stumbling through the home birth debate, continued

In my previous post i set the stage, as i see it, for the contemporary home birth debate, which for all its bluster orbits around only a few key disagreements. These appear to be whether (1) planning for home versus hospital delivery and (2) certified nurse-midwifery versus certified professional midwifery matter to the safety of the birth process.

Disclaimer: If you’re looking for advice, go somewhere else.

perinatal and neonatal mortality

To give context to these contentions, and to get a feel for the kind of research they rely upon, let’s review some of the most widely-cited studies that deal with the question of fetus/infant mortality. (The following four studies are illustrative of the sources of disagreement between home and hospital birth advocates, but as a sample they should not be taken as representative of the broader literature.)

• An analysis of eight years of birth registries in Washington State (Pang et al, 2002) found twice the neonatal death rate among deliveries planned for home as for hospitals.
• A prospective cohort study of all (several thousand) planned home births attended to by CPMs specifically (Johnson–Davis, 2005) found intrapartum and neonatal mortality rates consistent with those of hospital births.
• The gargantuan Dutch nationwide cohort study (de Jonge et al, 2009) analyzed hundreds of thousands of deliveries but found no significant difference in perinatal or neonatal mortality.
• The “Wax paper” (Wax et al, 2010), a meta-analysis, examined a dozen previous studies and found that the neonatal death rate for planned home births is double that for planned hospital births—triple if cases of birth defects are factored out.

Each of these papers illustrates a problem or limitation important to interpreting these kinds of results. I’ll take them in order:

The Washington study has been criticized, for instance by Susan Hodges at Citizens for Midwifery and by Gail Hart and Judy Slome Cohain at Midwifery Today, for using an unconventional definition of “planned” home birth: The authors were unable to determine from records which births were planned for delivery where, so they decided their own criteria to serve as a proxy—in particular excluding births before 34 weeks’ gestation. Hodges and others objected to this criterion, advocating a more widely-agreed 37-week mark instead, on the grounds that earlier home deliveries than these are likely to have been unplanned. They acknowledge, however, that “later in the paper [Pang et al] claim that omitting births prior to 37 weeks did not affect the relative rates of neonatal mortality”. (Admittedly i don’t understand Hodges’ objection to the last home birth criterion of having “had a midwife, nurse or physician listed as either the birth attendant or certifier on the birth certificate”; they object that this criterion would not distinguish “on the way” deliveries that were planned for the hospital from those that were planned for the home, but the reverse holds as well, and it’s unclear which of these is more common.)

Of 5418 deliveries, the CPM study recorded 14 deaths, detailed in the paper. Omitting those due to stillbirths and birth defects, they arrive at a combined intrapartum–neonatal death rate of 11/5418 = 2.0 per thousand. The authors then contrasted this rate to those reported in several other studies, which hovered between 0.7 and 3.0 per thousand in hospitals. While meaningful, this comparison is problematic, since methodologies and data sources may vary. Elsewhere in the paper, however, the authors compared their measured rates of interventions against “data from birth certificates for all 3 360 868 singleton, vertex births at 37 weeks or more gestation in the United States in 2000, as reported by the National Center for Health Statistics”, for which they cite the 12 February 2002 issue of NVSS. This raises the question of why they didn’t compare their measured mortality rates against NVSS numbers, which were contained in the 28 August issue from the same year. Home birth critic Amy Tuteur performed her own comparison, which i’ve tried to reproduce. The raw numbers, uncorrected for birth defects, complications, and twins (or more) offer a hospital baseline of 0.9 deaths per thousand. Including the three birth defect–related deaths and assuming an unchanged denominator (i.e. that all birth defects result in death, or that not very many occur at all) yields a combined death rate of 14/5418 = 2.6 per thousand. This is about thrice the baseline. Tuteur goes into more detail here.

The authors of the Dutch study used a different metric for neonatal mortality than others: rather than the 28-day period following birth, they truncated the time to 7 days. Nonetheless, the consistently close rates of intrapartum and neonatal death between the home and the hospital cohorts provides good reason to suspect that 28-day neonatal data would also conform. This study was also (naturally) performed in the Netherlands, where home birth has become the norm: almost twice as many deliveries were planned for home as were planned for the hospital. Home births have been embraced by and integrated into the Dutch health care infrastructure, whereas much of that infrastructure (home visits, rapid hospital accessibility) is not available in the United States.

The Wax paper became controversial immediately upon its publication. Within a year, the journal that published it, the American Journal of Obstetrics & Gynecology (AJOG), published several letters challenging the methodology and results along with some replies by Wax and coauthors. Erika Check Hayden discusses the major objections here and here, which i’ll summarize:

• Some statistics appeared to be off, and the authors had used a flawed meta-analysis calculator.
• The Washington study had been included in the neonatal mortality calculation.
• The Dutch study had been excluded from the neonatal mortality calculation.

As discussed in the Editors’ comment to the letters, an independent review panel convened by AJOG were shown the paper and the objections. They deemed most of the objections subjective, but did try to reproduce several calculations (addressing the first major objection). They obtained similar but not quite equal numbers and equivalent statements of statistical significance.

