Health Insurance May Be Improving—But Not for Individuals with Mental Illness

Health Services Research, April, 2000 by Roland Sturm, Kenneth Wells

Objective. To explore the question of how insurance coverage has changed among individuals with mental problems compared to the general population in the last two years.

Data Sources. HealthCare for Communities, a national survey to track health system changes.

Principal Findings. The percentage of uninsured persons in the general population has not changed very much, and more respondents believe that health insurance coverage has improved rather than deteriorated over the years 1996 to 1998. However, among individuals with probable mental health disorders, more have lost insurance in those two years than have gained it and more report decreases in health benefits. Individuals with worse mental health consistently report a deterioration of access to care compared to individuals with better mental health.

Conclusions. Substantial activity has taken place in state and federal legislation to increase the mental health benefits offered by health insurance. Although this activity could have improved health insurance especially for individuals with mental illness, such persons continue to fare significantly worse than the general population.

Key Words. Mental health, health insurance, depression, insurance coverage

The number of individuals without health insurance increased from 1979 to 1997 (Kronick and Gilmer 1999; Carrasquillo et al. 1999; U.S. Census Bureau 1997). This has been attributed to declines in the number of employers offering insurance or in lower employer contributions (resulting in fewer employees electing coverage due to higher costs) (Fronstin and Snider 1996; Cooper and Schone 1997; Ginsburg, Gabel, and Hunt 1998). Because individuals with major psychiatric disorders are at high risk for poverty (and least likely to be able to afford plans that require large contributions) or unemployment, persons with such disorders may be particularly vulnerable to becoming uninsured. Even if the improved economy is beginning to reduce the likelihood of becoming uninsured, such benefits may not accrue to persons with mental illness. What are the most recent developments in insurance coverage and perceived access to care for persons with mental illness, and how do such coverage and access compare to that seen in the general population? We investigate these questions using data from a national household survey that was completed in December 1998.

Persons who are mentally ill are of particular interest because the past decade has witnessed declining insurance coverage for mental health care, while at the same time, rapid advances have been made in efficacious treatments and clinical practice guidelines for major psychiatric disorders, such as affective and anxiety disorders and schizophrenia (Hay Group 1998; Wells et al. 1996; Depression Guideline Panel 1993; Kessler, Andrews, Mroczek, et al. 1999; Hirschfeld, Keller, Panico, et al. 1997). Although health plans seldom limit the number of covered outpatient visits or inpatient days for major medical illnesses, the typical employer-sponsored plan in 1996 imposed several limits on mental health care, including limits on visits or days, or on annual or lifetime dollars (Sturm and McCulloch 1998). Similar discrepancies in coverage for medical and mental illnesses are evident in deductibles, copayments, and coinsurance rates. These increasing discrepancies have resulted in an almost 50 percent drop in the m ental health share of total health care costs paid by employment-based insurance over the past ten years (Hay Group 1998).

In response, the 1990s have brought federal and state legislative activity in the form of insurance mandates that require mental health care coverage to maintain the same level as medical care benefits ("parity"). The enactment of a modest federal mental health parity bill in 1996 was followed by more ambitious state activity. More than 30 states introduced parity legislation in 1997 alone (Sturm and Pacula 1999). Although patient advocacy groups have praised their passage, it is unclear whether or not parity laws represent a major improvement either in benefits or in access to appropriate care. In particular, some have been concerned that employers may drop health care coverage altogether, or may drop mental health coverage, to avoid increased costs under parity legislation (Custer 1998; Jensen and Morrisey 1999; Sturm 1997).

DATA AND METHODS

We analyze data from HealthCare for Communities (HCC), a national survey funded by the Robert Wood Johnson Foundation (Sturm, Gresenz, Sherbourne, et al. 1999). HCC was designed to identify variations in health care and to track health care changes over time. The HCC household survey reinterviewed participants in the Community Tracking Study (CTS) (Kemper, Blumenthal, Corrigan, et al. 1996) about 18 months after their initial interview. The HCC household sample was selected from a random sample of 30,375 adult CTS telephone respondents (the HCC sampling frame), out of which 14,985 were selected for an expected completion of 10,000 interviews. We obtained 9,585 eligible responses (64 percent response rate). Two populations are of particular interest in HCC: individuals with likely alcohol/drug/mental health (ADM) problems and poor individuals. We used information from the baseline CTS interview to oversample individuals with low income or high psychological distress; or mental health specialty use from the HCC sampling frame. This increases the design effects for estimates on all individuals, but it provides more precise national estimates for these subgroups. To permit oversampling, the sample was stratified by three factors: (1) low income (family income [leq] $20,000) versus higher-income (family income [greater than] $20,000); (2) psychological distress, based on the two mental health items included in the SF-12 instrument (Ware, Kosinski, and Keller 1996); and (3) mental health specialty outpatient use in the past 12 months versus no use. Weights to adjust for sampling design and nonresponse were developed to obtain nationally representative estimates. A full description of the study design has been published in Sturm, Gresenz, Sherbourne, et al. (1999); additional information specific to HCC is available at http://www.hsrcenter.org.