成人脊柱畸形手术中常用健康相关生活质量量表达成最小临床重要差异的可部署预测模型的开发
本研究基于两个前瞻性多中心成人脊柱畸形(ASD)队列的回顾性分析,训练8种预测算法,在术前或术后基线至1年和2年四个时间点预测患者报告结局(PROs)达成最小临床重要差异(MCID)的概率,最终纳入570例患者。模型R值范围为20%至45%,平均绝对误差(MAE)范围为8%至15%,术前基线PROs较差的患者平均改善最大。
要点
- 模型预测ASD手术后1年和2年PROs达成MCID的概率,R值20%–45%,MAE 8%–15%。
- 术前基线PROs较差的患者平均改善最大。
- 手术医生和中心不是模型的重要组成,对预测1年和2年PROs解释的方差很小。
核心结论
- 在迄今最大规模、前瞻性、手术、多中心、随访2年的队列中,提出了准确且一致的预测ASD手术后达成临床相关改善概率的方法。
- 该研究对共享决策、手术计划和术后咨询具有显著临床意义。
研究设计与样本
- 回顾性分析前瞻性收集的多中心成人脊柱畸形(ASD)数据库。
- 查询两个前瞻性观察性ASD队列,纳入有术前基线及术后1年和2年脊柱侧凸研究学会-22(SRS-22)、Oswestry功能障碍指数(ODI)和Short Form-36(SF-36)数据的患者。
- 使用75个变量训练模型,包括人口学特征、基线PROs和可修改的手术参数。
- 在四个时间范围训练8种预测算法:术前或术后基线至1年、术前或术后基线至2年。
- 通过80%/20%随机拆分完成外部验证,在训练样本内进行五折交叉验证。
主要发现
- 共纳入570例患者。
- 选择MAE最低的模型;根据预测结局不同,R值范围为20%至45%,MAE范围为8%至15%。
- 术前基线PROs较差的患者平均改善最大。
- 手术医生和中心不是模型的重要组成,对预测1年和2年PROs解释的方差很小。
关键数值与推荐
- 精确度测量:MAE和R值。
- 选择MAE最低的模型;R值20%–45%,MAE 8%–15%,取决于预测结局。
- 证据等级:4级。
临床怎么用
- 作者指出该研究对共享决策、手术计划和术后咨询具有显著临床意义。
局限性
- 证据等级为4级。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
STUDY DESIGN: Retrospective analysis of prospectively-collected, multicenter adult spinal deformity (ASD) databases. OBJECTIVE: To predict the likelihood of reaching minimum clinically important differences in patient-reported outcomes after ASD surgery. SUMMARY OF BACKGROUND DATA: ASD surgeries are costly procedures that do not always provide the desired benefit. In some series only 50% of patients achieve minimum clinically important differences in patient-reported outcomes (PROs). Predictive modeling may be useful in shared-decision making and surgical planning processes. The goal of this study was to model the probability of achieving minimum clinically important differences change in PROs at 1 and 2 years after surgery. METHODS: Two prospective observational ASD cohorts were queried. Patients with Scoliosis Research Society-22, Oswestry Disability Index , and Short Form-36 data at preoperative baseline and at 1 and 2 years after surgery were included. Seventy-five variables were used in the training of the models including demographics, baseline PROs, and modifiable surgical parameters. Eight predictive algorithms were trained at four-time horizons: preoperative or postoperative baseline to 1 year and preoperative or postoperative baseline to 2 years. External validation was accomplished via an 80%/20% random split. Five-fold cross validation within the training sample was performed. Precision was measured as the mean average error (MAE) and R values. RESULTS: Five hundred seventy patients were included in the analysis. Models with the lowest MAE were selected; R values ranged from 20% to 45% and MAE ranged from 8% to 15% depending upon the predicted outcome. Patients with worse preoperative baseline PROs achieved the greatest mean improvements. Surgeon and site were not important components of the models, explaining little variance in the predicted 1- and 2-year PROs. CONCLUSION: We present an accurate and consistent way of predicting the probability for achieving clinically relevant improvement after ASD surgery in the largest-to-date prospective operative multicenter cohort with 2-year follow-up. This study has significant clinical implications for shared decision making, surgical planning, and postoperative counseling. LEVEL OF EVIDENCE: 4.
引用
- Ames CP, Smith JS, Pellisé F, et al. Development of Deployable Predictive Models for Minimal Clinically Important Difference Achievement Across the Commonly Used Health-related Quality of Life Instruments in Adult Spinal Deformity Surgery. Spine (Phila Pa 1976). 2019. PMID: 30896589. doi:10.1097/BRS.0000000000003031
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。