Kaiser Permanente Bernard J. Tyson School of Medicine, Pasadena, CA, USA; The Permanente Medical Group, Oakland, CA, USA.
California Northstate University School of Medicine, Elk Grove, CA, USA.
Am J Emerg Med. 2023 May;67:168-175. doi: 10.1016/j.ajem.2023.02.025. Epub 2023 Feb 24.
Computed tomography (CT) is performed in over 90% of patients diagnosed with ureteral stones, but only 10% of patients presenting to the emergency department (ED) with acute flank pain are hospitalized for a clinically important stone or non-stone diagnosis. Hydronephrosis can be accurately detected using point-of-care ultrasound and is a key predictor of ureteral stone and risk of subsequent complications. The absence of hydronephrosis is insufficient to exclude a stone. We created a sensitive clinical decision rule to predict clinically important ureteral stones. We hypothesized that this rule could identify patients at low risk for this outcome.
We conducted a retrospective cohort study in a random sample of 4000 adults who presented to one of 21 Kaiser Permanente Northern California EDs and underwent a CT for suspected ureteral stone from 1/1/2016 to 12/31/2020. The primary outcome was clinically important stone, defined as stone resulting in hospitalization or urologic procedure within 60 days. We used recursive partition analysis to generate a clinical decision rule predicting the outcome. We estimated the C-statistic (area under the curve), plotted the receiver operating characteristic (ROC) curve for the model, and calculated sensitivity, specificity, and predictive values of the model based on a risk threshold of 2%.
Among 4000 patients, 354 (8.9%) had a clinically important stone. Our partition model resulted in four terminal nodes with risks ranging from 0.4% to 21.8%. The area under the ROC curve was 0.81 (95% CI 0.80, 0.83). Using a 2% risk cut point, a clinical decision tree including hydronephrosis, hematuria, and a history of prior stones predicted complicated stones with sensitivity 95.5% (95% CI 92.8%-97.4%), specificity 59.9% (95% CI 58.3%-61.5%), positive predictive value 18.8% (95% CI 18.1%-19.5%), and negative predictive value 99.3% (95% CI 98.8%-99.6%).
Application of this clinical decision rule to imaging decisions would have led to 63% fewer CT scans with a miss rate of 0.4%. A limitation was the application of our decision rule only to patients who underwent CT for suspected ureteral stone. Thus, this rule would not apply to patients who were thought to have ureteral colic but did not receive a CT because ultrasound or history were sufficient for diagnosis. These results could inform future prospective validation studies.
在被诊断为输尿管结石的患者中，超过 90%接受了计算机断层扫描（CT）检查，但只有 10%在急诊科（ED）出现急性腰痛的患者因临床重要的结石或非结石诊断而住院。床边超声可准确检测肾积水，是预测输尿管结石和随后并发症风险的关键指标。无肾积水并不能排除结石。我们制定了一种敏感的临床决策规则来预测具有临床重要意义的输尿管结石。我们假设该规则可识别出发生这种结果风险较低的患者。
我们对 2016 年 1 月 1 日至 2020 年 12 月 31 日期间在加利福尼亚州北部 Kaiser Permanente 的 21 个急诊科之一就诊且因疑似输尿管结石接受 CT 检查的 4000 名成年人进行了回顾性队列研究。主要结局为临床重要的结石，定义为在 60 天内导致住院或泌尿科治疗的结石。我们使用递归分区分析生成预测该结果的临床决策规则。我们估计了 C 统计量（曲线下面积），绘制了模型的接收器操作特征（ROC）曲线，并根据风险阈值为 2%计算了模型的敏感性、特异性和预测值。
在 4000 名患者中，有 354 名（8.9%）患有临床重要的结石。我们的分区模型得出了四个终端节点，风险范围为 0.4%至 21.8%。ROC 曲线下面积为 0.81（95%CI 0.80，0.83）。使用 2%的风险切点，包括肾积水、血尿和既往结石史的临床决策树可预测复杂结石，其敏感性为 95.5%（95%CI 92.8%-97.4%），特异性为 59.9%（95%CI 58.3%-61.5%），阳性预测值为 18.8%（95%CI 18.1%-19.5%），阴性预测值为 99.3%（95%CI 98.8%-99.6%）。
将此临床决策规则应用于影像学决策，将导致 CT 扫描减少 63%，漏诊率为 0.4%。限制因素是我们的决策规则仅应用于因疑似输尿管结石而接受 CT 检查的患者。因此，该规则不适用于被认为患有输尿管绞痛但未接受 CT 检查的患者，因为超声或病史足以做出诊断。这些结果可以为未来的前瞻性验证研究提供信息。
