The Tragic Death Toll of Homebirth
More than 10,000 American women each year choose planned homebirth with a homebirth midwife in the mistaken belief that it is a safe choice. In fact, homebirth with a homebirth midwife is the most dangerous form of planned birth in the US.
In 2003 the US standard birth certificate form was revised to include place of birth and attendant at birth. In both the 2003 and 2004 Linked Birth Infant Death Statistics, mention was made of this data, but it was not included in the reports. Now the CDC has made the entire dataset available for review and the statistics for homebirth are quite remarkable. Homebirth increases the risk of neonatal death to double or triple the neonatal death rate at hospital birth.
As this chart shows, the neonatal mortality rate for DEM (direct entry midwife, another name for homebirth midwife) assisted homebirth is almost double the neonatal mortality rate for hospital birth with an MD. This is all the more remarkable when you consider that the hospital group contains women of all risk levels, with all possible pregnancy complications, and all pre-existing medical conditions. An even better comparison would be with the neonatal mortality rates for CNM assisted hospital birth. The risk profile of CNM hospital patients is slightly higher than that of DEM patients, but CNMs do not care for high risk patients. Compared to CNM assisted hospital birth, DEM assisted homebirth has TRIPLE the neonatal mortality rate.
The chart shows the data for 2003-2004, but the data for 2005 has recently become available. Homebirth death continues to be far higher than death in the hospital for comparable risk women. In 2005 the neonatal death rates were CNM in hospital 0.51/1000, MD in hospital 0.63/1000 and DEM attended homebirth 1.4/1000.
No wonder the Midwives Alliance of North American (MANA), the trade union for homebirth midwives, is suppressing their safety statistics. From 2001-2008, they have collected the single largest repository of data on homebirth. The data is publicly available, but only to those who can prove they will use them for the “advancement” of midwifery, and even then, a legal non-disclosure agreement must be signed as part of the process. MANA’s data may very well confirm that homebirth with a DEM has triple the neonatal mortality rate of hospital birth for comparable risk women in the same year.
What is also notable is that the results are consistent with all existing scientific studies, including the Johnson and Daviss study (Outcomes of planned home births with certified professional midwives: large prospective study in North America). Johnson and Daviss actually showed that homebirth with a CPM has a neonatal mortality rate almost triple that of hospital birth for low risk women. The latest statistics are the most recent and most reliable confirmation of that fact.
There really is no question about it. Homebirth with a homebirth midwife dramatically increases the risk of neonatal death.
Posted in: Obstetrics & gynecology
Leave a Comment (257) ↓
I said “Intervention rates are lower for planned home births than for planned hospital births in low-risk women (Janssen et al 2002 – the Canadian study – and others).”
Amy replied “Among homebirth advocates, an “intervention” is anything that homebirth advocates don’t approve of. For example, medication for pain relief is an “intervention” but acupuncture for pain relief is not. Electronic fetal monitoring is an “intervention,” but auscultation with a fetoscope or doppler is not.”
This is yet another red herring. Amy has read the paper Janssen et al paper and should know better. For those who haven’t, let me share a relevant summary of Table 2. For brevity, I’ll just give the percent of women who had a given intervention. I’m also excluding rows where no statistical comparison is reported in the table. HB=homebirth group; PA=hospital, physician-assisted (no midwife); MA=hospital, midwife-assisted. The hospital midwives are the same people (with the same training) as the homebirth midwives.
Epidural analgesia or anesthesia: HB=7.7%, PA=27.6%, MA=26.3%
Narcotic analgesia: HB=2.7%, PA=34.7%, MA=13.1%
Spinal anesthesia: HB=3.0%, PA=9.6%, MA=7.5%
General anesthesia: HB=0.7%, PA=2.7%, MA=1.1%
Electronic fetal monitoring
Any: HB=14.7%, PA=82.6%, MA=58.0%
External: HB=14.7%, PA=82.2%, MA=57.3%
Internal: HB=1.6%, PA=10.1%, MA=7.7%
Induction (with oxytocin or prostaglandins): HB=4.3%, PA=22.3%, MA=14.0%
Augmentation of labour
Artificial membrane rupture: HB=15.8%, PA=37.0%, MA=27.1%
With oxytocin or prostaglandins: HB=6.4%, PA=16.8%, MA=19.1%
Episiotomy
Any: HB=3.8%, PA=15.3%, MA=10.9%
Mediolateral: HB=1.7%, PA=10.8%, MA=8.1%
Caesarean section: HB=6.4%, PA=18.2%, MA=11.9%
Nulliparous CS rate: HB=11.2%, PA=21.5%, MA=15.4%
Multiparous CS rate: HB=2.2%, PA=15.1%, MA=7.1%
Multiparous CS rate (no previous CS): HB=0.9%, PA=4.2%, MA=3.9%
Primary indication for CS
Breech: HB=0.8%, PA=0%, MA=0%
Dystocia or cephalopelvic disproportion: HB=2.0%, PA=5.4%, MA=7.0%
Fetal distress: HB=1.3%, PA=3.6%, MA=2.1%
Repeat CS: HB=0%, PA=4.2%, MA=0.2%
Abruptio placentae: HB=0%, PA=0%, MA=0.4%
Placenta previa: HB=0%, PA=0.5%, MA=0.2%
Other: HB=1.4%, PA=1.7%, MA=0.9%
Malposition/malpresentation: HB=0.8%, PA=2.7%, MA=1.2%
Genital herpes: HB=0.1%, PA=0%, MA=0%
Remember that HB means planned homebirth – those weren’t caesarean sections performed on the kitchen table!
The authors compared the homebirth group separately to the physician-assisted and the midwife-assisted hospital groups. Out of all that, the only differences that were not statistically significant (after Bonferoni correction for multiple comparisons) were:
General anesthesia: HB against MA
Nulliparous CS rate: HB against MA
Multiparous CS rate (no previous CS): HB against PA or MA
Primary indication for CS:
Most of these differences were non-significant. However, most also had too few cases (often less than ten) to allow for robust statistical comparison.
So, whether you look at pain-relief, fetal monitoring, induction rates, induction and augmentation, episiotomy, or CS, they all tend to occur more frequently in a hospital setting than a home setting, even when subjects are matched for all possible confounding risk factors. Contrary to what Amy suggested, I assert that this study supports my claim that intervention rates, measured as objectively as possible, are higher among hospital births than among home births.
Some women may not care; others might prefer to reduce their risk of chemical or surgical interventions. Given that mortality rates are not different between the two groups, there is no need to label such women as irrational. You may choose differently yourself, but they are not being anti-science in making such a choice, given this evidence. (Provided they are at low risk, have well-trained midwives, etc.)
TimMills:
“Given that mortality rates are not different between the two groups”
Actually, Janssen was forced to publicly retract the claim that the mortality rates were the same, which was the least she could do when you consider that the homebirth group had two perinatal deaths and the hospital birth groups had none.
