發展暨驗證視聽整合工作記憶測驗於思覺失調症患者之心理計量特性

Abstract

背景與目的：工作記憶 (working memory, WM) 為個體短暫儲存並操弄訊息的認知能力。WM缺損常見於思覺失調症患者，並造成其難以回復病前功能及職能角色。情節緩衝 (episodic buffer) 為工作記憶的次系統之一，其功能為整合「語音」、「視空間」與「長期記憶」成為有限數量的組塊 (chunk) 或情節 (episodes)。情節緩衝系統負責同時處理多重來源的訊息，患有WM缺損的思覺失調症患者在處理此類複雜訊息時常會出現困難，推測患者的WM損傷可能是因為情節緩衝系統受損而致，因此評估思覺失調症患者的情節緩衝WM能力有其必要性。目前常用於思覺失調症患者的情節緩衝WM 評估工具有四項限制：(一) 缺乏以「情節形式」為主的測驗內容；(二) 測驗結果未能提供患者情節緩衝WM較差的原因；(三) 測驗內容具語言/文化隔閡，難以適用於台灣的患者；(四) 心理計量特性缺乏驗證。上述限制使得臨床及研究人員難以掌握思覺失調症患者情節緩衝WM的缺損原因與程度，嚴重影響情節緩衝WM的研究與臨床評估/治療效能。為解決上述情節緩衝 WM 測驗的限制，本研究目的有二：(一) 測驗發展：依據Baddeley之多系統WM模式，發展一套適用於思覺失調症患者的視聽整合工作記憶測驗 (visuospatial-verbal binding working memory test, VIVEB)；(二) 心理計量特性驗證：驗證VIVEB 應用於思覺失調症患者的再測信度、隨機測量誤差、最小可偵測變化值 (minimal detectable change, MDC)、練習效應、收斂效度、生態效度以及區辨效度。 研究方法：本研究包含二個階段，第一階段為「發展 VIVEB」，包含三個步驟：(一) 參考 Baddeley 之多系統 WM 模式中的情節緩衝WM概念以設計初版VIVEB 的題目；(二) 專家內容效度驗證以及題庫臨床測試；(三) 選題及建構最終版VIVEB。本研究第二階段為「驗證 VIVEB 的心理計量特性」。信度部分，研究者以組內相關係數 (intraclass correlation coefficient) 檢驗患者前後測VIVEB總分的再測信度，以評估標準誤 (standard error of measurement, SEM) 及SEM% 分析隨機測量誤差大小，並以MDC及MDC% 計算評估結果超過真實改變的參考標準，最後以效應值 (effect size) 檢驗練習效應程度。效度部分，收斂效度驗證上，研究者以 Pearson’s r 分別檢驗 VIVEB總分與5項效標測驗的相關程度，效標測驗為： Montreal Cognitive Assessment (MoCA)、Spatial Span Test (SS)、Digit Span Test (DS)、Computerized Digit Vigilance Test (C-DVT)、Lawton Instrumental Activities of Daily Living scale (LIADL)。區辨效度驗證上，研究者依據MoCA將受試者分為高、低認知等二組 (分數 ≥ 26 or ≤ 25)，使用獨立t檢定 (independent t test) 檢驗二組受試者的前測VIVEB總分是否有統計上顯著差異。生態效度驗證上，研究者以 Pearson’s r 分別檢驗前測VIVEB總分與LIADL的相關性以驗證生態效度。 結果：本研究第一階段 VIVEB 題庫經 3 回合之專家內容效度驗證，42 位思 覺失調症患者之臨床測試，研究者刪除 4 題不符合預期難度排序的題目，以組成最終版 VIVEB（共8題）。本研究第二階段驗證 VIVEB 信度之結果發現，「語音儲存」與「視空間儲存」向度之再測信度不佳、隨機測量誤差大但具有低度的練習效應；「整合視聽訊息」與「操弄整合後的訊息」向度具有可接受之再測信度、隨機測量誤差以及低度的練習效應。收斂效度驗證結果發現 VIVEB 之「語音儲存」與「視空間儲存」分數與效標測驗分數之 |Pearson’s r| 為 0.12－0.21，此結果顯示二個儲存向度分數之收斂效度不佳。VIVEB 之「整合視聽訊息」與「操弄整合後的訊息」向度與效標測驗分數之 |Pearson’s r| 為0.04－0.41，此結果顯示主要向度分數之收斂效度為不佳至良好。生態效度驗證結果顯示，VIVEB 之4向度與生態效標測驗分數之 |Pearson’s r| 為0.01－0.08，此結果顯示 VIVEB 具有不佳之生態效度。區辨效度驗證結果顯示，高、低認知功能組別之受試者於視空間儲存的Cohen’s d為 0.31，其餘3個向度的 Cohen’s d 為 0.89－0.93，此結果顯示 VIVEB大致具良好之區辨效度。 結論：VIVEB為依據 Baddeley 提出之情節緩衝工作記憶概念，以「情節形式」為主的多向度測驗。本研究初步驗證 VIVEB 於思覺失調症患者之心理計量特性，結果發現 VIVEB 主要向度分數之信度可接受、區辨效度良好、收斂效度為不佳至良好、生態效度不佳；VIVEB儲存向度具有較不佳之信效度。因此， VIVEB之心理計量特性尚不足以支持其應用於臨床及研究使用，研究者建議可修改儲存向度之測驗方式，並藉由更完整之效度驗證（如因素效度驗證），以釐清 VIVEB 各向度題目所評量之建構。

