台灣WAIS-III中文版算術、記憶廣度測驗及其組合估算工作記憶指數在臨床上之適用性：回溯性研究

The Clinical Applicability of the Arithmetic, Digit Span Subtests of the Taiwan WAIS-III, and their Composite in Reflecting the Working Memory Index: A Retrospective Study

受測驗題材適用性之限制，國內臨床在取得WAIS-III中文版之工作記憶指數時，仍只慣用算術與記憶廣度兩項分測驗組合進行估算，其有效性有待檢驗。此外，由於算術分測驗相當受語文理解能力干擾，而記憶廣度分測驗之順、逆序背誦廣度值在檢測腦部是否病變的有效性也仍有爭議，因此本研究目的在檢驗台灣WAIS-III中文版工作記憶指數、單一算術與記憶廣度分測驗量尺分數、及記憶廣度分測驗之各項廣度值能否有效反映工作記憶。本研究以回溯研究方法，將179名各類中樞神經病變患者之腦造影結果當為參照效標，WAIS-III中文版之各項工作記憶相關測驗分數當為檢驗標的，進行ROC （Receiver Operating Characteristic）分析，結果顯示僅由算術與記憶廣度分測驗組合所推估之工作記憶指數的特異性、陽性預測值、及陰性預測值皆在94%以上，但其敏感度僅有63%。單一算術與單一記憶廣度分測驗之敏感度皆為50%，但在陽性預測值上，算術分測驗為67%，記憶廣度分測驗則為86%。記憶廣度分測驗之順、逆序背誦廣度值的敏感度皆在58%以下，陰性預測值則皆在92%以上。綜合上述，顯見WAIS-III中文版中之工作記憶指數、單一算術及單一記憶廣度分測驗、或記憶廣度分測驗之各項廣度值在區分患者有無工作記憶功能異常上皆擁有良好之區辨力，但對偵測工作記憶功能異常患者的敏感度普遍不佳，建議臨床工作者應儘量取得算術、記憶廣度、與數－字序列等三項分測驗組合所估算之工作記憶指數；若在有限條件下，臨床工作者僅能以算術與記憶廣度兩項分測驗之組合所估算之工作記憶指數來評估受試者工作記憶，應留意假陰性比例上升的問題。有鑑於本研究受限於回溯性研究方法而延伸出的問題，我們建議採取前瞻性後續研究，重新探討這個議題有其必要性。

There are many issues in regards to whether or not the composite of Arithmetic and Digit Span subtests used to determine the working memory index of the Taiwan WAIS-III is compatible with that of the English WAIS-III, and whether the individual Arithmetic and Digit Span subtests, and the span scores of the Digit Span subtest are adequate for measuring working memory. This study was thus to make an attempt to examine these issues. In this retrospective study, receiver operating characteristic (ROC) analyses were employed, using brain imaging data of 179 patients with various CNS diseases as the gold standard, to identify the working- memory- related scores in the Taiwan WAIS-III. The results, as estimated by a composite of the Arithmetic and the Digit Span subtests, revealed all the specificity, negative predictive value, and positive predictive value of the Taiwan WAIS-III working memory index were each at or above 94%, yet the sensitivity was only 63%. The sensitivities of the Arithmetic and the Digit Span subtests were even poorer, measuring only 50%. However, the positive predictive value of the Digit Span subtest was 86% while the Arithmetic subtest was 67%. In addition, the forward and backward span scores of the Digit Span subtest had good negative predictive value (all of them were above 92%), but poor sensitivity (all of them were below 58%). Based on the present findings, it appears that all the Taiwan WAIS-III working memory index, the single Arithmetic and the Digit Span subtests, and the span scores of the Digit Span subtest could adequately discriminate individuals without working memory impairment, but had poor sensitivity on detecting the patients who evidenced working memory impairment. Accordingly, it is suggested that clinicians should gauge the working memory index by a composite of the Arithmetic, the Digit Span, and the Letter-Number Sequencing subtests if possible. Otherwise, a caveat should be taken with care to interpret the working memory index, as estimated by a composite of the Arithmetic and the Digit Span subtests, because the chance of the increasing false negative rate becomes ineludible. Respecting the limitations of retrospective study, however, it is necessary to make a prospective study to re-approach this issue.