使用Rasch分析協助醫師篩選勞工健檢異常值之應用研究

Using Rasch Method to Analyze Abnormal Results of Annual Labors' Health Examination

錢才瑋(Tsair-Wei Chien) ； 邱千芳(Chien-Fang Chiu) ； 王文中(Wen-Chung Wang) ； 蘇世斌(Shih-Bin Su)

《中華職業醫學雜誌》 14卷1期 (2007/01) Pp. 29-42

https://doi.org/10.30027/CJOM.200701.0004

摘要

In large scale annual labor health examination, we usually use traditional reference standards of laboratory test analyzed by computerized program to discriminate the abnormal results which may have a lot of random errors derived from difference between observed scores and true ability.
In this study we adopted the objective and interval characteristics of Rasch model to analyze the 11,830 labors' health examination data of a company with five factories in southern Taiwan. The Winsteps software was used to examine data whether fit Rasch model and match a single construct in order to (1)estimate health index abilities for examinees, (2)analyze differences within and between groups, (3)screen out special unusual examined results with Fit statistics to help health management systems setting up.
The results revealed that the 19 but one items fitted the model's expectation fairly well. The test for predicting kidney disease, cancer and abnormal chest X-ray threaten the female below 25 years old, predict mark of liver and gout diseases impact on the male, those age between 35 and 45 are suspected with high possibility of coronary heart disease, labors with age ranged from 25 to 45 have the potential liver disease, and diabetes potential affect employees of factory B. The principle of the older age and the less health index was evidenced in this study. Liver diseases threaten much to those labors who suffer less health index measured by Rasch analysis. The tri-loop map developed in this paper may help physicians to screen out 38.59% of examinees with potential health problems to be followed up.

關鍵字

勞工健檢 ； Rasch分析 ； 單一向度 ； DIF ； Fit統計

並列摘要

In large scale annual labor health examination, we usually use traditional reference standards of laboratory test analyzed by computerized program to discriminate the abnormal results which may have a lot of random errors derived from difference between observed scores and true ability.
In this study we adopted the objective and interval characteristics of Rasch model to analyze the 11,830 labors' health examination data of a company with five factories in southern Taiwan. The Winsteps software was used to examine data whether fit Rasch model and match a single construct in order to (1)estimate health index abilities for examinees, (2)analyze differences within and between groups, (3)screen out special unusual examined results with Fit statistics to help health management systems setting up.
The results revealed that the 19 but one items fitted the model's expectation fairly well. The test for predicting kidney disease, cancer and abnormal chest X-ray threaten the female below 25 years old, predict mark of liver and gout diseases impact on the male, those age between 35 and 45 are suspected with high possibility of coronary heart disease, labors with age ranged from 25 to 45 have the potential liver disease, and diabetes potential affect employees of factory B. The principle of the older age and the less health index was evidenced in this study. Liver diseases threaten much to those labors who suffer less health index measured by Rasch analysis. The tri-loop map developed in this paper may help physicians to screen out 38.59% of examinees with potential health problems to be followed up.

並列關鍵字

labor physical examination ； Rasch Analysis ； single construct ； DIF ； Fit statistics