Development Of A Predictive Model For Readmission Of Major Depressive Disorder Patients With Suicide Attempts

Objective: A chronic disease known as major depressive disorder(MDD)is characterized by a significant probability of relapse.In combination with previous studies,various factors affecting their relapse may include biological,psychological,and social factors.Suicide attempts,as one of the possible influencing factors,are often the most devastating for MDD patients,so it is important to explore the factors affecting relapse in MDD patients with suicide attempts.As a good indicator of relapse dynamics,this study set the occurrence or non-occurrence of readmission as an outcome event of the study,aiming to explore the factors associated with readmission in first-time hospitalized MDD patients with suicidal attempts and attempting to develop a predictive model eventually.Methods:1.Establish the data set: The study enrolled 3977 MDD patients.By reviewing the patient visit records in the hospital information system(HIS),biochemical parameters,demographic characteristics,the clinical course,symptom characteristics,family history,concomitant somatic disorders,and tobacco and alcohol habits were acquired.The data was randomly split into a training set and a validation set in R software.2.Filter variables: Statistical descriptions and univariate logistic regression analysis were performed on the training set.The multivariate logistic analysis incorporated factors that were linked to the outcome in univariate analysis to create a stepwise regression model(with a P threshold<0.2).3.Visualization and evaluation of predictive models: Then visualize it with Nomogram.In the next step to evaluate the model,we successively evaluated the model with the training set and validation set data by three directions from the discrimination,calibration,and clinical decision curve analysis,respectively.Results:1.There was a significant difference in the incidence of readmission between first-hospitalized MDD patients with suicide attempts and those without suicide attempts,and the former is higher than the latter(χ2=7.338,P=0.007).2.Length of stay,course of disease,somatic symptoms,psychosis symptoms,anxiety symptoms,alcohol/tobacco dependence,family history of psychiatric disease,continued use of antidepressants combined with antipsychotics or not after discharge,and total bilirubin(TBIL)were included to build a predictive model in R software called “model.step”.3.The AUCROC was 0.783 for the training set and 0.778 for the validation set.The Hosmer-Lemesho(H-L)test for calibration was P= 0.3521 for the training set and P= 0.8497 for the validation set.Conclusion: In conclusion,the study showed that the incidence of readmission was significantly higher in first-hospitalized MDD patients with suicide attempts than in first-hospitalized MDD patients without suicide attempts.The model built by these factors has acceptable discrimination and calibration,and also has a certain clinical applicability.For better clinical applicability and a better prognosis for MDD patients,the model still has to be further tuned after many validations in several centers.