目的 构建产后16～20周骨盆带疼痛(pelvic girdle pain，PGP)快速预测评分模型，为临床提供快速判断PGP风险的简易工具。方法 便利抽样法选取漳州、无锡市2所三级甲等医院复诊的产后16～20周产妇为研究对象。筛选PGP影响因素后构建风险预测模型并进行验证。结果 二元Logistic回归分析显示:当前体质量指数≥24、孕前下腰痛、孕期下腰痛、婴儿主要照顾者为产妇、最近1周哺乳姿势为坐姿、经产妇、剖宫产均为产后PGP影响因素。建模组模型曲线下面积(area under the curve，AUC)为0。788，最佳截断值5。5，敏感度74。88％，特异性66。67％。院内验证组AUC为0。790，敏感度73。91％，特异性68。53％。院外验证组AUC为0。760，敏感度70。21％，特异性66。67％。0～2分为PGP极低风险组，3～5分低风险组，6～10分为高风险组，11～15为极高风险组。结论 构建的产后16～20周PGP风险预测评分模型有较好的区分度和校准度，方便临床对产妇PGP风险进行快速评估。
Construction and Validation of A Rapid Risk Prediction Scoring Model for Pelvic Girdle Pain 16-20 Weeks Postpartum
Objective To construct a rapid risk prediction scoring model for pelvic girdle pain(PGP)16 to 20 weeks postpartum,in order to provide a simple tool for clinical diagnosis of PGP risk.Methods Conven-ience sampling method was used to select re-examined postpartum women from a tertiary A hospital in Zhangzhou and Wuxi,respectively.After screening the influencing factors of PGP,a risk prediction model was constructed,and the predictive efficiency of the model was verified.Results Binary logistic regression analysis showed that influencing factors of postpartum PGP included body mass index(BMI)24,lower back pain before and during pregnancy,primary caregiver of the baby being the mother,sitting position during the last week of breastfeeding,multiparous women,and cesarean section.The area under the curve(AUC)of the model,optimal cut-off value,sensitivity,and specificity in the modeling group were 0.788,5.5,74.88％,and 66.67％,respectively.AUC,sensitivity,and specificity in the internal hospital validation group were 0.790,73.91％,and 68.53％,respectively,and those in the external hospital validation group were 0.760,70.21％,and 66.67％,respectively.Those with 0-2 points,3-5 points,6-10 points and 11-15 points were divided into very low risk group,low risk group,high risk group,and very high risk group of PGP.Conclusions The model constructed has good discrimination and calibration,making it convenient for rapid assessment of PGP risk in clinical medicine.
万方数据
维普
