gms | German Medical Science

Introduction: Decision-making processes of patient treatment are not trivial, particularly in complex disease patterns. This applies especially to head and neck malignancies. Numerous information entities (IEs) are integrated into decision-making, oftentimes only in the brain of the health professional. Decision support tools can help in improving transparency and comprehensibility of complex situations and processes concerning treatment selection or during surgical interventions. Besides enhanced traceability of decisions, the aim is to include all known IEs of a tumor patient in the decision-making process and therefore provide a high quality standard of treatment. The first question in this context is how therapeutic decision processes are performed, on which IEs decisions are based and which tumor board decision emerges.

Methods: Due to the variety and complexity of IEs in decision-making, we concentrated on the modeling of therapy planning of laryngeal carcinoma. To detect crucial IEs for treatment decisions of patients suffering from laryngeal carcinoma, the following steps were conducted: 1) reviewing literature (e.g. [1]) and selecting relevant IEs, 2) categorizing and summarizing IEs into thematic groups, 3) detecting dependencies between single IEs, and 4) evaluating the relevance of each IE in the decision-making process. Simultaneously to literature review, decisions of the head and neck tumor board of the University of Leipzig concerning cases of laryngeal carcinoma in 2010 to 2012 were analyzed retrospectively. Furthermore, the process of decision-making and the subsequent board decision was recorded and evaluated regarding the cases discussed in the head and neck board starting in 2013.

Results: Basis for detection of IEs regarding laryngeal carcinoma, the guidelines of major cancer associations proved to be valuable. Also, well-documented standard works of head and neck oncology were consulted. From this database, about 120 IEs were identified, e. g. age, Karnofsky performance status scale, or extracapsular spread of lymph node metastasis. The IEs were divided into the following 4 categories: A) general patient information and risk factors, B) comorbidities, C) tumor-dependent data, and D) adverse treatment events. Through graphical presentation, the single IEs were correlated. Each IE was classified and labeled according to the putative relevance in the decision-making process. Approximately 70 out of 120 IEs were evaluated to be of medium- or high-grade relevance.

Discussion: Because of the multitude of IEs and the subsequent limited clarity, IEs of medium- or high-grade relevance were primarily addressed. The great complexity derives from the multitude of IEs not least on the basis of novel research results, that partially play no or a minor role in clinical practice. Therefore, the evaluation of the clinical relevance of the IEs was of predominant importance. However, that provokes the risk of underestimating the objective significance of single IEs. The next step towards a decision support tool is to model the medium- and high-grade relevant IEs and their dependencies. For this purpose, the usefulness of Multi Entity Bayesian Networks [2] will be studied. IEs concerning laryngeal carcinoma will be expanded and other tumor entities modeled subsequently to improve the therapy of patients suffering from head and neck malignancies.