Abstract
Children with severe aplastic anemia (SAA) face heterogeneous prognoses after immunosuppressive therapy (IST). There are few models that can predict the long-term outcomes of IST for these patients. The objective of this paper is to develop a more effective prediction model for SAA prognosis based on clinical electronic medical records from 203 children with newly diagnosed SAA. In the early stage, a novel model for long-term outcomes of SAA patients with IST was developed using machine-learning techniques. Among the indicators related to long-term efficacy, white blood cell count, lymphocyte count, absolute reticulocyte count, lymphocyte ratio in bone-marrow smears, C-reactive protein, and the level of IL-6, IL-8 and vitamin B12 in the early stage are strongly correlated with long-term efficacy (P < .05). Taken together, we analyzed the long-term outcomes of rabbit anti-thymocyte globulin and cyclosporine therapy for children with SAA through machine-learning techniques, which may shorten the observation period of therapeutic effects and reduce treatment costs and time.