背景與目的：現階段國內常使用直接轉譯國外量表之方式來評估，在文化差異下，量表直譯後是
否仍適用並未加以探討，本研究目的在於探討國內臨床上使用直譯國外肩功能適性量表（FLEXSF）為療效評估工具之合適性，並就題目難易度順序來討論兩者之間差異。方法：本研究先將FLEX-SF量表中文化後共取得32道題目，再由227名肩部功能失調且接受物理治療之病患完成作答。以電腦軟體統計方法WINSTEPS （Mesa Press, Chicago, IL, USA）排序每道題目難易程度，以項目分離信度（item separation reliability）檢視項目間之一致性；以合適度統計量（fit statistic）來檢測題目題意之反應型式是否符合羅序模式要求。結果：在心理計量特性檢測方面，再測信度係數ICC為0.97，Cronbach’s α係數分別為0.95、0.93；與台灣版SF-36問卷之生理成分分數之間相關係數為 rho=0.72（p<0.01）；與心理成分分數之間的相關係數為rho=0.35（p=0.08）。第28道題目因近合適度係數大於1.4，表示該題可能具有偏差，在刪減該題後發現項目分離度（7.36→ 7.18）與項目分離信度（0.98）並無太大影響。在難易程度與原版比較，發現中文化之難易順序有明顯不同。結論：東西方文化上認知與生活習俗所存在之差異，對於具有線性、連續關係之IRT量表亦會產生影響，為了適切評估台灣地區肩部功能失調患者之現況，對中文化FLEX-SF量表做適度調整確實有其必要。（物理治療 2009;34(3):177-184）

Background: In language and cultural differences, the question is not explored for the appropriate in Taiwan by using the directly-translated foreign questionnaire to evaluate. The purposes of this study are to discuss the appropriate for using the FLEX-SF Chinese format by direct translation in Taiwan and the deference of items difficulty. Method: Chinese Item pool (CFLEX-SF) was collected by the objective, standardizing translation process for FLEX-SF. The 227 subjects with shoulder dysfunction who underwent physical therapy completed the CFLEX-SF. To arrange the difficult degree of Items and analysis item calibration difficulty and fit statistic of each item, data was assessed by Rasch-Model computer program WINSTEPS (Mesa Press, Chicago, IL, USA). Internal consistency was assessed using item separation reliability which likes Cronbach alpha (α) coefficient. The Infit and Outfit statistics were used to assess whether item response pattern fits the Rasch model. Results: The internal consistency is adequate at both times with Cronbach’s alpha coefficient at 0.95 and 0.93, test-retest reliability is adequate with intraclass correlation coefficient (ICC) at 0.97, correlation coefficient were compared with Physical Component Scale (PCS) and Mental Component Scale (MCS) of the Taiwan’s SF-36 questionnaire are rho=0.72 (p<0.01) and rho=0.35 (p=0.08) respectively. Item 28 has Infit statistics outside this range (>1.4), indicating that the item maybe have misfit. This item has greater variation in its response pattern which does not match the subject’s ability. However, item separation (7.36 →7.18) and Item separation reliability (0.98) had not remarkable change through the removal of poorly fitting item. There are obvious differential to compare the item difficulty sequence with FLEX-SF. Conclusion: The discrepancy between cultural perception and onventionality are existence, can influence the linear, continuous characteristic of IRT questionnaire. In order to adequately assess the condition of Taiwan’s patients with shoulder dysfunction, it’s necessary to modify appropriately for the CFLEX-SF. (FJPT 2009;34(3):177-184)