Comparing Neonatal Mortality Rates for CNMs and Other Midwives
Plate from the first printed textbook of midwifery, c. 1513
I’ve been following with interest the discussion of homebirth midwife education at The Skeptical OB.
In her post, Dr. Tuteur repeats her oft-made claim that direct-entry midwives (DEMs) are poorly trained and incompetent compared with MDs, certified nurse-midwives (CNMs) and midwives in other wealthy countries.
Many claims for the safety of CPM-attended homebirth are the frequently cited Johnson & Daviss study, in which the authors concluded that planned homebirth with a CPM had intrapartum and neonatal mortality rates comparable to hospital birth (and which Dr. Tuteur has critiqued), but I was curious to see if there was any other evidence.
I took a closer look at the CDC’s so-called WONDER Database (cited in my recent post on the ACNM), which contains information on births and deaths of infants born in the U.S. between 1995 and 2005, querying the database for early neonatal mortality rates among different providers for low-risk women.
I hadn’t planned on posting about it, but what I found was so startling that I thought I would offer it up for discussion.
In every case, mortality rates for the “other midwife” category is considerably worse than for other categories. When compared with those for CNMs (the group with patients most closely matched for risk level), other midwives’ mortality rates are at least three times higher.
Query Parameters
In an attempt to eliminate neonatal deaths due to issues not generally related to quality of obstetrical care, I used the following parameters:
- Early neonatal deaths = < 1 hour to 6 days of age
- White women, ages 20-44
- 37+ weeks gestation
- Singleton pregnancy
- Prenatal care beginning in months 1-6
- Birth weight 2500+ grams
- Deaths due to “certain conditions originating in the perinatal period”
To eliminate (insofar as possible) unassisted births, I also eliminated births within the following parameters:
- Place of birth listed as “unknown or not stated”
- Medical attendant listed as “unknown or not stated” or “other”
Note: I did not query the database for 1995-1998 because it does not allow queries by attendant type, and uses ICD-9 codes for cause of death, which may not match up with later ICD-10 codes. The CDC separates databases for 1999-2002 and 2003-2005 because the latter group includes additional categories for maternal race.
I have denoted with an * where the numerators are <20, and the morality rate does not meet National Center for Healthcare Statistics standards for reliability or precision.
| CDC WONDER 1999-2002: Early neonatal deaths—low risk women—perinatal causes | |||
| Attendant | Deaths | Births | Mortality Rate/1,000 |
| CNM | 63 | 706,886 | 0.09 |
| MD | 1,187 | 8,045,250 | 0.15 |
| DO | 63 | 418,081 | 0.15 |
| Other Midwife | 19 | 48,438 | 0.39* |
| CDC WONDER 2003-2005: Early neonatal deaths—low risk women—perinatal causes | |||
| Attendant | Deaths | Births | Mortality Rate/1,000 |
| CNM | 33 | 446,015 | 0.07 |
| MD | 645 | 4,949,740 | 0.13 |
| DO | 30 | 283,693 | 0.11 |
| Other Midwife | 7 | 27,226 | 0.26* |
One claim frequently made by proponents of midwifery and out-of-hospital birth is that physicians and CNMs who practice in hospitals are too quick to induce postdates babies, so I also calculated rates for neonates born at 41+ weeks gestation.
Mortality rates for postdates neonates were significantly higher for non-CNM midwives than for any other attendants.
| CDC WONDER 1999-2002: Early neonatal deaths–perinatal causes–41+ weeks gestation (otherwise low-risk) | |||
| Attendant | Deaths | Births | Mortality Rate/1,000 |
| CNM | 21 | 170,074 | 0.12 |
| MD | 275 | 1,572,702 | 0.17 |
| DO | 15 | 82,964 | 0.18* |
| Other Midwife | 7 | 13,626 | 0.51* |
| CDC WONDER 2003-2005: Early neonatal deaths–perinatal causes–41+ weeks gestation (otherwise low-risk) | |||
| Attendant | Deaths | Births | Mortality Rate/1,000 |
| CNM | 9 | 96,753 | 0.09* |
| MD | 119 | 847,675 | 0.14 |
| DO | 7 | 48,861 | 0.14* |
| Other Midwife | 5 | 7,577 | 0.66* |
Conclusions?
