BMJ Home Birth Study Questions
 Understanding Birth Better


 English  Français  Español 

BMJ Home Birth Study Questions

Please do not post this information as is on your website as we may update it from time to time. We welcome you to make a link to it, however, and would appreciate you contacting us if you do so.

In this section we report:

Risks Similar in Careful Comparison of the CPM2000 and the 2004 U.S.
Neonatal Mortality among Term Births to non-Hispanic White Women

And answer the following questions:

1.   Why did you choose to publish this study in the British Medical Journal?

2. Why did you choose to compute an intrapartum and neonatal mortality rate rather than a perinatal mortality rate (PMR)? How were birth defects, premature births, and stillbirths reported?

3. What was the basis of your conclusion about the safety of home versus low risk hospital birth?

4.  Why does the Washington home birth study have different conclusions than almost all other articles on home birth?

5. Are you comparing apples to oranges?

6. Why did you choose not to provide confidence intervals for the intrapartum and neonatal death rates?

7.   Why did you choose not to provide details about other outcomes?

8.  Could the funders have biased the study in any way?

Risks Similar in Careful Comparison of the CPM2000 and the 2004 U.S.
Neonatal Mortality among Term Births to non-Hispanic White Women

Kenneth C. Johnson, PhD and Betty-Anne Daviss, MA, RM


We compare the CPM2000 neonatal mortality rate among planned homebirths to the U.S. National Institutes of Health (NIH) neonatal mortality rate for births in hospital to U.S. non-Hispanic white women of 37 weeks plus gestation in the year 2004. Adjustments are made to ensure that the comparison is as close as possible to comparing like with like. This includes removal from the CPM2000 death rate of intrapartum mortality, 3 deaths involving lethal birth defects unlikely to have been carried to term in the hospital population, and 1 death and 286 births among African-American and Hispanic women. After making the necessary adjustments that were possible, the neonatal death rate in both datasets was just under 1 death per 1000. Our conclusions remains unchanged from those in the original article -- the neonatal mortality rate for low risk women in North America using certified professional midwives is similar to that for low risk women in hospital in the U.S., and the intervention rates are much lower. We note that the premature birth rate for the NIH non-Hispanic white births in hospital was 11.3%, more than double the rate for the women cared for by Certified Professional Midwives (CPMs). Higher prematurity is a serious concern for the U.S. hospital births, because prematurity is associated with much higher perinatal mortality and morbidity.

Comparing like with like

The issue of comparing neonatal mortality between different datasets requires caution—it is important to be comparing like with like. In the BMJ 20005 article we provided the following caution regarding such comparisons as a footnote to Table 3:

"The Table is presented for general comparison only. Direct comparison of relative mortality between individual studies is ill advised, as many rates are unstable because of small numbers of deaths, study designs may differ (retrospective versus prospective, assessment and definition of low risk, etc.), the ability to capture and extract late neonatal mortality differs between studies, and significant differences may exist in populations studied with respect to factors such as socioeconomic status, distribution of parity, and risk screening criteria used. For example, see the study by Schlenzka. Although the crude mortality for low risk babies weighing over 2500 g intended at home was 2.4 per 1000 and intended in hospital was 1.9 per 1000, when standard methods were employed to adjust for differences in risk profiles of the two groups (indirect standardization and logistic regression), both methods showed slightly lower risk for intended home births." (Johnson and Daviss, 2005)

Comparison to U.S. NIH birth data

The U.S. NIH provides data on all live births in the US. As with any other comparisons, however, for intrapartum and neonatal mortality we were careful in our article in the BMJ to caution against direct comparisons with the NIH as it was not a comparison of like with like. Since our article was submitted for publication in 2004, the NIH has published analysis more closely comparable than was available at that time, and some have tried to use it as a comparison. While we still do not offer the comparison as a completely direct one, as it is the closest we have and the comparison is occurring regardless of our cautions, we offer the following adjustments that have to be made to provide the comparison of the CPM2000 analysis in as accurate a manner as is possible with the published NIH analysis.

In doing the analysis, we came to the same conclusion as we did in the original article: the neonatal and intrapartum deaths for low risk women in North America using certified professional midwives are similar to both the NIH and other studies on low risk women in hospital in the U.S., and the intervention rates are lower.

(Note: Definitions of mortality categorization are provided at the end of this report.)

Here are some issues:

a) Adjustment needed to the NIH 2004 Neonatal Mortality Rate to more closely approximate the rate for the year 2000

The reported NIH data for the U.S. for 2004 based on linkage of live birth and death certificates was 0.76 neonatal deaths/1000 births for Non-Hispanic white neonates of 37 weeks plus gestation. (Matthews et al., 2007) (The NIH only started to publish analysis of gestation –specific neonatal mortality rates by ethnic type for the 2004 data, after the BMJ article was submitted for publication.) Adjustments for the decreasing death rate among term babies documented by NIH would have placed the unpublished NIH 2000 rate slightly higher at about 0.91 neonatal deaths/1000 live births (based on the changes in death rates for babies of 2500 grams or more between these two time periods).(Matthews et al., 2002, Matthews et al., 2007)

b) Exclusions required for the CPM2000 data

A crude comparison of the CPM2000 death rate to the neonatal mortality rate among U.S. Non-Hispanic White women with 37 week plus births would also require the following exclusions:

· 5 intrapartum deaths need to be removed as the NIH data report only on live births and thus include only neonatal deaths

· 1 home birth neonatal death that was among the 286 Hispanic and African-American births in the dataset. Both the death and 286 births need to be removed from the comparison as they did not fit the non-Hispanic white women category provided by the NIH.