What can be made of all this? I draw the following provisional conclusions.

First, midwife certification matters. By their own standards of best evidence, Johnson and Daviss found an approximately tripled rate of neonatal death in home births attended to by certified professional midwives than in hospital births—which numbers should also include any planned home births that ended in emergency hospital trips. Meanwhile, when Wax et al excluded studies that did not specify midwives’ certification, or that specifically included non-CNM/CMs, the neonatal mortality rate lost its statistical significance; that is, the smaller batch of studies could not distinguish between “same effect” (as with non-nurse midwives included) and “no effect” (same rates as hospital plans).*

This is not a completely settled issue. Tuteur accuses MANA of restricting access to their own data, which could much better answer this question than existing analyses have. MANA claims that its data is readily available to whomever wishes to meaningfully analyze it, and cites good reasons for it to remain proprietary (unlike, for instance, the CDC’s birth/infant death records. Having studied proprietary data myself, i sympathize). The fact remains, however, that the only research to have used MANA’s data was Johnson–Daviss in 2005.**

It also makes sense a priori that the certifications confer different risks. CNMs are registered nurses, with all the training and experience that that entails. In comparison, CPM training is remarkably limited: a commenter at denialism blog points to the specifications (found in NARM’s certification standards, page 38) of a single written exam that, together with apprenticeship, is the threshold for CPM certification. Like L.Ac. (Licensed Acupuncturist) and ND (Doctor of Naturopathic Medicine), CPM seems in part to be an obfuscatory glorifier for a self-regulated community struggling to appear legitimate by modern medical standards. Admittedly, this limited training and apprenticeship make attendance by a CPM a far better option than going it solo. And NARM stresses that their midwifery certification is unique in requiring out-of-hospital experience.

Second, there is no good reason to believe that home births are better for newborns. While higher mortality rates may be associated to home births solely due to differences in certification, attempts to account for these differences, as in the Wax paper, have not yielded strong evidence for or against planning for home birth with a CNM over hospital birth. Meanwhile, home births in the United States are not directly comparable to those in the Netherlands; a Wax-like meta-analysis that included the Dutch study would have been overwhelmed by it, and produced essentially no interesting information about home births generally, much less in other countries (like ours). There is contention over whether the available evidence indicates better odds at hospitals or no discernible difference, but so far as i can understand the literature it is inconclusive.

Anyway, the home birth movement itself puts much more focus on the experience of the parent. They cite a stiff, distant, and otherwise inhospitable hospital environment and an excessive and unnecessary use of expensive interventions, which i’ll get to next time. Whether for good or bad reasons, though, they are right to put the emphasis where they do. Less a conclusion than a realization, borne of a search for meaning amidst the monotonous sorting of mortality statistics, is that the extent to which perinatal and neonatal mortality matters depends on the expectant parent.

There are two components here. One is that, regardless of anything—the consciousness or agency of the fetus/infant, the emotional investment of other people—the pregnant person alone has the prerogative to decide the circumstances under which they will carry and deliver. This follows from the principle of bodily integrity that also, so far as i can tell, seals the reasoned and reality-based abortion debate. The principle is not absolute, and the comparisons of pregnancy to other scenarios are imperfect, but ultimately what reasoned objections are leveled don’t stick. The other component is that fetus/infant survivability is only an ethical concern in the first place to the extent that it affects the people involved in the pregnancy. The most reasonable point to begin conferring elements of personhood upon a fetus/infant, again so far as i can tell, is birth itself—the flood of stimuli that allows it to begin pattern-seeking and develop consciousness.

This is of course not to say that expectant parents should resort to CPM-led home births only when they undervalue their child’s survival. Rather, if the hospital environment, for whatever reason (previous trauma, emotional investment in ideological opposition), would compromise the parent’s psychological health (and omitting for now, perhaps in a false dichotomy, the birth center option), then only the parent is poised to judge whether a slightly reduced risk of infant mortality is worth submitting themselves to this environment.*** The caveat, as always, is that this judgment should be informed. That obligation falls to the parent, the establishment, and the home birth movement.

* Statistically, the numbers are damning. Viewing p₁ and p₂ as the probabilities of infant mortality in the case of CNM and CPM attendance, a two-sample proportion comparison in R for the one-sided alternative that p₁<p₂ yielded a p-value of 0.00003. (I had trouble computing a posterior for p₁<p₂ using a beta-binomial prior. If someone can show me how to do this in R, i’d be grateful.) Some authors have mentioned, however, that with so numerically few instances of death the calculation of confidence intervals and p-values can be misleading. I suspect that this is because those few instances may carry confounding factors that undermine the underlying statistical assumptions—for example, if more than a couple of unsuccessful deliveries were led by CPMs from the same training program, and this program were not representative of CPM programs generally.

** In view of the legitimate uncertainty and high-profile debate surrounding these topics, it strains credulity to believe that no qualified researchers have appealed to MANA for access to their data.

*** Psychological strain on the parent may itself impact the survivability of an infant, but in the absence of data this consideration is mostly speculative.