How safe is home birth, really?

A lot of interest has been given recently to the safety of home birth attended by Certified Professional Midwifes (CPMs) and other lay- and nurse- midwives. A recent position statement from the American College of Obstetricians and Gynecologists (ACOG) endorsed by the AMA states that: 

. . . the safest setting for labor, delivery, and the immediate postpartum period is in the hospital, or a birthing center within a hospital complex . . . 

This statement has been vilified by many different groups who advocate for home birth, and even recently in an editorial by Nancy Lowe in the Journal of Obstetric, Gynecologic & Neonatal Nursing (2009;38:1). But is this statement inaccurate? Certainly the vast majority of low-risk births attended or even unattended at home will end well for both mother and baby. The graphic below shows some problems which can occur during labor, along with interventions and possible consequences of these problems. These are listed from most common to least common with the appropriate incidences as a percentage. Note that many of these interventions can only be provided in the hospital setting (e.g., cesarean section) and many of these problems cannot be anticipated prior to labor (e.g., abruption, shoulder dystocia, or anaphylactoid syndrome of pregnancy): 

As far as the safety of home birth is concerned, home birth advocates frequently cite a study from the British Medical Journal in 2005 (330:1416) by Johnson and Daviss. Indeed, this paper is the strongest study yet and is referenced directly or indirectly in both professional and lay literature frequently (see Ricki Lake's movie, The Business of Being Born) to establish the safety of home birth in North America. 

This paper analyzed 98% of home births in Canada and the United States in the year 2000 which were attended by CPMs. They analyzed major outcomes and transfers to hospitals and concluded:

Planned home births for low risk women in North America using certified professional midwives was associated with lower rates of medical interventions but similar intrapartum and neonatal mortality to that of low risk hospital births in the United States. 

But the data in the study has been shrewdly manipulated with surprisingly very little academic scrutiny. Of 5418 births of women who intended to deliver at home, they found that 12.1% were transferred to the hospital for various reasons. They also claimed 1.7 neonatal deaths/1000 planned home births and state that this compares favorably to hospital births in North America. 

However, they excluded babies with life-threatening congenital anomalies (though these infants were not excluded from the comparison group) and they did not compare the home-births to hospital births occurring in the US in the year 2000, even though that data was and is readily available from the CDC and they even cite other data from the same CDC report in their paper. Instead, the largest data set regarding hospital births used in this study comes from 1989 in California, with no matching of patient characteristics. They also excluded deaths occurring in breech deliveries at home (even though this was an intent-to-treat study). 

If we use their data from the study and include deaths from breeches, the rate rises to 2.0/1000. If we add back anomalous infants, the death rate rises to 3.32/1000. Using CDC data regarding all births in the hospital in the United States in the year 2000, and including only infants weighing > 2500 grams (99% of the home births in the report were > 2500 grams), then we find the following rates: 

Even this comparison, which shows 0.65/1000 neonatal deaths in the hospital compared to 2.0/1000 at home among non-anomalous infants, is not entirely accurate. The babies born in the hospital in this data set include all-comers, both low-risk and high-risk, healthy and not healthy. Yet, the babies born at home were necessarily a low-risk, healthy population (or else the CPMs would not have endeavored to deliver these women at home in the first place). The point is, a randomized trial using matched, healthy patient populations of only low-risk pregnancies would likely reveal a rate of neonatal death much lower than 0.65/1000 in the hospital.

What does this all mean? Well, there were 3,427 neonatal deaths in the United States among this demographic (term women) in the year 2000. If the same cohort of women had home births instead (at least if that was the intent-to-treat), then there would have been 12,443 neonatal deaths in the year 2000, which is an excess of 9,016 deaths. This calculation is based on the data reported in Johnson and Daviss' study! To put this in perspective, in the year 2002 in the US, there were only 12,008 deaths from all causes among children aged 1-14. So we would effectively double the death rate of children under the age of 14 in the United States. 

Of course, these projections do not account for other significant outcomes, such as the rates of permanent neurological injury from shoulder dystocia or hypoxia (Erb's or Cerebral Palsy), or post-neonatal mortality which tends to rise proportionately with neonatal mortality (in other words, this is the excess death caused in the first month of life). 

So the statement of ACOG and the AMA that the safest place to have a baby in the US is in the hospital is true. In fact, using Johnson and Daviss' paper, it is at least 3x as safe (0.65 vs 2.0/1000) using neonatal mortality alone as an outcome. Still, home-birth will always have strong anecdotal support since, fortunately, more than 99% of home or hospital births end well.