Integrated Long-Term Care Services Center, Kaohsiung Municipal Ta-Tung Hospital, Kaohsiung Medical University, Kaohsiung, Taiwan.
Department of Nursing, Kaohsiung Medical University Hospital, Kaohsiung Medical University, Kaohsiung, Taiwan.
Kaohsiung J Med Sci. 2024 Apr;40(4):404-412. doi: 10.1002/kjm2.12807. Epub 2024 Feb 16.
This retrospective study was conducted at a medical center in southern Taiwan to assess the accuracy of the Hendrich II Fall Risk Model (HIIFRM) in predicting falls. Sensitivity, specificity, accuracy, and optimal cutoff points were analyzed using receiver operating characteristic (ROC) curves. Data analysis was conducted using information from the electronic medical record and patient safety reporting systems, capturing 303 fall events and 47,146 non-fall events. Results revealed that at the standard threshold of HIIFRM score ≥5, the median score in the fall group was significantly higher than in the non-fall group. The top three units with HIIFRM scores exceeding 5 were the internal medicine (50.6%), surgical (26.5%), and oncology wards (14.1%), indicating a higher risk of falls in these areas. ROC analysis showed an HIIFRM sensitivity of 29.5% and specificity of 86.3%. The area under the curve (AUC) was 0.57, indicating limited discriminative ability in predicting falls. At a lower cutoff score (≥2), the AUC was 0.75 (95% confidence interval: 0.666-0.706; p < 0.0001), suggesting acceptable discriminative ability in predicting falls, with an additional identification of 101 fall events. This study emphasizes the importance of selecting an appropriate cutoff score when using the HIIFRM as a fall risk assessment tool. The findings have implications for fall prevention strategies and patient care in clinical settings, potentially leading to improved outcomes and patient safety.
本回顾性研究在台湾南部的一家医疗中心进行，旨在评估 Hendrich II 跌倒风险模型（HIIFRM）预测跌倒的准确性。使用接收者操作特性（ROC）曲线分析了敏感性、特异性、准确性和最佳截断点。数据分析使用电子病历和患者安全报告系统的信息进行，共捕获了 303 起跌倒事件和 47146 起非跌倒事件。结果表明，在 HIIFRM 评分≥5 的标准阈值下，跌倒组的中位数评分明显高于非跌倒组。HIIFRM 评分超过 5 的前三个科室是内科（50.6%）、外科（26.5%）和肿瘤科（14.1%），这表明这些区域跌倒风险较高。ROC 分析显示 HIIFRM 的敏感性为 29.5%，特异性为 86.3%。曲线下面积（AUC）为 0.57，表明在预测跌倒方面的区分能力有限。在较低的截断评分（≥2）下，AUC 为 0.75（95%置信区间：0.666-0.706；p<0.0001），表明在预测跌倒方面具有可接受的区分能力，额外识别了 101 起跌倒事件。本研究强调了在使用 HIIFRM 作为跌倒风险评估工具时选择适当截断评分的重要性。这些发现对临床环境中的跌倒预防策略和患者护理具有重要意义，可能会改善结果和患者安全。
