慢性腰痛临床预测模型建立与验证

Establishment and Validation of a Clinical Prediction Model for Chronic Low Back Pain

目的 探究慢性腰痛相关致病因素,建立并验证慢性腰痛临床预测模型。方法 采用自行设计的慢性腰痛危险因素调查问卷,于2022年5-10月,对四川省成都市火车南站、龙州路等10个社区卫生服务中心共760名门诊患者进行问卷调查,收集相关信息。使用LASSO回归筛选潜在危险因素,以Logistic回归建立慢性腰痛临床预测模型,并验证模型准确性。通过ROC曲线、AUC值来判断模型的区分度,HL拟合优度检验、校准曲线图和计算C指数来判断模型的准确度。用Bootstrap法对模型进行内部验证,检验模型的内部稳定性,并对模型进行外部验证。结果 760名研究对象中有慢性腰痛434人,患病率为56.8%,通过LASSO回归从27个潜在因素中筛选出7个相关因素,其中6个是慢性腰痛发生的独立影响因素;据此构建的预测模型区分度C指数为0.815(95%CI:0.772~0.857)、准确度C指数为0.778(95%CI:0.772~0.857)。结论 以患者年龄、工作要求、桌椅舒适度、社会支持程度、运动强度、三餐是否规律构建的慢性腰痛预测模型,能够有效预测患者慢性腰痛的患病风险。

Objective To investigate the incidence and related pathogenic factors of chronic low back pain(CLBP)and to establish and validate a clinical prediction model.Methods From May to October in 2022,a self-designed risk factor questionnaire for chronic low back pain was used to collect relevant information from outpatients in 10 community health service centers in Chengdu,Sichuan province,including south railway station and longzhou road.LASSO regression was used to screen potential risk factors,and Logistic regression was used to establish a clinical prediction model for CLBP.The degree of fit,accuracy,discrimination of the prediction model were further validated by HL goodness of fit test,the C-index,calibration curve.Internal validation was performed using the Bootstrap to test the internal stability of the model and to externally validate the model after 5 months.Results There were 434 people with chronic low back pain among 760 study participants,with a prevalence rate of 56.8%.27 potential factors were screened by LASSO regression into 7 relevant factors,of which 6 were independent influences on the occurrence of CLBP.The prediction model constructed accordingly had a differentiation C-index of 0.815(95%CI:0.772-0.857)and an accuracy C-index of 0.778(95%CI:0.772-0.857).Conclusion The CLBP prediction model constructed by patients'age,work requirements,comfort of table and chair,social support,exercise intensity,and regularity of three meals can effectively predict whether patients have the risk of chronic low back pain.