摘要：在台灣使用「台北市學前兒童發展檢核表第二版」(簡稱TaipeiⅡ)以早期發現遲緩兒童已有多年，然其預測效度尚未知。為了解TaipeiⅡ四至六歲分量表對入小學後適應行為之預測效度，乃進行此研究。方法：檢視過去建立TaipeiⅡ之研究資料庫，選擇符合：(1)具有四至六歲分量表；(2)居住台北市或新北市；(3)2007 年初就讀小學1-3 年級者，以姓名、生日、電話與學區資料及特殊教育通報網比對，追蹤其是否特教通報學生，並以文蘭適應量表對非特教通報學生測試，以了解其有無適應行為問題，共得310名。就學適應狀況之預測效標有兩個標準，標準甲為一般班學生文蘭適應量表總量表百分等級≦16 或已列入特教通報學生(即包括確定與疑似有適應問題者)；標準乙為一般班學生文蘭量表總量表百分等級＜5或已列入特教通報學生(即確定有適應問題者)。以接受者操作特徵(ROC)曲線與交叉表分析：(1)篩選預測正確機率；(2)TaipeiⅡ正總分與「文蘭適應行為量表」標準分數之相關係數；(3)切截策略B 篩檢表現與就學適應程度之預測效度指標值；(4)多層次概率比。結果：篩選預測正確機率(ROC 曲線下面積)為0.82-0.95 (p＜0.001)；以TaipeiⅡ切截策略B預測就學適應，陽性概率比為5.1-7.5，陰性概率比為0.08-0.48；於多切截點之多層次陽性概率比皆大於20，具高度預測診斷影響力；與「文蘭適應行為量表」標準分數顯著相關(rs=0.50-0.57)。結論：TaipeiⅡ四至六歲分量表具可接受之預測效度，多切截點之多層次概率比可協助臨床者進行二階段篩檢之決策。

Abstract: The use of Taipei Preschooler Developmental Checklist 2nd version (TaipeiⅡ) for early detection of developmental delay for children has been conducted for many years in Taiwan. However, the information of its predictive validity was unknown. Purpose: to investigate the predictive validity of the TaipeiⅡ on the adaptive behavior problem at school age. Methods: The original dataset for developing the TaipeiⅡ was investigated. Children who had been screened with one subscale of the TaipeiⅡ at 4-, 5-, or 6-year age, entered the primary school (grade 1-3) at the beginning of 2007 and living in Taipei area were the target population. We successfully followed up 310 children. The name, birth date, telephone of the target children were used to match the dataset of the Special Education Information Network to obtain the information of special educationneeds. For those who did not use special education resources, the Vineland Adaptive Behavior Scale (VABS) was used to define the problem of adaptive behaviors. Two predictive criteria were used to analyze the predictive validities of the TaipeiⅡ. Criteria I was students under special education or regular class students with percentile rank of VABS less than 16 (including both suspect and definite adaptive problems) and CriteriaⅡ was students under special education or regular class students with percentile rank of VABS less than 5 (definite adaptive problems only). The predictive validity indices included: the area under the receiver operating characteristic curve (AUC), sensitivities, specificities, multilevel likelihood ratios and the correlation coefficients between TaipeiⅡ scores and VABS total scores. Results: The predictive validities of the TaipeiⅡ on school adaptive problem were acceptable with AUC＝0.82-0.95 (p＜0.001). The positive likelihood ratios of TaipeiⅡ with multiple cutoff strategies were greater than 20 indicating high diagnostic impact. Using cut-off strategy B (failure item numbers of 2 or 1 failure star item) to predict adaptive problems at school age, positive likelihood ratios were 5.1-7.5 and negative likelihood ratios of 0.08-0.48. The TaipeiⅡ scores were significantly correlated with the VABS standard scores (rs＝0.50-0.57). Our result indicate that the predictive validities of TaipeiⅡ subscales for children aged 4-6 year were acceptable. The multiple likelihood ratios of multiple cutoff points could help clinicians make clinical decision more precisely.