237. The impact of artificial intelligence RPA driven clinical auxiliary diagnostic system on misdiagnosis rate of patients with mental disorders

Abstract Background The diagnosis of mental disorders highly relies on the subjective judgment and experience of clinical doctors. Traditional diagnostic processes have problems such as incomplete information integration and inconsistent application of diagnostic criteria, which may lead to an increased misdiagnosis rate. In recent years, artificial intelligence (AI) and robot process automation (RPA) technologies have provided a new path for optimizing medical processes. The clinical auxiliary diagnostic system driven by AI-RPA can reduce human errors by automatically collecting and analyzing multimodal data, and using machine learning algorithms to provide standardized diagnostic recommendations. The study aims to evaluate the impact of the AI-RPA assisted diagnostic system on the accuracy of diagnosing common mental disorders in real clinical scenarios, and quantify its changes in misdiagnosis rates, with the aim of optimizing the diagnostic process, improving diagnostic efficiency and accuracy. Methods The study adopted a prospective, multicenter, randomized controlled trial design. 1200 suspected patients with mental disorders who were seeking initial treatment were recruited from the psychiatric clinics of three tertiary hospitals. The patients were randomly divided into an intervention group (n = 600) and a control group (n = 600). In the intervention group's diagnosis and treatment process, doctors use an integrated AI-RPA assisted diagnostic system, which can automate interview guidance, real-time record and analyze patients' verbal and nonverbal characteristics, integrate historical medical records and scale scores, and generate preliminary diagnostic probability reports for doctors' reference. The control group used a traditional diagnostic process completely led by doctors. Using the standard diagnosis established by independent experts based on the Diagnostic and Statistical Manual of Mental Disorders and the International Classification of Diseases, combined with longitudinal follow-up information, 6 months later as a reference, the initial misdiagnosis rate of the two groups was compared. The study used SPSS 26.0 for statistical analysis, mainly using chi square test to compare the differences in misdiagnosis rates. Results The research results showed that the overall initial misdiagnosis rate of the intervention group was 8.5%, significantly lower than the control group's 15.2%, and the difference was statistically significant (p<.01). The analysis of different types of mental disorders shows that the system had the most significant reduction in misdiagnosis rate for early identification of bipolar disorder and schizophrenia spectrum disorders (p<.01), while there was no significant improvement in misdiagnosis rate for anxiety disorders (p>.05). In addition, the average diagnostic decision-making time of the intervention group doctors was shortened by 18%, and in highly complex cases, doctors' confidence scores in self diagnosis were significantly improved (p<.05). Discussion The research results indicate that the AI-RPA driven clinical auxiliary diagnostic system can effectively reduce the overall misdiagnosis rate of patients with mental disorders. Its mechanism of action may lie in the system providing more comprehensive and objective data integration and pattern recognition, reducing cognitive biases and information omissions. Future research will explore the best mode of human-machine collaboration to ensure that such technologies become reliable tools for improving the accessibility and quality of mental health services.