For a wide variety of reasons[1] , this data cannot be used to draw valid conclusions about the competence of any group of providers.
One particularly important caveat is that the numbers for the “other midwife” category are so small that a single death can disproportionately influence the mortality rate.
Nevertheless, I find it interesting that the numbers consistently point to poorer outcomes for non CNM-midwives (I ran several additional queries using other parameters that revealed similar results, but chose not to include them here because they included factors that either placed the mothers at higher risk, or causes of death largely unrelated to birth.)
What it suggests is not necessarily that non-CNM midwives are incompetent, but that it is incumbent upon the organizations that advocate for them to conduct or sponsor additional, high-quality research into potential disparities in quality of care, both among differently-trained midwives, and between non-CNM midwives and other providers.
- 1. This is raw data, and can’t be compared with a meticulously designed and conducted study for a wide variety of reasons, including:
- It is based on birth and death certificate information, which can be unreliable, particularly with regard to cause of death.
- It contains only information about actual place of birth rather than intended place of birth, which would be important in determining which deaths were due to unintended out-of-hospital birth, and which deaths occurred in hospital after transfer from homebirth or other out-of-hospital settings.
- It provides fairly simple parameters with which to eliminate potential confounding factors.
- The “other midwife” category includes all non-CNM midwives, from CPMs to lay midwives.
- Finally, and most importantly, this is based on my own review of the database. I am not a scientist, and have little formal education in statistics. Anything I say here should be taken as nothing more than the observations of an interested lay person.
Wendy,
“FWIW, the vast majority of midwives in the CPM study were apprentice-trained.” — where did you find that info? I couldn’t locate it in the study. I only found references to CPMs in general.
I don’t want to speak for her, but I believe Dr. Tuteur’s objections to the CPM2000 study are based on more than her belief that their undisclosed association with MANA biased the results. Her primary objection seems to be that they used an older set of data (ranging from the 1970s to the 1990s) for the hospital-birth group than for the homebirth group (2000 only) to compare mortality rates. The explanation they offer on their website–that “administrative records” are not ideal for accurately determining mortality rates–may be true, but it does not change the fact that the mortality data they use is skewed against hospital birth.
The fact that they used the “administrative data” from the year 2000 to calculate hospital intervention rates for comparison with those at homebirth makes their defense seem disingenuous to me. Moreover, in their follow-up, in which they ostensibly attempted to correct this disparity by substituting 2004 NIH data for the older data, they also adjusted the data by removing deaths from congenital anomalies from the CPM group but not from the hospital group, explaining that the hospital cohort would be less likely to carry fetuses with congenital anomalies to term. I will confess to being unsophisticated in matters of statistical analysis, but I have to wonder why they did not simply remove congenital anomalies from both cohorts.
Moreover, the CPM study does not address differences of outcome (if any exist) between CPMs and CNMs, nor do the various studies to which you refer, with the exception of Janssen, 1994, which looked at outcomes for CNMs and licensed midwives (not CPMs, as the credential did not yet exist) in Washington between 1981 and 1990.
As you say, several decades’ worth of research (both good and bad) on homebirth has been done; indeed, most of the studies cited in Johnson & Daviss are at least 15-20 years old, and many of the more recent ones are not from the U.S. I’m not sure how much validity these have in comparing CPM and CNM outcomes today, but I’m interested to hear what you think about it.
To echo what you wrote, I very much hope a good study comes out of the MANA data.
One of the great things about a study like this is that you have excellent data for CPMs. You can compare it to any other study you like. Ideally, the quality of the comparison data would be at least as good as the CPM data, but that’s been rather hard to come by. Analyses like the ones that you and Amy have done are really quite a stretch, for all the reasons that you have outlined. Also, tests for statistical significance have not been performed on your analysis. Someone has done those tests on Dr. Tuteur’s analysis and found her differences to lack statistical signficance. (That has not stopped her from posting her analysis all over the web, which I find to be not only “disingenuous” but deliberately misleading to fail to include the fact that it lacks statistical significance.)
