DOI
https://doi.org/10.1023/A:1009802528071

Summary
In this article, the unsupervised artificial neural network (ANN) analysis to classify performance from complex clinical performance datasets was presented. The strategy was demonstrated for the case performance data the National Board of Medical Examiners (NBME) Computer-based Clinical Scenarios (CCS). The results were compared with the expert-rater model. The ANN-based models were able to extract the same patterns as the expert-rater model, thus, both models demonstrated consistent results. The visualization by the means of search path mapping enabled to perform a deeper analysis of the clustering results and determine the sensitivity to quantitative and qualitative test selections. The authors conclude that ANN-based approaches in combination with parametric methods can have additional value in problem-solving.

Relevance for Complex Systems Knowledge
The model provided in this article is dedicated to assessing the complex performance of the physician-patient management skills. The different diagnostic strategies including tests, medications, consultations and procedures are used in the constructive data modelling to build performance models using artificial intelligence methods. These models are later applied to model a computer-simulated patient in a realistic fashion during the assessment. The decisions made by the examinee are later analyzed in various aspects using unsupervised learning or artificial neural networks. This article is related to complex systems in two manners. Firstly, students’ competence to evaluate complex system such as patients’ medical health is developed by analyzing various parameters. Secondly, students’ performance during the diagnosis process is also considered as a complex behavior which consists of made decisions and steps taken.

Point of Strength
The point of strength is the provided application of unsupervised learning model and ANN to create a representing model in the cognitive process and asses a sequence of students’ steps.