Human's article announced: "Children with health insurance, studies have shown, are less likely than uninsured kids to end up in emergency rooms, more likely to get key vaccinations, and less likely to be absent from school."
My friend Ari Armstrong is a columnist for the Grand Junction Free Press, and also the publisher of the Colorado Freedom Report weblog. Armstrong used to be a senior fellow at the Independence Institute and, although he's no longer formally affiliated with the institute, he still writes for us from time to time. Armstrong e-mailed Human and politely asked if she could send him the names of two or three of the studies she had in mind.
She refused. E-mail exchanges continued, as Armstrong copied Jason Salzman, my counterpart on these pages, and me. Eventually, Human did send Armstrong a list of five studies, along with a note: "I won't be doing this for you again . . . you can do it yourself, and I don't have time to repeat these types of searches for everyone who asks."
One wonders why Human had to spend time searching for the citations, since presumably she knew about each of the studies before she wrote the article. The official standards of the Post so require.
None of five studies Human cited after the fact support her article's statement about what "studies have shown" regarding the effects of insurance on emergency room use, vaccinations and school absences. Indeed four of the five studies she cited do not even address those topics (Cousineau, Medical Care, 2008; Skinner, BMC Health Services Research, 2007; Ward, CA: A Cancer Journal for Clinicians, 2008; Morbidity & Mortality Weekly Report, Sept. 7, 2007).
One study cited by Human was relevant, and it directly contradicted her article's claim. The study looked at the effect of providing SCHIP coverage (subsidized insurance for children whose families have too much income to qualify for Medicaid). Emergency room usage "did not change," the study found. (Szilagyi, Pediatrics, 2004).
I e-mailed Human some questions, and got a response from the Post's public affairs editor, Chuck Murphy. Via Murphy, Human supplied two more citations to substantiate her article's claim about emergency rooms. One of the studies was irrelevant, a 2002 Centers for Disease Control and Prevention report which reported no data about frequency of emergency room use for the insured and uninsured.
Human supplied another study which did support her claim. William Johnson and Mary Rimsza investigated Yuma County, Ariz., and found that uninsured children there use emergency rooms more often. The Johnson and Rimsza article, published in Pediatrics in 2004, forthrightly acknowledged that four other studies have found that taxpayer-funded insurance for children actually increases emergency room usage, and a fifth study finds that there is no effect. Johnson and Rimsza suggested that results were different in Arizona because the state's medical welfare program links recipients to pediatricians, and having a pediatrician drastically reduces ER visits for both the insured and uninsured.
So Human's pronouncement in her Post article - "Children with health insurance, studies have shown, are less likely than uninsured kids to end up in emergency rooms" - turns out to be not entirely accurate. A large body of research contradicts her claim, and that research is in the very studies which Human pointed to when she was challenged to support her claims.
In the last two years, the phrase "studies have shown" has appeared in staff-written pieces 31 times in the Rocky Mountain News, and 36 times in the Post. About half the time the phrase is used in a direct quote, or in another way which tells the reader the source of the information. For example, "According to professor Roy Hinkley, studies have shown that minnows . . . "
But the other half of the time, the dailies used "studies have shown" with no source. The unattributed locution was especially common in Post editorials, and in health and nutrition coverage in both papers.
The phrase ill-serves readers who want to learn more about a subject, but who are left in the dark about where to look. The phrase can be used to falsely declare scholarly consensus about a subject. And the phrase can be a crutch for a writer who feels "sure" about a supposed fact, but who doesn't want to take the time to verify it.
If "studies" are important enough to mention in an article, they should usually should be named in the article. Print space is scarce, but a quick citation can be offered in three to five words. For online versions, full citations can be supplied at the end of the article.