修訂版學步期自閉症檢核表在不同場域篩檢自閉症類群障礙症幼兒的正確率

Detection of Autism Spectrum Disorder in Different Settings: Accuracy of the Modified Checklist for Autism in Toddlers

修訂版學步期自閉症檢核表（Modified Checklist for Autism in Toddlers, M-CHAT）為國際上常見的自閉症類群障礙症（autism spectrum disorder, ASD）幼兒篩檢工具。然而，M-CHAT在臺灣的研究不多。因此，本研究探究M-CHAT在臺灣的不同場域（如：基層診所與綜合醫院）篩檢ASD幼兒的正確率。18-29個月的幼兒共272名，包括：ASD幼兒74名、發展遲緩（developmental delay, DD）幼兒87名與一般發展（typically developing, TD）幼兒111名，由主要照顧者填寫23題「是／非」選項的M-CHAT。以卡方檢定（chi-squared test）比較ASD、DD與TD幼兒在M-CHAT題目失敗率的差異，單因子變異數分析（analysis of variance, ANOVA）檢驗三組幼兒在總分及不同簡版分數的差異。然後，以訊號偵測理論（signal detection theory, SDT）決定M-CHAT的切截分數，並以判別函數分析（discriminant function analysis, DFA）找出M-CHAT的結構矩陣，決定關鍵題目。在綜合醫院場域使用時，M-CHAT篩檢ASD與DD幼兒，以任4題失敗作為切截分數，敏感度與特異度皆為.72。使用不同簡版時的敏感度為.62－.72，特異度為.85－.91，簡版14題（Brief 14）篩檢率正確率最佳。在基層診所場域使用時，M-CHAT篩檢ASD與TD幼兒，以任3題失敗作為切截分數，敏感度為.81，特異度為.83。使用不同簡版時的敏感度為62－.72，特異度為.95－.99，一樣是簡版14題篩檢正確率最佳。M-CHAT用在綜合醫院場域篩檢ASD幼兒正確率尚可，用在基層診所場域篩檢ASD幼兒正確率良好。本研究結果支持M-CHAT可以在臺灣用於ASD幼兒的早期篩檢，特別是在基層診所場域。然而，在不同場域使用M-CHAT，需要使用不同的切截分數。

Purpose: The Modified Checklist for Autism in Toddlers (M-CHAT) is a wellresearched screening tool that was originally developed for detecting autism spectrum disorder (ASD) in toddlers. However, few studies have examined the accuracy of the M-CHAT for detecting ASD in the Taiwanese population. Thus, this study examined the accuracy of the M-CHAT for detecting ASD in Taiwanese toddlers in different settings (i.e., primary clinics and general hospitals). Methods: The M-CHAT comprises 23 questions with yes/no responses, and it is designed to detect ASD in toddlers aged 16-30 months. In this study, the M-CHAT was administered to 272 caregivers of toddlers aged 18-29 months. The 272 caregivers comprised those of 74 toddlers with ASD, 87 toddlers with developmental delay (DD), and 111 toddlers who were typically developing (TD). The chi-square test was used to examine the failure rate of each M-CHAT item among the ASD, DD, and TD groups. Analysis of variance was also used to examine the sensitivity and specificity of the total scores of the 23 M-CHAT items and the various short forms among the three groups of participants. In addition, signal detection theory was used to determine the optimal cutoffs of the M-CHAT, and discriminant function analysis was conducted to obtain the structure matrix for deciding the critical items. Results/Findings: When the M-CHAT was used for discriminating between the ASD and DD groups in general hospitals, sensitivity of .72 and specificity of .72 were obtained. The cutoff was the failure of any 4 of the 23 M-CHAT items. For the different short forms of the M-CHAT, sensitivities of .62- .72 and specificities of .85- .91 were obtained; “Brief 14” was the best short form of the M-CHAT. When the M-CHAT was used for discriminating between ASD and TD groups in primary clinics, sensitivity of .81 and specificity of .83 were obtained. The cutoff was the failure of any 3 of the 23 M-CHAT items. For different short forms of the M-CHAT, sensitivities of .62- .72 and specificities of .95- .99 were obtained; “Brief 14” was the best short form of the M-CHAT. Conclusions/Implications: The accuracy of the M-CHAT was adequate in general hospital settings and favorable in primary clinic settings. The study results suggest that the M-CHAT can be used as a screening tool for detecting ASD in toddlers in Taiwan, especially in primary clinics. However, different cutoffs in the form of different M-CHAT scores should be used depending on setting (primary clinic or general hospital).