發展思覺失調症患者之電腦化多階段工作記憶測驗組

Abstract

背景與目的：工作記憶 (working memory, WM) 為個體短暫儲存並操縱訊息之能力。WM缺損為思覺失調症患者核心認知缺損之一，影響患者之思覺失調症狀、功能性表現及生活品質。然而，目前常用於思覺失調症患者之WM評量工具有三項限制：（一）未完整評量Baddeley「多系統WM模式」之所有WM系統；（二）紙筆/桌上型測驗之施測者負擔較大；（三）心理計量特性 (psychometric properties) 驗證結果不佳或缺乏驗證。上述限制使得臨床及研究人員難全面地掌握思覺失調症患者之WM狀態，嚴重影響WM之研究與臨床評估/治療效能。為解決上述WM測驗之限制，本研究之目的為：依據Baddeley之「多系統WM模式」，以「多向度羅序模型」發展一套電腦化多階段WM測驗：電腦化多階段工作記憶測驗組 (Computerized Multistage Testing of Working Memory, COWMEM)，以完整地評量各WM系統。此外研究者驗證COWMEM之心理計量特性，包含信度、同時效度、收斂效度及已知族群效度。 方法：本研究包含二個階段，第一階段為「發展COWMEM」，包含六個步驟：（一）參考Baddeley之多系統WM模式以設計COWMEM之6個向度題目，包含：視空間系統、語音系統、訊息整合系統、及中央執行系統之3種注意力：集中性注意力、轉換性注意力及分配性注意力；（二）專家內容效度驗證；（三）電腦施測平台建構；（四）題庫測試於病情穩定之思覺失調症患者；（五）題庫驗證：以單向度羅序分析各WM向度題庫，並刪除違反羅序模型假設之題目；及（六）選題以建構COWMEM。本研究第二階段為「驗證COWMEM之信度與效度」，以本研究第一階段收集之患者資料進行分析。信度部分，研究者計算COWMEM之受測者個別信度及所有受測者平均信度。同時效度驗證上，研究者以 Pearson’s r 分別檢驗COWMEM之6個WM向度分數與6個一一對應之效標測驗 [Spatial Span-backward, Digit Span-backward, California Verbal Learning Test-10 minutes delayed recall, Test of Everyday Attention (TEA)-Elevator counting, TEA-Visual elevator, and TEA-Telephone search dual task] 分數之相關性。收斂效度以 Pearson’s r 檢驗6個WM向度分數間之相關程度。已知族群效度驗證上，研究者依據整體認知功能量表 Montreal Cognitive Assessment 將受測者分為高、低認知等二組 (分數 ≥ 26 or ≤ 25)，再計算效應值 (Cohen’s d) 以檢驗二組之COWMEM分數差異程度。 結果：本研究第一階段COWMEM題庫經3回合之專家內容效度驗證，149位思覺失調症患者之臨床測試，題庫驗證刪除5題不符合羅序模型或研究者假設之題目，於最後階段挑選6題篩檢題，並於6個WM向度題庫中分別挑選高及低難度測驗組之題目各15題，以組成最終版COWMEM（共186題），每位受測者作答COWMEM時須完成6題篩檢題及各WM系統之1組高或低難度測驗組（共96題）。本研究第二階段驗證COWMEM信度之結果發現，6個WM向度平均信度為0.70~0.86，此結果顯示團體層級信度可接受。個別信度驗證結果顯示52.0%~100.0%之受測者個別信度 ≥ 0.70，表示個別信度可接受。同時效度驗證結果發現COWMEM之中央執行系統3種注意力分數與效標測驗分數之 |Pearson’s rs| 為0.21~0.22，此結果顯示COWMEM中3種注意力分數之同時效度不佳。COWMEM其它3個WM向度（視空間系統、語音系統及訊息整合系統）與效標測驗分數之 Pearson’s rs 為0.50~0.52，此結果顯示上述三個系統之同時效度中等。COWMEM之收斂效度發現6個WM向度分數間之 Pearson’s rs 為0.33~0.87，此結果顯示COWMEM之收斂效度良好。COWMEM之已知族群效度驗證結果顯示，高、低認知功能組別之受測者於6個WM向度之 Cohen’s d 為0.72~1.06，此結果顯示COWMEM具中等至良好之已知族群效度。 結論：COWMEM可能為第一個依據Baddeley多系統WM模式發展之電腦化WM測驗組，可完整評量6個WM向度。本研究初步驗證COWMEM於思覺失調症患者之心理計量特性，結果發現COWMEM中6個WM向度分數之信度可接受、收斂效度良好、已知族群效度中等至良好。同時效度驗證結果發現，COWMEM之視空間系統、語音系統及訊息整合系統分數具可接受之同時效度，但中央執行系統中3種注意力分數之同時效度尚不明確。COWMEM適用於思覺失調症患者團體層級之各WM系統能力比較，例如：研究中實驗組及對照組於各WM向度能力比較，或臨床上不同嚴重程度或病程之患者族群於各WM向度能力比較。