One of the issues that was highlighted by the recent Netherlands study of a half million planned homebirths is that a major element of their success is a smooth and coordinated transport system where the midwife retains continuity of care from home to hospital. In the U.S., homebirth midwives are often met with outright hostility by hospital providers when they bring someone in. This must affect the decision-making process of the midwife at home when a client is starting to show signs of possibly needing transport, and in turn, it must affect outcomes. All this is to say that even if we had great data on the hospital side to compare with the CPM data, there are elements inherent in the American medical system that create (some) poor outcomes for both sides: hospitals can make it hard for the homebirth providers, and midwives can make it hard for the hospital providers. I hope to see the day when all maternity care providers can work together with the best intentions for mothers and babies at the forefront.
The category of “Other Midwife” doesn’t differentiate between Certified Professional Midwives, who are educated and trained to provide safe home birth care, and so-called direct entry midwives, who may or may not be adequately trained and are often known for taking high-risk women because they are not held to any standards of care or training at all.
Wisconsin CPMs analyzed the birth certificate and CPM 2000 study data for their state and found that the neonatal mortality rate for CPMs in the year 2000 was zero per thousand births, compared to 11.63 per thousand for “Other Midwives.”
The point is that anyone can call themselves a “direct-entry midwife” and check the “Other Midwife” category on a birth certificate regardless of their level of education, training or skill, which vary widely, thereby making it a meaningless title. Data that doesn’t differentiate between CPMs and DEMs is equally meaningless.
Wisconsin Doula:
“Data that doesn’t differentiate between CPMs and DEMs is equally meaningless.”
While I’m not sure it’s completely “meaningless,” I agree that the differentiation is important in determining not only differences in outcomes between CPMs and other midwives, but between CPMs and CNMs. That’s why I mentioned it in my post.
The only group (that I am aware of–please correct me if I’m wrong) that has collected a large body of statistics that does make this differentiation is MANA, and they have not yet released their data. Given their support for The Big Push for Midwives and the MAMA Campaign, I find it troubling that they have not done so in support of their efforts.
The MANA statistics are voluntarily submitted by any type of midwife who chooses to enter her data, including CNMs, CPMs, CMs and DEMs.
I don’t know if the MANA database will differentiate among the various categories of midwives, but given that it’s voluntarily submitted and self-selected, I wouldn’t put much stock into it. It’s very unlikely that the data will include statistics from DEMs who accept high-risk clients and drive up mortality rates as a result.
For example, I seriously doubt that “midwives” from sects such as the Church of the First Born or the Twelve Tribes, which have left a trail of dead mothers and babies in their wake, are going to be submitting their statistics to the MANA database. Yet so-called midwives from these and many other sects that do not believe in any form of intervention besides prayer are included in the “Other Midwife” category on birth certificates.
This is a good point. My own primary interest is to compare outcomes between CNMs and CPMs.
Were it to turn out that CPMs as a group have outcomes comparable to CNMs, I would be quite comfortable advocating for expanded recognition of the CPM credential, in addition to making it the minimum required standard nationwide. (Assuming such a feat is feasible.)
Nevertheless, I also believe it behooves CPMs to investigate whether or not the different training paths (PEP vs. midwifery program) makes any difference. Ongoing evaluation of training programs is an essential part of health care policy.
Non-CNM midwives are at a distinct disadvantage in that, because their training programs are not integrated into university settings, as are CNM programs, they do not have the same opportunity to participate in research. In my opinion, this is one of the biggest shortcomings in the way the CPM evolved. It leaves the onus of funding research largely on organizations like MANA, whose resources and interests are quite different from those of a large university.
“”the numbers for the “other midwife” category are so small that a single death can disproportionately influence the mortality rate.”"
So then, If you knew the comparison was flawed, why post it? What you should have said was that you tried to find evidence to substantiate Tutuer’s claims and found none based on sound data.
I posted it because it suggests that DEMs and their advocates should take a much closer look at their safety record before they attempt to convince legislatures and the public that they should be recognized as competent health care providers.
The study most often used to support CPMs’ claims uses a much smaller sample size (5,418) than any of those I listed, from a single year (2000). Moreover, It seems reasonable to call attention to information that MAY contradict the conclusions from that study.
I thought it was interesting that every query I ran–with raw data and all its shortcomings acknowledged–yielded a similar result. I posted it because I think anyone interested in this topic might find it worth noting as a topic for further discussion.
