Tuesday, August 21st 2012
What Do You Do When There is No Best Dataset? A follow-up on pregnancy and rape statistics
Trigger warning: discussion of violence against women.
I’ve received a number of questions over email and Twitter about the statistics I mention at the end of my last post. Here is what I wrote:
“So, rape and consensual sex have the same pregnancy rate. This means that of those 64,080 rapes in 2004-2005, minus the 15% of rapes that are of children under the age of 12 which gets us to 54,468 rapes of almost all reproductively-aged women, somewhere between 1,689 (3.1%) to 2,723 (5%) pregnancies from rape could have occurred in that year alone. Somewhere around half of those women were probably using some form of hormonal contraception, so let’s hope the numbers are even lower. Unfortunately, access to emergency contraception is still a challenge for rape survivors who go to hospitals, particularly Catholic hospitals, to receive treatment.”
The question is: why did I not use the Holmes et al (1996) estimate that there are 32,000 pregnancies from rape each year? Why use the RAINN stats which are clearly more conservative?
Both of these sets of numbers – the Holmes et al and the RAINN – are estimates. It’s worse than that. The RAINN values are estimates from the 2005 National Crime Victimization Survey, which I then estimated contained 85% reproductively-aged adults from an estimate of the percentage of rape survivors that are children (15%) from the 2004 National Crime Victimization Survey (also on the RAINN site). I then had to arrive at a range based on two different studies about the rate of pregnancy from a single act of intercourse (Holmes et al versus Wilcox et al). I’m not wedded to these numbers, it just seemed like a good place to start.
The Holmes et al estimates from their pool of 3031 adult women from the third wave of their research, that there is a 5% pregnancy rate from rape. The reason I didn’t use their extrapolation out to 32,000 pregnancies a year was that I found the discussion section a little unclear, and I didn’t quite understand how they arrived at that number. In their methods they called their study the National Women’s Study, and say that they interviewed 4008 women in the first wave. Each wave had fewer women. They said they noted 616 instances of rape from the 3031 respondents in wave three. How then, do they then say:
“Analysis of the National Women’s Survey raw data (without statistical weighting required for determining representative population estimates) indicates that there were 34 cases of rape-related pregnancy. A total of 30 women reported one rape-related pregnancy and two additional women reported two rape-related pregnancies. Of the 34 cases 21% occurred when the victim was aged 12 to 15 years, 27% occurred among women aged 16 and 17, and 52% occurred after age 18.”
Do you see how I’m confused? I thought there were 20 instances of rape-related pregnancy, and here it says 34. I think they are maybe referring to their first citation: Kilpatrick DG, Edmunds CN, Seymour AK. Rape in America: a report to the nation. Charleston (SC): The National Victim Center and the National Crime Victims Research and Treatment Center at the Medical University of South Carolina, 1992 Apr. But that’s not what they cite.
In any case, both of these sets of numbers are estimates. Both are likely based on decent rationale, but I had guessed that the answer lay somewhere between these two numbers.
It gets worse
I decided to crunch some numbers myself. Using US Census Bureau data for July 1996 (arbitrary, but I wanted to choose the same year as publication of the Holmes et al), there were 60,704,500 women between 15-44 years of age. If 13% of them were raped, then the range of possible pregnancies is between 244,639 (3.1% pregnancy rate from Wilcox et al (2001)) and 394,579 (5% pregnancy rate). Semen is found in about half of rapes (Riggs et al., 2000), about half of women use hormonal contraception. So let’s quarter those numbers.
That still works out to 61,160 to 98,645 pregnancies from rape.
Ok, let’s factor in the use of emergency contraception (EC). I’m getting desperate now. In one survey, 85.9% of emergency departments provide EC counseling, and 87.7% of those will administer the first dose right there (Azikiwe et al., 2005). If all of the women who were raped actually went to emergency departments, if all of them who were offered EC took it, and if it worked for ALL of them, then…
…it’s only 15,082 to 24,129 pregnancies from rape. If rape survivors always went to emergency departments for treatment, which they rarely do.
An estimate is an estimate. None of these numbers captures the reality of the situation, except that they are all unacceptably big. Rather than accept any one of these values as the right one, perhaps we need to just decide that any number greater than zero is too many.
AN UPDATE (8/22/2012, 9:28am CST):
Based on some feedback I’ve received that the 13% was likely the incidence over 3 years according to the Holmes et al study, we could cut the yearly incidence in third again. So that would be 5,027 to 8,043 pregnancies from rape each year, if everything goes perfectly under the rather unrealistic conditions I’ve set above (regarding access to EC).
References
Azikiwe N, Wright J, Cheng T, D’Angelo LJ. 2005. Management of rape victims (regarding STD treatment and pregnancy prevention): Do academic emergency departments practice what they preach? Journal of Adolescent Health 36(5):446-448.
Holmes MM, Resnick HS, Kilpatrick DG, Best CL. 1996. Rape-related pregnancy: Estimates and descriptive characteristics from a national sample of women. American Journal of Obstetrics and Gynecology 175(2):320-325.
Riggs N, Houry D, Long G, Markovchick V, Feldhaus K. 2000. Analysis of 1,076 cases of sexual assault. Ann Emerg Med 35:358-362.
Wilcox AJ, Dunson DB, Weinberg CR, Trussell J, Baird DD. 2001. Likelihood of conception with a single act of intercourse: providing benchmark rates for assessment of post-coital contraceptives. Contraception 63(4):211-215.
