Journal of International Psychiatry

Objective: to establish a relapse model of schizophrenia patients and analyze the influencing relapse factors of schizophrenia patients with random forest method. Methods: collect the outpatient records, inpatient information, community follow-up notes and other informations of schizophrenic patients, structurally process the data, use Python to analyze the data, establish a prediction model of relapse. Results: the influencing factors with statistical significance were the number of visits every year, stimulating events, self-care ability and learning ability (P < 0.05). After random forest analysis, the number of visits, stimulation events, personal life care and learning ability are important trade-off factors in the prediction model. The recurrence prediction model has a hit rate of 46.41% ,The accuracy rate was 84.4%, and a coverage rate of 91.23%. Conclusion: the prediction model can effectively predict the relapse of schizophrenia and provide early warning information.

Key words: Schizophrenia; relapse; random forest; influencing factors; predictive model