Introduction Practicing physical activity (PA) on a regular basis is an important support for people with type 1 diabetes (T1D). However, exercise may induce in them hypoglycaemic events during or after it. One major consequence of this is that, to limit this risk, many people with T1D tend to avoid performing PA. The availability of modern continuous glucose-monitoring (CGM) devices is potentially a great asset for reducing the chances of hypoglycaemia (HP) due to PA. Several algorithms have already been proposed to predict HP in subjects with T1D. However, not many of them are specifically focused on HP induced by exercise. Among those, many involve a large number of covariates making the applicability more difficult, and none uses CGM values available during the training session. Objectives We study the problem of predicting hypoglycaemia events in subjects with T1D during PA. The final aim is to produce algorithms enabling a person with T1D to perform a planned PA session without experiencing HP. Method One of the two algorithms we developed uses the CGM data in an initial part of a PA session. A parametric model is fitted to the data and then used to predict a possible HP during the remaining part of the session. Our second algorithm uses the CGM value at the start of a session. It also relies on statistical information about the average rate of decrease of the aforementioned model, as derived from a previously measured CGM data during PA. Then, the algorithm estimates the probability of HP during the planned PA session. Both algorithms have a very simple structure and therefore are of wide applicability. Results The application of the two algorithms to a very large dataset shows their very good ability to predict HP during PA in people with T1D.

Objectives: Cardiovascular diseases (CVDs) represent a major risk for people with type 1 diabetes (T1D). Our aim here is to develop a new methodology that overcomes some of the problems and limitations of existing risk calculators. First, they are rarely tailored to people with T1D and, in general, they do not deal with missing values for any risk factor. Moreover, they do not take into account information on risk factors dependencies, which is often available from medical experts. Method: This study introduces a Bayesian Belief Network (BBN) model to quantify CVD risk in individuals with T1D. The developed methodology is applied to a large T1D dataset and its performances are assessed. A simulation study is also carried out to quantify the parameter estimation properties. Results: The performances of individual risk estimation, as measured by the area under the ROC curve and by the C-index, are about 0.75 for both real and simulated data with comparable sample sizes. Conclusions: We observe a good predictive ability of the proposed methodology with accurate parameter estimation. The BBN approach takes into account causal relationships between variables, providing a comprehensive description of the system. This makes it possible to derive useful tools for optimising intervention.

Future occurrence of a disease can be highly influenced by some specific risk factors. This work presents a comprehensive approach to quantify the event probability as a function of each separate risk factor by means of a parametric model. The proposed methodology is mainly described and applied here in the case of a linear model, but the non-linear case is also addressed. To improve estimation accuracy, three distinct methods are developed and their results are integrated. One of them is Bayesian, based on a non-informative prior. Each of the other two, uses aggregation of sample elements based on their factor values, which is optimized by means of a different specific criterion. For one of these two, optimization is performed by Simulated Annealing. The methodology presented is applicable across various diseases but here we quantify the risk for cardiovascular diseases in subjects with type 1 diabetes. The results obtained combining the three different methods show accurate estimates of cardiovascular risk variation rates for the factors considered. Furthermore, the detection of a biological activation phenomenon for one of the factors is also illustrated. To quantify the performances of the proposed methodology and to compare them with those from a known method used for this type of models, a large simulation study is done, whose results are illustrated here.

Hepatitis E is considered an emerging foodborne disease in Europe. Several types of foods are implicated in the transmission of the hepatitis E virus (HEV) to humans, in particular, pork and wild boar products. We developed a parametric stochastic model to estimate the risk of foodborne exposure to HEV in the Italian population and to rank the relevance of pork products with and without liver (PL and PNL, respectively), leafy vegetables, shellfish and raw milk in HEV transmission. Original data on HEV prevalence in different foods were obtained from a recent sampling study conducted in Italy at the retail level. Other data were obtained by publicly available sources and published literature. The model output indicated that the consumption of PNL was associated with the highest number of HEV infections in the population. However, the sensitivity analysis showed that slight variations in the consumption of PL led to an increase in the number of HEV infections much higher than PNL, suggesting that PL at an individual level are the top risky food. Uncertainty analysis underlined that further characterization of the pork products preparation and better assessment of consumption data at a regional level is critical information for fine-tuning the most risky implicated food items in Italy.