Stories for Outcasts

Debunking Anti-Homebirth Rhetoric

Thursday, October 10, 2013

"There is no honor in vaginal birth and it is despicable for anyone to insist that there is." -Amy Tuteur

("The only thing that makes you an honorable parent is loving your child. However to equate submission to the medical establishment during pregnancy and birth with feminist ideals is insane, and a slap in the face to women who have fought for independence and self-determination." -Curtis)

Since my wife, Liberty, and I planned for our daughter Adelaide's birth in 2011, I've become a passionate supporter of midwifery and naturalized childbirth. I've written about this before in my Blogger site, here.

After Emrys was born this year, I felt even stronger support for midwifery, so much so that I want to contibute to the cause with more than just advocacy to my circle of friends. Unfortunately, my main skill in software development doesn't translate very well into that world, so I spent some time searching for good fits for me in homebirth circles. While searching, I came across the most curious thing: An anti-homebirth advocate, spewing vitriol and rhetoric that would put Rush Limbaugh to shame.

Her name is Dr. Amy Tuteur, and she runs such sites as homebirth death statistics (links to abstracts of scientific studies showing home birth is more dangerous than hospital birth), Hurt by Homebirth (stories purportedly submitted by people involved in home births gone bad), and The Skeptical OB (general inflammatory statements and terms... "2 out of 3 babies who die at homebirth could be saved in a hospital", "quacktivists", "lactivists", "Why won't MANA release its death rates!!?!?")

There is a slate.com article that covers her brand of fearmongering fairly well, so I won't delve too deeply into her attempts to sow FUD, however a pair of studies linked to on her "death statistics" blog in this entry I found pretty interesting, and led me to download lots of birth datasets from the CDC's National Center for Health Statistics site, and comb through them with my go-to analysis tools of perl and Excel.

After poking around for some time, I found that the stats shown in the studies seemed to be accurate, but that the questions they were answering were the wrong ones. The questions were:

  1. Of live births that are full term with single babies in 2008, do home births have a higher percentage of Apgar scores less than 7?
  2. Of live births that are full term in 2005, do home births have a higher percentage of infant mortality.

The answers to both are yes. Just barely. And the difference is even more slight if your set of home births only contains those attended to by certified midwives. And the answers are both no if you choose different sets, like multiple births, or births to non-white mothers. More importantly, if you start looking at the entire spectrum of gestation times instead of narrowing the search to full-term, you'll see something pretty shocking: A much larger percentage of standard medicalized pregnancies result in pre-term babies, infant mortality, and stillbirths than do moms under the care of midwives.

If your prenatal care is given to you by a midwife, you are an order of magnitude more likely to bring the baby to term, to have a perfect Apgar score of 10, and for the baby to live past infancy. If you discount the tragedies of moms who miscarry or deliver stillborn babies, and narrow the focus of a study to birth types that tend to be good in hospitals, and choose years where home birth had poorer than typical results, only then can you paint home birth in a bad statistical light. Only slightly. Only by ignoring tragedies. And only by squinting.

The question should be "if I get pregnant, am I more likely to have a good outcome with the prenatal and birthing care of a midwife, or the standard medical community and hospitals? The answer is unambiguous: you're better off with a midwife.

Let's start by looking at the studies the way the are, then widen the searches to see the dirt hiding on the other side, and then at the close of this article we'll consider what happens to the percentages if you include fetal deaths.

The first study abstract is number 65 on this pdf file. Here's an abstract of the abstract (emphases mine):

Neonatal outcomes associated with intended place of birth: birth centers and home birth compared to hospitals

STUDY DESIGN: This was a retrospective cohort study of singleton live births that occurred in 2008 in the U.S. that had specified birthing facility information. Deliveries were categorized by location of occurrence: hospitals, birthing centers, or intended home births.

RESULTS: There were 2,296,953 singleton, live, term births meeting study criteria; of these, 10,726 (0.47%) delivered at birthing centers and 12,433 (0.54%) had intended home births. While the risk of cesarean delivery was much lower for women who delivered/or intend to deliver outside of hospitals (0.02-4% vs. 24%), the odds of 5-minute Apgar score < 7 and neonatal seizure was significantly higher for intended home births compared to hospital birth.

