Abstract
Missing data is a problem found in real-world datasets that has a considerable impact on the learning process of classifiers. Although extensive work has been done in this field, the MNAR mechanism still remains a challenge for the existing imputation methods, mainly because it is not related with any observed information. Focusing on healthcare contexts, MNAR is present in multiple scenarios such as clinical trials where the participants may be quitting the study for reasons related to the outcome that is being measured. This work proposes an approach that uses different sources of information from the same healthcare context to improve the imputation quality and classification performance for datasets with missing data under MNAR. The experiment was performed with several databases from the medical context and the results show that the use of multiple sources of data has a positive impact in the imputation error and classification performance. © 2019, Springer Nature Switzerland AG.

Abstract

Abstract

Abstract

Abstract
Background Ulcerative colitis [UC] is a chronic inflammatory disease often accompanied by severe and distressing symptoms that, in some patients, might require a surgical intervention [colectomy]. This study aimed at determining the risk of experiencing progressive disease or requiring colectomy. Material and Methods This was a multicentre study: patients' data [n = 1481] were retrieved from the Portuguese database of inflammatory bowel disease patients. Bayesian networks and logistic regression were used to build risk matrices concerning the outcomes of interest. Results The derivation cohort included a total of 1210 patients, of whom 6% required a colectomy and 37% had progressive disease [over a median follow-up period of 12 syears]. The risk matrices show that previously hospitalised patients with extensive disease, who are not on immunomodulators and who are refractory to corticosteroid treatment, are the ones at the highest risk of undergoing a colectomy [88%]; whereas male patients, with extensive disease and less than 40 years old at diagnosis, are the ones at the highest risk of experiencing progressive disease [72%]. These results were internally and externally validated, and the AUC [area under the curve] of the ROC [receiver operating characteristic] analysis for the derivation cohort yielded a high discriminative power [92% for colectomy and 72% for progressive disease]. Conclusions This study allowed the construction of risk matrices that can be used to accurately predict a UC patient's likelihood of requiring a colectomy or of facing progressive disease, and can be used to individualise therapeutic strategies.

Abstract
Healthcare institutions face serious security challenges, namely confidentiality, integrity and availability of patient's data due to the amounts of sensitive data collected on Health Information Systems (HIS) and the complex data management processes in health care. This paper describes scenarios of undue HIS access by staff in healthcare institutions, use cases (UC) that model the activities on HIS and identify the variables on audit trails (AT) logs that can be used to detect illegitimate actions on patients' data. Firstly, a survey was conducted through discussion meetings with Information Systems Director (ISD), Data Protection Officer (DPO) and a jurist to discuss their concerns about patient data access, followed by interviews to professionals from healthcare institutions to gather information about their routines and HIS access practices. Then, undue access scenarios were described and UC of activities on HIS which allow their detection were modelled. Lastly, necessary log variables were identified in order to produce algorithms for illegitimate accesses detection. UC and variables selected were matched with the specific requirements of Ministers Council Resolution (MCR) nr.41/2018 which provides guidelines for technology to be compliant with General Data Protection Regulations (GDPR). Discussions with ISD, DPO and the jurist, and professionals' interviews allowed us to describe nine scenarios of undue access. For each scenario we modelled one UC. 32 variables from different type of logs were identified for illegitimate access detection, of which 14 are mandatory according to MCR nr. 41/2018. Despite we might have some limitations related to poor HIS log quality, the mandatory data that logs must comply will be very useful for the development of UC presented. In addition, it is possible to request systems' vendors the improvement of logs' data to meet the detail we propose for this model, which may be very useful to comply not only with GDPR requirements but also with the Standard "Management of Information" (MOI.11) of Joint Commission International Standards for Hospitals (JCI) certification. As future work, we intend to develop the algorithms for the UC modelled, that will detect suspicious activities and produce alarmistic in their presence, testing them in real environment of a hospital to help Information Systems department and DPO on investigation and prevention of data breaches. (C) 2019 The Authors. Published by Elsevier B.V.