比較精神科住院病人與Wilson-Sims跌倒風險評估量表之臨床診斷效果

Comparison of Clinical Diagnostic Effectiveness Between Psychiatric Inpatient and Wilson-Sims Fall Risk Assessment Tools

背景：精神科住院病人跌倒率高，但缺乏有效驗證之跌倒評估量表來輔助醫療人員正確判斷。目的：比較精神科住院病人跌倒風險評估量表（Psychiatric Inpatient Fall Risk Assessment Tool，PIFRAT）與Wilson-Sims Fall Risk Assessment Tool（WSFRAT）之敏感度、特異度、準確度差異並預測精神科住院病人之跌倒危險因素。方法：本研究2016年10月1日至2017年3月10日於台灣北部某精神科醫院，針對新病人住院第1－7天進行跌倒評估資料收集。採描述性統計、邏輯斯迴歸分析、信效度檢測、工具效能檢測及ROC曲線（receiver operating characteristic curve）等分析資料。結果：兩種跌倒評估表敏感度不佳（WSFRAT 57.1%，PIFRAT 50.0%）；特異度WSFRAT（79.6%）優於PIFRAT（70.4%）；準確度WSFRAT（76.9%）較PIFRAT（67.9%）佳。ROC曲線分析PIFRAT之AUC（area under the curve）=.602，95% CI［0.48，0.73］，根據Youden指數，最佳切點為7.5分，可達88.8%特異度及39.3%敏感度，若要提升敏感度至96.4%，則切點必須調整為1.5分。WSFRAT AUC=.625，最佳切點為3.5分可達82.1%敏感度。多變量邏輯斯迴歸顯示，過去曾跌倒者再跌倒的風險顯著高於無跌倒經驗者；男性病人（OR=2.57，95% CI［1.11，5.94］）、身體活動困難（OR=3.43；95% CI［1.40，8.41］）、虛弱無力（OR=3.03；95% CI［1.08，8.49］）等達統計顯著。結論／實務應用：研究發現四個重要的跌倒危險因素，未來臨床照護人員應提高警覺，發展跌倒相關防範措施。本研究結果亦可作為未來發展精神科跌倒評估量表之參考。 Background: The incidence of falls is very high among psychiatric inpatients. However, the lack of an effective, validated psychiatric inpatient fall risk assessment tool inhibits the ability of medical staffs to make correct judgments. Purpose: The purposes of this study were to compare the sensitivity, specificity, and accuracy of the psychiatric inpatient fall risk assessment tool (PIFRAT) and the Wilson-Sims fall risk assessment tool (WSFRAT) and to predict the fall risk factors in PIFRAT and WSFRAT for psychiatric inpatients. Methods: Study data were collected from 2016/10/01 to 2017/03/10. Fall assessment data were collected from new patients during their 1^(st) through 7^(th) days after admission to a psychiatry unit in northern Taiwan. Data were analyzed using descriptive analysis, logistic regression analysis, reliability and validity testing, tool effective testing, and receiver operating characteristic (ROC) curve analysis. Results: Both of the fall risk assessment tools exhibited low sensitivity (WSFRAT 57.1%, PIFRAT 50%), the specificity of WSFRAT (79.6%) was higher than that of PIFRAT (70.4%), and the accuracy of WSFRAT (76.9%) was higher than that of PIFRAT (67.9%). The ROC curve analysis revealed that the AUC of PIFRAT was .602 (95% CI [0.48, 0.73]). According to the Youden index, the best cutoff level is 7.5 points, in which the specificity is 88.8% and the sensitivity is 39.3%. To increase the sensitivity to 96.4%, the cutoff level must be set to 1.5 points. Moreover, the AUC of WSFRAT was .625 and the highest sensitivity was 82.1% when the cutoff point was set to 3.5 points. Further, multivariate logistic regression analysis revealed that fall risk was significantly higher among patients who had previously fallen than among those had not. Male gender (OR = 2.57, 95% CI [1.11, 5.94]), physical activity difficulties (OR = 3.43; 95% CI [1.40, 8.41]), and weakness (OR = 3.03; 95% CI [1.08, 8.49]) were each significantly associated with fall risk. Conclusions/Implications for Practice: This study identified four critical risk factors for falls. In the future, clinical healthcare professionals should be more aware of these factors and develop related fall-prevention interventions. The findings may serve as references for the future development of psychiatric fall assessment tools.