Background and purposes: Episodic buffer (EB) is one of the components of the multiple-component working memory (WM) model, which was proposed by Baddeley. EB is capable of binding information from subsidiary systems (phonological loop, Visuospatial sketchpad) and from long-term memory into chunks or episodes. EB processes multidimensional information simultaneously, and schizophrenia patients with WM deficits often display difficulties when processing such complex information. Therefore, it is necessary to evaluate episodic buffer WM deficits in patients with schizophrenia. However, EBWM assessment tools commonly used in patients with schizophrenia have four limitations: (1) lack of test content based on 'episodes'; (2) test results fail to provide reasons for EBWM deficits; (3) test content has language/cultural barriers; (4) lack of validation of psychometric properties. These limitations make it difficult for clinicians and researchers to ascertain the reason and extent of EBWM deficiency in patients with schizophrenia which may severely affect the clinical treatment efficacy of EBWM. To deal with the limitations of the EBWM tests, the first aim of this study was to develop a visuospatial-verbal binding working memory test (VIVEB) in patients with schizophrenia based on Baddeley’s multiple-component model. The second aim of this study was to examine the psychometric properties of the VIVEB in patients with schizophrenia, including reliability [test-retest reliability, random measurement error, minimal detectable change (MDC), practice effect] and validity (convergence validity, discriminant validity and ecological validity) Methods: This study was comprised of 2 phases. Phase 1 contains 3 steps to develop the VIVEB: (1) designing the first version of the VIVEB referring to the concept of episodic buffer WM based on Baddeley’s multiple-component WM model; (2) verifying expert (content) validity and administering the item bank in patients with schizophrenia; (3) selecting items from the item bank to establish the final version of the VIVEB. Phase 2 of the study was to verify the psychometric characteristics of the VIVEB. To examine the reliability of the VIVEB, researchers recruited schizophrenia patients to be assessed twice by the VIVEB (4 weeks apart), and used intraclass correlation coefficient to investigate the test-retest reliability of the total scores of the VIVEB. In addition, researchers used standard error of measurement (SEM) and SEM% to analyze the random measurement error. Moreover, researchers calculated MDC and MDC% to examine the threshold beyond real change, and used the effect size to examine the level of practice effect. To examine the validity of the VIVEB, researchers verified the correlation between the VIVEB and criterion tests. For convergence validity, researchers used Pearson's r to validate the correlation between the total scores of the VIVEB and the four criterion tests. The criterion tests were: Spatial Span Test (SS), Digit Span Test (DS), and Computerized Digit Vigilance Test (C-DVT). For discriminant validity, the patients were divided into 2 groups (good or poor cognition) according to the patients’ scores of the Montreal Cognitive Assessment (MoCA) (≥ 26 or ≤ 25). Next, researchers used effect size (Cohen’s d) to examine the difference in the total scores of the VIVEB between the 2 groups. For ecological validity, researchers used Pearson's r to validate the correlation between the total scores of the VIVEB and the Lawton Instrumental Activities of Daily Living scale. Results: In Phase 1, the item banks of the VIVEB were validated through 3 rounds of examination for expert (content) validity. The VIVEB was administered on 42 patients with schizophrenia, and 4 items which did not fit the sorting difficulty by expected were deleted. Finally, the final version of VIVEB contained 8 items. In Phase 2, the results of reliability indicate that verbal and visual-spatial storage dimensions of the VIVEB have poor test-retest reliability and large random measure error, however, they have trivial practice effect. The binding and manipulation dimensions of VIVEB have acceptable test-retest reliability and random measure error, and they also have trivial practice effect. The results of the convergent validity indicate that the VIVEB has poor-to-fair convergent validity in the binding and manipulation dimensions (Pearson’s r = 0.04-0.41), but poor convergent validity in the other 2 dimensions (verbal and visual-spatial storing; Pearson’s r = 0.12-0.21). The results of ecological validity show that VIVEB has poor ecological validity in 4 dimensions (Pearson’s r = 0.01-0.08). The results of discriminant validity indicate that VIVEB has good discriminant validity in 4 dimensions (Cohen’s d = 0.89-0.93). Conclusion: The VIVEB may be the first episodic buffer WM test which assesses the 4 dimensions of episodic buffer comprehensively by “episodes”. The validations of psychometric properties indicate that binding and manipulation dimensions have acceptable reliability, poor-to-fair convergent validity, good discriminant validity and poor ecological validity. However, verbal and visual-spatial storage dimensions have good discriminant validity but poor reliability, convergent validity, and ecological validity. Therefore, these psychometric properties do not adequately support the use of VIVEB in clinical and research setting. Researchers may modify the testing method of the storage dimensions. In addition, future research can use more comprehensively methods of validity verification, such as factor validity verification, to clarify the structure of the 4 dimensions of the VIVEB.