【摘要】 目的鉴于医院获得性肺炎(NP)的高发病率及危害性,本研究对太原地区NP的流行趋势进行了评价和预测。方法收集太原地区十二年NP发病资料,以NP发生的危险因素:年龄≥60岁、基础疾病、联合使用两种以上的抗生素、入住ICU、住院时间≥21天、有创性机械通气或雾化吸氧为独立的6个危险因素作为网络输入,发病率为网络输出,构建BP人工神经网络模型。结果在评价基础上,对2006年12个月的NP发病率进行了预测,与实际发病率比较,符合率较高,预测误差较小,同时对2007年1~3月的发病率进行了预测。结论NP的发生受到很多因素的影响,各因素之间又存在着十分复杂的相互作用与不确定性,而BP人工神经网络有许多优点可以解决这些变量间关系不能精确地用函数表达的复杂问题。实践证明,该模型是可行的,同时有其普遍性,可推广到卫生领域其他疾病的评价和预测体系中。更多还原

【Abstract】 Objective Owing to the high morbidity and harmfulness of NP,this research has been appraised and predicted to the epidemic trend of NP in Taiyuan area.Methods Collecting case materials of NP for 12 years in Taiyuan area,the independence 6 risk factors of NP were as follows:Age is older than or equal to 60 years old,underlying disease,using more than two kinds of antibiotics,the therapy intensive care unit,more than 21 days in hospital stays,the use of cascade humidification system.The BP neural network model was built with the risk factors as the input and the NP incidence of each month as the output.Results On the basis of appraising,the NP morbidity of 12 months of 2006 had been predicted,compared with the actual morbidity,the coincidence rate was relatively high, error predicted was relatively little,the morbidity in January-March of 2007 had been predicted at the same time.Conclusion The emergence of NP is influenced by a lot of factors,there are quite complicated interaction and uncertainty among these factors,BP artificial neural network has a lot of advantage can solve the complicated problems with the relationship between the variables that function cannot express accurately.Practice has proved,this model is feasible and has its universality at the same time,it can be popularized to the appraisal of other diseases of domain of hygiene and in the prediction system.更多还原