Construction Of A Risk Prediction Model For Recurrent Falls In Patients With Parkinson’s Disease

Objectives1.To integrate and analyze the influential factors of recurrent falls in Parkinson’s disease patients.2.To construct a risk prediction model for Parkinson’s disease patients with recurrent falls,and present it in the form of nomogram,and evaluate the predictive performance of the model.Methods1.Extraction of factors influencing recurrent falls in Parkinson’s disease patients.Through the method of evidence-based medicine,computer was used to comprehensively search the Chinese and English databases,find relevant literatures on the influencing factors of Parkinson’s disease patients repeated falls,conduct quality evaluation,data extraction and meta-analysis of the included literatures,and preliminatively identify the influencing factors related to Parkinson’s disease patients repeated falls.On the basis of literature selection,the group discussion finally formed the investigation program of the influencing factors of repeated falls in Parkinson’s disease patients.2.Construction of a risk prediction model for recurrent falls in Parkinson’s disease patients.In this cross-sectional study,based on the influencing factors obtained in part I,patients from a third-class Class A neurological hospital in Tianjin were selected as the research objects,basic information of patients and related data of influencing factors were collected,and patients’ repeated falls in the past year(the number of falls ≥2 times)were used as outcome indicators.Logistic regression model was used to construct the risk prediction model of repeated falls in Parkinson’s disease patients,and it was represented by nomogram.The prediction performance of the model was evaluated by area AUC under Receviver Operating Characteristic Curve,calibration diagram,sensitivity and specificity.Results1.Results of meta-analysis A total of 13 articles were included in this part,including 9cohort studies and 4 cross-sectional studies,involving a total of 1625 PD patients.The results of quantitative analysis of 17 factors were as follows: age: MD=1.50,95%CI(1,1.99);Gender:OR=0.85,95%CI(0.66,1.08);Body Mass Index: MD=-1.52,95%CI(-1.88,-1.16);Course of disease: MD=3.56,95%CI(3.26,3.86);Disease stage: MD=0.42,95%CI(0.36,0.49);Motion fluctuation: OR=1.80,95%CI(1.11,2.93);Frozen gait: OR=3.71,95%CI(2.17,6.33);Activities of daily life: MD=-11.35,95%CI(-14.47,-8.23);Balance confidence: MD=-16.44,95%CI(-19.39,-13.49);Cognitive level: MD=-0.92,95%CI(-1.86,0.02);Unified Parkinson’s disease Rating ScaleⅢ: MD=8.22,95%CI(6.07,10.37);Levodopa Equivalent Dose:MD=245.49,95%CI(173.77,317.20);International Fear of Falling Scale score: MD=9.35,95%CI(5.80,12.90);Geriatric depression Scale score: MD=0.77,95%CI(-0.25,1.78);Berg Balance Scale score: MD=-5.90,95%CI(-10.19,-1.61);Functional Stretching Test:MD=-4.78,95%CI(-5.87,-3.69);Timing standing up and walking test: MD=3.83,95%CI(2.27,5.38).2.The results of the risk prediction model of the nomogram A total of 230 PD patients were included in this study,among whom 53 had recurrent falls,and 177 had no falls or had one fall,the incidence of repeated falls was 23%.In univariate analysis,there were statistically significant differences in frozen gait,movement fluctuation,orthostatic hypotension,benzodiazepines,depression,course of disease,H-Y stage,levodopa equivalent dose,confidence score of specific activity balance,and timing standing and walking test between two groups(P < 0.05).The final course of disease(OR=1.118,95%CI=1.007～1.240),Orthostatic Hypotension(OR=2.683,95%CI=1.250～5.759),levodopa equivalent dose(OR=1.004,95%CI=1.001～1.007),motor fluctuation(OR=2.335,95%CI=1.051～5.187),The specific activity balance confidence score(OR=0.965,95%CI =0.934～0.988),moderate depression(OR=3.567,95%CI=1.035～12.296)entered the model.The AUC of the model was0.806,95%CI was(743,0.868),the sensitivity and specificity of the model were 84.9% and61.6%,respectively.The calibration chart showed that the actual value was basically consistent with the predicted value,and the clinical decision curve suggested that the model had certain clinical application value.Conclusions1.Meta analysis showed that age,course of Parkinson’s disease,body mass index and disease stage,motion fluctuations,freezing gait,unified Parkinson disease rating scale Ⅲ,equivalent dose levodopa,balance,confidence,falling fear,daily life activity,time to stand up,walk test,functional test and extend Berg balance test has influence on Parkinson’s disease falls over and over again Statistical significance;Gender,cognitive level and depression had no statistical significance on recurrent falls in Parkinson’s disease patients.2.The predictors of recurrent falls risk in Parkinson’s disease patients were course of disease,orthostatic hypotension,levodopa equivalent dose,movement fluctuation,balance confidence and moderate depression.The model fits well and has good differentiation,specificity,sensitivity and clinical applicability.