The Canadian Medical Association Journal published 7 letters to the editor critical of the study’s statistical methods and the conclusions of the study, specifically the fact that the homebirth and hospital groups differed in risk level and that the homebirth group had 2 perinatal deaths and the hospital group had none. The CMAJ offered Janssen an opportunity to reply to her critics. Her response includes the following (CMAJ, June 11, 2002; 166 (12)):
“Although we tried to ensure that comparison groups met eligibility criteria for home birth, women who choose home birth differ from those who select hospital birth in both measurable and unmeasurable ways…”
Furthermore:
“The purpose of our study was not to determine which method of care was better, home vs. hospital, but rather to assess whether, at the 2-year interval, home birth was safe enough to continue to be offered as a choice for women in the context of ongoing evaluation.”
So the author herself was forced to acknowledge that the study cannot be used to support the safety of homebirth.
Amy Tuteur writes:
“But cEFM does not require a woman to stay motionless on her back. That’s just another canard made up by homebirth advocates.”
“Canard” means, in many people’s vocabulary, “Fact I don’t like.” I personally know a woman who had delivered two children in her native country without medication; when she had her third in an American hospital, she was indeed required to stay in bed, and because she could not walk to relieve the pain of contractions, needed analgesia. Fortunately, that did not further lead to the need for a C-section.
Dr. Amy’s argumentative style looks to me like the health version of global warming denialism: conflate values conflicts with facts, reject facts and research results you don’t like, demonize researchers who published them, and heap vitriol on nonscientists who disagree with you. It’s a popular style around here. Ironically, for many of your beliefs you’ve got enough facts on your side, or at least not completely or provably against you, that you could make a credible case for donig things your way with civil, logical, relatively unbiased arguments. Why not try it?
In discussing the Chamberlain study above, Dr Tuteur says
“Which means that the groups were not comparable and no conclusion can be drawn.”
The study was a case-control. The pairs were matched on a wide variety of attibutes; in the aggregate, the cohorts differed only marginally in a few factors.
But if any minor defect in cohort matching is sufficient to invalidate a result, then I don’t understand why Dr Tuteur (in this thread and now the newer one as well) seems to assign any import to the use of “state and national statistics”. In particular, her cohort analysis from the CDC Wonder database fails on a much larger scale than any of the published studies which she criticizes.
The largest flaw is the assumption that restricting the selection of hospital births to 37+ weeks of gestation, 2500+gram BW, white women, maternal age 20- 44, and further restricting to CNM-attended births, somehow yields a “low-risk” group that can be statistically compared to the self-selected cohort of women who choose homebirth with a non-nurse midwife. Dr Tuteur claims “The risk profile of CNM hospital patients is slightly higher than that of DEM patients” but this is an assertion without evidence. It is true that in an ideal world, all home birth clients ought to be at low risk. But the retrospective statistics capture results from the real world, and in the real world in the United States, some women choose home birth even though their risk profiles might be less than pristine.
Let me anticipate your response that this is evidence that home-birth midwives are incompetent because they cannot or do not screen their clients appropriately. Leave aside the ethical dilemma presented by abandoning a client who resists a recommended referral or refuses transfer; those are issues to pursue in another thread. They have no relevance at the moment to the issue of proper statistical matching. In order to draw a valid conclusion, you need to select a comparative cohort that matches the home-birth cohort as it actually exists in real life.
I assert that your choice of cohort matching is flawed in serious ways.
First, take note of the fact that in the CNM hospital cohort, approximately 5% (25,000 of 560,000) of your group consists of grand-multiparas. (women who have already had at least 4 children). In the out-of-hospital, other midwife group, the proportion is 21% (7500 of 35,000). This is on its face an independent confounding factor, which you have neither acknowledged nor adjusted for. But it also strongly suggests other demographic differences which were mentioned upthread by heyunyi, the fact that specific religious subgroups (Amish, Mennonite, certain offshoots of LDS and Seventh Day Adventists, and others) are far overrepresented in the homebirth group. While religion on its own is not a medical risk factor, isolated cultural subgroups can and do have many lifestyle factors that impact overall health.
Any serious analysis of raw birth statistics must at a minimum acknowledge the existence of such factors, in order to correctly match them.
Second, let’s take a closer look at the hospital cohort as well, to see what hidden confounders may be present. For rough analysis purposes, I have selected the state data for Indiana and Pennsylvania to compare against each other. For the record, this is not cherry picking. I selected these states because on the surface, you might expect them to have similar results. They are both middle-size states, with populations that are roughly balanced between large urban centers and many rural communities; there are few truly remote areas where distance to medical care is a large factor; among the white population already selected for, the ethnic profiles seem relatively comparable; income and socioeconomic factors are also relatively comparable. Now, I am not trying to make the case that these populations are totally identical, I am just pointing out that they are more similar than, for instance Massachusetts vs North Dakota or Mississippi vs Wisconsin.
That said, let’s look at neonatal mortality among hospital births, all providers (already limited by the same exclusion criteria in the original post) – the last column is neonatal mortality per 1000:
Indiana (18) In Hospital Total 132 168,019 0.79
Pennsylvania (42) In Hospital Total 140 268,838 0.52
(For the record, this difference is statistically significant at the 95% confidence level – CIs are 0.66 – 0.92 for IN and 0.44 – 0.60 for PA, in other words, the difference in neonatal death rate cannot be attributed to chance).
There are three plausible explanations for this difference. (1) Differences in quality of care (2) differences in underlying risk level (3) some combination of the above.
Now, if some hypothetical person (not I) were to apply the same standard of evidence as Dr Tuteur uses in her analysis, that person might say that the excess deaths in Indiana were the result of incompetent hospital practitioners, or incompetent hospital practices, since our prior restrictions (BW, gestation etc) have allegedly leveled out risk. While we can’t rule that out entirely, it would seem highly unlikely that the overall quality of hospital care in Indiana can explain this result, since hospital practitioners in Indiana are trained to the same standards as elsewhere.
Perhaps the Indiana statistics look worse due to contamination by home birth transfers? That explanation fails, because Pennsylvania has a higher proportion of out-of-hospital births than Indiana (3.2% vs 2.4%), so if hospital mortality rates were affected by home-birth transfers, it would affect Pennsylvania results more than Indiana’s.
So we are left with the conclusion that even after restricting birth statistics by race, age, gestational age, and birth weight, we still do not have a homogeneous level of underlying risk factors. At this point, we don’t have the tools to identify exactly what those factors are; as a resident of Indiana, I think it would be important to do further research. If it were possible for Indiana hospitals to achieve the same results as Pennsylvania hospitals, then approximately 50 deaths could have been prevented in Indiana over 3 years, which is incidentally a larger number than the number of deaths attributed nationally to homebirth over that same time.
