基于神经网络算法下肩关节术后患者院外长期照顾需求预测因素分析

Predictive factors analysis of long-term care demand of patients after shoulder surgery based on neural network algorithm

目的分析基于神经网络算法下肩关节术后患者院外长期照顾需求预测因素。方法回顾性选取2020年1月—2021年12月期间于秦皇岛市第一医院进行肩关节手术治疗的146例患者为研究对象,根据术后患者院外长期照顾需求分为需求组和无需求组,收集患者的临床资料,通过Logistic多元回归分析及深层神经网络模型(DNN)分析对患者院外长期照顾需求的影响因素进行预测,并比较两者之间预测效能的差异。结果需求组年龄、锚钉数及住院时间显著大于无需求组,肩关节功能UCLA评分显著高于无需求组,肌力显著低于无需求组(P<0.05)。Logistic回归模型后结果显示,一般资料中年龄、锚钉数、住院时间、UCLA评分、肌力是影响患者院外长期照顾需求的独立危险因素(均P<0.05)。DNN模型中训练样本的平均损失率为(4.82±0.58)%,测试样本的平均损失率为(7.65±3.64)%;DNN模型中训练样本的平均准确率为(96.36±2.65)%,测试样本的平均准确率为(91.17±1.04)%。DNN模型诊断效能显著优于Logistic回归分析模型(P<0.05)。结论DNN模型对于肩关节术后患者院外长期照顾需求具有较好的预测性,可以将之运用于临床院外长期照顾需求的预测中。

Objective To analyze the predictive factors of long-term care demand of patients after shoulder surgery based on neural network algorithm.Methods A total of 146 patients who underwent shoulder joint surgery in our hospital from January 2020 to December 2021 were retrospectively selected as the research objects.Patients were divided into demand group and non-demand group according to their long-term care needs outside the hospital after surgery.The clinical data of patients were collected,and the influencing factors of patients'long-term care needs outside the hospital were predicted by Logistic multiple regression analysis and deep neural network model(DNN)analysis,and the difference of prediction efficiency between the two was compared.Results The age,number of anchors and length of hospital stay in the demand group were significantly higher than those in the no-demand group,the UCLA score of shoulder joint function was significantly higher than that in the no-demand group,and and muscle strength was significantly lower than that in the no-demand group(P<0.05).The Logistic regression model showed that age,number of anchors,length of hospital stay,UCLA score and muscle strength were independent risk factors affecting patients'long-term care needs outside the hospital(P<0.05).In the DNN model,the average loss rate of training samples and test samples was(2.82±0.58)%and(7.65±3.64)%,respectively.The average accuracy of the training samples in DNN model was(96.36±2.65)%,and the average accuracy of the test samples was(91.17±1.04)%.The diagnostic performance of DNN model was significantly better than that of Logistic regression model(P<0.05).Conclusion The DNN model has a good predictability for the long-term out-of-hospital care needs of patients after shoulder surgery,which can be used to predict the long-term out-of-hospital care needs of clinical patients.