In medical education, the medical interview is a fundamental skills that medical students need to learn. Practicing interviews with standardized patients (SPs), who simulate the behavior of actual patients, is known to be effective for developing medical interview skills. However, preparing qualified SPs who can act as patients in strict accordance with given clinical scenarios is both time- and cost-intensive. To address this issue, many studies have proposed virtual simulated patients (VSPs) utilizing large language models (LLMs). Conventional VSPs, however, often generate excessive responses, which are undesirable. Moreover, these methods typically rely on cloud-based LLMs, which raises significant concerns about data leakage and operational costs when used in certain educational and testing contexts. To overcome these challenges, this study proposes an approach to develop VSPs equipped with mechanisms to suppress excessive responses by utilizing an open-source LLM. Through experiments with actual data, we demonstrate that the proposed method significantly reduces excessive responses.