But I digress; what is clear is that any serious cohort analysis for underlying risk factors is lacking in the orginal presentation at the top of the post.
And there is a third major flaw in this analysis introduced by restricting the cohort to CNM births. CNMs are not evenly distributed about the country. CNM-attended births as a proportion of all vaginal births range from 0.7% in Arkansas to 35.9% in New Mexico. http://www.midwife.org/siteFiles/news/TrendsinCNMBirthsfromJF07JMWH.pdf
Generally speaking, CNM births are more common in the Northeast and west coast, and least common in the midwest and mountain states. The result is that when you restrict your hospital cohort to CNM births only, you have introduced a serious geographical skewing, oversampling from areas where the underlying risk factors are smallest, and drastically undersampling from the regions where they are greater. Therefore any national cohort of CNM births is inherently biased toward good outcomes.
And finally, there is the unexamined issue of data quality. In a closer look at the IN/PA results referenced above, I noticed a specific peculiarity in the out-of-hospital data.
Indiana (18) Not in Hospital Other 8 2,023
Indiana (18) Not in Hospital Other Midwife 1 205
Pennsylvania (42) Not in Hospital Other 2 634
Pennsylvania (42) Not in Hospital Other Midwife 8 3,820
I’m not focusing on death rates now, the numerators are too small to be useful. I am focusing on the number of reported births. Now the “other” category of attendant is clearly a catch-all. It includes births which were unattended, attended only by untrained bystanders, husbands, partners, relatives, whatever. At least some of these were unplanned. But what catches my eye is the huge difference in the way the “other midwife vs other attendant” numbers are hugely unbalanced when comparing IN to PA. Indiana has ten times as many births recorded by “other” than by “other midwife” (CPM or other DEM). In Pennsylvania, the numbers are skewed 5 to 1 in the OTHER direction. I don’t claim to have a complete picture of home birth in Indiana, but I am close enough to it to know that Indiana is not a hotbed of “trendy” unattended childbirth (UC), and I highly doubt that a huge number of Indiana women who intend hospital birth are having trouble getting to a hospital on time. What I do know is that the legal environment for non-nurse midwives in Indiana is very hostile. What I strongly suspect, from the above numbers, is that there is a considerable number of midwife-attended homebirths in Indiana, where the birth certificate data, later registered by the family, does not reflect the midwife’s attendance. We can only speculate whether the “true” neonatal death rate would be better or worse with more accurate reporting. And we can only speculate to what extent this may be true in other areas. Any discussions of the legal and ethical import of this can be put aside for now, because again I am focusing on the validity of the statistical analysis. It is clear (at least to me) that the accuracy of the original dataset is questionable, and given that the homebirth numbers are relatively small, we should be somewhat skeptical of accepting the calculated mortality rates as accurate.
And I haven’t even touched on the issue of homebirth-tranfer-to-hospital vs CNM-transfer-to-MD. A serious statistical analysis would at least discuss some plausible estimate of the number of cases which are missing due to these factors.
So in summary, Dr Tuteur’s analysis of CDC statistics is flawed in these ways:
It is based on data where the underlying data quality is questionable for one cohort.
It fails to identify or discuss demographic factors which make the homebirth cohort unique.
It fails to evaluate or adjust for known confounding factors.
It uses a geographically skewed hospital cohort for comparison.
It fails to account for missing data due to transfer of care.
It is clear that the original statistical presentation in this post falls far short of anything resembling scientific evidence. And, in my opinion, it is unprofessional to use throw-away lines to dismiss the results of other serious researchers, and at the same time feature this sort of crude statistical analysis as being relevant. What’s also missing is a serious discussion of the legal and ethical obstacles involved in doing any valid scientific research on homebirth in the US.
If we are truly discussing science based medicine, then we should expect all evidence provided to meet the standards of science. (Thank you TimMills for your presentation of the Cochrane summary). I would hope that when we see future posts regarding specific childbirth practices, they will spend more time on an unbiased review of published valid research as a starting point, rather than a much later unannotated offhand reference to “all existing scientific studies”.
And I hope that this doesn’t come across as nitpicking or carping, but I think that in this forum it would be advisable to avoid the informal conflation of correlation and causation – i.e.
“homebirth with a CPM has a neonatal mortality rate almost triple that of hospital birth” as opposed to “Homebirth with a homebirth midwife dramatically increases the risk of neonatal death.” To a scientifically-minded reader, these two statements are not equivalent, and you could use more careful language to express your conclusions.
A comprehensive review of homebirth studies by NICE (National Institute of Health and Clinical Excellence) concluded that the hospital group was a higher risk group.
As regards the national statistics, you are claiming that the homebirth group was higher risk than the hospital group but you have not demonstrated it. Moreover the results are consistent with the Johnson & Daviss studies of CPMs.
Finally, MANA has the resultsof 20,000 CPM homebirth. A recent announcement declared that the results will not be released to anyone who can’t prove an ideological commitment to using the results in a way that is acceptable to MANA.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
A comprehensive review of homebirth studies by NICE (National Institute of Health and Clinical Excellence) concluded that the hospital group was a higher risk group.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
And your evidence that this is applicable to the US population in 2003 – 2005 is?
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>As regards the national statistics, you are claiming that the homebirth group was higher risk than the hospital group but you have not demonstrated it. Moreover the results are consistent with the Johnson & Daviss studies of CPMs.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
I have not made that claim. I have made the observation that you have not properly accounted for confounding factors, and I have noted that you are applying a double standard by criticizing the cohort matching in other studies, and following a much much much more lax approach to cohort matching in your own analysis.
The Johnson & Daviss study suffers from the same shortcoming, neither their original paper nor your commentary has properly identified a correct hospital cohort group for direct comparison. At least their paper discussed the obstacles involved in selecting such a cohort.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Finally, MANA has the results of 20,000 CPM homebirth. A recent announcement declared that the results will not be released to anyone who can’t prove an ideological commitment to using the results in a way that is acceptable to MANA.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
If you had access to those statistics, please describe how you would choose a comparable hospital group of births for comparison. If you will be starting with vital health statistics from the CDC, you need to explain how you would account for and adjust for all of the factors that I noted above, and hopefully as a professional you would go far deeper into statistical analysis.
IndianaFran:
“And your evidence that this is applicable to the US population in 2003 – 2005 is?”
It’s not applicable to the US population. I’m sorry I didn’t make it clear. It was NICE’s criticism of the Chamberlain study.
“I have made the observation that you have not properly accounted for confounding factors”
You haven’t shown that they are confounding factors.
“The Johnson & Daviss study suffers from the same shortcoming”
No, the J&D study suffers from a different, and more serious shortcoming. The authors didn’t compare the neonatal mortality rate in the homebirth group with the neonatal mortality rate in the hospital group that THEY USED for comparison of interventions.
