We investigate the nexus between mental health outcomes in nursing home residents and the use of physical restraints. Previous studies in this area used limited statistical tests such as correlations and t-tests, that could not account for potential biases, such as residents who become mentally disturbed may be most likely to be restrained. We use propensity matching models that are less susceptible to this bias and data from the Minimum Data Set, representing approximately 2,000 residents over a period of 6 years. Our results clearly show that restrained residents are more likely to become more impaired with respect to cognitive performance, depression, and social engagement. We conclude that if facilities reduce restraint use then the prevalence of resident mental health problems will also likely decline.