Case classification in measles surveillance system under the Two-level Logistic Model
10.3760/cma.j.issn.0254-6450.2014.01.014
- VernacularTitle:利用两水平logistic模型探讨麻疹监测系统病例分类
- Author:
Handong LI
1
;
Rui AO
;
Lin PENG
;
Di LI
;
Juying ZHANG
Author Information
1. 四川大学华西公共卫生学院卫生统计学教研室
- Keywords:
Measles;
Surveillance;
Two-level logistic regression;
ROC curve
- From:
Chinese Journal of Epidemiology
2014;35(1):57-60
- CountryChina
- Language:Chinese
-
Abstract:
Objective To study the prevalence rates of Rash and Febrile Illnesses (RFIs) including measles,rubella,scarlet fever,exanthema subitum and the differences among measles and other RFIs to tentatively formulate the logistic regression model through clinical manifestation.Methods All the suspected cases of measles,rubella,scarlet fever,exanthema subitum reported by the county/prefecture lever hospitals at four counties were collected during March 2011 to February 2012.When setting laboratory confirmed measles as dependent variable and existed symptoms as independent variable,a logistic regression model was formulated and optimal operational point (OOP) chosen,according to the ROC curve.Results A total number of 551 cases were collected but the consistency of measles diagnosis between clinical and laboratory was not satisfied,with Kappa value=0.349,same to the diagnosis of rubella.As for the result from the two-lever logistic regression model,symptoms that related to the confirmation of measles would include cough (OR=5.75),conjunctivitis (OR=3.00),Koplik spot (OR =7.52),lymphadenectasis (OR =0.07),rash after fever (OR=0.07).The area under ROC curve was 0.97 and the optimal operational point was 0.249.Conclusion A logistic regression model was formulated using the clinical symptoms which was resulted in better performance on prediction.As the sample size of this survey was small,the expansion on the scale of investigation and laboratory testings were needed before the types and components of measles-related RFIs be clarified.