I've italicized a few items here:

Gist of the first study

The claim is that having an Apgar score less than 7 is significantly greater at a home birth. With the specific filter used in the study, the chance was about 2 and a half percent, compared to one and a quarter for hospital birth. The abstract shows a birthing center rate of just under 1%, but focuses on hospital and homebirth comparisons instead. The study also does not mention that home births have a much higher percentage of Apgar scores of 10 (40% vs. 4% in hospitals), or that midwife attended births have a lower percentage of babies with an Apgar of 4 or less. As the study was funded by the Society for Maternal-Fetal Medicine (SMFM), whose member requirements include being an obstetrician, a slant in this direction is expected.

The claim that 2.51% of homebirth Apgars of full term, single deliveries are under 7, compared to 1.23% for hospital births is correct. Including only home births that had a midwife attendant drops the percentage of under 7s to 2.15%. Should the Midwives Alliance of North America (MANA) and local midwifery groups make efforts to reduce the percentage of babies with lower Apgar scores? Absolutely. Is home birth then inherently more dangerous? No. Overall the numbers show that babies have a higher percentage of better outcomes with midwives at home than at a hospital.

While a Certified Professional Midwife (CPM) can't perform an emergency cesarean or administer pitocin when it is truly necessary, the better treatment they show moms in labor pays off in spades: Mom is more relaxed, able to move around, eat and drink to keep her strength up, and typical births have a baby in much better condition than after the trauma a mom goes through at a hospital.

Before I supply evidence for my outlandish claims, let's take a look at the data definitions, and the tools I have to analyze it. Should you wish to run your own analysis, all the CDC vital stats files are available at the CDC. If you want to avoid a crash course in searching data with Linux commandline tools, feel free to skip down to the charts.

CDC File format

The CDC vital stats files contain fixed length records for easy processing on mainframes or SAS programs. The records use numeric codes to specify data about the baby, the gestation time, the parents (race, age, marital status, previous children), prenatal care, etc. With over 4 million births per year in the US, the uncompressed files are pretty huge. The 2008 file is over 3 gigs, for instance. Because of this, the CDC serves the files zipped.

The data set I downloaded was the "Period Linked Birth-Infant Death" file for 2008, as this also had data on infant mortality, which I used when analyzing the second study. The file name is "LinkPE08US.zip".

According to the docs, the USDENPUB file contains details on births, USNUMPUB contains infant deaths from those births that occurred in the same year, and USUNMPUB contains infant deaths that could not be linked with that years birth data. The nomenclature is DEN for denominator, NUM for numerator, as various mortality probabilities can be determined by applying the same search to both files, and dividing the mortalities by the births that match the search.

Fields of interest

This shows the column numbers (0 indexed - the column numbers mentioned in the docs for these same fields are all 1 greater than the numbers below) of fields we are interested in. Unless otherwise noted, fields are one character.

  6 = Line format. S = 1989 report standard, A = 2003 standard (separates intended from accidental home births)

The first thing to note I mentioned above, namely this file contains records in one of two formats, where only 27 states are using the newer format. Also, the documentation is wrong about the codes used in the actual file. They are S and A, as shown above, instead of U and R, as shown in the docs.

 40 = Birth place. 1 = hospital, 2 = birth center, 3 = intended homebirth

There are more options for birth blace, namely unintended homebirth, homebirth - intent unknown, clinic, other, and unknown. However, we are only interested in the values above. In the 1989 format file, the field after this one is used to identify birth place, and only 1 option for homebirth is present.

 88 = Mother's age. (2 chars)
409 = Attendant. 1 = MD, 3 = CNM, 4 = "Other midwife"

Yes, the Certified Professional Midwife gets no love here, lumping them in with a generic "other" category, which may include non-certified midwives practicing in a very grey legal area, and skewing the results unfavorably. The data collected by MANA may have better granularity. Phase two of my number crunching will be to see if I can get a researcher account with them. Until then, nurse midwife and "other" is all I have to search on.

414 = Apgar score. 00 - 10, or 99 for unknown (two chars)
422 = Single or multiple birth. 1 - 5, where quints or greater = 5
450 = Gestation weeks. 17 - 47, or 99 for unknown (two chars)

Command line tools

The Linux unzip program can make a particular zip file member's data available to another program by means of the pipe character (|). For example, this command sends the text of the denominator file to the "wc" program (word count), which is using the "-l" switch to count the number of lines:

?: unzip -p LinkPE08US.zip VS08LINK.USDENPUB | wc -l
4255188
?:

In other words, there were over 4 million births in the US in 2008. I want to use unzip in this way so that I don't have to explode each year's data files, taking up multiple gigs each, and taking a large bite out of my hard drive's free space.