· 3 neonatal deaths caused by fatal birth defects need to be removed. All three of these deaths would have occurred regardless of whether the birth was planned at initiation of labour to be in hospital or at home. We are only adjusting for "when" the death would likely have happened for better comparison of like with like for the term neonatal death rates. Had these three birth defect deaths occurred among the hospital population in the present medical culture, they would have been far more likely than not to have been induced or terminated before term. The first was a baby who was dwarf with associated lethal anomalies, discovered on ultrasound but declined for hospital birth by the Amish couple, as they did not want to be induced or have the routine autopsy done at their hospital. The second was a baby with Trisomy 13 born to a woman of 42 years. This is picked up with routine genetic testing, recommended especially for women over 35, and if the mother has declined genetic testing, usually picked up on ultrasound. The woman simply chose not to have any testing. The third was a baby with Acrocallosal Syndrome (multiple severe anomalies) that would have been picked up on ultrasound, born to a couple that disagreed with routine ultrasound and did not have one. These three women chose to let nature take its course, in spite of their awareness of medical knowledge, and this changed the timing of the eventual deaths but not the outcome.

The exclusions are summarized in Table 1.

Table 1: CPM2000 neonatal death rate adjusted for crude comparison to National Institutes of Health Non-Hispanic White >37 weeks neonatal death rate

Description of Deaths in the CPM2000 Study

Total Deaths

Death Rate




Total Deaths (including 3 birth defect deaths): 5 intrapartum, 9 neonatal deaths

14 deaths among 5418 births





Deaths removed from CPM2000 to make data comparable to NIH rate:



- 5 intrapartum



- 3 birth defect deaths



-1 death among 286 Hispanic/ African- American



"Comparable" Crude Death Rate

5 deaths among 5,132 births

0.97 neonatal deaths/1,000 live births

c) Crude Comparison Yields Similar Risk Estimates

Thus a crude comparison of the comparable rates for non-Hispanic white >37 week babies in hospital in the year 2000 would be about 0.91 neonatal deaths/1000 live births and for the CPM2000 0.97 deaths per 1000 live births, virtually indistinguishable.

d) Further Comparison Refinements Required

Even among low-risk women, risk will vary within subgroups of this population, so a proper comparison requires adjustment for the differences in the populations one is trying to compare. Adjustment for differences in the characteristics of the two populations (distribution of socioeconomic status, parity, risk screening, percentage of older mothers, percent Amish, percent grand multips, etc.) would be required to compare more closely. This adjustment is only possible when both datasets have collected the same information using a similar methodology. The NIH report does not provide rates of neonatal deaths for these subgroups (for example, Amish, grand multips, etc) among the non-Hispanic white 37 weeks gestation plus women, to allow us to make these adjustments for comparison to the CPM2000 dataset.

As an example of how the differences in even low-risk populations can affect the comparison, we provided the example of the study by Schlenzka in California in the section above "Comparing Like with Like."

e) The Larger Issue – Perinatal Mortality and the High Rate of Prematurity in U.S. Hospital Births

Our study focussed on term intrapartum and neonatal death rather than death before labour in order to understand the difference in outcomes between women who choose at the initiation of labour to birth in home or, conversely, to birth in hospital. The larger issue is the perinatal mortality associated with different caregivers, which includes pre-labour as well as intrapartum and neonatal deaths.

It is relevant to note that 11.3% of non-Hispanic white U.S. live births in hospital in the NIH stats happen before 37 weeks. (I.e. 11.3% of non-Hispanic White U.S. births are premature.) This is substantially different from the data among the CPMs in the year 2000 in the BMJ article, where there is an approximate 4% prematurity rate. (This rate is based on the 1.1% of women who delivered at home before 37 weeks, the 58 (1 %) of women risked out because of prematurity; an estimated half of the 205 women who were risked out for complications before labour (2%) and the 3 neonatal deaths that involved fatal birth defects that would have likely been induced or terminated before term in the hospital population.) We draw attention to this difference because the infant mortality is much higher in premature births. Even for women who deliver between 34 and 36 weeks the infant mortality is 3 times as high as it is for 37 weeks plus. Births between 34 and 36 weeks represent 2/3 of the 11.3% premature births in white non-Hispanic births in the U.S.

f) Fewer Premature Babies with CPMs

Thus either there really are fewer premature births with CPMs, or gestation reporting is different in hospital and some of the babies identified as premature in hospital were actually term or would have been included as term births (and deaths) rather than preterm in the CPM2000 project. A possible reason for this from anecdotal evidence is that defining risk can take place subsequent to delivery. Some babies who by date were term might have been re-categorized, after delivery and a newborn exam, as preterm in search for an explanation for their demise.

With either explanation for the difference between the CPM and NIH rates of prematurity, this has a serious impact on neonatal mortality estimates and comparisons that need to be carefully considered and would very likely remove any excess risk in the CPM2000 in comparison to the US non-Hispanic births from the NIH data. This is also why comparison of perinatal mortality rather than neonatal mortality is usually preferred because it includes premature births and gives a better overall risk picture. The nature of our study, however, called for intrapartum and neonatal mortality in order that the risks associated with planned homebirth could be isolated--i.e., mortality that occurred from the time labour began and a homebirth was still planned. Thus we provide the extra information about the antenatal status of clients of CPMs here, to provide the more complete picture.

