Guesstimator
"We are the source the government uses, the legislature uses, the advocacy groups use."Rick Brown
|
Under director Rick Brown, the UCLA Center for Health Policy Research has become the primary source in California for information on health-insurance coverage. "We are the source the government uses, the legislature uses, the advocacy groups use," says Brown, who came to UCLA in 1979. "If people want to know anything about health-insurance coverage and want to know about the uninsured, they turn to us." Ironically, Brown has never taken a public-health course in his life. "I got my doctorate at Berkeley in education and sociology and then got interested in public health through my work in the sociology of professions," he says. Eventually, he concentrated his attention on issues related to health-care access for low-income and other disadvantaged people - and, thus, became the insurance-data guru. Brown and his colleagues are the first to concede they don't have all the answers, but that doesn't stop them from attacking problems. In fact, Brown and his team are most interested in decision- and policy-making situations in which the relevant information doesn't exist. To solve this problem, Brown has honed the use of a statistical concept known as synthetic data, or synthetic estimates, which he defines as the statistical equivalent of "highly educated guesses." Over the years, for instance, Brown and his colleagues have been using large databases and government surveys to come up with data that could affect policy decisions on health-insurance coverage. But the surveys, while extensive, are not nearly comprehensive enough for informed decision making on a county or state level. "Imagine you're the county health department in Santa Barbara and you have a responsibility to meet the needs of your uninsured residents," says Brown. "Or you're a network of community clinics or hospitals and you want to know how many people will need your services but will not be able to pay for them. You may come to us for help."
|
Traditionally, there have been two ways to answer these questions: One is to look for existing information through surveys like the "Current Population Survey," which is a product of the U.S. Census Bureau and the Bureau of Statistics. The other method is to do a survey. In most cases, however, the former is not comprehensive enough and the latter is usually too costly and time consuming. "We can identify people who reside in the County of Santa Barbara," he explains, "but that sample size is very small and probably too small to come up with a reliable estimate. So as a researcher, I would say, 'Well, let's find some way to do an estimate that draws on the information we do have in order to provide a stable estimate for the information we don't.' This is where synthetic estimates come in. They are a way of trying to answer questions for which there are no available data, using the best-available data you do have." In the Santa Barbara example, the way to generate synthetic data is to take other counties for which sufficient data is available and create a statistical model for who's insured and who's not. This estimate might be based on four demographic variables: ethnicity, marital status, income and age. The model then provides a probability for whether any single individual is likely to have health insurance depending upon those four variables. These probabilities can then be used to derive a countywide estimate of uninsured individuals in Santa Barbara County (or subdistricts) that can be used with a high degree of probability in making decisions. But as more data accumulate - and computer technology makes it increasingly available - the question ultimately becomes whether synthetic data will still be necessary. Indeed it will, says Brown. Since information will never be perfect, more information should just make the estimates all that much better. - G.T. |
||
