长期住院精神分裂症病人骨质疏松风险预测模型的构建及验证

Construction and validation of osteoporosis risk prediction model in long-term hospitalized patients with schizophrenia

目的建立长期住院精神分裂症病人骨质疏松风险预测模型,并进行验证。方法选择2018年1月至2021年10月在眉山市中医医院长期住院的精神分裂症病人436例进行回顾性分析。收集可能引起骨质疏松的相关因素,根据病人骨量情况分为两组,比较两组病人基本资料及生化指标,并以多因素logistic回归分析长期住院精神分裂症病人骨质疏松的影响因素,以此建立列线图模型并进行验证。结果多因素logistic回归分析结果显示:性别、年龄、服药年限、联合使用精神病药物、血钙、骨折史及血清泌乳素为长期住院精神分裂症病人骨质疏松的独立性影响因素(P<0.05)。以上述指标建立的列线图模型具有较高的区分度[ROC曲线下面积为0.70,95%CI:(0.65,0.75)]及预测准确度(平均绝对误差为0.02,预测曲线与标准曲线基本拟合)。结论长期住院精神分裂症病人骨质疏松主要受病人性别、年龄、服药年限等7个因素的影响,基于上述因素构建的列线图模型具有较高的区分度及预测准确度。

Objective To establish an osteoporosis risk prediction model in long-term hospitalized patients with schizophrenia,and to verify it.Methods A retrospective analysis was performed on 436 patients with schizophrenia who had been hospitalized in Meishan Hospital of Traditional Chinese Medicine from January 2018 to October 2021.The related factors that may cause osteoporosis were collected,and the patients were divided into 2 groups according to their bone mass.The basic data and biochemical indexes of the two groups of patients were compared,and multivariate logistic regression was used to analyze the influencing factors of osteoporosis in longterm hospitalized patients with schizophrenia,so as to establish a nomogram model and verify it.Results Multivariate logistic regression analysis showed that gender,age,duration of medication,combined use of psychiatric drugs,blood calcium,fracture history and serum prolactin were the independent influencing factors for osteoporosis in long-term hospitalized schizophrenia patients(P<0.05).The nomogram model established in this way has a high degree of discrimination[area under the ROC curve is 0.70,95%CI:(0.65,0.75)]and prediction accuracy(mean absolute error was 0.02,the prediction curve and the standard curve were basically fitted).Conclusion Osteoporosis in long-term hospitalized patients with schizophrenia is mainly affected by seven factors,including gender,age,and years of medication.The nomogram model constructed based on the above factors has high discrimination and prediction accuracy.