A Bayesian network for modelling blood glucose concentration and exercise in type 1 diabetes

摘要: This article presents a new statistical approach to analysing the effects of everyday physical activity on blood glucose concentration in people with type 1 diabetes. A physiologically based model dynamics is developed cope frequently sampled data food, insulin and habitual activity; then converted Bayesian network account for measurement error variability physiological processes. simulation study conducted determine feasibility using Markov chain Monte Carlo methods simultaneous estimation all parameters prediction concentration. Although there are problems parameter identification minority cases, most can be estimated without bias. Predictive performance unaffected by misspecification insensitive misleading prior distributions. highlights important practical theoretical issues not previously addressed quest an artificial pancreas as treatment The proposed represent paradigm analysis deterministic mathematical models