Wednesday, June 13, 2007

The Farm Study: a classic case of deception

I never tire of pointing out that there are no studies that show homebirth to be as safe as hospital birth. There are, however, studies that CLAIM to show that homebirth is as safe as hospital birth. None of them do because they rely on the deliberate use of the WRONG group for comparison. By deliberately choosing the wrong group (always a group with higher neonatal or perinatal mortality than the right group) the authors can make homebirth look safer than it really is.

My favorite example is the BMJ 2005 study by Johnson and Daviss which compared neonatal death at homebirth in 2000 to a bunch of out of date papers instead of comparing it to hospital birth in 2000. The correct comparison shows that homebirth has a neonatal mortality rate 3-4 times higher than hospital birth for low risk women.

The Farm study (Durand, 1992) is even worse. The Farm study compares homebirth to the 1980 US National Natality-National Fetal Mortality Survey and concludes that homebirth is safer. However, the documentation for the NN-NFMS reveals that it is a deliberately non-representative sample of births in 1980 and was never meant to be used as a measure of neonatal mortality. Durand calculated a perinatal mortality rate for homebirth at the Farm was 10/1000 and compared it to the perinatal mortality rate for the NN-NFMS of 13/1000. However, the neonatal mortality rate at the Farm was 8.2/1000 and the neonatal mortality rate for the entire US in 1980 was 8.5/1000. It was only 7.6/1000 for white women and was substantially lower for low risk white women at term, probably in the range of 3.8/1000 or less. So, rather than showing that homebirth is as safe as hospital birth, the Farm study showed the the neonatal mortality rate at homebirth is more than double that in the hospital.

What is the 1980 US National Natality-National Fetal Mortality Survey and does it give us accurate information about neonatal death rates? The 1980 US National Natality-National Fetal Mortality Survey comes with copious data about the methodology of the NN-NFMS and this data leaves absolutely no doubt that the NN-NFMS does NOT yield neonatal mortality rates for hospital birth in 1980. Let's look at some important reasons why:

The NN-NFMS was not designed to yield mortality data for 1980. It had an entirely different purpose. The NN-NFMS was designed to obtain health data that was not on the original birth certificate in order to analyze that data. So, for example, the NN-NFMS asked about the use of ultrasound during pregnancy, which does not appear on the birth certificate. The investigators also collected data that did appear on the birth certificate in order to determine the accuracy of the NN-NFMS data.

No other scientific paper uses the mortality data on the NN-NFMS. (This is not suprising, since, as explained above, the mortality data is not accurate).

The NN-NFMS deliberately oversampled high risk births. This fact is acknowledged within the Durand paper, but it is not explained. It means that the sample used in the NN-NFMS has a higher risk level than the population in general, and is much higher risk than any lower risk group. As expected, the neonatal mortality rate in the NN-NFMS sample is HIGHER than the overall neonatal mortality rate for the entire country in 1980.

The NN-NFMS is known to be a tiny (0.3% of births) NON-REPRESENTATIVE fraction of the deliveries in 1980. That is deliberate on the part of the authors, but it means that under no circumstances can the NN-NFMS mortality data be substituted for the 1980 birth certificate data.

In summary, Durand compared death at homebirth at the Farm with neonatal death rates from a tiny, non-representative fraction of deliveries in 1980 instead of comparing the Farm to know mortality rates from hospital birth in 1980. The correct comparison shows that homebirth at the Farm had a neonatal death rate more than twice as high as hospital birth for low risk white women. Once again, the only way to make homebirth look good is to use deception.

You can find out much more about the NN-NFMS by reading these publications:

Response Characteristics 1980 National Natality and Fetal Mortality Surveys

Comparability of Reporting Between the Birth Certificate and the 1980 National Natality Survey

For more information about neonatal mortality rates in 1980:

Infant, neonatal, and postneonatal mortality rates by race and sex: United States, 1940, 1950, 1960, 1970, and 1975-2000

Labels:

0 Old Comments: