Enhancing Healthcare Quality with Reinforcement Learning Modeling

Optimizing healthcare efficiency is a major concern today. Currently, medical diagnostics can be inaccurate, and when translated to medical treatment strategies, like choosing if to operate or not, and which drag type and optimal dose, it might affect the medical treatment effectivity. Predictive models can be used for the design of treatment plan, which is a challenging task due to large number of parameters and the nonlinearity of the problem. A major challenge in these models is to model the natural subject response - spontaneous and in respond to the treatment. In this work, we suggest design of predictive models that can incorporate in optimal manner the human internal mechanism for his/her healthcare maintenance to a medical expert system by using reinforcement learning model with natural and medical agents. The personal and public medical records, medical knowledge, sensory inputs, including brain recording when available, can be used not just as constraints to the medical agent, but to estimate the natural agent states and reward. This new model suggestion has a potential to improve healthcare quality, with better control of the desired level of medical intervention.