As we reported in our discussion of the BMJ article, there was no such difference for perinatal mortality between planned out of hospital and hospital births for all of California in 1989-1990 in a study that was able to establish similar low risk profiles because birth and death certificates in California include intended place of birth and these certificates had been linked to hospital discharge abstracts for 1989-1990 for a cesarean section study. This is one of the reasons why we identified the Schlenzka study as an important adjunct to our study.

g) Comparison of the CPM2000 Study to the MacDorman Study

A study by MacDorman et al. (MacDorman et al. 2006) found that for women with no risk factors, the perinatal mortality was 0.7 deaths per thousand. It has been suggested as a comparison group for examining risk in the CPM2000 study. However that would not be a comparison of like with like. It is not appropriate as a comparison group to the BMJ study results because it excluded not only births with risk factors discovered before labour but also those births where a risk factor occurred during labour. The CPM2000 study did not exclude women with risk factors that occurred during labour. Ida Darragh’s response to one blogger is succinct and accurate:

"…you continue to claim that the MacDorman study published in Birth indicates a much lower hospital neonatal mortality rate than the BMJ study of CPM births. The MacDorman study looked at only one thing - the difference in mortality rate based on cesarean or vaginal delivery. The BMJ study did screen out high-risk factors that were identified before labor began, but the MacDorman study screened out any risk factor that was identified during labor, including prolonged rupture of membranes, prolonged labor, and meconium. They looked at outcomes related only to method of delivery. The BMJ study looked at all outcomes of births attended by CPMs once labor began, and included any transports for risk factors in labor. The low mortality rate you quote was also only for the vaginal births; the cesarean births in the MacDorman study had a higher neonatal mortality rate. The BMJ mortality rate included both cesarean and vaginal deliveries. It is not unexpected that the mortality rate for vaginal births with no risk factors at any time would be lower than the rate that includes labor transports and cesarean births. As was clearly stated in the BMJ study, it is impossible to compare research outcomes without using the same criteria for evaluation."


Intrapartum Mortality - baby who died during labour (before birth)

Neonatal Mortality – live born baby that died in the first 28 days of life

Early Neonatal Mortality – live born baby that died up to 7 days of life

Late Neonatal Mortality – live born baby that died between 7 and 28 days of life

Infant Death - live born baby that died in the first year of life

Perinatal Mortality - deaths between 20 or 22 or 28 weeks gestation and 7 days of life


Johnson, K.C., Daviss, B.A., 2005. Outcomes of planned home births with certified professional midwives: large prospective study in North America. BMJ 330, 1416.

MacDorman, M.F., Declercq, E., Menacker, F., Malloy, M.H., 2006. Infant and neonatal mortality for primary cesarean and vaginal births to women with "no indicated risk," United States, 1998-2001 birth cohorts. Birth 33, 175-182.

Mathews TJ, Menacker F, MacDorman MF. 2002. Infant mortality statistics from the 2000 period linked birth/infant death data set. National vital statistics reports; vol 50 no 12. Hyattsville, Maryland: National Center for Health Statistics.

Mathews TJ, MacDorman MF. 2007 Infant mortality statistics from the 2004 period linked birth/infant death data set. National vital statistics reports; vol 55 no 15. Hyattsville, MD: National Center for Health Statistics.

Content Last Updated: November 20, 2007

Answers to Questions About "Outcomes of planned home births with certified professional midwives:  large prospective study in North America"

Kenneth C. Johnson, Betty-Anne Daviss

Please do not post this information as is on your website as we may update it from time to time. We welcome you to make a link to it, however, and would appreciate you contacting us if you do so.

The following questions have been asked by one or more midwives, physicians, legislators, or consumers about the CPM2000 Study published in the British Medical Journal (June 18, 2005), which is accessible for free on the web at

Although for the most part, these questions were addressed in the article itself, and a re-read should clarify most of them, some may find a question and answer format is easier to navigate.  As well, some added information which could not be squeezed into the word restrictions of a medical journal can be helpful.  We think it critical that  the article be well understood because it has attracted enormous attention not just in North America but worldwide, partly due to the fact that it is the largest prospective home birth study done to date, possibly because it is largely based in the U.S , and contextually because the debate over home birth and hospital birth has grown in sophistication since the early eighties and the article captures the best available methodology for a national study of the U.S. (with a glimpse of Canadians who have the same certification.)


1.   Why did you choose to publish this study in the British Medical Journal?


There was great interest in our study as it was drawing to a close and we received advice from several different parties about the best strategy for publication.  One of the obstetricians/epidemiologists who had originated Effective Care in Pregnancy and Childbirth at the National Perinatal Epidemiology Unit (NPEU) in Oxford, suggested we send the article to the British Medical Journal because of their web-based system and prestigious reputation.  However, American colleagues wanted us to submit to the Journal of the American Medical Association (JAMA), to see if it would spark some interest among American physicians.  Therefore, we submitted first to JAMA.  The article was returned quickly without being sent out to reviewers, and comment from only one editor – that the JAMA readership would not be interested in an article on home birth.   This was regardless of the calibre of the methodology and size of the study.


We turned to our advisory council, including obstetric, public health, epidemiology, and nurse midwifery consultants, to debate the relative merits of submitting a home birth article to the Lancet, the New England Journal of Medicine (NEJM), the American Journal of Public Health (AJPH) or the British Medical Journal (BMJ).  It was thought that there was high probability that the NEJM and the Lancet, even if they sent the article on for review, would respond to this topic with similar lack of interest as a priority item to publish. We assumed that the American Public Health Association’s journal would be interested because the study’s methodology was approved and applauded by epidemiologists in the organization, and subsequent to presentation of the preliminary findings at the APHA conference in 2001, a resolution was passed in the Maternal Child Health Division to increase out-of-hospital birth done by direct entry midwives (American Public Health Association 2001).  However, at that time it took more than a year to reach print in the APHA journal and it does not have the same distribution or reputation as the BMJ.


