應用ROC 曲線分析驗證精簡版住院病人跌倒危險因子評估工具之準確度研究

Using ROC Curve Analysis to Assess the Accuracy of Short Form Inpatient Fall Risk Assessment Tool

本研究設計為次級資料分析法，目的在驗證精簡版自擬住院病人跌倒危險評估工具之準確度及預測因子，此七項跌倒危險因子為「年齡大於65歲或小於15歲」、「跌倒史」、「意識紊亂、持續性或間斷意識認知障礙」、「頭暈」、「軟弱」、「頻尿、 腹瀉或需協助如廁」、「需要提供輔具」，經receiver operating characteristic curve與logistic regression統計分析，得知此7題精簡版量表之area under curve=.90，最佳切點總分3分以上為跌倒高危險病人，預測敏感性74.07%，特異性86.93%，準確度86.26%，概似比19.01；經logisticregression鑑定高風險跌倒病人之發生跌倒勝算比為其他住院病人的17倍，工具精簡且測量效能更優於原17題量表。故建議以此精簡版工具進行住院病人跌倒危險評估，可精實護理人員操作流程，並作為病人安全政策修正之實證依據。

The study was a secondary data analysis aimed to assess the effectiveness of the compacted fall risk assessment tool and the predictors designed for inpatient. The seven fall risk factors are: a) age above 65 or below 15; b) fall history; c) consciousness disturbance, continuous or temporary cognitive impairment; d) dizziness; e) weakness; f) frequent urination, diarrhea or requiring toilet assistance; and g) requiring assistive devices. The data were analyzed by receiver operating characteristic curve (ROC) and logistic regression. The result of the analysis showed an AUC = .90, the best cut-off point was when the fall risk assessment score of 3 points or higher, the sensitivity was 74.07%, the the specificity was 86.93%, accuracy was 86.26%, and the likelihood ratio was 19.01. Furthermore, the logistic regression had identified that the odds ratio of falling among high risk fall inpatients was 17 times higher than other inpatients. The study provided a strong evidence that our fall risk assessment tool for inpatient, when compared with the original 17 items scale, is far simpler to operate and with higher efficiency. The implementation of the new fall risk assessment tool can simplify the operating procedures of nurses, and serves as an empirical evidence to the revision of current patient safety policy, and the basis of new multifaceted fall prevention program.