In order to solve the problem that the existing pattern mining methods do not make full use of the correlation between the abnormality of check items in the health examination data and specific diseases, this paper proposes a multi-objective Top-k frequent pattern mining approach oriented for health examination data. First, according to the characteristics of health examination data, two indicators of abnormality and coverage are proposed, and with these metrics, the Top-k frequent pattern mining is modeled as a multi-objective optimization problem. Then, an efficient evolutionary multi-objective optimization algorithm is designed to solve the problem, in which a preference-based population initialization strategy and a two-layer update strategy oriented to patterns and items are respectively proposed. The experimental results show that the achieved Top-k frequent patterns not only effectively reflect the correlation with the specific diseases, but also provide a variety of patterns, which gives an important reference for health management.