That comparison shows that homebirth has nearly triple the neonatal death rate of hospital birth. They tried to hide that by comparing homebirth in 2000 with a bunch of out of date hospital studies extending back to 1969.
“If you had access to those statistics, please describe how you would choose a comparable hospital group of births for comparison.”
That’s easy. I’d use the same parameters they used in choosing homebirth patients, low risk patients with no pre-existing medical conditions and no pregnancy complications.
I can’t do any comparison, because MANA is insisting on an ideological litmus test for showing the data. They are explicit that they will not show the data to anyone who cannot be counted on in advance to use the data to support homebirth midwifery. Why would they need to do that unless they know that the data shows homebirth with a CPM is not as safe as hospital birth?
Frankly, I think withholding the data and applying an ideological litmus test to researchers is unethical. It deprives women of the information they need to make an informed choice about homebirth.
Do you really need evidence that grandmultiparity is an independent confounding factor?
here it is (I apologize for formatting ugliness)
Live Birth Order Deaths Births Death Rate Per 1,000
One child born alive to mother 1,640 2,657,850 0.62
Two children born alive to mother 1,386 2,576,366 0.54
Three children born alive to mother 776 1,357,767 0.57
Four children born alive to mother 370 505,147 0.73
Five children born alive to mother 146 170,360 0.86
Six or more children born alive to mother 142 120,835 1.18
Unknown or not stated 28 23,032 1.22
Total 4,488 7,411,357 0.61
After the second birth, there is a linearly increasing rate of neonatal death. The sixth and subsequent children have twice the incidence of neonatal death as the second child. A cohort that contains 20% grand multiparas cannot be directly compared to a cohort that contains 5%.
If you want to argue that this is not a known confounding factor, then I would suggest that you reread this post that you yourself made on Dec 26, 2006
http://homebirthdebate.blogspot.com/search/label/Johnson%20and%20Daviss
Here is an excerpt:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
For a characteristic to be a confounder in a particular study, it must meet two criteria. The first is that it must be related to the outcome in terms of prognosis or susceptibility…
The second criterion that defines a confounder is that the distribution of the characteristic is different in the groups being compared.
The article highlights three questions that must be asked to identify confounders in a cohort study:
Has there been a systematic effort to identify and measure potential confounders?
Is there information on how the potential confounders are distributed between the comparison groups?
What methods are used to assess differences in the distribution of potential confounders?
>>>>>>>>>>>>>>>>>>>>>>>>>>>>end of excerpt
You can go ahead and adjust your calculations for this factor, but you can’t claim that there are no others, because it is clear that you have not even looked for them. I see no evidence here that you have performed even the most basic statistical analysis of this CDC data.
And you haven’t even tried to address the issue of geographic skewing I brought up. I demonstrated that there is significant state-by-state variation in neonatal death rates (not related to place of birth or birth attendant), and that using CNM data distorts the state-by-state balance in a significant way.
So if you were to propose using the same seriously deficient statistical analysis methods to evaluate the home birth data currently owned by MANA, well, I wouldn’t trust you to derive a meaningful result either.
IndianaFran:
“Live Birth Order Deaths Births Death Rate Per 1,000
One child born alive to mother 1,640 2,657,850 0.62 …”
We don’t need to adjust, we can do a direct comparison.
We can directly compare the death rates of CNM hospital birth with CPM homebirth for each level of parity.
1 child CNM 0.48/1000 CPM 1.40/1000
2 children CNM 0.30/1000 CPM 1.50/1000
3 children CNM 0.42/1000 CPM 0.46/1000
4 children CNM 0.42/1000 CPM 2.10/1000
5 children CNM 0.67/1000 CPM 1.20/1000
6+children CNM 0.28/1000 CPM 0.70/1000
As you can see, in almost every category homebirth with a CPM has triple or higher neonatal mortality compared with CNM hospital birth. That’s a pretty devastating indictment.
Your response is totally unresponsive to my original observation.
I have demonstrated the presence of at least one confounding factor in the CDC cohorts you have compared.
Therefore, you cannot claim that the cohorts are comparable.
Therefore, you cannot claim (based on the CDC data) that “homebirth has xx times the neonatal death rate of comparable risk hospital births”.
Your subanalysis does not magically make the cohorts equal.
It confirms the fact that high parity is a confounding factor – no matter who is taking care of these women, increasing parity above two increases the neonatal death rate.
(The result you show for the highest parity group for CNMs is likely an anomaly due to small numbers, the hospital-MD rate is 1.24. )
In fact, speaking of small numbers, it’s interesting that you have chosen to present only calculated mortality rates and not the counts. Is that perhaps because of this caveat which accompanies the CDC database:
“Rates are suppressed when there are fewer than 20 deaths in the numerator, because the figure does not meet the NCHS standard of reliability or precision.”
A closer looks shows that of the 12 calculated death rates you presented above, only 3 of them actually meet that standard, the rest were suppressed.
There’s nothing “wrong” with recalculating those rates by hand, and presenting them, but when you say that “we can directly compare” you are on very shaky ground based on statistical methodology.
But you knew that, since the CDC inquiry brings the actual rates up as “suppressed”. Some hypothetical person (not I) who might be prone to divining motives behind other people’s behavior, might actually come to the conclusion that you were trying to “hide” something by presenting the chart in this way. Others might simply ascribe this to a lack of familiarity with basis statistical concepts like valid sample size.
IndianaFran:
“Your response is totally unresponsive to my original observation.”
Hardly. You can slice and dice the data any way you want, but the end result is still the same. When you compare like with like, homebirth has a higher rate of neonatal death.
Merely announcing that there is a potentially confounding variable is not enough to discredit the overall result.
In the first place, you never showed that parity is a confounding variable. The fact that the neonatal death rate among all hospital births varies with parity does not necessarily make it a confounding variable. Indeed, there is scientific evidence that neonatal death may vary with parity in some circumstances because parity is a proxy for both age and social class.
As a general matter, women of higher parity are older (and therefore at greater risk) than women of lower parity. In addition, high parity in our society is generally associated with lower socioeconomic class, another risk factor for poor perinatal outcome.
If you look at the comparison of neonatal mortality rates with parity for CNM patients as I presented above, you will see that there is no association of neonatal mortality rates with parity.
Most importantly, if neonatal mortality rates are higher for homebirth at every level of parity, adjusting both populations for parity is still going to show that homebirth has a higher neonatal death rate.
Therefore, your claim that parity is a confounding factor is not justified by the data, comparing homebirth with low risk hospital birth at every level of parity shows homebirth has an increased rate of neonatal death, and even if parity were a confounding factor, adjusting for it would still show that homebirth has an increased rate of neonatal death.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>IndianaFran said”
“If you had access to those statistics, please describe how you would choose a comparable hospital group of births for comparison.”
