Friday, June 06, 2008

A homebirth midwife has a problem

It seems that my posts about the dramatically increased risk of neonatal death at DEM homebirth revealed by the Linked Birth Infant Death 2003-2004 datasets, has created a problem for Lisa at Homebirth: Midwife Mutiny. In an effort to dismiss the obvious, she has "re-analyzed" the statistics. There is one big problem, though. She lacks basic knowledge of statistics so her conclusions are completely wrong.

Here's how she starts her "re-analysis":
So these are the filters I applied. Everything else was left as default:

Gestation age of birth is 37+ weeks (premature babies do need specialized care not available at home)
Single or twin birth only (multiple births are higher risk and frequently result in fatalities)
A Vaginal delivery (believe it or not the figures do show some Caesarians at home - Angelina Jolie wannabes perhaps)
Age of infant at death 0-27 days (neonatal period)
Figures include 2003 & 2004
Right away there are several serious problems that reflect a lack of understanding of statistics.

1. The results cannot be restricted to vaginal deliveries because that would mean that only SUCCESSFUL homebirths are being counted. The unsuccessful homebirths are automatically discarded by removing C-sections.

2. Race must be taken into account, because it is a risk factor. Less than 1% of homebirths involve African American women. If you don't remove them from the hospital group (and the homebirth group), they will skew the statistics of the hospital group.

3. Multiple births, including twins, must be removed from both groups. That's because multiple births, even twins, are a risk factor and the proportion of twins in the hospital group is much greater than the in homebirth group. Leaving twins in skews the statistics of the hospital group.

Lisa committed three major errors, all of which serve to either raise the risk level of the hospital group, or remove bad outcomes from the homebirth group. The statistics in her first chart are all wrong.

Then she goes on to commit further errors:
Doctors - It doesn't appear very reassuring to have a Doctor at a homebirth. You can understand why they prefer to stay in the hospital and slice and dice. It's what they are trained to do!
Apparently it does not occur to Lisa that some doctors actually do back up homebirth midwives and they do so by attending the homebirth. Once again, their patients are more likely to be high risk, or they are more likely to be there because complications have developed.

Now let's take a look at the hospitals. The only thing I have changed in the filters is to include caesarian births
In other words, here Lisa has decided to add back in the the highest of the high risk: placenta previa, abruptions, pre-eclampsia, gestational diabetes, pre-existing medical conditions, etc. She does not seem to understand that in order to have a valid comparison, the risk levels MUST be the same.

Lisa asks:
So What Does It All Mean?
What does it mean? It means that statistical analysis should not be done by people who don't understand statistics. I don't think that Lisa did this deliberately or cynically. She played around with the data until she got it to look the way she wanted, and she doesn't understand that she violated multiple rules of basic statistical analysis. We come back again to the same place: homebirth midwives spread misinformation because they lack the appropriate education to understand many basic concepts, and they don't even realize that they don't know.

0 Old Comments: