labor nurse has been reborn and shares her experiences as a new nurse-midwife, woman, and blogger

Saturday, August 29, 2009

Statistical Prowess

For the research minded folk out there, I have to pick your brain. When I was in school, I loved performing research queries and reading the articles my school's search engine would provide(which was extremely user friendly, unlike my current hospital library...). I would frequently end up with thousands of pages of research articles, meta-analysis, and the like. I would even hope that I would be a main researcher during my career. Now, however, I am just so overwhelmed trying to balance life as a new nurse midwife with regular day to day crap (read: housework, errands, gym, and maybe a little relaxation) that I can't seem to do anything beyond the bare minimum. But I digress...

What I really wanted to say was that frequently many of the articles and published research in obstetrics have small participant numbers. I guess this is where meta-analysis comes in, but many times the participant numbers in obstetrical research are less than 500. A patient I was meeting with recently is a statistician at a local university and laughs at the "statistical power" of obstetric research. Her argument is that the conclusions we frequently draw from the literature are invalid given how small the participant numbers are. She points out that many times research on obstetrical care is also limited by the small geographic regions it's performed- for instance a group of 350 women receiving care at just one university hospital- and yet we extrapolate this to all women. And frankly, many times this is true when looking at it from a statistician's paradigm.

I think there are a lot of potential problems with performing research in obstetrics. This topic could provide fodder for entire program courses in midwifery and obstetrics so I realize I am not even skimming the surface. First off, the "gold standard" of research is difficult to perform ethically in pregnant women. Randomized controlled trials (RCTs) are the top tier of research, yet we are limited in performing it in this population. Ideally, wouldn't it be great if we could randomize women to home birth vs hospital birth? What a fantasy, I know, I know.... but it would be hard to argue the statistical power of a large nation wide RCT on the subject and quiet the subject once and for all. So we settle for smaller, less statistically powerful data or argue over the biases or errors (real or imagined) in what is out there.

Now I am just rambling, but what this leads me to is that if a trained midwife (me) is overwhelmed with this both professionally and personally, how can we expect the general public to draw sound conclusions from what's out there? I know there are plenty of midwives and women's health care advocates out there who have a really awesome grasp on the role of obstetrical research (see: Amy Romano, Henci Goer) in the care of women, and I highly admire that. And I'm jealous.


momstinfoilhat said...

Power in a study is the ability to make claims of significance. It depends on the number of subjects, but also the type of analysis you will be doing, the types of variables you are using, and the difference in effect you are looking for (e.g. a 10% reduction in cesareans, or a 30% reduction in episiotomy).

Some interventions and outcomes in birth can ethically be researched in RCTs, depending on the practice. For example, RCTs can be done on restricting nutrition and fluids in labor, or upright positions in the second stage, or many other practices. Some practices are obvious not ethically amenable to RCTs, but they still can be the focus of pilot studies, case control studies, quasi experimental studies and other study types.

Megan said...

I was just going to comment the same thing! I'm an epidemiologist, and I've found statistically significant results with fewer than 100 cases, because the effect was large.

If the analysis is thoughtfully and correctly done, small sample size is not a deal-breaker. It's true that it can be difficult to extrapolate results to apply to the population at large, especially when there are such large geographic differences in obstetric care in the US.