A recent study published in the journal Obstetrics & Gynecology is creating a collective orgasm in the Sandra Fluke wing of the Democrat party. This study purports to show that free contraception reduces not only pregnancies but abortions. This study is a prime example of the politically motivated crap that appears in low impact scientific journals. It is research by press release and agitprop masquerading as science.

The study is a monument to tautology. We often hear that 30 percent of medical costs occur in the last year of life and that most traffic accidents occur within a mile of home. The reasons for this is obvious. Most people don’t receive expensive medical care unless they are near death and about 100% of your driving is done with a mile or so of home. The results of the study are hardly surprising. Women using contraception tend to have a lower rate of pregnancies and because they aren’t pregnant there is no need to have an abortion. I think most of us knew, or at least suspected, that to be the case.

The problems with the study are manifest. The two that reach out and smack you are:

First, there is no control group. The entire cohort in the study received free contraception. A similar group should have received the same contraceptive services and paid for the service. Without the control group we have no way of knowing if “free contraception” had anything to do with the outcome.

The self-selection bias in the study cohort calls everything in the study into question. From the study:

Participants were recruited from the two abortion facilities in the St. Louis region and through provider referral, advertisements, and word of mouth.

Michael New writing in National Review identifies many more problems:

2). Limited impact on repeat abortion rate: The study makes much of the fact that between 2006 and 2010 there was a statistically significant decline in the repeat abortion rate in St. Louis City and County. This may well be true. However, the results indicate that the repeat abortion rate fell from about 48 percent in 2006 to about 45 percent in 2010 — hardly a dramatic decline.

3) Exaggerated impact on overall abortion rate: The authors also make much of the fact that the number of abortions performed at Reproductive Health Services on women who resided in St. Louis City and County declined by 20.6 percent between 2008 and 2010. However, Reproductive Health Services is not the only abortion provider in the St. Louis area. Furthermore, only a small percentage of St. Louis area women took part in the program.
4) The weighting mechanism overstates effectiveness of contraception program: Program participants were not a random sample of women residing in the St. Louis area. They were more likely to be African-American, young, and low-income. As such, the authors weigh the data to compare birth rates and abortion rates of program participants to birth rates and abortion rates of a similar demographic cohort. Consequently, these contraceptive methods likely appeared more effective than they actually were — because they were being used by a demographic with both relatively high birth rates and abortion rates.

5) The results are not generalizable to a large population: The authors state that IUDs are more popular in Europe than they are in the United States. There are a variety of reasons for this. However, one factor the authors overlook is that many physicians in the United States are unwilling to insert IUDs because of liability issues. Indeed, IUDs users have an increased risk of pelvic inflammatory disease and perforation of the uterus. Also, if a woman using an IUD wants to get pregnant, her IUD would have to be removed by a physician. For this reason, even if these long-term methods were available at no cost, it is not clear that many women would choose to use them.

And the problems continue. For instance, according to the pro-abortion Guttmacher Institute, 5.5% of women select the IUD as a contraceptive method and 1.1% use an implant. The study cohort differed radically:

Participants chose the following contraceptive methods at baseline: 46% levonorgestrel IUD, 12% copper IUD, 17% subdermal implant, 9% OCPs, 7% contraceptive vaginal ring, 7% depot medroxyprogesterone acetate, and 2% contraceptive patch.

In the study 77% of the cohort used methods used by 6.6% of all women and those methods are the most effective methods of contraception.

It is doubtful that any of this is news to anyone, even the team who participated in the study. It was published in a low impact journal (Impact 4.75) rather than appearing in Journal of the American Medical Association (Impact 30). The lack of statistical rigor would have prevented it ever being accepted.

What makes this study particularly odious is that the intent is to influence the 2012 elections. Lydia McGrew at Whats’s Wrong With The World points out this nugget:

Before getting into criticisms of the methodology and conclusions of the study, I want to ponder something that I haven’t seen discussed elsewhere. The celebratory article on the study from NBC says this:

The results were so dramatic, in fact, that Peipert asked the journal of Obstetrics & Gynecology to publish the study before the Nov. 6 presidential election, knowing that the Affordable Care Act, and its reproductive health provisions, are major issues in the campaign.

“It just has so many implications for our society,” he told NBC News.

Since when are professional scientific journals openly and proudly attempting to influence political elections by their decisions about publication? Isn’t there something a tad unprofessional about that? More than a tad? The study’s author, Jeffrey Peipert, tells a news organization unabashedly that he expressly asked the journal, Obstetrics and Gynecology, to hasten the publication of his article in order to influence this fall’s presidential election.

The most this study could ever demonstrate is that a given cohort of women who are given highly effective means of birth control have lower incidence of pregnancy than a comparable non-participating population. No inference about cost can be made from this study at all. Unfortunately, because of the lack of a control group and the cavalier selection of a study population we really know nothing at all that we didn’t know before.