Clinical efficacy evaluation method using dominance-based rough sets and temporal hybrid information systems

Confronted with diverse treatment schemes for chronic diseases, the quantitative evaluation of their efficacy has become a key research topic in clinical management practice. In reality, the information about patient state often changes dynamically depending on the treatment scheme adopted. This dynamic may alternate between better, worse, and no change in the patient’s condition, which corresponds to the treatment schemes being evaluated as effective, ineffective, or ambiguous. The efficacy evaluation problem is essentially a classification problem with three classes. Based on real clinical scenarios, this study focuses on a class of efficacy evaluation problems with multi-granularity, time series, and hybrid information. In the current research, although many efficacy evaluation methods have been constructed, they still have some limitations in obtaining accurate interpretable evaluation results. On the one hand, the black-box nature of machine learning makes the clinical efficacy evaluation process difficult to interpret. On the other hand, the existing studies cannot achieve personalized efficacy evaluation with three classes. Notably, granular computing is a useful tool for dividing the efficacy of treatment schemes into three classes. To this end, we attempt to develop a novel efficacy evaluation method from the perspective of granular computing to address the considered problems. Specifically, the concept of multi-granularity temporal hybrid decision systems is first introduced to facilitate the representation of decision information. To capture changes in patient information at different time nodes, we define temporal dominance relations to accomplish the granulation of complex information. Furthermore, two temporal dominance-based rough sets are introduced, and their associated properties are explored. Given the applicability of temporal dominance-based fuzzy probabilistic rough sets, we subsequently present a classification method using grey correlation analysis and the given decision rules to evaluate the efficacy of the treatment schemes. Based on this, key attributes affected by each treatment scheme are identified. Through an empirical case of rheumatoid arthritis, we verify the feasibility of this method. The results of the comparison and sensitivity analysis demonstrate the superiority and stability of the proposed method. Overall, this study extends classical dominance-based rough set approaches and their real-world applications, to provide decision support for clinical efficacy evaluation.