Application of PrefixSpan Algorithms for Disease Pattern Analysis

Abstract

PrefixSpan is a pattern-growth method for mining sequential patterns, and it is employed in this research for identifying disease trajectory patterns based on frequent subsequence analysis. One of the most beneficial features of this algorithm is the maintainable characteristics of original data order, especially for effectively and efficiently searching sequential patterns within a huge database. In this study, a medical database was adopted for disease transition analysis, and seven chronic diseases including diabetes, hyperlipidemia, hypertension, cerebrovascular disease, kidney disease, heart failure, and chronic obstructive pulmonary disease were mainly considered. By employing PrefixSpan algorithms, the statistical results of various combinations of chronic diseases with specific orders could be observed and compared. The results shows that patients suffered from hypertension (HTN) and followed by hyperlipidemia (DP) possess the most proportion among all subjects with a percentage of 37% (89,058/241,017). All statistical results of different combinations of seven chronic diseases, transition order, and proportional ranking were shown and discussed.