Why Use System-Level Data Sets?

Data from nationally representative or other well-defined large-scale populations and events are a valuable resource for research on mental illness and disability. They are most appropriate for conducting epidemiologic research, studies of health care utilization and outcomes, and investigations of patient or provider characteristics. These data sets come from medical and behavioral health care practitioners, provider organizations, federal and private payors, and registries such as death records. System-level data sets are complex and multi-faceted, which can make them challenging to understand and analyze. However, these types of data are the best sources of information on important topics such as morbidity and mortality, utilization and costs of care, patient outcomes, and service user experiences.​

How Are They Used?
There are three major uses of system-level data sets in mental health research:
 1. Development or testing of research hypotheses
  • Do people with diagnoses of mental illness receive treatment in accordance with standards of care?
  • Is there an association between mental illness and specific physical health conditions?
 2. Mental health service use assessments
  • What types and amounts of treatment do people with mental illness receive and at what cost?
  • Are there demographic or provider differences in utilization of services?
 3. Policy and program planning
  • What is the prevalence of treated mental illness? 
  • ​Is there a trend in types of treatment received over time?

What Do Researchers Need to Consider?
There are limitations inherent in the use of system-level data sets. For example, often only recipients of care are included. In other instances, populations are sampled rather than studied in total, creating biases related to non-random inclusion or exclusion. Some data sets are longitudinal while others are cross-sectional. There often is variation in the ways that health conditions are measured, which can depend on the training of coders and nature of local practice organizations. Health service utilization data can be selective, according to the availability of treatment resources and reimbursable procedures. On the other hand, these data are a rich source of information on patterns of service utilization and characteristics of people treated for mental illness and other behavioral health conditions.

How is Mental Illness Identified in System-Level Data Sets?
System-level data sets often are derived from claims or administrative data, and commonly use International Statistical Classification of Diseases and Related Health Problems (ICD) diagnosis codes. Behavioral health providers and researchers may be more familiar with the Diagnostic and Statistical Manual of Mental Disorders (DSM) diagnosis codes. However, since 2014, all health care providers, including behavioral health providers, covered by the Health Insurance Portability and Accountability Act (HIPAA) are required to use the International Statistical Classification of Diseases and Related Health Problems (ICD) coding systems for billing and reimbursement. In many cases, the DSM and ICD codes are similar if not identical. In addition, cross-walks between the two systems are available in billing software packages and online.

How to Decide Which Data Set to Use?
This section of the compendium helps researchers understand and choose from system-level data sets created for public use and those that contain proprietary information. Often these include diagnoses or other identifiers of people with behavioral health conditions as well as health and mental health service utilization and costs. We focus primarily on public use data sets, but also have included government and proprietary data that are restricted, requiring researchers to pay for data use licenses and/or submit formal applications to access protected information.​