Department of Orthopaedic Surgery, Rubin Institute for Advanced Orthopaedics, Baltimore, MD.
Department of Biomedical and Health Informatics, Children's Hospital of Philadelphia, Philadelphia, PA.
J Arthroplasty. 2021 Jul;36(7):2541-2545. doi: 10.1016/j.arth.2021.02.067. Epub 2021 Mar 3.
Preoperative calculation of treatment failure risk in patients undergoing surgery for periprosthetic joint infection (PJI) is imperative to allow for medical optimization and targeted prevention. A preoperative prognostic model for PJI treatment failure was previously developed, and this study sought to externally validate the model.
A retrospective review was performed of 380 PJIs treated at two institutions. The model was used to calculate the risk of treatment failure, and receiver operating characteristic curves were generated to calculate the area under the curve (AUC) for each institution.
When applying this model to institution 1, an AUC of 0.795 (95% confidence interval [CI]: 0.693-0.897) was found, whereas institution 2 had an AUC of 0.592 (95% CI: 0.502-0.683). Comparing all institutions in which the model had been applied to, we found institution 2 represented a significantly sicker population and different infection profile.
In this cohort study, we externally validated the prior published model for institution 1. However, institution 2 had a decreased AUC using the prior model and represented a sicker and less homogenous cohort compared with institution 1. When matching for chronicity of the infection, the AUC of the model was not affected. This study highlights the impact of comorbidities and their distributions on PJI prognosis and brings to question the clinical utility of the algorithm which requires further external validation.
对于接受人工关节周围感染（PJI）手术的患者，术前计算治疗失败风险至关重要，这有助于进行医疗优化和有针对性的预防。先前已经开发了一种用于预测 PJI 治疗失败的术前预后模型，本研究旨在对该模型进行外部验证。
对两家机构治疗的 380 例 PJI 进行了回顾性研究。使用该模型计算治疗失败的风险，并生成每个机构的接收者操作特征曲线以计算曲线下面积（AUC）。
将该模型应用于机构 1 时，AUC 为 0.795（95%置信区间[CI]：0.693-0.897），而机构 2 的 AUC 为 0.592（95%CI：0.502-0.683）。比较应用该模型的所有机构，我们发现机构 2 代表了一个病情更严重、感染类型不同的人群。
在这项队列研究中，我们对先前发表的机构 1 模型进行了外部验证。然而，与机构 1 相比，机构 2 应用先前模型的 AUC 降低，代表了一个更严重且更异质的患者群体。当对感染的慢性程度进行匹配时，该模型的 AUC 不受影响。本研究强调了合并症及其分布对 PJI 预后的影响，并对该算法的临床实用性提出了质疑，需要进一步进行外部验证。
