「修正版中風病患姿勢控制評估量表」之心理計量特性

Abstract

背景與目的：本研究藉由加入轉身及踏階等四項中高難度測試項目於原版「中風病患姿勢控制評估量表」（Postural Assessment Scale for Stroke Patients，簡稱 PASS）中，形成「修正版中風病患姿勢控制評估量表（簡稱mPASS）」，以期改善原量表於輕度中風病人的天花板效應及高原現象等問題。本研究的目的主要在探討新加入的測試項目之區辨效度，及「修正版中風病患姿勢控制評估量表」之區辨效度、信度、內在一致性、同時與預測效度、及反應性。 方法：本研究分成三部分。第一部份徵召63位首次缺血性中風病患（平均年齡61.5歲），依據病患入院時之美國國家衛生研究院中風量表（National Institute Health Stroke Scale，簡稱NIHSS）分數分成輕度（NIHSS＜4）、中度（4≦NIHSS≦15）及重度（NIHSS＞15）中風病患三組，並蒐集病患第14天之mPASS分數。以Mann-Whitney分析三組病患於新加入的四個測試項目（簡稱mPASS-4）與原PASS量表中三個中或高難度測試項目的得分差異；用主要成分分析（principal components analysis）探討mPASS的組成因素。第二部分由第一部份受試者中分層隨機取出不同嚴重度的中風病患各5位，由7位評估者分別於間隔兩週的不同時間點觀看病患進行mPASS施測的錄影帶資料並評分。利用ICC值及weighted kappa值來分析評估者內及評估者間信度；另以Cronbach alpha分析mPASS的內在一致性。第三部分另徵召39位首次中風病患，進行世代追蹤研究。計有27位病患完成第十四天、第三十天及第九十天等三次追蹤測試。評估工具包括mPASS、伯格式平衡量表(Berg Balance Scale (BBS))、軀幹控制測試(Trunk Control Test (TCT))、巴氏量表(Barthel Index (BI))、修正版任金量表(modified Ranking Scale (mRS))及三公尺計時起走測試(Timed “Up and Go” Test (TUGT))。以斯皮爾曼相關係數分析各時間點mPASS、PASS、BBS及TCT得分的相關性，探討mPASS的同時效度。再以斯皮爾曼相關係數分析第14天及第30天的mPASS得分與第九十天BBS、BI、mRS及TUGT得分的相關性，探討mPASS的預測效度。另以ROC 曲線(receiver operating characteristics curve)分析以第14天mPASS的得分預測第90天mRS得分的敏感度(sensitivity)與特異度(specificity)。 結果：第一部份之結果顯示：新加入的四個測試項目中，其平均分數（1.3至1.5）均介於原量表的 B12（中難度）（1.8）及A4、A5項目（高難度）（1.1、0.8）平均分數間，且僅mPASS-4及mPASS的得分在輕度與中度中風病患之間具有意義的差異(p< 0.05)；PASS 及B12、A4、A5項目的得分在輕度與中度中風病患之間均無差異(p> 0.05)。主要成分分析結果顯示mPASS具有兩種面向，mPASS-4主要屬於「在無扶持下站姿之姿勢控制」面向。第二部分之結果顯示mPASS具極高之評估者內與評估者間信度及內在一致性。七位評估者內mPASS總分的ICC值介於0.9992至1.0000之間，mPASS各項的weighted kappa值為0.833至1.000；七位評估者間mPASS總分的ICC值介於0.9992至0.9999，mPASS各項的weighted kappa值為0.858至1.000。七位評估者mPASS各項之間的Cronbach α值為0.9610至0.9625。第三部分之結果顯示mPASS具極佳之同時與預測效度及反應性。第十四天、三十天、與第九十天三個時間點的mPASS得分與同時間點的PASS及BBS的得分相關係數分別為0.973至0.983及0.950至0.980。第十四天mPASS的得分與第九十天BBS、TUGT、BI及mRS得分的相關係數分別為0.835、-0.795、0.767及-0.843；第三十天mPASS的得分與第九十天BBS、TUGT、BI及mRS得分的相關係數分別為0.889、-0.847、0.748及-0.820。此外，第14天的mPASS得分若大於35分可預測第90天mRS有良好失能程度的改善(mRS≦ 2)，ROC下方面積為0.915，敏感度為81.2%，特異度為90.9%。在反應性方面，mPASS於第14至30天、第30至90天及第14至90天的效應值分別為1.19、0.76及1.22；在不同嚴重度的各時期病人之mPASS的效應值亦均為中度（0.73）至大(> 0.8)的效應。 討論與結論：mPASS-4及mPASS具有區辨效度，且mPASS亦具有良好的心理計量特性，包括高的評估者內信度、評估者間信度、內在一致性、同時效度、預測效度及反應性。唯本研究中，中風嚴重度個案數較少，需增加其收案個數來進一步驗證mPASS於第90天產生天花板效應的問題。

