Despite extensive research on hierarchical structures and community detection in complex networks, the interplay between these two mesoscopic features remains largely unexplored. Existing methods, including modularity optimization and dynamic approaches, often struggle to accurately identify ground-truth communities, particularly in networks with ambiguous community boundaries. In this paper, we address these challenges by introducing a novel hierarchical structure metric that captures higher-order adjacency relationships. Building on this metric, we propose the hierarchy-based community detection (HCD) algorithm, which incorporates a hierarchy-based node similarity (HS) measure. Unlike traditional similarity measures based solely on direct neighbors or edge density, the HS measure identifies structural centers within communities, enabling more precise and interpretable community detection. Extensive evaluations on diverse biological, social, and citation networks demonstrate that HCD achieves a notable 11.94% improvement in normalized mutual information (NMI) over state-of-the-art methods, highlighting its effectiveness in networks with ground-truth communities. Furthermore, HCD exhibits remarkable robustness, accurately identifying communities even in networks with weak boundary significance. By bridging the gap between hierarchical structure analysis and community detection, HCD offers a powerful tool for examining complex systems across diverse domains, from biology to social networks and knowledge graphs.