HIV相关神经认知障碍列线图风险预测模型构建与验证

The construction and validation of a nomogram risk prediction model for HIV-associated neurocognitive disorders

目的构建并验证HIV相关神经认知障碍(HAND)列线图风险预测模型,以期为早期识别高危患者及针对性预防干预提供有效的参考工具。方法采用横断面研究方法,2021年6月12日至11月30日现场问卷调查泸州市江阳区疾病预防控制中心前来随访的222例HIV/AIDS患者为建模组,调查内容包括自行设计一般资料调查表、蒙特利尔认知评估量表、患者健康问卷,基于其多因素Logistic回归分析结果构建HAND风险预测模型,应用R 4.1.2软件进行列线图可视化呈现,并于2021年12月1日至2022年1月31日采用与前期研究一致的方法收集泸县玉蟾街道社区卫生服务中心前来随访的95例HIV/AIDS患者为验证组对模型进行外部验证。分别通过ROC曲线及曲线下面积、Hosmer-Lemeshow(H-L)检验与校准曲线、决策曲线分析(DCA)评价模型的区分度、校准度与临床实用性。结果共纳入分析317例HIV/AIDS患者,168例判定为有认知障碍,HAND总体发生率为53.0%。构建的HAND列线图风险预测模型在建模组和验证组ROC曲线下面积分别为0.960(95%CI:0.936~0.983)与0.897(95%CI:0.821~0.951),校准曲线均显示出较好的一致性,H-L检验P均>0.05,模型预测准确率分别为90.5%与84.2%,DCA分别在3%~98%及22%~98%之间时均有较高的正收益率。结论本研究构建的HAND列线图风险预测模型预测效能较好,可用于早期个体化预测HIV/AIDS患者发生HAND的概率。建议重点关注HAND风险预测概率≥0.536的HIV/AIDS患者是否发生HAND,并采取针对性的预防干预措施。

Objective To construct and validate a nomogram risk prediction model for HIV-associated neurocognitive disorders(HAND),and to provide an effective reference for early identification of high-risk patients and targeted prevention and intervention.Methods Using the cross-sectional study method,222 HIV/AIDS patients who were followed-up by the Jiangyang Center for Disease Control and Prevention in Luzhou City from June 12,2021 to November 30,2021 were interviewed as the development group.The questionnaire included self-designed general information items,the Montreal Cognitive Assessment and the Patient Health Questionnaire.The HAND risk prediction model was constructed based on the results of multivariate logistic regression analysis,and the R 4.1.2 software was used to visualize the nomogram.Using the method consistent with the previous study,95 HIV/AIDS patients from Yuchan Street Community Health Service Center of Lu County were collected as the validation group to externally verify the model from December 1,2021 to January 31,2022.Discrimination,calibration and clinical validity of the model were evaluated by receiver operating characteristic(ROC)curve and area under curve,Hosmer-Lemeshow(H-L)test and calibration curves,and decision curve analysis(DCA),respectively.Results A total of 317 HIV/AIDS patients were included in the analysis,and 168 patients had cognitive disorders.The overall incidence of HAND was 53.0%.For the development and validation groups,the area under the ROC curve of the model was 0.960(95%CI:0.936-0.983)and 0.897(95%CI:0.821-0.951),respectively;the calibration curves showed a good consistency;the P values of the H-L test were all>0.05;the predictive accuracy of the model was 90.5%and 84.2%,respectively;the DCA had high positive yield between 3%-98%and 22%-98%,respectively.Conclusions The nomogram risk prediction model of HAND constructed in this study has good predictive efficiency,and can be used to individually predict the probability of HAND in HIV/AIDS patients at early stages.It is suggested to pay more attention to the occurrence of HAND in HIV/AIDS patients whose predicted probability of HAND≥0.536 and take targeted preventive intervention measures.