Background and purposes: In this study, four balance testing items of moderate-high difficulty level were added to the original Postural Assessment Scale for Stroke Patients (PASS) to form the modified PASS (mPASS), in an attempt to reduce the ceiling effects and plateau phenomenon found in studies using the PASS to assess postural control ability in patients with stroke. The purposes of this study were to investigate the discriminant validity of the four new items (mPASS-4) and the mPASS, as well as the reliability, internal consistency, concurrent and predictive validity, and responsiveness of the mPASS. Methods: The study was divided into three parts. In Part I of the study, 63 patients with stroke (mean age= 61.5 yrs) participated and were divided into the mild (National Institutes of Health Stroke Scale (NIHSS) < 4), moderate (4≦NIHSS≦15), and severe (NIHSS＞15) subgroups. All participants performed the mPASS on the 14th day post stroke onset (D14). The Mann-Whitney statistics was used to investigate the differences in mPASS-4, mPASS total, and PASS total scores among the three subgroups. The principal components analysis (PCA) was used to investigate the primary factors underlying the mPASS. In Part II of the study, 5 patients of each severity level of stroke were randomly selected from the 63 patients in Part 1. Seven independent raters evaluated the mPASS performance of these 15 patients recored on videotapes twice with a two-week interval. The ICCs(2,1) and weighted kappa analyses were used to investigate the intra- and inter-rater reliability for the total mPASS scores and item scores, respectively. The Cronbach alpha was analyzed to investigate the internal consistency of the mPASS. In Part III of the study, a cohort design was used and another 39 patients with stroke participated, among which 27 of them completed the longitudinal follow-up assessments of the mPASS, Berg Balance Scale (BBS), Trunk Control Test (TCT), Barthel Index (BI), modified Rankin Scale (mRS), and the Timed “Up and Go” Test (TUGT) on D14, D30, and D90. The Spearman correlation coefficients were used to analyze the concurrent validity (mPASS versus PASS, BBS, and TCT on D14, D30 and D90) and predictive validity (mPASS on D14 and D30 versus BBS, BI, mRS, and TUGT on D90). In addition, the receiver operating characteristics curve (ROC curve) analysis was used to investigate the sensitivity and specificity of using the mPASS score on D14 in predicting the mRS score on D90. Results: In Part I, the results showed that the mean score of the mPASS-4 (1.3-1.5) ranged inbetween that of the B12 (moderate difficulty level, mean= 1.8) and that of the A4 and A5 (high difficulty level, mean= 1.1 and 0.8, respectively). The mPASS-4 and total mPASS scores, but not the B12, A4, A5, and the total PASS scores, showed significant differences (p < 0.05) between patients with mild versus moderate severity of stroke. Results of the PCA revealed that the mPASS had two domains and the mPASS-4 belonged to the “balance control in standing without support” domain. Results of the Part II of the study showed excellent inter- and intra-rater reliability, and internal consistency of the mPASS. The ICCs of the total mPASS scores for intrarater reliability ranged between 0.9992 and 1, and those for interrater reliability ranged between 0.9992 and 0.9999. The weighted kappa of all mPASS item scores for intrarater reliability ranged from 0.833 to 1.000 and from 0.858 to 1.000 for interrater reliability. The Chronba alpha values among all mPASS items ranged from 0.9610 to 0.9625. Results of the Part III showed that the mPASS had good concurrent and predictive validity, and responsiveness. The total mPASS scores performed on D14, D30, and D90 were highly correlated with thoses of the PASS and BBS performed on the same testing dates (rs = 0.950~0.983, p < 0.05). The total mPASS scores performed on D14 and D30 were both moderately to highly correlated to the scores on BBS, TUGT, BI and mRS performed on D90 (rs = 0.835, -0.795, 0.767, and -0.843, respectively, for D14; rs = 0.889, -0.847, 0.748, and -0.820, respectively, for D30; all p < 0.05). In addition, results of the ROC analysis showed that a cutoff score set at 35 point on mPASS could significantly predict an mRS score of equal to or below 2 on D90 with the sensitivity being 81.2%, specificity being 90.9%, and the area under the ROC curve being 0.915. Regarding the responsiveness, the results showed that the effect sizes of the changes in mPASS scores from D14 to D30, D30 to D90, and D14 to D90 were 1.19, 0.76, and 1.22, respectively for all participants. The effect sizes of these changes for patients with different severity also were moderate (0.73) to large (>0.80). Discussion and Conclusions: The mPASS-4 and mPASS had good discriminant validity, and mPASS also had good psychometric properties, including high intrarater reliability, interrater reliability, internal consistency, concurrent validity, predictive validity and responsiveness. However, further studies are needed to determine whether the ceiling effect of mPASS on D90 would still exist if a greater portion of patients with severe stroke were recruited.