The Perl scripting language has some fairly powerful commandline features, including being able to easily iterate over each line of a file, incrementing counters or hash values as it goes, and dump the output at the end. As a general function type I'm referring to as a "filter", the perl program can receive a CDC file and count all the records that match a certain parameter.

The general command format for a file filter is this:

perl -lne '<filter>'

In the "-lne" switch, the "e" means "execute the commands in the quotes that follow", "n" means "over every line coming from the pipeline, or every line in the file name following the quotes", and the "l" means "lop off line-end characters from the input, and add a line end character every time you print some text".

Let's go over a complete command for a very simple filter: count the number of records in the 2003 format in the denominator file:

$cnt++ if substr($_, 6, 1) eq "A"}{print $cnt

Here we're declaring a variable "cnt" (count), and incrementing it (++) if the substring of the current line of input ($_) at column 6 for 1 character long is equal to the string "A". The "}{" is shorthand for an "END" block, meaning everything before "}" gets run on each line of the file, and everything after "{" gets run once after the last line is processed. Finally, the value of "cnt" is displayed. Here's the whole command, including the unzip command that feeds the script, and the output:

?: unzip -p LinkPE08US.zip VS08LINK.USDENPUB | perl -lne '$cnt++ if substr($_, 6, 1) eq "A"}{print $cnt'
2761407
?:

So there are around 2.7 million records in the file using the 2003 format. With that simple example out of the way, let's look at a real search, showing us the probable records used in the actual SMFM study. From this search onward, I'll just show the filter followed by the output, and omit the full command line. This search and many that follow uses regular expressions (the business with "=~" and "/.../"), which I won't attempt to explain as part of this post. If you're unfamiliar with them and fancy yourself a nerd, go look them up! They're super handy.

$cnt++ if substr($_, 6, 1) eq "A" && substr($_, 40, 1) =~ /[123]/ && substr($_, 414, 2) ne "99" && substr($_, 422, 1) eq "1" && substr($_, 450, 2) =~ /3[7-9]|4[0-2]/}{print $cnt

2286562

We searched for lines in the 2003 format (column 6 = A), where the birth place (col 40) was any of 1, 2, or 3, the Apgar score was not 99 (unknown), the number of babies delivered (col 422) was 1, and the gestation (col 450) was either 37-39 or 40-42. There were 2,286,562 such records, a close match to the 2,296,953 records in the study. The specific selection criteria (and specifics on how outliers were excluded, which the abstract mentions) aren't available to me, so I'm sticking with these records. The resulting distribution of Apgar scores matches very closely what the study found. (Also explicitly filtering on the "A" in column 6 is redundant if we're also using the column 40 birth place, which is always blank in the 1989 standard, so for future filters, we'll leave that off when it's not explicitly necessary.)

Now, for creating a list of Apgar distributions, we can tell perl to create a hash (a primitive collection type populat in scripting languages) of counts for each Apgar score. These values can then be thrown into a simple spreadsheet to graph the counts, and show the percentage of deliveries that were at that score or below:

$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "1"  && substr($_, 422, 1) eq "1" && substr($_, 450, 2) =~ /3[7-9]|4[0-2]/}{print $h{$_} || 0 for "00" .. "10"

This creates a hash h, with keys for each of the found Apgar scores, if the birth place is 1, the number of babies delivered is 1, and the gestation is 37-42 weeks. After the hash is populated, the END block iterates from 00 to 10 using perl's magic string sequencer (the ".." operator), and prints out the hash value with that key... or 0 if the key isn't defined. The results are:

403
1271
1970
2500
3622
6851
11562
35822
248394
1881141
70445

Running the same filter, but for intended home births:

$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "3"  && substr($_, 422, 1) eq "1" && substr($_, 450, 2) =~ /3[7-9]|4[0-2]/}{print $h{$_} || 0 for "00" .. "10"

18
5
11
17
17
103
131
284
1204
4797
5281

Plugging those values into an Excel chart gives us this:

The shapes of the two graphs is fairly consistent, regardless of what filter is used. Both delivery places are essentially flatlined until the Apgar score is 7, they jump a little at 8, and spike at 9. An Apgar score of 10 is substantially different between home and hospital births. 45% of homebirths have a perfectly pink, breathing, responsive, and flexing baby with a strong heart. Statistically, a score of 10 is the mode for home birth, where a score of 9 is the mode for hospital births. This is true across all searches.

