Acute Gastrointestinal Injury Grading Models In Critically Ill Patients On The Basis Of Gastrointestinal Sound Monitoring

Object:To survey the situation of the acute gastrointestinal injury (AGI) in critically ill patients, explore the influential factors of the AGI, and monitor and analyze the gastrointestinal sounds (GIS) in critically ill patients, finally developed the AGI grades instrument of critically ill patients, which can help medical staffs better understand the situation of their gastrointestinal functions and make optimizing planning to improve them.Method:1.A prospective survey was carried out to understand the AGI of 509 critically ill patients by purposive sampling in ICU of 8 tertiary hospitals of Beijing.The self-made questionnaire according to the guidelines provided by the ESICM was used as the instrument. Statistical analyses including Kruskal-Wallis test, SNK test,x2 test and generalized estimating equation were performed to explore the influential factors of AGI of critically ill patients.2. Based on the former study in the GIS monitor method, the binary channel GIS monitor system was used to gather and analyze the GIS of 60 critically ill patients. Meanwhile, the AGI grades were evaluated. Statistical analyses were performed to explore the characteristics of the GIS of critically ill patients. The primary AGI grades models were separately developed by generalized estimating equation and back propagation artificial neural network.3. The possible clinical influential factors of AGI in critically ill patients were gathered. Integrated with the above primary AGI grades models, more comprehensive AGI grades models were developed by generalized estimating equation and back propagation artificial neural network.Results:1. The result of the prospective survey revealed that the incidence of AGI was high in critically ill patients. The patients of AGI grade I accounted for 54.18% of the total, patients above AGI grade II accounted for 45.82%.The most common gastrointestinal symptoms were the decrease of GIS and the absence of GIS. Early enteral feeding was started in 38.9% of patients. The generalized estimating equation showed that the ICU days, laxative drugs, mechanical ventilation and shock were risk factors to the AGI grade, enteral feeding was a protective factor to the AGI grade.2. There was significant correlate in the parameters of GIS of two channels. Also, the parameters were correlated with the time. And the number, percentage of time, mean power, maximum power and maximum time of GIS wave in the channel located at the stomach were negative correlated with the AGI grades. The generalized estimating equation showed that the first principal component of the GIS and time were significantly associated with the AGI grades. The accuracy rate of the primary generalized estimating equation model to classify acute gastrointestinal injury grades was 64.1%. The back propagation artificial neural network analyzed the five parameters of GIS and proposed an AGI grades network model including 9 hide layers, with an accuracy rate of 70.83%.3. The generalized estimating equation entered the sedatives, catecholamines, enteral feeding, SOFA scores, GCS scores, the first principal component of the GIS and time as the dependent variables by stepwise method. The accuracy rate of the generalized estimating equation model to classify acute gastrointestinal injury grades was 86%.The back propagation artificial neural network analyzed the five parameters of GIS, lactic acid, SOFA scores, catecholamines and mechanical ventilation, took the first 5 principal components of the principal component analysis, finally proposed a AGI grades network model including 14 hide layers, with a accuracy rate of 82.61%.Conclusion:The incidence of AGI was high in critically ill patients. The ICU days, laxative drugs, mechanical ventilation and shock were risk factors to the AGI grades, enteral feeding was a protective factor to AGI grades. The digitized GIS monitor system was proposed to quantificationally evaluate their grades of critically ill patients. The primary model to classify AGI grades with the GIS,and the comprehensive model to classify AGI grades with the GIS and clinical factors were developed. The models could help improving the accuracy of the classification of AGI grades and better evaluating the AGI of critically ill patients.