Many Central Virginians are not receiving the public mental health services they need, according to new research by University of Virginia economics professor Steven Stern. His new study focuses on people who can’t afford private mental health services, and thus must rely on Virginia’s regional community service boards, called Region Ten in Central Virginia.

Stern estimates that anywhere from 5.8 percent to 33.2 percent of those in the local region without a college degree – a rough proxy for low income – have mental health issues ranging from more common issues like anxiety, depression and substance abuse problems to more acute issues like schizophrenia, paranoia, bipolar or personality disorders.

The paper, “Estimating Local Prevalence of Mental Health Problems,” was published today in the journal Health Services and Outcomes Research Methodology.

Those who cannot afford private mental health services rely on public service providers like Region Ten, which treated an average of 4,813 clients annually between 2008 and 2010. However, Stern estimates that number is between 13 percent and 50 percent of the actual local need – either 9,634 or 38,022 people, depending on the national survey used to estimate demand.

“While the estimated number of people needing public mental health services varies greatly with the survey used, it is clear that there is a large shortage of services for those in need,” Stern writes in the study.

Robert Johnson, executive director of Region Ten, agreed with the study findings. “Stern’s research highlights several methodologies, all of which demonstrate that the need for services in our area far outweighs the ability of any one agency to satisfy that need, due to inadequate funding, service cost, eligibility restrictions, etc.,” Johnson said.

Region Ten’s service area includes Charlottesville and the counties of Albemarle, Greene, Fluvanna, Louisa and Nelson, with a total adult population of 173,642, according to 2006 census estimates used in Stern’s paper.

The supply of private mental health providers in the region is “actually quite large,” Stern writes, but for people without the financial resources to pay for such care, there is a “severe shortage” of available services.

Prior research has found that some people with mental health needs choose not to seek help, but for many of those locally without a college degree, the main barrier to getting treatment is cost and the lack of public service supply, Stern said.

Existing national surveys on mental health issues lack localized geographical information, so Stern estimates the local prevalence of mental health issues by modeling the intersections of five data sets: three national surveys that address mental health issues, combined with two sets of census data – one national and one from the Charlottesville metropolitan statistical area.

“Any methodology that relies on estimates from national datasets must either accept on faith that one particular survey is more appropriate than others or confront the issue of large variation across surveys in reported rates for different mental health diagnoses,” Stern writes in his conclusion. “Since, as an outsider with no vested interest in one screening tool versus another, I have no reason to ‘believe’ in one over others, I view this problem as a critical issue for the field to resolve.”

Details on Study Data Sources

Stern estimates the prevalence of mental health problems in the general U.S. population using three sources: the National Health Interview Survey (1995) (NHIS), the National Comorbidity Survey-Replication (2001) (NCS-R), and the National Survey of Alcohol, Drug and Mental Health Problems (2000–01) (NSADMHP). Stern selects a relevant sample from each data set, comprised of 54,754 observations, 9,283 observations and 5,747 observations, respectively.

He combines that data with two U.S. Census data sets: the Public Use Microdata Sample (PUMS) for the Charlottesville metropolitan statistical area (2000), a 1 percent cross-section sample from the 2000 Census for the local region, with 7,919 observations; and the Community Tracking Survey (2003), a cross-section of 53,138 observations of individuals from 60 communities across the United States.

Stern finds that the estimated demand for public mental health services depends substantially on which national data is used. The estimate of someone having a mental health problem in the Charlottesville MSA from the NHIS data is between 5.8 percent and 10.5 percent, while, from the NCS-R data, it is between 27.4 percent and 33.2 percent. These estimates are different because 10 percent of the observations in the NHIS report having a mental health problem, while 29 percent of those in the NCS-R report having a mental health problem. “In effect, the variation of local prevalence estimates is caused by order-of-magnitude variation in the underlying national survey data,” Stern said.

Using the conservative NHIS estimate of the number of people needing mental health care services, 9,634 Central Virginians need help, meaning Region Ten served about 50 percent of the actual demand (4,813 out of 9,634). Using the NCS-R estimate, 38,022 local residents have mental health needs, meaning Region Ten served only 13 percent (4,813 of 38,022).

The lack of any “gold standard” measure of mental health is one key hurdle that complicates the task of estimating regional mental health needs, Stern notes, with surveys varying widely in how they classify, diagnose or even identify mental health problems including dementia and panic attacks.

Even when the same categorization is used, there is wide variation in reported prevalence. For example, in the NCS-R, 19.1 percent of respondents report having anxiety problems, while, for the other two surveys, anxiety prevalence ranges between 3.8 percent and 6 percent. There is also wide variation in reports of depression and substance abuse problems. The NHIS relies on self-reports, while the NSADMHP and NCS-R rely on answers to screening questions frequently used in diagnosis, but even among the latter two surveys, the reported prevalence rates vary quite a lot.