背景：使用自擬「住院兒童跌倒危險評估表」，2009～2012年住院兒童跌倒發生率（千分之0.35－0.5），均高於台灣臨床成效指標（Taiwan Clinical Performance Indicator, TCPI）及本院監控閾值（千分之0.3）。本研究目的在探討「住院兒童跌倒危險評估表」之敏感性、特異性及相似比，冀望能找出住院病童跌倒危險因子，有效預防病童跌倒。方法：本研究採回溯性病例對照，以二階段展開。第一階段：回顧2011年～2013年住院後第一次評估舊表「兒童跌倒危險因子評估表」、跌倒記錄，將六項危險因子與跌倒與否進行卡方檢定、邏輯斯迴歸分析，確認相關之危險因子。第二階段：2015年9月∼2016年2月於兒科病房全面實施測試新表「新兒童跌倒危險因子評估表」，刪除不相關危險因子，收集資料驗證新表之敏感性及特異性。結果：第一階段由醫療資訊系統下載共計40,703人，以邏輯斯迴歸分析發現於年齡、性別、用藥情形、跌倒經驗等四項危險因子與跌倒有相關（p＜.05），評估分數≥3分的跌倒發生率是＜3分的3.3倍。第二階段使用「新兒童跌倒危險因子評估表」，下載共計5,816人，以ROC曲線分析≥4分敏感度91.7%、特異性51.1%，曲線下面積AUC 0.656，以邏輯斯迴歸分析，≥4分的跌倒發生率是＜4分的9.6倍。結論／實務運用：「新兒童跌倒危險因子評估表」包含年齡、性別、用藥情形、跌倒經驗等四項危險因子，刪除不相關二項「活動力」「體能狀況」，全面資訊化運用於本院體系兒科病童。

Background: When using the self-established fall risk scale, the fall rate of hospitalized children from 2009-2012 was .35- .5% which was higher than Taiwan Clinical Performance Indicator and hospital threshold ( .3%). The purpose of this study was to test the sensitivity, specificity, and likelihood ratios of current risk scale for hospitalized children to find out risk factors for better children fall prevention. Methods: Two stages of retrospective chart review were proceeded. On the first stage, 2011-2013, fall record data from hospital information system using old fall risk scale were collected and six risk factors were used to analyze to identify fall risk factors by using chi-square test and multiple logistic regressions. On the second stage, September 1, 2015 - February 28, 2016, new fall risk scale was used for all pediatric units to assess its sensitivity, specificity, and likelihood ratios. Results: On the first stage, patient fall data of 40,703 patients were downloaded from hospital information system. Data were analyzed using logistic regression. Age, sex, medication, fall experience were associated with fall (p＜ .05). The occurrence of patient with a fall score ≥3 was 3.3 times higher than those with a score of ＜3. On the second stage, patient fall data of 5,816 patients were downloaded by using new revision fall risk scale. The results revealed that the sensitivity of a score ≥4 was 91.7%, specificity was 51.1%, and the area under curve was .656 by using ROC curve. The occurrence of patients with a fall score ≥4 was 9.6 times higher than those with a score of ＜4. Activity and physical condition were deleted to come up with a four-factor fall risk scale. Each factor weighted differently. Conclusion/ Implications for Practice: By deleting activity and physical condition, the revised fall risk scale with age, sex, medication, and fall experience, is currently used in all pediatric units.