Clinical decision support system based on artificial intelligence and the patient’s subjective intention treatment model: A randomized controlled clinical trial.

10596 Background: To achieve intelligent diagnosis and treatment of tumors through the application of artificial intelligence and to meet the needs of areas lacking medical resources, a new treatment model based on the patient's subjective intention (PSI) was proposed, and a clinical decision support system (CDSS) called Doctors In Hands (DIH) was designed. Here, we report the DIH trial with the PSI treatment model to evaluate its efficiency. Methods: In this randomized controlled open-label trial (ChiCTR2400094469), patients from the departments of oncology, internal medicine, and surgery signed informed consent forms and were randomly divided into the clinician group, the CDSS plus clinician group and the CDSS group. The PSI treatment model was applied in the CDSS group. The primary endpoint was the sensitivity of alternative diagnoses and treatment options. The secondary endpoints were patient satisfaction and the specificity of alternative diagnoses and treatment options in different departments and groups. A multimodal large model based on the novel PSI treatment model was used as the intelligent platform for diagnosis and treatment data analysis and human‒computer interaction. Results: A total of 120 patients and 9 doctors were enrolled and randomly divided into three groups. In the clinician group, the sensitivity and specificity of diagnosis were 0.82 and 0.76, the Youden index was 0.58. In the CDSS group, the sensitivity and specificity were 0.85 and 0.80, the Youden index was 0.65. For the CDSS plus clinician group, the sensitivity and specificity were 0.90 and 0.87, the Youden index was 0.77. Subgroup analysis of patients according to treatment strategy revealed that the satisfaction of patients in the PSI-based CDSS group was significantly greater than that of patients in the clinician group (92% vs. 86%, P < 0.01), but there was no significant difference between the CDSS group and the CDSS plus clinician group (92% vs. 93%, P = 0.37). Subgroup analysis of the PSI-based CDSS group according to diagnostic strategy revealed that the diagnostic accuracy for patients from the department of oncology was significantly greater than that for patients from the department of internal medicine (0.88 vs. 0.81, P < 0.05) but was not significantly different from that for patients from the department of surgery (0.88 vs. 0.87, P = 0.14). Conclusions: DIH has good performance in the diagnosis of tumors and development of treatment plans. In the PSI model, patients can make independent choices regarding disease treatment plans according to the objective clinical or basic research evidence provided by DIH to reduce the duration of consultation and save medical resources. The clinical significance of this trial is that the CDSS can promote the sharing of medical resources and improve the efficiency and quality of clinical work in areas lacking medical institutions. Clinical trial information: ChiCTR2400094469 .