Physician Review Websites allow users to subjectively evaluate self-experienced health services. As these are private impressions, users provide deep insights into their lives while sharing their experiences after a visit to their doctors. Thereby, users accidentally disclose information on the Internet, what poses a serious threat to users' privacy and may lead to unforeseeable consequences. In this paper, we introduce our tool "Text Broom" that detects privacy breaches. For this purpose, we combine methods of Natural Language Processing such as Named Entity Recognition with domain-specific Machine Learning approaches in order to achieve a wide coverage of recognized violations of privacy.