Automatic detection of crosscutting concerns helps comprehension and refactoring of large-scale software system. Some of the detection techniques use occurrence frequency of program entities to detect concerns; however, the detected concerns include generic idioms such as a loop using iterator. The generic idioms are less interesting than applicationor domain-specific concerns in many cases. Therefore, some techniques use heuristics to filter the concerns, where developers should adjust the filter carefully or actual concerns may be accidentally filtered out. In this paper, we propose a metric named universality that represents how widely a class is used in applications. Using the universality, we filter concerns comprising only universallyused classes. We apply the method to coding patterns that include crosscutting concerns in various open source applications and discuss the result.