Amy Tuteur, MD replied:
That’s easy. I’d use the same parameters they used in choosing homebirth patients, low risk patients with no pre-existing medical conditions and no pregnancy complications.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Your tossoff answer here again shows that you don’t fully get it. This is not an easy question to answer. If it were easy, then science would have arrived at a definitive answer.
ACOG says in their position paper
http://www.acog.org/from_home/publications/press_releases/nr02-06-08-2.cfm
“It should be emphasized that studies comparing the safety and outcome of births in hospitals with those occurring in other settings in the US are limited and have not been scientifically rigorous. ” They don’t say “the scientific evidence is clear”.
ACNM says
http://www.acnm.org/siteFiles/education/Direct_Entry_Midwifery_Nov_05.pdf
“Despite a plethora of published papers on related topics, there are very few studies that isolate DEM practice as
a variable. The outcomes documented in the literature are mixed, with some investigators reporting better than
average outcomes among women attended by direct entry midwives, while others document better outcomes
provided by physicians and/or nurse-midwives. It is difficult to draw a conclusion from these studies, since they
are few in number, have a number of design problems, provide limited information regarding the multiplicity of
educational models for preparing DEMs, and often do not address the most current models. For those who are
seeking evidence to support a particular model of education or regulation for DEMs, it is clear that more
definitive research is needed on the relationship between educational pathways and clinical outcomes.”
You seem to be the only “authority” who believes that science has reached a settled conclusion.
The issue of cohort matching is, indeed, a difficult one.
This is particularly true in the US, where home birth providers have faced legal and political pressures that are unique. As a result, the very small number of women who choose home birth here consists of a motley assortment of “odd ducks”. While “some” of them are middle-to-upper-class who identify largely with mainstream culture, there are also many who are some sort of cultural dropout – the fringe religious groups mentioned earlier, people who self-identify as counterculture for various reasons, women with economic hurdles to hospital birth, and women whose prior birth experience has left them with emotional trauma. (It is irrelevant whether you think they “ought” to feel traumatized).
Finding a group of hospital clients who “matches” this heterogeneous group is statistically difficult, if not impossible.
Indeed, even defining the characteristics of the home birth group is problematic, because of the inaccuracies in recording birth attendant, especially in states where direct-entry midwifery is illegal.
So, go ahead and tell us that this is an easy question. It only shows that you are interested in finding a short-cut to the answer, not in using the difficult tools of science to get there.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
If you look at the comparison of neonatal mortality rates with parity for CNM patients as I presented above, you will see that there is no association of neonatal mortality rates with parity.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>.
No, what you can see is that among the higher parity levels, the numbers are too small to draw any conclusions.
That’s why I presented the data for all women (within the 20-44 age group, etc) by parity, because the numbers related to parity meet the statistical levels for reliability and precision.
You are the one who is “slicing and dicing” the data into statistically questionable territory.
In fact, most CNMs likely refer most high-parity clients to MDs. The small numbers who remain in the CNM group are not representative of high-parity women in general.
“Indeed, there is scientific evidence that neonatal death may vary with parity in some circumstances because parity is a proxy for both age and social class.”
I have made the claim that parity is a confounding factor. I have NOT made the claim that parity is an independent confounding factor. Of course it is related to age and class.
If the cohorts were already proven to be equivalent in age and social class, then indeed parity might already be fully accounted for, and would not correctly be assessed as a confounder. But you haven’t even tried to show any attempt at matching for age or socioeconomic status.
To start, you lumped all women age 20 – 44 into one risk class for your analysis. Have you even attempted to verify that the homebirth and hospital CNM cohorts are balanced similarly within the age categories (20-24, 25-29, etc)? Can you show us your work?
And since you have also brought up the issue of socioeconomic class, there is no way to assert that the two cohorts you have selected for are matched in socioeconomic profile. That data point is not collected, and therefore cannot be assessed. Therefore you cannot assert that the groups are comparable. The differences in parity level, in fact, strongly imply that socioeconomic profiles would not match if they were available.
I ask you here to answer these three questions
“Has there been a systematic effort to identify and measure potential confounders?
Is there information on how the potential confounders are distributed between the comparison groups?
What methods are used to assess differences in the distribution of potential confounders?”
IndianaFran,
You need to decide what you are trying to argue.
First you claim that parity is a confounder.
I point out that when you control for parity, homebirth has a higher neonatal death rate.
Then you claim that we can’t rely on the death rates for homebirth at high parity because there are too few women in each parity cohort.
But if there are too few women, how can you claim that the data indicates that parity is a confounding factor? You certainly are not entitled to extrapolate from a higher risk group since we are talking only about low risk birth.
Moreover, as I keep pointing out, this dataset is not the only source of information. The data is consistent with the Johnson and Daviss study, the most carefully done study of American homebirth to date. When you look at the neonatal death rate for all CPM attended homebirths in North America in 2000 and compare it to the death rate for low risk hospital birth in 2000, you find that homebirth has nearly triple the neonatal death rate.
In fact, with the exception of the recent Dutch and Canadian studies, where the training of midwives are quite different from homebirth midwives in the US, there is no scientific study or dataset that shows homebirth to be as safe as hospital birth.
Finally, there is the situation with the MANA (the official organization of homebirth midwives) dataset. We don’t know what’s in it, or whether the data is valid. We do know that MANA is restricting access to ONLY those who can prove, in advance, that they will use the data to support the homebirth movement. Even these people must sign a legal non-disclosure agreement promising not to share the data with anyone else.
If you have to prove that you will use the data only to support the homebirth movement, the database must contain data that undermines the homebirth movement. I don’t know specifically what that data is, but it is not unreasonable to speculate that the largest database of its kind shows that homebirth with a CPM is not as safe as hospital birth. Why else would MANA need to hide it from the general public. I can’t think of a legitimate reason; can you?
“You need to decide what you are trying to argue.”
OK, let me make this very clear.
I am arguing that the CDC statistics which you have presented at the top of your original post, cannot be claimed to contribute to any definition of Science-Based Medicine because there has been insufficient statistical analysis of the data to meet the standards of contemporary science.
I have pointed out several specific flaws in the cohort matching.
Some of these (geographic skewing) you have made no attempt to answer.
Your answers to other questions (in my opinion) reveals the fact that you did not perform even minimal statistical testing to verify cohort matching before you presented these “results”.
The question of what other studies show remains open to debate. That debate is a valid way to attempt to find a science-based answer.
My focus has been on the results which you yourself have presented, as an introduction to the topic of home birth in this forum. I infer that you believe that this data adds evidence to the scientific debate. I contend that this does not qualify as scientific evidence.
IndianaFran,
I’d be the first to acknowledge that the dataset is not perfect. However, its biggest problem biases strongly toward homebirth. In other words, this dataset makes homebirth looks safer than it is. That’s because the homebirth to hospital transfers are not included in the homebirth group.
