| Objective To explore the risk factors of cancer inpatients fall,build universal cancer inpatients fall risk prediction model and map nomogram,in order to accurately predict cancer inpatients fall finale,help clinical medical staff quickly identify patients at risk of falls,to make effective preventive measures,provide the theoretical basis to protect the safety of patients.Methods ① Through literature review,expert meeting method,combined with clinical practice,and comprehensive analysis,the "fall Risk factors Data Collection Table for inpatients with cancer"was formulated.The survey contents included general information,cancer-related information,laboratory results,drug use,clinical symptoms,use of auxiliary devices,family company,etc.②Cancer inpatients fall risk prediction model building and validation:This study adopts the research method of retrospective nested case-control study,according to the inclusion and exclusion criteria,120 inpatients who fell in a third-class tumor specialty hospital in Zhejiang Province from January 1,2018 to September 30,2021 were selected.At the same time,the ratio of 1:4 in the ward and admission time(±2 days)matched 480 patients who did not fall in the same period as the control group.A total of 600 patients were collected.Using SPSS26.0 software to randomly take the 7:3 ratio as the training set(A total of 420 cases,of which 87 fell(20.7%),333(79.3%)did not fall)and verification set(A total of 180 cases,of which 33 fell(18.3%),147(81.7%)did not fall)of the model,the one-way analysis and binary logistic regression analysis were carried out in the training set,combined with the partial regression coefficient β value and intercept of the independent risk factors,the regression equation was constructed.The prediction model of the fall risk factors of cancer hospitalized patients was formed.The nomograms was drawn in R language.Receiver Operating Characteristic(ROC),Area Under Curve(AUC)and Hosmer-Lemeshow test are used to evaluate the calibration and differentiation of the model,and the predictive power of the model is evaluated using indicators such as specificity and sensitivity.③Application evaluation of fall risk prediction model:The prediction model is applied to the data of the verification set,and the prediction accuracy of the model in the validation set is calculated by(true positive rate+true negative rate)/total number of cases in the verification set,and draw the ROC curve.Results ① Multivariate analysis showed that permanent residence,activities of daily living(Barthel index),history of falls in the past year,treatment methods,serum sodium,serum potassium,antiemetics,gait imbalance,postural hypotension,nocturia and family accompany were independent predictors of falls in cancer inpatients.②The model equation is Logit(P)=0.84+0.882*Habitual place of residence+2.854*History of falls in the past year-1.429*Daily activity ability(BI)(1)-0.623*Daily activity ability(BI)(2)-2.049*Treatment modality(1)1.401*Treatment modal ity(2)-1.895*Treatment modality(3)-0.09*Treatment modality(4)+1.744*Gait imbalance+1.697*Serum sodium+1.342*Serum potassium+2.074*antiemetic drug+2.645*orthostatic hypotension+2.098*increased nocturia-2.309*family accompaniment.The H-L test of the model is P>0.05,the area under curve was 0.931(95%CI:0.901～0.960),it indicates that the model has good calibration and differentiation.③In the verification set,there were 33 cases of actual falls,147 cases of actual non-occurrences,24 cases of model predicted falls,156 cases of model prediction of falls did not occur.The accuracy of the model in the validation set is 88.3%and the area under curve was 0.907(95%CI:0.856～0.958).Conclusion The fall risk prediction model of cancer inpatients constructed in this study has good predictive ability,which can provide a theoretical basis for personalized management of fall risk groups in cancer inpatients. |
PDF Full Text Request