Risk Prediction Of Sleep Disturbance Among Nurses In Hospice Care Pilot Wards Based On Logistic Regression And Artificial Neural Networks

ObjectiveTo understand the prevalence of nurses ’sleep disturbance and influencing factors,establishing risk prediction models based on Logistic regression and artificial neural network,and evaluating the predictive efficacy of the two models in sleep disturbance which provided a simple and effective tool for early identification of sleep disturbance among nurses,and to provide reference for nurses’ sleep disturbance intervention in the future.MethodsThe convenience sampling method was used in the study.A total of 335 nurses in the hospice care pilot ward from March 2021 to July 2022 were selected.The self designed sociodemographic data questionnaire,the Maslach Burnout Inventory-General Survey(MBI-GS),the Fatigue Scale-14(FS-14),the Perceived Social Support Scale(PSSS),the Pittsburgh Sleep Quality Index(PSQI)were used to investigate the basic condition and sleep quality of the nurses in the hospice care pilot ward.Through the simple random sampling method,about 70%of the nurses was selected as the modeling set,the remaining 30%of the nurses was selected as the validation set.Univariate and binary logistic regression multivariate analysis used SPSS25.0 statistical software.The nomogram of multivariate logistic regression model was drawn using R4.1.3 software,constructed artificial neural network prediction model with the help of Multi Layer Perceptron(MLP)in the neural network.The Receiver Operating Characteristics(ROC)curves of the modeling set and the validation set were drawn separately,and the prediction effect of the model was evaluated according to the Area Under the Curve(AUC).Results1.A total of 235 nurses were included in the modeling set,of which 152 nurses had sleep disturbance,and the prevalence of sleep disturbance was 64.68%;A total of 100 nurses were included in the validation set,of which 62 nurses had sleep disturbance,and the prevalence of sleep disturbance was 62.00%.2.Age(χ~2=28.736),fertility(χ~2=11.316),cultural level(χ~2=11.062),night shift frequency(χ~2=6.421),physical activity time(χ~2=4.975),fatigue(t=-7.807),chronic diseases(χ~2=35.010),alcohol(χ~2=4.391),job burnout(Z=12.917),perceived social support(t=10.540)showed statistically significant differences in the single factor analysis results of sleep disturbance in nurses in pilot hospice care wards(P<0.05);and age(OR=4.945),physical exercise time(OR=0.264),chronic diseases(OR=3.038),job burnout(OR=4.275),fatigue(OR=1.597)and perceived social support(OR=0.905)were independent influencing factors of sleep disturbance of nurses in palliative care pilot ward(P<0.05).3.The logistic regression prediction model was LogitP=1.155+1.598×Age>40 years old-1.332× Physical exercise time(>1 hour/day)+1.111×Chronic disease(yes)+1.453 X Job burnout(severe)+0.468 X Fatigue-0.100×Perceived social support;According to the subjects’ age,chronic diseases,fatigue,perceived social support,physical exercise time and job burnout,the scores corresponding to each variable were read out from the nomogram to obtain the specific probability of sleep disturbance in the study subjects.MLP neural network model:the input layer contains 21 neurons,the hidden layer contains 6 neurons,and the output layer contains 2 neurons.4.In the modeling set,the sensitivity,specificity,area under ROC curve and Youden index of the Logistic regression model were 88.46%,82.13%,0.943(95%CI:0.914～0.971,P<0.001),70.59%,respectively.The sensitivity,specificity,area under ROC curve and Youden index of MLP neural network model were 90.60%,80.23%,0.954(95%CI.0.929～0.978,P<0.001),70.83%,respectively,and the Z test of AUC of the two models showed Z=0.576,P>0.05;In the validation set,the sensitivity,specificity,area under the ROC curve and Youden index of the Logistic regression model were 86.15%,82.86%,0.924(95%CI:0.872～0.975,P<0.001),69.01%,respectively.The sensitivity,specificity,area under ROC curve and Youden index of MLP neural network model were 88.89%,83.78%,0.957(95%CI:0.923～0.991,P<0.001),72.67%,respectively,and the Z test of the AUC of the two models showed Z=1.044,P>0.05.Conclusions1.The prevalence of sleep disturbance was high among nurses in hospice care pilot wards.2.Risk factors affecting the occurrence of sleep disturbance among nurses in the palliative care pilot ward include age,chronic diseases,job burnout and fatigue;The protective factors affecting the occurrence of sleep disturbance among nurses in the palliative care pilot ward were physical exercise time and perceived social support.The Logistic regression model and the artificial neural network model jointly screened out five major influencing factors of sleep disturbance among nurses in the palliative care pilot ward:age,physical exercise time,job burnout,fatigue and perceived social support.3.In this study,the sleep disturbance risk prediction model for nurses in palliative care pilot ward constructed based on Logistic regression and artificial neural network has high sensitivity and specificity,and has good prediction efficiency.4.The prediction model constructed in this study includes simple and easy to obtain indicators,which has certain guiding significance for early identification of sleep disturbance and the formulation of interventions,and is suitable for promotion and use in the nursing.