偵測問卷調查受試者亂答填卷之研究：以醫師滿意度調查為例

Detecting Abnormalities of Questionnaire Surveys: Online Satisfaction Analyses onto Physicians

錢才瑋(Tsair-Wei Chien) ； 王文中(Wen-Chung Wang) ； 蘇世斌(Shih-Bin Su) ； 陳俞成(Yu-Cheng Chen)

《醫療資訊雜誌》 16卷3期 (2007/09) Pp. 71-82

https://doi.org/10.29829/TJTAMI.200709.0006

摘要

醫院新制評鑑要求醫院做各式樣的問卷調查，也規範要對醫師進行滿意度評估。本研究期以有別於傳統問卷調查的分析系統，提供統計科學與資訊系統結合的管理工具：病患於網路線上回饋滿意度調查、偵測出胡亂作答的受試者問卷、利用網路科技即時分析出調查的摘述結果、節省醫師時間以閱讀有效的問卷。利用試題反應理論的Rasch分析，首先對250位門診患者調查其23道對醫師滿意的題目，以李克特的5點計分，分析滿意度題目是否是研究同一個共同特徵的單向度（fit均方統計界於0.6及1.4間）。續而將估計參數置於網頁，供就醫後的患者填答，然後線上傳遞給醫師。以標準化Z分數及fit統計量檢測填卷者是否亂答，再利用KIDAMP顯示網路滿意度的答題反應，醫師即刻可獲知病患滿意的反應情形，有別於傳統問卷調查無法即時偵測出胡亂作答的受試者反應及其調查的分析結果。

關鍵字

Rasch分析 ； 試題反應理論 ； 李克特量尺 ； 滿意度調查 ； KIDAMP

並列摘要

The latest version of hospital accreditation rules stipulated that every hospital should conduct various kinds of questionnaire surveys to obtain necessary information for improving quality of service. The satisfaction survey onto physicians' service is also included in the regulation of hospital accreditation. Owing to the abnormalities like fumbling or careless responses from examinees, the invalid answers will tumble our explanations and interpretations for surveys. In this study, we firstly adopted the objective and interval characteristics of the Rasch model to analyze a satisfaction survey on physicians’ service in a hospital in southern Taiwan. The software Winsteps was used to examine whether the data fit Rasch model and match a single construct. Items with outfit or infit MNSQ (mean square error) not in the range between .06 and 1.4 are usually deemed as potential misfits to the Rasch model and considered for deletion from the assessment sequence. Successive Rasch analyses were performed until a final set of items was obtained that met the model's fit requirements. The results of those item calibrations were programmed into a web site to render surveys on internet, and obtain instant responses from patients. A KIDMAP performance diagnosis was designed to help physicians recognize which response endorsed by patients should be trusted or neglected in accordance with the Rasch fit statistics. The module developed in this study could meet the needs of users with functionality and user-friendliness in helping physicians save their time and know their customers' real responses.

並列關鍵字

Rasch analysis ； item response theory ； Likert-type scale ； satisfaction survey ； KIDMAP