We returned to the original decision to submit to the BMJ because:

1) It is one of the most highly respected medical journals with high standards for study acceptance and is a preferred place for publication (the BMJ publishes less than 10% of the papers it receives).

2) It publishes articles it thinks are important and timely, and had previously published on home birth without apparent concern their readers would lack interest;

3) The BMJ reviews and publishes manuscripts quickly;

4 The BMJ is one of the few journals which provides articles freely available on the web to everyone, allowing anyone to get a copy to review it;


5) The BMJ is one of a small number of journals that maintains the Rapid Response feature which allows the readership to post letters to the editor quickly – this is a great benefit to the readership of an article, compared to the traditional average of 6 months before letters to the editor appear in print. (See *Note 1 at the end of this question.)

6) Rapid Responses also allow for a greater diversity of opinion than traditional letters because the BMJ posts all letters of substance whereas rarely are more than 3 or 4 letters to the editor chosen to be published in print and the choices are totally at the editors discretion about which if any letter will be published.


In the end, it appears the BMJ was the best place to publish it because the article has been accessed by over 25,000 individual readers and according to the BMJ continues to be accessed at the rate of about 1500 times per month (Personal correspondence with British Medical Journal). It was the third most accessed article in the first week of publication in the BMJ for the year 2005.(Top Ten List for 2005, BMJ)


*NOTE 1: For example, consider the NEJM publishing of the Lydon-Rochelle et al. study (2001) of VBAC and the accompanying editorial. It was six months before 6 letters to the editor were published in the NEJM and 2 in the BMJ (Daviss 2001; Johnson and Gaskin 2001) --  all highly critical of the study and/or editorial – but only after hospital protocols had changed as result of the original publication and had caused considerable damage to choice of VBAC for women in America.  Furthermore, when we presented on the information in the letters to the NEJM, in our tertiary care setting in Ottawa, none of the obstetricians, midwives or obstetrical nurses were even aware of the letters, let alone knowledgeable of their content, yet our hospital had  based their informed consent for patients on the this and one other article alone.  VBAC at our hospital as well as the majority across North American appear to have similarly acted on the information from this article, evidenced by another drop on the graph in the downward trend away from offering VBAC in North America.


2. Why did you choose to compute an intrapartum and neonatal mortality rate rather than a perinatal mortality rate (PMR)? How were birth defects, premature births, and stillbirths reported?


Perinatal mortality rate (PMR) refers to the sum of deaths of fetuses and neonates per thousand births, but there are variations in the definition because jurisdictions can include or exclude early fetal death (22 weeks to 28 weeks) and/or late neonatal death (7 to 28 days after birth). Thus the WHO’s definition "Deaths occurring during late pregnancy (at 22 completed weeks gestation and over), during childbirth and up to seven completed days of life" is not universally used.


The perinatal mortality rate (PMR) debate -- how and whether to use it to compare practitioners and place of birth -- ensued at the NPEU in Oxford when we were there in 1991, on the tail of the Australian home birth studies.  PMR is relevant if one wants to compare the effectiveness of midwifery vs. physician care or the difference in care between jurisdictions both during the prenatal period as well as during the intrapartum and postpartum period.  However, it is not the most relevant measure for examining the risk associated with place of birth, which was the focus of our article in the BMJ. 

Intrapartum and neonatal mortality better address the clinical question of risk associated with planned out-of-hospital and hospital birth, by isolating the risk related to delivery itself.     Therefore, while we collected PMR, we reported in this particular article on intrapartum and neonatal mortality.  We chose deaths up to 6 weeks of age, because this includes both early and late neonatal death (up to seven and 28 days respectively) and it is possible to tally this information because 6 weeks is the common time of the last visit with the midwife.  Unfortunately, this information is not captured consistently among physician-attended births because of the fragmentation of care subsequent to birth in hospital.  Mothers generally go to family physicians, general practitioners, and pediatricians after their birth for baby care rather than obstetricians.  However, the subsequent difficulty in finding comparison groups does not suggest it is not important.


The focus was home births, and therefore low-risk, normal births.  It is not standard to count premature babies in mortality computations for low-risk women.  Premature births are considered high risk for the newborn and therefore generally ruled out for the option of home birth.  However, some premature births did occur at home and were included in the study, on the rare occasions when mothers slightly premature chose to remain at home, or there were doubts about the gestation.     


While stillbirths prior to labor are eliminated from comparisons of home birth vs. hospital outcomes because they are not part of intrapartum or neonatal mortality, we did report stillbirths after 20 weeks in Figure 1.  We also reported the four women who chose to deliver at home in spite of learning prior to the delivery that they had had a pre-labour stillbirth. However, they were not included in calculations of intrapartum and neonatal mortality as they do not fall into that category.


The occurrence of fatal birth defects is not associated with place of birth, and is not standardly reported when making comparisons among differing practices and so those deaths were reported in the article, but removed from final calculations just as they were from comparison studies.

3. What was the basis of your conclusion about the safety of home versus low risk hospital birth?


We came to the conclusion that "Planned home birth for low risk women in North America using certified professional midwives was associated with lower rates of medical intervention but similar intrapartum and neonatal mortality to that of low risk hospital births in the United States."  (BMJ abstract).  This was based primarily on our study results, as we had chosen the most refined methodology possible for a study across the U.S. –a large enough sample size, a prospective methodology with required accountability for recertification(with inclusion of Canadian midwives with the same certification),and incorporation of a validation study.  It became evident that our study expanded the weight of evidence that already existed and we were careful to include an examination of the pertinent literature, both American studies of low risk hospital birth and other North American studies of out of hospital birth.  Our findings are also consistent with the international literature on home birth.  Consistency in results across studies with different designs and in different settings yields more confidence in results. 