Background and purposes: Working memory (WM) can be defined as an individual’s ability to maintain and manipulate information temporarily. WM deficit is a core syndrome of cognitive deficits for patients with schizophrenia. WM deficit may affect other syndromes of schizophrenia, functional outcomes, and quality of life. However, current frequently used WM tests have 3 limitations, including: (1) incomprehensive assessment of the WM components of the Baddeley’s multiple-component model; (2) heavy administration burden of the paper-and-pencil tests for assessors; and (3) unsound / less-examined psychometric properties. Due to these 4 limitations, researchers and clinicians cannot efficiently and accurately measure the WM deficits in patients, which hampers the efficiency and efficacy in the research and treatment on WM. To deal with the limitations of the WM tests, the first aim of this study was to develop a Computerized Multistage Testing of Working Memory (COWMEM) in patients with schizophrenia based on the Baddeley’s multiple-component model. The second aim of this study was to examine the preliminary psychometric properties of the COWMEM in patients with schizophrenia, including reliability and validity (concurrent validity, convergent validity, and known-groups validity). Methods: This study is comprised of 2 phases. Phase 1 contained 6 steps to develop the COWMEM: (1) designing the items of the COWMEM’s 6 item banks (domains): phonological loop, visuospatial sketchpad, episodic buffer, and central executive’s 3 types of attentional controls (focused attention, switching attention, and divided attention); (2) examining the content validity; (3) establishing the administration platform of the item banks; (4) administering the item banks in patients with schizophrenia; (5) validating the item banks: deleting the items which did not fit the unidimenal Rasch model in each item bank; and (6) selecting items from the item banks to establich the COWMEM. In Phase 2, the data collected in Phase 1 were used to validate the psychometric properties of the COWMEM. To examine the reliability of the COWMEM, individual patient’s Rasch reliability and the average Rasch reliability of the patients were calculated. To validate the concurrent validity, Pearson’s r was calculated to examine the correlations between the 6 domain scores of the COWMEM and the scores of the corresponding external criteria [Spatial Span-backward, Digit Span-backward, California Verbal Learning Test-10 minutes delayed recall, Test of Everyday Attention (TEA)-Elevator counting, TEA-Visual elevator, and TEA-Telephone search dual task]. To validate the convergent validity, the correlations among the COWMEM’s 6 domain scores were examined using Pearson’s r. To validate the known-groups validity, the patients were divided into 2 groups (good or poor cognition) according to the patients’ scores of the Montreal Cognitive Assessment (≥ 26 or ≤ 25). Next, in each domain of the COWMEM, the extent of difference of the mean domain scores between the 2 groups was examined using effect size (Cohen’s d). Results: In Phase 1, the item banks of the COWMEM were validated through 3 rounds of examination for content validity, administered on 149 patients with schizophrenia, and 5 items which did not fit the Rasch model were deleted. Finally, the COWMEM contained 6 routing items and 2 item sets (difficult or easy; 15 items in each item set) in each of the 6 domains (a total of 186 items). In the administration of the COWMEM, each patient had to complete the 6 routing items and 1 item set (which are chosen according to the patient’s performance on the routing items) in each of the 6 domains (a total of 96 items). In Phase 2, the results of the reliability indicate that the COWMEM has acceptable individual reliability (52.0%~100.0% of patients had Rasch reliability ≥ 0.70) and average reliability (0.70~0.86 across the 6 domains). The results of the concurrent validity indicate that the COWMEM has acceptable concurrent validity in the 3 domains (phonological loop, visuospatial sketchpad, and episodic buffer; Pearson’s rs = 0.50~0.52) but poor concurrent valditiy in the other 3 domains (central executive’s 3 types of attentional controls; Pearson’s rs = 0.21~0.22). The results of the convergent validity indicate that the COWMEM has good convergent validity (Pearson’s rs = 0.33~0.87). The results of the known-groups validity indicate that the COWMEM has moderate-to-good known-groups validity (Cohen’s ds = 0.72~1.06). Conclusion: The COWMEM may be the first computerized WM test for comprehensively assessing 6 domains of WM. The preliminary examination of psychometric properties indicate that the COWMEM has acceptable reliability, good convergent validity, moderate-to-good known-groups validity. Regarding the concurrent validity, the 3 domain scores (phonological loop, visuospatial sketchpad, and episodic buffer) of the COWMEM have acceptable concurrent validity, but the concurrent validity of the other 3 domain scores (central executive’s 3 types of attentional controls) are unclear. The COWMEM can be used to compare the scores of the 6 WM domains among groups of patients with schizophrenia (e.g., control vs. treatment group in research or comparisons between groups of patients with different severities or stages in clinical settings).