Objective To determine the factors associated with aggressive behaviors of patients with schizophrenia by gender in communities in Shanghai， and further develop and validate the prediction model.

Methods A total of 7 955 community patients with schizophrenia were investigated in Xuhui District， Hongkou District and Jiading District of Shanghai. Baseline information was collected from April 2018 and follow-up was conducted every 3 months for 6 months. Multivariate logistic regression was used to calculate the odd ratio （OR） and 95% CI， and determine the factors associated with aggressive behaviors of patients. The risk score for each patient was developed based on the β coefficient， and the best cut-off value was determined by the Youden index. For the models， predictive ability was determined using area under the curve （AUC） of receiver operator characteristic curve （ROC curve）， and internal validation ability was evaluated by the ten-fold cross validation method.

Results There were 3 563 males in this study with an average age of （54.83±13.72） years old， and the incidence of aggressive behaviors was 2.55%.There were 4 392 females with an average age of （57.20±14.98） years， and the incidence of aggressive behaviors was 2.64%. For male patients with schizophrenia， single/divorced status （OR=2.04， 95%CI： 1.15‒3.61）， low economic status （OR=2.79， 95%CI： 1.71‒4.54）， irregular medication （OR=4.35， 95%CI：2.23‒8.47）， no medication （OR=1.83， 95%CI：1.03‒3.26）， incomplete/no insight （OR=1.97， 95%CI：0.99‒3.94）， adverse drug reaction （OR=2.61， 95%CI：1.27‒5.37）， psychiatric symptoms involving violence （OR=2.06， 95%CI：1.01‒4.18）， history of aggression （OR=5.29， 95%CI：2.33‒11.98） and recent stress events （OR=8.36， 95%CI：4.13‒16.92） were associated with aggressive behaviors. In contrast， for female patients， age less than 60 years （50‒59 years， OR=2.09， 95%CI： 1.13‒3.87； 40‒49 years， OR=2.74， 95%CI： 1.46‒5.17； 30‒39 years，OR=2.88， 95%CI： 1.48‒5.60； 18‒29 years， OR=5.71， 95%CI： 2.44‒13.37）， educational level of high school and above （senior high school， OR=3.30， 95%CI： 1.46‒7.49； college and university， OR=2.88，95%CI： 1.21‒6.82）， unemployed status （OR=1.81， 95%CI=1.17‒2.82）， irregular medication （OR=7.87， 95%CI：4.75‒13.05）， no medication （OR=2.11， 95%CI：1.24‒3.62）， adverse drug reaction （OR=2.75， 95%CI：1.50‒5.04）， psychiatric symptoms involving violence （OR=3.08， 95%CI：1.77‒5.37）， social function （OR=3.51， 95%CI：2.07‒5.94） and recent stress events （OR=5.92， 95%CI： 2.82‒12.44） were risk factors. In both male and female， the prediction models for aggressive behaviors of community patients with schizophrenia had strong predictive ability （AUC=0.779， 95%CI： 0.725‒0.834； AUC=0.822， 95%CI： 0.780‒0.863）.

Conclusion This study suggests that diverse risk factors should be considered for community patients with schizophrenia by gender to prevent the aggressive behaviors.