發展適用於思覺失調症患者之電腦化情緒辨識影音測驗

背景：情緒辨識 (emotional recognition) 是人們判讀他人情緒之能力，通常包含快樂、悲傷、生氣、厭惡、害怕、驚訝、輕蔑，以及平靜/無情緒等8種情緒類別之辨識。思覺失調症患者普遍罹患情緒辨識損傷，影響其人際互動以及社會功能。目前常用之情緒辨識測驗因評估向度不完整、缺乏個別情緒類別指標、心理計量特性大多未知與缺乏華人刺激物等限制，難以提供完整、精準且適用於國人之情緒辨識評估。目的：發展適用於思覺失調症患者之電腦化情緒辨識影音測驗 (Computerized Emotional Recognition Video Test, CERVT)，並驗證其心理計量特性，包含建構效度、羅序–受測者信度、再測信度、隨機測量誤差、練習效應，以及已知族群效度。方法：本研究分為二階段：（一）以4步驟發展CERVT：(1) 建立初版測驗：自「台灣地區華人情緒與相關心理生理資料庫—基本情緒聲調」擇取包含8種情緒類別之影音刺激物，以組成初版測驗；(2) 臨床施測初版測驗：施測於思覺失調症患者及健康成人；(3) 分析並決定CERVT最終版測驗（含驗證建構效度、信度及已知族群效度）：針對初版測驗分析多向度羅序模型之適配度，挑選測量特性良好之題目作為最終版測驗；(4) 發展電腦施測介面。（二）驗證CERVT之心理計量特性：患者接受2次測驗（間隔2週），以檢驗CERVT之再測信度、隨機測量誤差以及練習效應。研究結果：第一階段中，初版測驗共238題（8種情緒），並施測於446位受測者（含269位患者及177位健康成人）。刪除21題適配度不佳之題目後，剩餘217題之適配度良好 (infit and outfit mean square = 0.53–1.36)，且8種情緒向度之第一個殘差主成份特徵值小 (≤ 3.0)，支持個別單向度之假設。其中26題具嚴重性別差異試題功能 (differential item functioning, DIF)，為控制其性別偏差，故於施測這些題目時會依據受測者之性別給予不同分數。各向度皆具有良好至極佳的信度（平均信度 = 0.85–0.93），並有良好之已知族群效度 (Cohen’s d = -1.41–-0.75)。第二階段共75位患者參與，結果顯示CERVT各向度具可接受至良好的再測信度 (intraclass correlation coefficient = 0.66–0.81) 與稍高的隨機測量誤差 (percentage of standard error of measurement = 11.0%–19.2%)。結論：CERVT可評估8種情緒辨識能力，且有可接受之信效度，具備潛力可應用於臨床與研究，以提供完整、準確且有效之情緒辨識評估。

關鍵字

思覺失調症；情緒辨識；電腦化測驗；心理計量特性

並列摘要

Background: Emotional recognition (ER) is an individual’s ability to identify others’ emotional status, which conceptually includes the recognitions of 8 basic emotions: happiness, sadness, anger, disgust, fear, surprise, contempt, and calm. Patients with schizophrenia tend to have deficits of ER that affect their interpersonal interaction and social function. However, the commonly-used ER measures have 4 problems (i.e., lack of items in Chinese, lack of individual scores of the 8 different emotions, unknown/poor psychometric properties, and incomprehensive item bank), which make it difficult to comprehensively evaluate ER. Purposes: To develop a computerized emotional recognition video test (CERVT), and validate its construct validity, Rasch reliabilities, test-retest reliability, random measurement error, practice effect, and known-groups validity. Methods: This study contained two phases. In the first phase, the CERVT was developed through 4 steps. (1) Selecting candidate items for the CERVT. The candidate items, which are the videos of performers across 8 emotions, were selected from a published database. (2) Testing of the candidate items. The selected items were administered on patients with schizophrenia and healthy adults for validations. (3) Determining the final items of the CERVT. Rasch model fits of the CERVT items were used to determine the final CERVT items and the unidimensionality (one kind of construct validity) of these items was examined. (4) Constructing the assessment system of the CERVT. In the second phase, we examined the other psychometric properties of the CERVT in patients with schizophrenia who completed the CERVT twice within a 2-week period. Results: In phase 1, 238 items were selected as the candidate items. These items were tested on 446 participants (including 269 patients and 177 healthy adults). After the removal of 21 misfit items and adjusting item difficulties for the 26 items with differential item functioning (DIF), the remaining 217 items were determined as the final CERVT items. All the CERVT items showed good model fits (infit and outfit mean square = 0.53–1.36) with small eigenvalues (< 3.0) based on the residual-based principal components analysis for each domain, so the unidimensionality of these items was held. The 8 domains of the CERVT had good to excellent Rasch reliabilities (0.85–0.93) and satisfactory known-groups validity (Cohen’s d = -1.41–-0.75). In phase 2, the 8 domains of the CERVT showed acceptable to good test-retest reliability (intraclass correlation coefficient = 0.66–0.81) and somewhat high random measurement error (percentage of standard measurement error = 11.0%–19.2%). Conclusions: The CERVT contains items of the 8 basic emotions with an individualized index for each score. Moreover, the CERVT showed acceptable reliability and validity. Thus, the CERVT has the potential to provide a comprehensive, reliable, and valid assessment of ER in clinical and research settings.