In the Johnson and Daviss study, for example, approximately 12% of the patients were transferred to the hospital. However, 8/14 deaths in the homebirth group were pronounced at the hospital. In a dataset such as the one we are analyzing, more than 50% of the homebirth deaths would be erroneously put into the hospital birth group.
You keep insisting that we should ignore this dataset because there must be some confounding factor that makes the homebirth group higher risk than the hospital birth group, but you haven’t provided evidence of such confounders.
Finally, what I consider the most important point is incontrovertible. Although professional homebirth advocates in websites and publications insist that the scientific evidence shows that homebirth is as safe as hospital birth, there is no such evidence supporting the safety of American homebirth and a considerable amount of evidence showing that homebirth increases the risk of neonatal death.
There is certainly no scientific justification for claiming that homebirth with an American homebirth midwife is as safe as hospital birth.
Tim Mills:
Epidural analgesia or anesthesia: HB=7.7%, PA=27.6%, MA=26.3%
7.7%of the home births had an epidural or anesthesia? Who did they get to do that?
“You keep insisting that we should ignore this dataset because there must be some confounding factor that makes the homebirth group higher risk than the hospital birth group, but you haven’t provided evidence of such confounders.”
I have not said that we should ignore this dataset. I have said that your cursory analysis of the data is not scientific.
I have asked you to answer this question, and the followups:
Has there been a systematic effort to identify and measure potential confounders?
It might be possible to tease out some meaningful results using the CDC dataset as a starting point. But the first hurdle is to find a hospital cohort that is truly comparable, and provide the evidence for comparability. Simply making an assertion about risk profiles and hoping that nobody looks any deeper is not science.
I really really hope that when Monday comes, we have some statistically minded feedback on this exchange.
“I’d be the first to acknowledge……….”
may I suggest that after 220+ comments over 10 days, this statement can be said to be objectively false?
TsuDhoNimh
7.7%of the home births had an epidural or anesthesia? Who did they get to do that?
_________
The study to which Tim Mills refers was intention to treat analysis of place of birth. His note on caesareans also applies to epidurals
“Remember that HB means planned homebirth – those weren’t caesarean sections performed on the kitchen table!” (i.e. a proportion of the home birth group transferred, and of those, some had epidurals).
“There is certainly no scientific justification for claiming that homebirth with an American homebirth midwife is as safe as hospital birth.”
I’m not sure why you have directed that comment to me; I have never made that claim.
The claim I have made can perhaps be stated
There is no scientific justification for claiming that homebirth with an American homebirth midwife is x times as dangerous as hospital birth, regardless of what value you supply for x.
“In the Johnson and Daviss study, for example, approximately 12% of the patients were transferred to the hospital. However, 8/14 deaths in the homebirth group were pronounced at the hospital. In a dataset such as the one we are analyzing, more than 50% of the homebirth deaths would be erroneously put into the hospital birth group.”
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
One last nitpick before I go to bed. The location where the death was pronounced has no relevance to how the birth location is registered on the birth certificate. For births which occur at home or in transit, where the baby later dies at the hospital, I’m pretty sure that the hospital administration/legal dept folks will insist that the birth certificate is correctly registered as an out-of-hospital birth, because they would not want that death to appear in their statistics. It’s far more likely that only the only transfers which are lost are the ones where the baby was actually delivered in the hospital after transporting the mother. By my count from the J/D paper, that is 2, not 8.
Perhaps someone here can check this out with a hospital admin contact? Maybe it varies with state health dept protocols?
IndianaFran:
“I’m pretty sure that the hospital administration/legal dept folks will insist that the birth certificate is correctly registered as an out-of-hospital birth, because they would not want that death to appear in their statistics.”
Absolutely not! The person who signs the birth certificate is the person who delivers the baby. If the baby is delivered at the hospital, the doctor attending is the one who signs. Hospitals don’t “register” births.
That’s why the statistics for homebirth are almost certainly worse that they appear in the dataset. Any hospital transfers before the baby is actually born appear as hospital births, even if the baby is born in the parking lot.
Moreover, and we have not yet discussed this, intrapartum stillbirths are not registered as births at all (only as fetal deaths)l and therefore would not appear in the dataset. The intrapartum stillbirth rate in the hospital if vanishingly small. According to almost all existing homebirth studies from a variety of countries, the intrapartum stillbirth rate at homebirth is much higher.
In other words, a baby rarely drops into the hands of a doctor or hospital based CNM already dead, since there is usually continuous monitoring. Unfortunately, the situation at homebirth is quite different.
@Plonit:
Thanks for picking up that point. I’m afraid I’ve lost the energy to pick at every little mis-statement that Amy or others make regarding the valid studies.
@All:
I’m going to have to bow out here. I’ve put the case as I see it (not “my” case, you realize, but the scientific case). Some points I have not made here can be found at the discussion (from a couple years back) at <a href="http://www.badscience.net/forum/viewtopic.php?f=3&t=2193"Bad Science. Any further energy I spend here is going to have a greatly diminishing return. I doubt any casual wanderer who stumbles on this blog entry is going to read all (at this point) 228 comments.
I encourage everyone to, where possible, look at the original literature. Or at least browse the abstracts. I assure you, it’s no more opaque than the discussion here, and at least it’s all peer-reviewed. That doesn’t mean it’s perfect – and I thank Amy for pointing out some of the later developments in some of these papers. But, as someone, somwhere in the mists of early comments above, said: in the absence of thorough personal knowledge on a topic, I’ll trust a published, peer-reviewed result over an unpublished blog entry, regardless of the writer’s credentials.
@Amy:
Thankyou for a bracing discussion. I look forward to seeing what else you have in store for us here at SBM.
Gah! Sorry about the bad link markup. Here it is again, if copying and pasting isn’t your style: Bad Science.
And I’m sorry about the arrogance of the statement “not my case, but the scientific case”. Perhaps I can blame it on the hours I spent up last night with the kids (both planned homebirths, one non-emergency transfer to hospital due to failure to progress, two very healthy babies, no regrets).
Anyway, carry on.
I absolutely agree that if the baby is delivered in the hospital after transport, that birth certificate will count as a hospital birth. As I read the descriptions of death cases in J/D, that is 2 cases.
I am questioning the other six cases where the baby was delivered by midwife at home or in transit, and later died at the hospital.
You implied that because the death was pronounced at the hospital, the birth would also be considered a hospital birth. I don’t think that is the case.
I am directing this question at the site editors, because I do not want to be perceived as a troll or a distraction:
Is it in any way inappropriate or harassment for me to ask Dr Tuteur, regarding her presentation of CDC statistics, this question:
Has there been a systematic effort to identify and measure potential confounders?
I have asked this question several times, and no direct answer has been given. Since the original data as presented in her post was originally composed over a year ago for another site, it ought to be a simple matter to answer this question with a yes or no.
Dr Tuteur has been reading and responding to other comments over the last few days.
Would it be inappropriate for me to press this matter further, and again ask for a direct answer to the question?