We used the term "similar" to describe how our findings measured up to the weight of evidence of other intrapartum and neonatal outcomes and "lower" to describe our study’s rates of medical intervention for the following reasons:


"SIMILAR" Intrapartum and Neonatal Mortality

The term "similar" risk was appropriate to use with regard to intrapartum and neonatal death in our study because our results were "alike but not identical" to outcomes in other studies of low risk hospital births and other out-of-hospital births.  It is not surprising that they were not identical, and only similar, because of the random variation that occurs when you are dealing with relatively small studies comparing rare outcomes, differences in study design, different populations, and hospital studies that are rarely able to collect neonatal mortality beyond a week.   As Judith Rooks says in "The Safety of Out of Hospital Birth In America" from the book,  Midwifery and Childbirth in America, one of the best treatises of home birth/hospital birth comparisons:


The studies of low-risk births attended by physicians in hospitals {see our BMJ article Table 4}…provide a general frame of reference for making judgments about the mortality rates reported in the studies of out-of-hospital births.  Each of the comparison studies created a low-risk pregnancy data set by excluding women with certain characteristics or outcomes from a larger database. (Rooks 1997:370)


Because we had seen how difficult it is to obtain a direct comparison, even within a provincial/state jurisdiction, and other studies had tried to create comparisons, but they inevitably ended up not having comparison groups that were collected with the same methodologies, we decided instead to present all the studies of greater than 500 births for a general frame of reference.


We carefully reviewed Rooks 40 page chapter which we had contributed to back in the 1990’s at the request of the author (Rooks 1997),investigating a variety of issues related to establishing the relative safety of home versus hospital birth.  She also reviews each of the studies listed in Table 4 of our BMJ article published up to 1997 when her book was published, focusing on the aspects of the design of each study and their implications for the evaluation and comparison of intrapartum and neonatal mortality.  For anyone with an interest in more deeply understanding the safety of home birth issue, this is highly recommended reading and will help one to understand more clearly the considerations that go into evaluating the perinatal risks observed and why we came to our conclusion in the BMJ report.


The need for the same careful consideration of each study that Judith Rooks provides in her Chapter on home birth is noted in our footnote to Table 4 of our BMJ article on home birth:


"Table is presented for general comparison only. Direct comparison of relative mortality between individual studies is ill advised, as many rates are unstable because of small numbers of deaths, study designs may differ (retrospective versus prospective, assessment and definition of low risk, etc.), the ability to capture and extract late neonatal mortality differs between studies, and significant differences may exist in populations studied with respect to factors such as socioeconomic status, distribution of parity, and risk screening criteria used. For example, see the study by Schlenzka. Although the crude mortality for low risk babies weighing over 2500 g intended at home was 2.4 per 1000 and intended in hospital was 1.9 per 1000, when standard methods were employed to adjust for differences in risk profiles of the two groups (indirect standardization and logistic regression), both methods showed slightly lower risk for intended home births."

Since the article was submitted for review to the BMJ, the NIH has published a report that provides neonatal death data categorized by ethnicity for which we adjusted our data in order to do a crude comparison. In doing the analysis, we came to the same conclusion as we did in the original article but added the NIH as a crude comparison for neonatal mortality.

"LOWER" Medical Interventions

We had the good fortune to conduct our study coincidentally during the overlapping time period that the first nation-wide Listening To Mothers (Declercq ER et al. 2002) survey was also carried out. This enabled us not just to obtain the standard data reported by health care professionals to the National Center for Health Statistics (NCHS)  for relatively low risk women (Martin JA 2002) , but the data reported by women across the U.S. in all risk categories with the exception of multiples.   For comparative purposes, obviously we found the low risk births as reported by the NCHS were better for comparing to our study largely of low risk women than the data from Listening to Mothers. The latter reported on only singleton births but included risk categories such as breech and premature births, but we laid them side by side on Table 3 to show the differences. 


In the case of the NCHS data, the interventions rate for low risk (single vertex babies at 37 weeks) were twice to ten times higher in planned hospital births compared to home births, even when isolating the single, vertex babies at term, which could have been spared the interventions in hospital.  The interventions rates are so overwhelmingly lower, it justified our use of the term "lower" without need for further discussion.  When looking at such large differences in such common events , the same precision required for intrapartum and neonatal mortality (both rare occurrences in either setting), is less necessary for adequate intervention comparisons.

4.  Why does the Washington home birth study have different conclusions than almost all other articles on home birth?
The highly publicized birth-certificate-based study in Washington State, (Pang JWY et al. 2002), reported increased risk for births in the home setting based on analysis of birth and death certificates. These findings are not only inconsistent with our study, but also unlike the weight of evidence of home birth articles in North America and worldwide.


Several studies conducted in the 1970s and 1980s established the wide variance in outcomes between planned vs. unplanned and attended vs. unattended home births:  in North Carolina researchers found a neonatal mortality of 3/1000 for planned home deliveries attended by a lay midwife,  30/1000 for unattended planned homebirths, and 120 neonatal deaths/1000 for unplanned homebirths (Burnett et al. 1980); in Kentucky they found a neonatal death rate of 3.5/1000 for planned vs. 72.7 per 1000 for unplanned homebirths (Hinds et al. 1985); and  Indiana found an estimated perinatal mortality of 18/1000 live births for the whole state compared to 45/1000 for women who received no prenatal care and gave birth at home without trained attendants (Kaunitz et al. 1984).  It has also been known since those early studies on home birth were done, that in almost all jurisdictions, birth certificates do not include any information about the intended site of birth, and their use creates the potential for inclusion of high-risk unplanned and unattended home births. The concern that this was the case with this study was highlighted when the study reported a transfer rate to hospital during labor from planned home births of 4.5% (279 transfers among 6,133 births). This is about half the transfer rate one would expect in a cohort of planned home births with trained midwives in North America.  The carefully conducted studies in the U.S.A. which include detailed data on each birth, consistently report hospital transfer rates of 8-10%. (Murphy and Fullerton 1998, Anderson 1995 and our BMJ article) during labour and low intrapartum and neonatal mortality. The low transfer rate of the Washington study hospital transfer rate suggests that many of the births identified by Pang et al. as "planned" may in fact not have been attended by health care professionals, as they generally have standards for transfer, and thus may not have been planned home births. (See Johnson K and Daviss BA, "Letter to the editor re Outcomes of Planned Home Births in Washington State:  1989-1996.  The Journal of the American College of Obstetricians and Gynecologists (101), 1. Jan. 2003.)


