Abstract
Bayesian networks are one of the most intuitive statistical models for both estimation, classification and prediction of patients' outcomes. However, the availability of inference software in clinical settings is still limited. This work presents preliminary steps towards the creation of simple web-based forms that can access a powerful Bayesian network inference engine, making the derived models usable at bedside by both the clinicians and the patients themselves.

Abstract

Abstract
In obstructive sleep apnea, respiratory effort is maintained but ventilation decreases/disappears because of the partial/total occlusion in the upper airway. It affects about 4% of men and 2% of women in the world population. The aim was to define an auxiliary diagnostic method that can support the decision to perform polysomnography (standard test), based on risk and diagnostic factors. Our sample performed polysomnography between January and May 2015. Two Bayesian classifiers were used to build the models: Naive Bayes (NB) and Tree augmented Naive Bayes (TAN), using all 39 variables or just a selection of 13. Area under the ROC curve, sensitivity, specificity, predictive values were evaluated using cross-validation. From a collected total of 241 patients, only 194 fulfill the inclusion criteria. 123 (63%) were male, with a mean age of 58 years old. 66 (34%) patients had a normal result and 128 (66%) a diagnostic of obstructive sleep apnea. The AUCs for each model were: NB39 - 72%; TAN39 - 79%; NB13 - 75% and TAN13 - 75%. The high (34%) proportion of normal results confirm the need for a pre-evaluation prior to polysomnography. The constant seeking of a validated model to screen patients with suspicion of obstructive sleep apnea is essential, especially at the level of primary care.

Abstract

Abstract
Crohn's disease is one type of inflammatory bowel disease whose incidence is currently increasing, subject to relapse and disabling, with unknown etiology, and usually diagnosed between the second and third decade of life. The aim of this work is to develop a Bayesian network tool to predict disabling and reoperation in patients with Crohn's disease subject to early surgery or immunosuppressors intake. Multi-centric study data from patients with surgery or immunosuppression in the first six months after diagnosis was used, focusing on the prognosis and the analysis of factors' interaction. Patients were grouped by the index episode: immunosuppressors intake, and surgery (stratified considering the use or not of immunosuppressors 6 months after surgery). Patient group was associated with disease behavior, upper gastrointestinal tract location (L4) and age at diagnosis, while disease extent was associated to perianal disease. For disabling, association between perianal disease and gender and location was also found. Association between gender and L4 was also found for reoperation. The cross-validated discriminative power of the models were high for both disabling (above 70%) and reoperation (above 80%). The generated models presented interesting insights on factor interaction and predictive ability for the prognosis, supporting their use in future clinical decision support systems.

Abstract
Clinical record integration and visualisation is one of the most important abilities of modern health information systems (HIS). Its use on clinical encounters plays a relevant role in the efficacy and efficiency of health care. One solution is to consider a virtual patient record (VPR), created by integrating all clinical records, which must collect documents from distributed departmental HIS. However, the amount of data currently being produced, stored and used in these settings is stressing information technology infrastructure: integrated VPR of central hospitals may gather millions of clinical documents, so accessing data becomes an issue. Our vision is that, making clinical reports to be stored either in primary (fast) or secondary (slower) storage devices according to their likelihood of visualisation can help manage the workload of these systems. The aim of this work was to develop a model that predicts the probability of visualisation, within 24h after production, of each clinical report in the VPR, so that reports less likely to be visualised in the following 24 hours can be stored in secondary devices. We studied log data from an existing virtual patient record (n=4975 reports) with information on report creation and report first-time visualisation dates, along with contextual information. Bayesian network classifiers were built and compared with logistic regression, revealing high discriminating power (AUC around 90%) and accuracy in predicting whether a report is going to be accessed in the 24 hours after creation.