The United States must improve its data infrastructure to ensure high-quality mental health care.

Use of and spending on mental health services in the United States more than doubled over the past two decades. In 2019, 19.2% of adults received mental health treatment (medications and/or counseling) at a cost of $135 billion. Yet, the United States has no data collection system to determine what proportion of the population benefited from treatment. Experts have for decades called for a learning behavioral health care system: a system that collects data on treatment services and outcomes to generate knowledge to improve practice. As the rates of suicide, depression, and drug overdoses in the United States continue to rise, the need for a learning health care system becomes even more pressing. In this paper, I suggest steps to move toward such a system. First, I describe the availability of data on mental health service use, mortality, symptoms, functioning, and quality of life. In the United States, the best sources of longitudinal information on mental health services received are Medicare, Medicaid, and private insurance claims and enrollment data. Federal and state agencies are starting to link these data to mortality information; however, these efforts need to be substantially expanded and include information on mental health symptoms, functioning, and quality of life. Finally, there must be greater efforts to make the data easier to access such as through standard data use agreements, online analytic tools, and data portals. Federal and state mental health policy leaders should be at the forefront of efforts to create a learning mental health care system.