I have presented several specific factors (geographic imbalance, maternal age, parity as a medical risk, parity as a marker for socioeconomic factors) which strongly lead me to question the degree to which scientific methods were applied to this data before publishing it at this site.
At what point is it justified to assume that an honest answer to the question might be “no”?
Without a direct answer to the question, are the site editors here comfortable with having the chart presented above in a large font, to be representative of a science-based approach to the question of home birth safety?
IndianaFran:
“I am questioning the other six cases where the baby was delivered by midwife at home or in transit, and later died at the hospital.”
Two cases were characterized by the authors as intrapartum deaths and no mention was made of transport of any kind. They would not have received birth certificates at all.
Four other babies died at the hospital, but that poses another problem. Ordinarily, the person who delivered the baby would sign the birth certificate. However, many homebirth midwives are practicing in violation of the law. Therefore, in the event of transport, many midwives disappear and/or parents claim that the baby was accidently delivered at home without a medical attendant.
We don’t know what happened in these four cases, but it is possible that some or even all of them wound up in a group different from that of other midwife attended homebirth.
There is really no question that the truth homebirth death rate is higher than that indicated by the database.
The issue of inaccuracy in birth certificate filings for out-of-hospital births has already been discussed here.
Due to inconsistent legal policies, we cannot say with any degree of accuracy, what number of bad outcomes or good outcomes are actually missing from the CDC database.
We also cannot say what number of bad outcomes or good outcomes are missing from the CNM data due to transfer to MD before delivery.
I have been consistently trying to say that making a comparison is far more complex than just grabbing 3 lines of summary data from the CDC Wonder database.
I’m getting really tired of all the quibbling. The bottom line is that there are certain unpredictable emergencies that require immediate response. Arrangements for quick transport can diminish but can’t eliminate the risk. If home birthers understand this and are willing to take that small risk, that’s OK with me. But if they are deciding on home births because of false information, that’s a shame. Amy has shown that false information has been circulating. Can’t we leave it at that?
The bottom line is that there are certain unpredictable emergencies that require immediate response.
++++++++++++
That would only be the “bottom line” if you assume that incidence of emergenices and complications are unaffected by place of birth. This may be a faulty assumption. Studies of place of birth consistently show much lower rates of intervention such as caesarean section with planned home birth. Given that we don’t do caesarean section for the fun of it, the rates of intervention may stand as a surrogate measure for the complications that prompt them.
It is obvious that “if something does go unexpectedly seriously wrong during labour at home, the outcome for the woman and baby could be worse than if they were in the obstetric unit with access to specialised care.” (NICE guidelines on intrapartum care). In fact, I would say, is likely to be worse.
It does not logically follow that the outcomes of home birth are worse overall.
A dubious proposition, as lower rates can signify either (a) fewer complications necessitating such intervention or (b) a reduced rate of intervention when those complications arise. Or a combination of both.
Please not the “may” in that sentence.
The point is that there is uncertainty about the relative safety home and hospital birth for women at low-risk of developing complications in labour.
If it turns out that (in certain contexts) that home birth can be as safe as hospital birth for low-risk women, then the most likely mechanism is that lower incidence of complications at home offsets the poorer outcomes when complications occur.
How do you explain the outcomes of the recent Dutch and Canadian studies?
http://www.cmaj.ca/cgi/content/abstract/181/6-7/377
http://www3.interscience.wiley.com/journal/122323202/abstract?CRETRY=1&SRETRY=0
Plonit:
“That would only be the “bottom line” if you assume that incidence of emergencies and complications are unaffected by place of birth.”
This claim has already been raised and discredited multiple times in this thread alone.
First, it is up to you to show that the rate of emergencies and complications is affected by the place of birth. Otherwise, we have no reason to make that assumption.
Second, and even more importantly, the hospital group ALREADY includes iatrogenic deaths. The homebirth death rate is still triple the hospital death rate. While it might be possible to further reduced the hospital death rate, that would only make homebirth look even worse by comparison.
“Studies of place of birth consistently show much lower rates of intervention such as caesarean section with planned home birth.”
That’s hardly something to crow about if it leads to increased neonatal deaths.
It’s the equivalent of a breast surgeon who decides to biopsy breast lumps that are golf ball size or greater and then boasts that he does very few “unnecessary” breast biopsies. If the death rate of his patients is three times higher that negates any “benefit” from doing fewer biopsies.
First, it is up to you to show that the rate of emergencies and complications is affected by the place of birth. Otherwise, we have no reason to make that assumption.
+++++++++
See Tim Mills post upthread, which quotes the recent Canadian study. The Dutch study shows similar results, as do others.
That’s hardly something to crow about if it leads to increased neonatal deaths.
+++++++++++
Which is not the case in the most robust studies to date (i.e. the recent Dutch and Canadian studies).
Plonit:
“See Tim Mills post upthread …”
I don’t see any quote that established that the risk of complications and emergencies is lower at homebirth. Perhaps you could repeat the relevant quote and present the relevant data.
And you haven’t explained why it even matters since the death toll of homebirth is still higher.
the hospital group ALREADY includes iatrogenic deaths.
+++++++++
Which “hospital group”? If you mean the hospital group in the CDC figures that you have presented, I think it is worth addressing IndianaFran’s question:
“Has there been a systematic effort to identify and measure potential confounders?”
Would you say that “fetal distress” is a complication?
Fetal distress: HB=1.3%, PA=3.6%, MA=2.1%
Plonit:
“Has there been a systematic effort to identify and measure potential confounders?”
It is a dataset, not a scientific paper. If anyone can show that there are confounders, they should do so. Merely speculating is not enough.
This thread already contains 240 comments and there is not a single one that shows ANY evidence that homebirth with an American homebirth midwife is as safe as hospital birth. No papers, no datasets, nothing.
People need to present actual data if they expect anyone to believe that homebirth in the US is as safe as hospital birth. Clearly it is not dispositive, but it is impressive that no one can provide anything to support that claim.
Plonit:
“Would you say that “fetal distress” is a complication?”
Not when babies are dropping lifeless into the hands of homebirth midwives who had no idea they were already dead. That suggests that fetal distress was undiagnosed in the homebirth group.
not a single one that shows ANY evidence that homebirth with an American homebirth midwife is as safe as hospital birth.
People need to present actual data if they expect anyone to believe that homebirth in the US is as safe as hospital birth.
++++++++++
This is a rather important qualification of the more absolute statements in your post that “Homebirth increases the risk of neonatal death to double or triple the neonatal death rate at hospital birth.” and that “Homebirth with a homebirth midwife dramatically increases the risk of neonatal death.”
This is becoming repetitive and annoying.
Enuf already!!!
Harriet:
Do you really think that it is quibbling to insist that when comparative statistics are presented as “scientific evidence”, those statistics should be based upon well-matched cohorts?
This is becoming repetitive and annoying.
