论文部分内容阅读
目的基于贝叶斯定理建立常见呼吸道传染病的分类判别模型,为传染病暴发疫情调查和实验室检测提供病因线索。方法通过查阅文献、历史疫情数据和暴发疫情调查报告,收集常见传染病的症状、体征、实验室检测结果、流行病学特征及发病数据。基于朴素贝叶斯分类算法原理,采用SAS 9.1.3软件建立分类判别模型,并分别用2013—2015年浙江省发生的流行性感冒、流行性腮腺炎、水痘和麻疹各2起疫情数据对模型的判别效果进行验证。结果 8起疫情的第一位次判别概率最低为20.00%、最高为100.00%、中位数为53.85%,前三位次判别概率最低为55.00%、最高为100.00%、中位数为98.34%。第一位次判别的灵敏度中位数为53.85%,特异度中位数为100.00%,阳性似然比最小为5.73、最大趋向无穷大;前三位次判别的灵敏度中位数为98.34%,特异度中位数为82.14%,阳性似然比最小为1.26、最大趋向无穷大。结论贝叶斯分类判别模型适用于常见呼吸道传染病的分类判别,判别效果达到实际工作要求,能够提高呼吸道传染病暴发疫情病因的早期判别能力。
OBJECTIVE To establish a discriminant model of common respiratory infectious diseases based on Bayes’ theorem, and provide etiological clues for investigation and laboratory testing of outbreaks of infectious diseases. Methods Through the literature review, historical epidemic data and outbreak investigation reports, the common infectious diseases were collected, the symptoms, signs, laboratory test results, epidemiological characteristics and incidence data. Based on the naive Bayesian classification algorithm, SAS 9.1.3 software was used to establish the classification discriminant model. The epidemic data of two epidemics including epidemic influenza, mumps, chickenpox and measles in Zhejiang province from 2013 to 2015 were analyzed respectively. The discriminant effect is verified. Results The lowest discriminant probability of the first eight episodes was 20.00%, the highest was 100.00% and the median was 53.85%. The lowest discriminant probability of the first three was 55.00%, the highest was 100.00% and the median was 98.34% . The sensitivity of the first order discrimination was 53.85%, the median of specificity was 100.00%, the positive likelihood ratio was the lowest, 5.73, and the maximum trend was infinity. The sensitivity of the first three determinations was 98.34% The median was 82.14%, the positive likelihood ratio was the lowest at 1.26, and the maximum tended to infinity. Conclusion The Bayesian classification discriminant model is suitable for the classification and identification of common respiratory infectious diseases. The discriminant effect meets the actual working requirements and can improve the early discriminating ability of the cause of outbreak of respiratory infectious diseases.