.2%. SEN of both models were much higher than the one of surgeons experience and intuition. Conclusion: Logistic model and CART model are the useful methods to differentiate strangulated intestinal obstruction(SIO) from the simple obstruction and can be applied in the prediction for SIO. (J Beijing Med Univ, 2000,32:183 186)