++++++++++++
Do you have a science-based (rather than aesthetic) argument to put here?
I’ve never commented on SBM before, mainly because I’ve felt that the arguments presented are sound in most of the entries that I’ve read.
Dr Tuteur’s disdain for the practice of midwifery is so palpable. To conclude that “home birth is not safe” based on raw data is so blatantly unscientific and has no place in a blog like this.
I live in Ontario, where midwifery is highly regulated and embraced by OB’s, where Registered Midwives have hospital privileges and where women have a choice of place of birth if they meet strict criteria. I’m curious to how Dr Tuteur would respond to this recent study:
http://www.cfpc.ca/local/user/files/%7BB51825B6-44FF-4F63-9413-2B8829E117D6%7D/Ontario%20Home%20Birth.pdf with almost 7000 women in each well matched cohort.
This is more about lack of training in the US for midwives rather than the “safety” of home birth across the board.
I come here for science, not vitriolic speculation.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Harriet said:
I’m getting really tired of all the quibbling. The bottom line is that there are certain unpredictable emergencies that require immediate response. Arrangements for quick transport can diminish but can’t eliminate the risk. If home birthers understand this and are willing to take that small risk, that’s OK with me. But if they are deciding on home births because of false information, that’s a shame. Amy has shown that false information has been circulating. Can’t we leave it at that?
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Well, I suppose we could. But we couldn’t do that and simultaneously claim that we are seeking a science-based answer to the question of birth location and safety.
A science-based approach requires you to look at the overall outcomes, not just the outcomes of the small minority who encounter a genuine emergency. And it requires you to look at those outcomes using rigorous science-based analysis.
Let me ask you this: if someone presented a “data set” similar to the original post, comparing neonatal death among women who consumed fish oil supplements during pregnancy with women who didn’t – and the author claimed that the data “showed” that lack of fish oil supplements “dramatically increases the rate of neonatal death”. Wouldn’t your first response be to question whether the two groups were equal in other ways? Wouldn’t your skeptical side be thinking that access to and use of various supplements is probably a marker for other health-conscious behaviors and socioeconomic status, and that the results were not likely causally related to the supplements? Why hasn’t your skeptic radar reacted the same way here?
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Amy said:
“Has there been a systematic effort to identify and measure potential confounders?”
It is a dataset, not a scientific paper.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Did I kinda sorta almost hear the word NO whispered between the lines?
So, I guess it’s safe to assume that the answer is indeed NO.
Interestingly, Amy says here:
“It is a dataset, not a scientific paper.”
And later she says on the correlation/causation thread:
“The CDC database is not the only scientific evidence that shows…..”
You can’t have it both ways. A “dataset” only becomes “scientific evidence” after it is processed through the tools of rigorous scientific analysis and evaluation. Which has clearly not been done here.
I am one of “those” women who chose homebirth (3 times) successfully, no complications.
I think if you want to “woo” women like me back to birthing our babies in the hospital, you need to change how things go in the maternity ward. From start to finish- from the first visits to the Doctor-OB/GYN/CNM for prenatal care to the actual birth itself.
Frankly I don’t like being treated like cattle,like my opnions don’t matter,that I couldn’t possibly be informed because well, I didn’t spend 10 years in medical school. I have heard so many horror stories from friends who have birthed in the hospital, it makes me sick to my stomach literally.
You, the Ob/Gyns are pushing (literally) women to find alternatives outside the hospitals. We are willing to take the “risk of having a homebirth” because we believe the “risks” outway the “risk” of being stripped of our dignity and respect.
A complete overhaul of how the whole process of having a baby in the hospital needs to be made and including womens opinions and what they want out of it need to be addressed. Until then I will continue to have my babies at home and spread the word about the amazing experience.
Returning the discussion back here from the “causation and correlation” post, Amy says:
++++++++++++++++++++++++++++++++++++++++
“It is also biologically plausible that the higher proportion of older mothers in the home birth group is responsible for some part of the differing results.”
But analyzing the data shows that there is no association between neonatal death rate and maternal age for ages 20-44.
“It is also statistically plausible that the results are affected by geographic skewing.”
But analyzing the data shows that there is no association between neonatal death rate and geographical region.
“It is also biologically plausible that socioeconomic differences explain some of the difference in results.”
But analyzing the data shows that there is no association between neonatal death a maternal education status in this group.
“It is also biologically plausible that the disproportionate presence of certain isolated ethnic groups (e.g. Amish) in the homebirth group can account for some part of the difference”
Even if you remove congenital anomalies entirely, the CNM group has a neonatal mortality rate of 0.26/1000 and the homebirth group has a neonatal mortality rate of 0.61/1000.
Moreover, congenital anomalies are not necessarily incompatible life. Babies born with significant cardiac anomalies can be saved with appropriate resuscitation and surgery. Of course, if such a baby is born at home, the chances of survival drop precipitously.
There is really no evidence that the CNM hospital group and the homebirth group differ in any way likely to affect neonatal mortality rates. No matter how you analyze the data, the homebirth group always has higher neonatal mortality than the hospital birth group.
++++++++++++++++++++++++++++++++++++++++++
It really strains any sense of credibility that after many days of completely evading the questions about the presence of confounders, after being presented with a list of specific differences in the cohorts, you now claim to have “analyzed” the data and ruled out any such effects. I’m not buying it. Any epidemiologist or public health expert would tell you that it is absurd to claim that two groups are equivalent in underlying risks when there is clear evidence that one group has a higher percentage of older mothers, a higher percentage of grandmultiparas, a higher percentage of low-education levels, a higher percentage of ethnic groups tending toward congenital anomalies, and a higher percentage of women from lower-income regions. Simply asserting that none of these factors has any meaningful effect on outcomes is, well, not at all scientific.
But let’s look at this claim from another perspective:
“There is really no evidence that the CNM hospital group and the homebirth group differ in any way likely to affect neonatal mortality rates. ”
If you believe this is true, then you are also saying that direct-entry midwives in the US are just as competent, meticulous, and prudent about evaluating their clients for risk factors as hospital-based CNMs are. Or that women who choose home birth are themselves capable of making a well-informed and educated self-evaluation of their own risk level and self-referring to hospital based care when appropriate. Can you agree with either of these statements?
IndianaFran:
“Can you agree with either of these statements?”
No.
We don’t have to venture into speculation. We have the data. We know what the data show.
I realize that you don’t wish to believe that homebirth increases the risk of neonatal death, and that you are casting about for some reason to ignore the data, but you haven’t found one yet.
I didn’t know that the range of science based medicine included long-distance psychic readings of what other people believe or wish to believe.
I believe in science-based evidence. I believe (with the majority consensus of experts) that at this time, the valid scientific evidence related to home birth safety is mixed and inconclusive. And I know that your personal “analysis” of the CDC dataset adds nothing genuinely scientific to that evidence base.