脉冲射频治疗慢性肩痛疗效预测模型开发与验证*

Development and Validation of a Predictive Model for the Efficacy of Pulsed Radiofrequency in Chronic Shoulder Pain

摘要:

目的：本研究旨在分析多种因素对脉冲射频(pulsed radiofrequency, PRF)治疗慢性肩痛(chronic shoulder pain, CSP)疗效的影响并构建预测模型，为临床制定个体化治疗方案参考。方法：选取2023年9月至2025年5月在武汉大学中南医院疼痛科住院的102例CSP患者，根据PRF后疗效分为有效组和无效组。通过单因素分析、相关性分析及LASSO回归筛选关键特征变量，再进行Logistic回归分析构建预测模型，并以列线图形式呈现。最后利用ROC曲线、决策曲线及校准度曲线对模型进行综合评价。结果：分析确定CSM评分、糖尿病、年龄、DBIL及TP为影响因素并纳入预测模型。构建的模型ROC曲线AUC为0.972，区分能力强；校准曲线显示预测结果与实际观察值高度一致；决策曲线分析显示模型在可接受风险阈值内有显著净获益。结论：本研究基于CSM评分、糖尿病、年龄、DBIL及TP等指标构建了PRF治疗CSP疗效预测模型，且具有高判别效能和良好临床适用性。

Abstract:

Objective: This study aims to analyze the impact of multiple factors on the efficacy of pulsed radiofrequency (PRF) treatment for chronic shoulder pain (CSP) and construct a predictive model to provide references for formulating individualized treatment plans in clinical practice. Methods: A total of 102 CSP patients admitted to the Department of Pain Medicine at Zhongnan Hospital of Wuhan University from September 2023 to May 2025 were selected and divided into effective and ineffective groups according to the therapeutic effects after PRF treatment. Key characteristic variables were screened through univariate analysis, correlation analysis, and LASSO regression, and then a predictive model was constructed using logistic regression analysis and presented in the form of a nomogram. Finally, the model was c omprehensively evaluated using ROC curves, decision curves, and calibration curves. Results: The analysis identified CSM scores, diabetes, age, and TP as influencing factors and incorporated them into the predictive model. The model demonstrated strong discriminative ability with an AUC of 0.972 in the ROC curve. The calibration curve showed a high consistency between the predicted results and the actual observations. The decision curve analysis indicated significant net benefit within the acceptable risk threshold. Conclusion: This study constructed a predictive model for the efficacy of PRF treatment for CSP based on CSM scores, diabetes, age, DBIL，and TP, which has high discriminative efficacy and good clinical applicability.