Joep Tummers’s research while affiliated with Wageningen University & Research and other places

In the field of healthcare, as well as many others, textual descriptions of events are logged. With the use of Natural Language Processing (NLP), these texts are used to train event prediction machine learning algorithms. In this review the aim was to assess the state-of-the-art within current literature concerning prediction of events on textual records. Thus, this study follows a standard Systematic Literature Review (SLR) process. Primary articles are selected from PubMed, IEEE and WebOfScience with a search query, and then exclusion and quality assessment criteria are used to select the articles that are relevant to this study. Published performance metrics for the prediction algorithms used in the studies were then extracted from the included articles and used to assess the different methods. The general-purpose neural network algorithms: Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) and Conditional Random Fields (CRF) demonstrate the highest F1-scores amongst all the methods in this review, of 98.5%, 98% and 90.13% respectively. The algorithms that were designed specifically for NLP such as word2vec and BERT also performed well with F1-scores of 88.93% and 91.50%. This review does not give a comparison between methods but gives an indication about which machine learning methods perform well according to the authors of the selected studies. Not enough performance results are published under comparable circumstances to give conclusive results about which methods perform the best. More research needs to be done in comparing algorithms on the same dataset to proof the performance of the methods.

Background Care for people with an Intellectual Disability (ID) is complex: multiple health care professionals are involved and use different Health Information Systems (HISs) to store medical and daily care information on the same individuals. The objective of this study is to identify the HISs needs of professionals in ID care by addressing the obstacles and challenges they meet in their current HISs. Methods We distributed an online questionnaire amongst Dutch ID care professionals via different professional associations and care providers. 328 respondents answered questions on their HISs. An inventory was made of HIS usage purposes, problems, satisfaction and desired features, with and without stratification on type of HIS and care professional. Results Typical in ID care, two types of HISs are being used that differ with respect to their features and users: Electronic Client Dossiers (ECDs) and Electronic Patient Dossiers (EPDs). In total, the respondents mentioned 52 unique HISs. Groups of care professionals differed in their satisfaction with ECDs only. Both HIS types present users with difficulties related to the specifics of care for people with an ID. Particularly the much needed communication between the many unique HISs was reported a major issue which implies major issues with inter-operability. Other problems seem design-related as well. Conclusion This study can be used to improve current HISs and design new HISs that take ID care professionals requirements into account.

Background Healthcare relies on health information systems (HISs) to support the care and receive reimbursement for the care provided. Healthcare providers experience many problems with their HISs due to improper architecture design. To support the design of a proper HIS architecture, a reference architecture (RA) can be used that meets the various stakeholder concerns of HISs. Therefore, the objective of this study is to develop and analyze an RA following well-established architecture design methods. Methods Domain analysis was performed to scope and model the domain of HISs. For the architecture design, we applied the views and beyond approach and designed the RA’s views based on the stakeholders and features from the domain analysis. We evaluated the RA with a case study. Results We derived the following four architecture views for HISs: The context diagram, decomposition view, layered view, and deployment view. Each view shows the architecture of the HIS from a different angle, suitable for various stakeholders. Based on a Japanese hospital information system study, we applied the RA and derived the application architecture. Conclusion We demonstrated that the methods of the software architecture design community could be used in the healthcare domain effectively and showed the applicability of the RA.

Background: The lockdown-measures in response to COVID-19 taken by long-term care organisations might have impacted problem behaviour and behavioural functioning of people with intellectual disability. This study tested changes in reported incidents, in particular regarding aggression, unexplained absence and, for contrast, medication errors. Methods: Metadata on weekly incident and near-incident reports from 2016 to June 2020 involving over 14 000 clients with mild to serious intellectual disability of 's Heeren Loo, a long-term care organisation for people with intellectual disability, were subjected to interrupted time series analysis, comparing the COVID-19 with the pre-COVID-19 period. Results: The imposition of lockdown-measures coincided with a significant drop in incidents (total, P < .001; aggression, P = .008; unexplained absences, P = .008; and medication errors, P < .001). Incidents in total (P = .001) and with aggression (P < .001) then climbed from this initial low level, while medication errors remained stably low (P = .94). Conclusion: The rise in incidents involving aggression, against the background of generally lowered reporting, underlines the need for pandemic control measures that are suitable for people with intellectual disability in long-term care.