To create a doubling of mortality from unplanned unattended home births, it would require only 15% of such births with a death rate of 12/1000 in unplanned or unattended home births and a death rate of 1.7/1000 in planned home births.  Given the experience in North Carolina, Kentucky, Indiana, and considering the risks involved with those who do not seek any form of care (Rooks  1997:390-391), and the religious and self-sustenance culture of this particular state on the West Coast, this may well be an underestimate of risk associated with unplanned homebirth. With an intrapartum and neonatal death rate like that found in the CPM2000 study of 1.7/1000 in planned home births, here is the math on 1000 births:


850 planned home births with 1.7/1000 death rate =  1.6 deaths
150 unplanned  with 12/1000 death rate               =  1.8 death
Total deaths for 1000 births                                  =  3.4 deaths


In this example with 15% of the out of hospital births unplanned an observed home birth risk could be observed that was doubled from 1.7 to 3.4 deaths/1,000 births.

For more details critiquing the Washington home birth study, see a thorough commentary on the Pang et al. study by Saraswathi Vedam (2003).   Her comments include the following:


"In fact there is no way to determine from the birth certificate if a birth recorded as having occurred at home or attempted at home was a planned home birth (of a preselected woman who had qualified attendants and access to consultation and referral), a precipitous or unattended home birth, or an intended home birth outside the health care system." (page 58)


"To add to the confusion, the investigators included all births occurring between 34 and 37 weeks gestation in their initial analysis, although prematurity is a universal exclusion from planned home birth.  They later restricted the analysis to babies of at least 2500 g and 37 weeks, but their abstract, tables, results, and conclusions are primarily based on the larger group." (page 58)

Under a "Truth in Reporting" subsection Vedam notes:


"The authors noted controversy in literature concerning the safety of home deliveries, but cited only 4 studies that they claim support their findings of increased risk of neonatal mortality and adverse outcomes. They failed to mention or discuss findings from the bulk of the recent national and international literature, which concluded that planned home birth is a safe and reasonable option (8,16,20,26–31). More importantly, they misrepresented the literature that they did cite." (page 60)


"Furthermore, Pang et al noted the study limitations of reliance on birth certificate data, inability to ensure assignment of study subjects to correct study groups (planned home versus planned hospital), and possible differences related to attendant type. They concluded that the results suggest increased risk of adverse neonatal and maternal outcomes, particularly among nulliparous women. Nonetheless, more light needs to be shed on this controversial topic before practitioners and expectant parents can be fairly counselled about the safety of planned home births__ (p 259). Yet, they still represented their findings in an alarming manner. Throughout the article they warned that planned home births presented twice the risk of neonatal death—a risk amounting to a difference of 1 in 1000 for the study, which is a worst-case scenario given the weakness of the study design" (page 61)


5. Are you comparing apples to oranges?


The science of "comparing apples to oranges" had a breakthrough in the same year that we began to collect our data for the study on home birth by Certified Professional Midwives. Refuting the outdated belief that a fruit comparison was impossible, "Comparing apples and oranges: a randomized prospective study," was published in the BMJ issue of 23-30 December 2000.  With its important conclusion, that apples and oranges are not only comparable but similar, the article begins by stating:

Many authors use the analogy of the putative inability to compare apples and oranges as a means of scornfully reviewing the work of others.


a)  Why the Randomized Controlled Trial Is NOT the Gold Standard for Home Birth Research

The phrase "comparing apples to oranges" has become a cliché increasingly used in the health sciences when attempting to categorize research as "invalid" if the methodology has not been a Randomized Controlled Trial.  While convenient because it avoids the need to understand comparison in epidemiology on a more sophisticated level, the attempt to put all studies in one fruit basket of methodology in this kind of research reductionism ignores some important research realities.


When searching for excellence by using RCTs, it is quickly forgotten that the true gold standard is a randomized controlled trial which is double-blinded, where neither the clinician nor the subject know which treatment is being received. While useful for randomly and blindly allocating patients in studies of drugs, to receiving either the medication or a placebo envelope, such double-blindedness can rarely be applied in obstetric or midwifery research.  This is because the performance of episiotomies, rupturing of the membranes, induction, or caesarean section, are impossible to do without both the recipient and the practitioner painfully aware when the patient has been randomly assigned to treatment.  Furthermore, such important research as whether caesareans cause more maternal mortality than vaginal birth is simply better carried out by methodology other than the RCT.