The numbers from the graph above are essentially what the study reported (my numbers are off by a fraction of a percent), however if we also filter home births by those attended by qualified midwives, we see an even better picture:

$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "3"  && substr($_, 422, 1) eq "1" && substr($_, 450, 2) =~ /3[7-9]|4[0-2]/ && substr($_, 409, 1) =~ /3|4/}{print $h{$_} || 0 for "00" .. "10"

2
4
7
13
14
83
81
175
724
4026
4346

This is where we first see something interesting about home births with a qualified attendant: The percent of Apgar scores under 5 is slightly less than at hospitals. At this level, the general "hypoxia" diagnosis, which under 7 can be correlated with slower development, poorer motor control through infancy, etc., is instead correlated with cerebral palsy.

Here are other filters where home birth beats hospitals:

Same filter as the original study, but home births are filtered by nurse midwives only:

$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "3"  && substr($_, 422, 1) eq "1" && substr($_, 450, 2) =~ /3[7-9]|4[0-2]/ && substr($_, 409, 1) eq "3" }{print $h{$_} || 0 for "00" .. "10"

1
0
1
4
3
7
20
42
255
1860
1065

There are a few values of 1 and 0 here, a product of such a small sample size. To be more accurate in what type of work CNMs perform in home births, here's the same search over the years 2005 - 2009:

I used the same filter as above, run on all 5 datasets using the pattern:

unzip -p LinkPE<YY>US.zip VS<YY>LINK.USDENPUB | perl -lne '<filter>'

...followed by throwing all the output into Excel and summing it. There are hundreds of more hip ways to do the same thing programmatically, but my thinking cap had started to fall off from a long stint of looking at these numbers.

Inverting the study. Hospital v. home birth w/midwife, with more than one birth, or preterm:

'$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "1"  && (substr($_, 422, 1) ne "1" || substr($_, 450, 2) !~ /3[7-9]|4[0-2]/)}{print $h{$_} || 0 for "00" .. "10"

vs.

$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "3"  && (substr($_, 422, 1) ne "1" || substr($_, 450, 2) !~ /3[7-9]|4[0-2]/) && substr($_, 409, 1) =~ /3|4/}{print $h{$_} || 0 for "00" .. "10"

I believe these results speak clearly to the type of prenatal care a midwife provides. Midwives encourage moms to be active and eat healthy, drink herbal teas for stomach issues, connect emotionally with the baby (which sounded like hocus pocus to me the first time I heard it... but biofeedback is a real thing, and one's emotional state determines a lot about one's health), and give lots of encouragement and praise (your belly looks great, thats awesome weight gain, I'm looking forward to meeting the baby) to reduce mom's stress level. And the outcomes reflect that.

Imagine combining that methodology with good medical equipment and a full staff of midwives capable of performing stronger interventions, but preferring to encourage and relax mom. What do you get? The birthing center results that the study glosses over:

$apgar = substr($_, 414, 2); $h{$apgar}++ if substr($_, 40, 1) eq "2"  && substr($_, 422, 1) eq "1" && substr($_, 450, 2) =~ /3[7-9]|4[0-2]/}{print $h{$_} || 0 for "00" .. "10"

2
3
7
6
11
25
55
180
818
5702
3904

Top to bottom, it's clear why the SMFM would not want to draw too much attention to that.

The second study

The second study's abstract is number 563 on this pdf file. From the abstract:

Does planned home birth affect neonatal mortality?

OBJECTIVE: To determine the relationship between planned home birth and neonatal mortality

STUDY DESIGN: This is a population based retrospective cohort study of all births between 37 and 42 weeks gestation using the National Health Center for Vital Statistics 2005 Linked Birth/Infant Death Cohort Data Set. The primary outcome was neonatal mortality and the primary predictor was planned home birth. The referent group for the regression model was births that occurred in a hospital.

RESULTS: There were 4,145,887 births included in the final analysis with 27,968 cases of infant mortality (6.69 cases per 1000 births) with 3,192,544 births and 7,620 neonatal deaths occurring between 37 and 42 weeks gestation. ... The unadjusted odds ratio (OR) for neonatal mortality among individuals having a planned home birth was 1.37 (95% CI 0.89, 2.10)

So, the study implies that we are 37% more likely to have an infant death if we have a planned home birth if our baby is born full term. Fortunately, we have the specific dataset used, which is publicly available, and exact numbers from various searches so we can verify our search results. Good start; let's jump right in.