Finally, I posted it because I think most readers are smart enough to decide for themselves to what degree the information is meaningful, given the shortcomings I listed.
The analysis that you did was not a “study” — as you pointed out, you crunched some raw data from a database of numbers pulled from birth and death certificates. The CPM study was designed to specifically look at outcomes of planned homebirths with Certified Professional Midwives and rigorously collected data for all such births in the year 2000. You mentioned that your “sample size” was larger than the CPM study, but there is a world of difference between the quality of data in what you did and what they did. The size itself is secondary — a large quantity of poor data does not trump a smaller quantity of excellent data.
You also stated a couple of times that it was interesting that every query you ran gave you similar results. I think that you have to consider the possibility that since you used the same database with the same exclusion criteria for your queries and the same flaws inherent in the data, it might be reasonable to expect the same trends in the results.
Finally, the concept of statistical significance has not been mentioned yet. Differences in outcomes can be calculated, but if they’re not statistically significant, then the differences could be due to random fluctuations in the data and not real at all. We can’t draw any conclusions or make any claims about the information if the differences are not statistically significant. This is one of the major flaws in Dr. Tuteur’s claims that she conveniently leaves out — the differences that she has found in her “analysis” are not statistically significant. This has been thoroughly hashed out on her old blog, but she either doesn’t understand it or wants to deliberately mislead people by refusing to acknowledge it.
I appreciate your willingness to be transparent about your analysis and the shortcomings that it has in drawing conclusions. I think that we have to be equally careful in the statements made about those trends or any comparisons when part of the data is much less clearly defined than the rest.
WendyCPM,
I agree with your comments. I would be very interested to see further, rigorous study of outcomes for direct-entry midwives (and especially between midwives with different training–e.g., CPMs vs. other midwives, apprenticeship-trained vs. didactic+clinical.) I don’t think any valid conclusions can be drawn about the safety of out-of-hospital birth with direct-entry midwives based on the information currently available. That is the point of the post.
I agree with Dr. Tuteur insofar as she points out the differences in training between CNMs and direct-entry midwives and the need for MANA to release its statistics or produce a study based on them. That data may yield the information needed to determine whether those educational differences have an effect on outcomes.
To clarify: in the additional queries I mentioned, I used a variety of exclusion criteria, different from those I used for the posted charts. Or are you referring to exclusion criteria inherent in the CDC’s databases?
I agree that it would be interesting to see some studies comparing different types of midwives. At this point, since licensure is still not available in all 50 states, the studies would be small and state-based or regional. FWIW, the vast majority of midwives in the CPM study were apprentice-trained.
You said “I don’t think any valid conclusions can be drawn about the safety of out-of-hospital birth with direct-entry midwives based on the information currently available.” The CPM study offers quite a bit of valuable information about the outcomes of planned home births with Certified Professional Midwives:
* c-section rates were below 4%
* rates of interventions were rock-bottom: epidurals 4.7%, episiotomy 2%, forceps 1%, vacuum extraction 0.6%
* intrapartum and neonatal mortality was 1.7/1000
* no mothers died
Dr. Tuteur takes issue with the comparisons made to low-risk hospital births in the study and would like to invalidate the entire thing over who the researchers were, but the data presented can be evaluated independently of the authors’ conclusions. This is prospectively-collected data using an extremely well-designed tool, and I don’t think it’s fair to say that “no valid conclusions can be drawn” about it.
We should always strive to renew the evidence base on ALL obstetric practices, and I commend your earnest attempts to explore the data to see what it says. There are several decades worth of research that has been done on the safety of homebirth, some well-designed and some poorly designed, but the vast majority of it supports the conclusion that planned homebirth for low-risk women with a qualified midwife is no more risky than birth in the hospital, but with far fewer interventions. I agree that more research needs to be done, and MANA has just closed a 4-year dataset of several thousand births that will be made available to researchers this year. We’re all looking forward to finding out what’s in it!
P.S. The data you used to run your queries has the same inherent bias regardless of what outcomes you are looking at. Intended place of birth is not captured, “other midwife” is not defined, and the general unreliability of birth and death certificate data — all of the things that you mentioned in the footnotes of your post. Planned place of birth and definition of “other midwife” disproportionately affect the midwife’s numbers.