Enter the complicated politics of doing home birth research.  As we have known for many years, and stated in the article in the BMJ, to date it has been impossible to randomly assign women to either home or hospital birth. If you could actually find women who didn’t care enough about planning where they would have one of the most important events of their life to participate in such a trial, there would be high suspicion that they were psychological outliers. As we said in the article, this had already been attempted in Britain without success (Dowswell T 1996), a country where home birth is not a foreign concept and the health care system embraces the skills of midwives without question.  It would also be extremely costly to create a trial large enough to have enough perinatal mortality to do comparison and finding funding even for retrospective studies is not easy to come by.  A practice such as home birth that cuts down on intervention is not popular among companies selling pharmaceuticals and medical instruments.  For further discussion about the strengths and limitations of RCTs for obstetrics and midwifery see Johnson 1997. 


As well, even in the future if it were possible to find a society in which it was possible to carry out an RCT for out-of-hospital research compared to in-hospital, the complexities of clinical judgement and decision-making around home birth would make it inappropriate for similar reasons to those that have presented themselves with the use of the RCT to compare vaginal to caesarean section for breech delivery has fallen into question. (See Kotaska 2004.)


b)  Finding National Comparisons

The principles of comparative analysis have to be well understood before claims are made about the legitimacy of any study.  As mentioned above in question #3, Judith Rooks describes the difficulties of comparing home and hospital birth in an eloquent treatise.  Only one state to date and one province have been able to do cohort studies of home birth with a defined, direct comparison group -- Schlenzka in California (see details below) and Janssen in B.C. However, as we stated in the article, the other national home birth studies --and the low risk hospital studies for that matter (see our Table 4 and read Rooks 1997: 370-371) -- did not have direct comparison groups that were collected with the same methodologies.  There are practical reasons for this. 

With the introduction of the CPM credential, we knew that we would be able to accomplish the task of a cross-the-country study on a defined group over a defined time period. We sought out comparisons from other studies for the most important comparable group between home and hospital -- the group of low risk mothers in each setting.  Data were secured through the NIH statistics for intervention and all other studies on out-of-hospital birth or low risk hospital birth that had at least 500 births (see Table 4 in the BMJ article). However, to directly compare our international study of all CPMs to a directly comparable group of hospital births in North America using the same prospective methodology as for the CPMs would have been extraordinarily difficult logistically, if it would even have been possible to get buy in from hospitals across the USA.  Even if it would ever be possible, it would be extremely costly and large scale funding for studying home birth is not readily available in the United States.  We carefully detailed the difficulties and shortcomings of using existing administrative records, particularly birth certificates, to form a directly-comparable national group, and decided against such comparison, with the comfort that Schlenzka in 1999 had managed it with one of the largest states in the union and would provide an important adjunct to what we had found nationally (see "c" below).


Just as apples and oranges are both grown in orchards or groves, have flowering trees, are considered a fruit, may be eaten and made into juice, are subject to damage by disease and insects, so too, the births in the cohorts to which we compared our cohort of home births were all births that occurred in the year 2000,   all involved women living in North America (98% in the U.S.), and all fit into the category of "low risk" for intrapartum and neonatal considerations as defined by the accepted definition of singleton, head down, and > or = 37 weeks.


It should not really be considered remarkable that each study that we found that was carried out on low risk births fell into a similar pattern of 1 to 3 deaths per thousand births despite fluctuations in the definitions of intrapartum and neonatal death.  Refer to question #3 to understand why that it is a respectable way to compare home and hospital birth and how we came to our conclusions. 

c)  The Important Adjunct to Our Study, as described in our BMJ article – A Direct Comparison Group in One of the Largest States in the Union With Ample Home Birth


The Stanford PHD thesis (Schlenzka P, Safety Of Alternative Approaches To Childbirth, [Unpublished Dissertation]. Palo Alto, CA: Department of Sociology, Stanford University, 1999), provided a rare opportunity to have a homebirth study with a  directly comparable group of low risk hospital births as we noted in the Discussion in the BMJ.  The Stanford PHD is an important compliment to our study, finding similar intrapartum and neonatal death rates as the CPM2000 study did for more than 3,000 planned home births and more than 800,000 low risk hospital births.  Schlenzka was able to establish a large defined retrospective cohort of planned home and hospital births with similar low risk profiles, because birth and death certificates in California include intended place of birth and these had been linked to hospital discharge abstracts for 1989-90 for a caesarean section study through a contract with the Rand Corporation.  When the author compared 3,385 planned home births with 806,402 low risk hospital births, he consistently found a non-significantly lower perinatal mortality in the home birth group. The results were consistent regardless of liberal or more restrictive criteria to define low risk, and whether or not the analysis involved simple standardization of rates or extensive adjustment for all potential risk variables collected.


The Schlenzka study is available for use at Stanford, through Interlibrary Loan to Shares partners ( and for purchase
from UMI/Proquest. UMI has the thesis available for download; the copy at the Interlibrary Loan is paper.

6. Why did you choose not to provide confidence intervals for the intrapartum and neonatal death rates?


We did not provide confidence intervals for the rates, taking guidance from other published studies, including the National Birth Center study (Rooks JP et al. 1989).  As in the birth centre study, our study represents the risk for CPMs for the defined period – not an estimate of it, but the actual rate for CPMS for the year 2000, so a confidence interval is not required, as the rate is not an estimate, based on a sample, but the actual rate for the year 2000.
7.   Why did you choose not to provide details about other outcomes?


We were limited to 2,000 words when we submitted our article for publication in the BMJ.  We chose to focus on interventions and mortality figures because those appeared to be most critical to legislators and public health interests.  We did not have the space to discuss perinatal mortality in sufficient depth. We will discuss other outcomes in future articles. 


8.  Could the funders have biased the study in any way?


Initial and primary funding was obtained from the Benjamin Spencer Fund, a small, private foundation with program interests in the environment, women and families and reproductive rights.  The Fund had no advance review of the BMJ article, receiving a copy of the published article as part of the final grant report.    


