Construction And Verification Of Sleep Disorder Prediction Model In Patients With Systemic Lupus Erythematosus

Objectives This study comprehensively explored the influencing factors of sleep disorders in patients with systemic lupus erythematosus（SLE）,and constructed a predictive model of sleep disorders in patients with SLE and internal verification,in order to quickly provide clinical medical workers with the direction of early intervention,and then improve the sleep quality of patients with SLE.MethodsThe convenient sampling method was used to select 284 inpatients from the Department of Rheumatology and Immunology of Anhui Provincial Hospital,the Department of Rheumatology and Immunology of the first affiliated Hospital of Anhui Medical University and the Department of Nephrology of the second affiliated Hospital of Anhui Medical University.The research variables and tools were selected through literature review and previous meta-analysis,Pittsburgh sleep quality index（PSQI）scale was used to evaluate sleep quality.At the same time,the data were collected by self-rating anxiety scale（SAS）,self-rating depression scale（SDS）,social support rating scale（SSRS）,lupus quality of life（SLEDAI-2000）,the distress disclosure index（DDI）,fatigue scale（FSS）,visual analogue score of pain and so on.SPSS 23.0 was used for data collation,and R 4.0.2 was used for single factor analysis,lasso regression analysis and prediction model construction.The prediction model was constructed by logisitic regression analysis and decision tree,and the data was divided into training data and test data.The model effect was verified by training data and test data.The prediction effect of the model was evaluated by the area,sensitivity,specificity and accuracy under the subject working characteristic curve.ResultsThe sleep quality of patients with SLE was generally poor,and the incidence of sleep disorders was 44%.The univariate results showed that there were significant differences in education level,marital status,family per capita monthly income,medical payment method,exercise frequency,medical treatment,age,ESR,anxiety,depression,fatigue and lupus quality of life in the study of sleep quality in patients with SLE,which became the predictive model variable pool for sleep study in patients with SLE.Lasso regression further screened marital status,family per capita monthly income,medical payment method,exercise frequency,medical treatment,anxiety,depression,fatigue,quality of life（physical health,planning,emotional health,self-image evaluation）were the influencing factors of sleep disorders in patients with SLE.Comparing the test data verification effect of the two prediction models,the sensitivity was the same,the best critical point of the maximum logistic regression specificity and Jordan index was higher,the AUC was lower,and the accuracy was higher.The results of the two models complement each other,the prediction effect was similar and each had its own advantages.Two predictive models were used to study sleep disorders in SLE patients.The detailed results were as follows: The results of logistic regression training data construction prediction model were as follows: fatigue,anxiety,exercise frequency（5-7 times / week）and depression.Among them,the sensitivity of the training data model was 0.649,the specificity was 0.850,the best critical point was 0.592,the area under the subject’s working characteristic curve AUC was 0.815,AUC 95% confidence interval was（0.758,0.872）,and the model accuracy was 0.762,95% confidence interval（0.699,0.817）.Using the training data prediction model,the test data prediction model showed that depression was the predictor,OR（1.517,40.736）,the sensitivity of the test data model was 0.419,the specificity was 0.923,the critical value was 0.607,AUC was 0.702,AUC95% confidence interval was（0.578,0.826）,the model accuracy was 0.700,and the accuracy was 95% confidence interval（0.579,0.804）.The predictors of the decision tree model were physical health,planning,family per capita monthly income,anxiety,depression,exercise frequency,self-image evaluation,medical treatment and emotional health.The final predictors of the decision tree model were physical health,planning,family per capita monthly income and anxiety.Among them,the sensitivity of the training data model was 0.468,the specificity was 0.908,the best critical value was 0.610,the model accuracy was 0.715,and the accuracy was 95%confidence interval（0.649,0.774）.The test data carries on the internal verification,the sensitivity was 0.419,the specificity was 0.872,the critical value was 0.358,the AUC was 0.721,the 95% confidence interval was（0.608,0.833）,the model prediction accuracy was 0.671,the accuracy was 95%,the confidence interval was（0.549,0.779）.ConclusionThe sleep quality of patients with SLE is poor,and sleep is affected by many factors..In this study,10 important predictive variables of sleep disorders in patients with SLE were obtained by combining two prediction models: fatigue,anxiety,exercise frequency,depression,physical health,planning,family per capita monthly income,self-image evaluation,medical treatment and emotional health.7 predictors were finally formed in the two prediction models: fatigue,anxiety,exercise frequency,depression,physical health,planning,family per capita monthly income.Among them,fatigue,physical health,anxiety,depression and exercise frequency were the predictive variables that should be paid more attention to.In addition,both logistic regression and decision tree model have good prediction effect,and there was little difference,but the accuracy of logistic regression was slightly higher than that of decision tree model.Research on the prediction model of sleep disorders in patients with SLE can focus on the sleep conditions of patients with SLE,so as to reduce the incidence of sleep disorders and improve sleep conditions.