Associate Professor | Poznan University of Technology, Poland
Szymon Wilk is an Associate Professor at the Poznan University of Technology (PUT) and an Adjunct Professor at the University of Ottawa (UofO). He graduated with a Ph.D. in knowledge-based decision support systems from PUT and defended a habilitation thesis on intergrative clinical decision support systems. His research interests include various aspects related to supporting clinical decisions with a special focus on constructing diagnostic and therapeutic models from imperfect (inconsistent, incomplete and imbalanced) data, mitigating adverse interactions between clinical practice guidelines (CPGs) in multi-morbid patients and customizing CPGs according to patient preferences, and assisting interdisciplinary healthcare teams in executing clinical workflows. Szymon is also a member of the MET (Mobile Emergency Triage) Group at UofO where he has been involved in design and development of an agent-based clinical decision support platform. The platform, after having been customized for several specific problems (e.g., evaluation of asthma severity), underwent a successful clinical trial at the Children’s Hospital of Eastern Ontario in Ottawa.
Clinical practice guidelines (CPGs) are defined as “systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances”. Interest in CPG development and use inspired the health informatics community to work on methods for structuring and computerizing CPG representations so they can be automatically processed. This line of research produced a number of guideline representation languages, such as GLIF or PROforma. However, almost all of the research on guidelines is focused on developing and executing CPGs designed for the management of a single clinical condition and lack of support for handling multi-morbidities limits practical adoption of traditional (paper-based) and computerized CPGs. The presented research addresses the above challenge and focuses on patients with multi-morbidities who are managed according to several diseases-specific CPGs. It proposes a mitigation framework for personalizing the applied. Specifically, we define the mitigation problem as the need to identify and address adverse interactions associated with applying multiple CPGs to the same patient, while taking into account the temporal characteristics of the activities in the CPGs and patient preferences related to prescribed treatments. To solve this problem, we propose the framework that provides a formal representation for the CPGs (considered as primary medical knowledge) and the secondary medical knowledge related to adverse interactions and patient preferences, and a mitigation algorithm to identify and address adverse interactions for a multi-morbid patient. The proposed framework assumes that CPGs applied to manage the patient are represented as actionable graphs and it uses this representation during subsequent processing of the CPGs. It adopts first-order logic (FOL) for the internal representation of the mitigation problem and related knowledge. Specifically, it introduces a mitigation-specific FOL language to describe core components of the mitigation problem (actionable graphs, available patient information), a suggested treatment, and secondary medical knowledge related to mitigating adverse interactions and considering patient preferences. The mitigation algorithm employs FOL-based reasoning techniques such as theorem proving and model finding and it generates a management scenario that describes a preferable and interaction-free treatment of a patient and the patient’s possible future state, or it provides a warning if such scenario does not exist. Application of the mitigation framework is illustrated with a clinical case study that describes the treatment of a patient suffering from chronic kidney disease and hypertension, who experiences a sudden onset of atrial fibrillation.