檳榔依賴量表（BNDS）的發展

Development of the Betel Nut Dependency Scale (BNDS)

鑑於過去檳榔依賴量表的編製在量化程序上的不足，本研究參考過去檳榔量表的試題，編製一份符合心理計量程序的檳榔依賴量表，並設定決斷分數以利臨床上快速篩檢可能的成癮者，進而達到減少罹患口腔癌的機率。樣本透過人力仲介公司經歷近一年五個月的招募，總共獲得257位目前仍在嚼食檳榔的社會人士作為正式樣本，其中，將前六個月招募而來的128人視為預試樣本。試題的來源共39題，乃收集過去國內2份研究檳榔依賴行為的報告以及參考國外知名的相關成癮量表而來，經專家效度的檢驗後，以37題進行後續的量化分析。以預試樣本資料進行探索性因素分析，採用主軸因素萃取法及斜交轉軸方式，在設定3項刪題標準後，獲得11題的3因素結構，分別為「渴求欲望」4題、「戒斷反應」4題、「使用習慣」3題，三個因素的特徵值共解釋63.10%的總變異，內部一致性α信度在 .73 ~ .89之間。接著，以正式樣本使用結構方程式模型進行一階三因素二階一因素模型的驗證性因素分析，在多數適配指標都良好的情況下，獲得本量表具有建構效度的訊息。本量表並提出效標關聯效度等效度訊息供參考。最後使用ROC（receiver operating characteristic）曲線法來探討可能的決斷分數，並以試題反應理論最大訊息量的概念決定出一個最穩定的決斷分數24分，作為界定檳榔依賴者的標準。

On behave of the limited development in betel nut dependence scale with adequate psychometric properties, the current study aimed to establish a clinical-relevant quick screening tool on betel nut dependence. The formal participants included 257 betel nut chewers, recruited within one year and five months by human resource agency. Amongst them, the 128 participants recruited in the primary 6 months period were regarded as the preliminary sample. There were 39 items adopted, referencing 2 domestic studies on betel nut dependence and a widely-used smoking addicton scale (the Questionnaire on Smoking Urges-Brief, QSU-Brief); 37 items were selected for quantified analyses after expert validity analysis. Data from the preliminary sample were investigated with exploratory factor analysis by principal axis factoring extraction and direct oblimin rotation method. Based on the 3 item-deletion standards that we set up, we obtained 11 items and 3-factors structure in the scale: craving and desire (4 items), withdrawal response (4 items) and tasting habits (3 items). The eigenvalues of the 3 factors accounted for 63.10% of total variances, and α coefficients of reliability were between .73 ~ .89. Then, data from formal sample were modeled with 3 first-order factors and 1 second-order factor by confirmatory factor analysis using structural equation modeling. Most of model-fit indexes showed good fitting results, suggesting optimal construct validity of the scale. The scale also offered criterion-related validity for reference. Finally, we used ROC (receiver operating characteristic) curve to explore possible cutoff scores. Based on the maximum information method of item response theory, we recommend the cut-off score of 24 to be the most stable approximation to determine potentials of betel nut dependence.