The study was not commissioned by any national or state midwifery group. However, we are greatly indebted to the North American Registry of Midwives for their role in requiring all CPMs to participate in the data collection as a requirement for renewal of their CPM credential.  Without NARM’s facilitation of CPM participation, we could not have met a key research requirement – participation by midwives with the same credential. This is important because the conclusions of the study would have been less compelling if the midwives were a diverse group with no commonality of credential, education, skills and site of practice.


NARM was not involved in the design or analysis of any of data collected and did not participate in editing the final article but to clarify the description of the CPM credentialing process.  In fact, the risk to NARM, should the outcomes reflect negatively on the care by CPMs, was substantial and significant. The NARM Board realized that their organization was taking a risk, especially since as researchers we were very transparent in our process and made it very apparent to the epidemiologists at the APHA what we intended to do and how we intended to carry it out.  However, NARM was at a point in its development where a study of the outcomes of CPMs was essential to demonstrate the safety of out-of-hospital birth with credentialed direct-entry midwives.


To further enhance the independent verification of the data, the mothers involved in the study were contacted independently and directly about outcomes and transfers to compare them to what the midwives had reported.  Finally, the obstetrician on the Advisory Council for the study wanted to be assured that autopsy or hospital reports would be obtained where possible so that causes of intrapartum and neonatal deaths would not be assessed only through the eyes of the midwives in attendance, but a separate professional party. We were obliged to strictly follow all these processes for the credibility of the study as professional researchers.


The Foundation of the Advancement of Midwifery (FAM), a 501(c) (3) public charity was formed after the data for our study had already been collected, the preliminary analysis had been done and we had presented the draft findings to the American Public Health Association. FAM provided a small grant to fund the gathering of the remaining client satisfaction survey data, phone for autopsy reports and make sure any stragglers were accounted for. FAM had no role in the analysis of the data.


American Public Health Association. 2001-3: increasing access to out-of-hospital maternity care services through state-regulated and nationally-certified direct-entry midwives. Am J Public Health 2002;92:453-5

Anderson RE, Murphy PA. Outcomes of 11,788 planned home births attended by certified nurse-midwives. A retrospective descriptive study. J Nurse Midwifery 1995;40:483-92.

Burnett CA III, Jones JA, Rooks J, Chen CH, Tyler CW Jr, Miller CA. Home delivery and neonatal mortality in North Carolina. JAMA 1980;244:2741-5.

Daviss BA. Study’s focus on induction v spontaneous labour neglects spontaneous delivery. BMJ 2001:323:1307 (1 december).

Declercq ER, Sakala C, Corry MP, Applebaum S, Risher P. Listening to mothers: report of the first national US survey of women’s childbearing experiences [monograph]. New York: Maternity Center Association, 2002. (accessed 3 Apr 2005).

Dowswell T, Thornton JG, Hewison J, Lilford RJ, Raisler J, Macfarlane A, et al. Should there be a trial of home versus hospital delivery in the United Kingdom? BMJ 1996;312:753-7.

Hinds MW, Bergeisen GH, Allen DT. Neonatal outcome in planned v unplanned outof- hospital births in Kentucky. JAMA 1985;253:1578-82.

Janssen PA, Lee SK, Ryan EM, Etches DJ, Farquharson DF, Peacock D, et al. Outcomes of planned home births versus planned hospital births after regulation of midwifery in British Columbia. CMAJ 2002;166:315-23.

Johnson KC, Daviss BA. Outcomes of planned home births in Washington State: 1989-1996. Obstet Gynecol 2003;101:198-200.

Johnson K. Randomized Controlled Trials as Authoritative Knowledge: Keeping An Ally from Becoming a Threat to North American Midwifery Practice. In ed: Robbie Davis-Floyd and Carolyn Sargent. Childbirth and Authoritative Knowledge. California: University of California Press, 1997:350-361.

Johnson K and Gaskin IM. Safety of single-layer suturing in caesarean setions must be proved. BMJ 2001:323:1307 (1 december).

Kauntiz AM, Spence C, Danielson TS, Rochat RW, Grimes DA. Perinatal and maternal mortality in a religious group avoiding obstetric care. Am J Obstet Gynecol. 1984 Dec 1;150(7):826-31.

Lydon-Rochelle M, Holt VL, Easterling TR, Martin DP. Risk of uterine rupture during labor among women with a prior cesarean delivery. N Engl J Med 2001; 345:3-8.

Martin JA, Hamilton BE, Ventura SJ, Mencaker F, Park MM. Births: final data for 2000. National vital statistics reports. Hyattsville, MD: National Center for Health Statistics, 2002;50(5).

Murphy PA, Fullerton J. Outcomes of intended home births in nurse-midwifery practice: a prospective descriptive study. Obstet Gynecol 1998;92:461-70.

Pang JWY,Heffelfinger JD, Huang GJ, Benedetti TJ, Weiss NS. Outcomes of Planned Home Birth In Washington State, Obstetrics & Gynecology, 2002 Aug;100(2):253-259,

Rooks JP, Weatherby NL, Ernst EK, Stapleton S, Rosen D, Rosenfield A. Outcomes of care in birth centers. The National Birth Center study. N Engl J Med 1989;321:1804-11.

Rooks JP. Safety of out-of-hospital births in the United States. In: Midwifery and childbirth in America. Philadelphia: Temple University Press, 1997:345-84.

Schlenzka P. Safety of alternative approaches to childbirth. PhD thesis, California: Stanford University, 1999.

Vedam, S. Home versus hospital birth: questioning the quality of the evidence on safety. Birth 2003,30(1), 57-63.

Version: 3.0.1
Top of page   |   Printer-friendly