Apply a filter to the numerator file with:

unzip -p LinkCO05US.zip VS05LKBC.USNUMPUB | perl -lne '<filter>'

Apply a filter to the denominator file with:

unzip -p LinkCO05US.zip VS05LKBC.DUSDENOM | perl -lne '<filter>'

Filter by 37-42 wks gestation estimate (columns 446-447)

Filter: $cnt++ if substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:      7620
Denominator: 3192544

Good, this exactly matches the study numbers. Now let's add in a filter by intended birth place:

Home birth
Filter: $cnt++ if substr($_, 40, 1) eq "3" && substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:      21
Denominator: 6,385
Ratio: .329%

Hospital birth
Filter: $cnt++ if substr($_, 40, 1) eq "1" && substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:       2,613
Denominator: 1,051,123
Ratio: .241%

329/241 = 1.37

So far the numbers match exactly what the study says they do. What happens, though, if we don't filter explicitly for full term babies?

Home birth
Filter: $cnt++ if substr($_, 40, 1) eq "3"}{print $cnt

Numerator:      27
Denominator: 6,786
Ratio: .398%

Hospital birth
Filter: $cnt++ if substr($_, 40, 1) eq "1"}{print $cnt

Numerator:       8,476
Denominator: 1,258,559
Ratio: .673%

This is pretty telling. With no gestation filter, the home birth mortality rate rises slightly, however the hospital birth mortality rate nearly triples! After seeing this, I wanted to know one thing: What percentage of births in both categories don't make it to term?

From the 2005 file (denominator only):

Home birth
Filter: if (substr($_, 40, 1) eq "3") {$tot++;  $cnt++ if substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}}{print 1-$cnt/$tot
Ratio: 5.91%

Hospital birth
Filter: if (substr($_, 40, 1) eq "1") {$tot++;  $cnt++ if substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}}{print 1-$cnt/$tot
Ratio: 11.56%

And what is the infant mortality rate of non-term babies in each category?

Home birth
Filter: $cnt++ if substr($_, 40, 1) eq "3" && substr($_, 445, 2) !~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:     6
Denominator: 401
Ratio: 1.50%

Hospital birth
Filter: $cnt++ if substr($_, 40, 1) eq "1" && substr($_, 445, 2) !~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:     5,794
Denominator: 145,507
Ratio: 3.98%

In short, if you are pregnant and getting hospital/obstetrics care, you are twice as likely to have a preterm baby than if you are seeing a midwife. And if you have a preterm baby at a hospital, it will be 2.5 times as likely to die in the first year than if you delivered a preterm baby at home with a midwife.

Now, let's add the tragedy of fetal deaths into the equations. The CDC Vital Stats page has a link for the records of miscarriages and stillbirths from 2005. Included in the records is the estimated gestation time, and many of those are 37-42 weeks, unfortunately. However, unlike the live birth and infant death datasets, this one doesn't break down the intended delivery place, but rather only the actual delivery place. This is in column 41 (again, 0 indexed, the docs put this at column 42), where 1 is a hospital, and 4 is a residence.

Because of this, to identify which were probably intentional homebirths gone bad, we must be creative in our search criteria, and then broaden the searches for live births and infant deaths to include data from the 23 states not included in the two studies we're critiquing. Since my argument is that the type of care is the most significant factor in outcomes, we'll combine searching by place with expected attendant type. E.g., delivery place = hospital + attendant = doctor vs. delivery place = residence + attendant = CNM or "other midwife".

First, let's apply these searches to the live birth/infant death datasets to make sure we get ratios in the same range as above. This is the same general criteria as the second study, namely the infant death ratio of full term babies.

Home birth
Filter: $cnt++ if substr($_, 41, 1) eq "4" && substr($_, 409, 1) =~ /3|4/ && substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:       47
Denominator: 13,250
Ratio: .355%

Hospital birth
Filter: $cnt++ if substr($_, 41, 1) eq "1" && substr($_, 409, 1) eq "1" && substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt

Numerator:       6,571
Denominator: 2,740,845
Ratio: .240%

Both ratios are within a couple hundredths of a percent of the study's ratios, so we can deduce we're searching similar criteria. Now, let's look at the same search applied only to fetal deaths only.

Home
Filter: $cnt++ if substr($_, 41, 1) eq "4" && substr($_, 409, 1) =~ /3|4/ && substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt
Fetal deaths: 1

Hospital
Filter: $cnt++ if substr($_, 41, 1) eq "1" && substr($_, 409, 1) eq "1" && substr($_, 445, 2) =~ /3[7-9]|4[0-2]/}{print $cnt
Fetal deaths: 3,035

And if we change the study's equation from infant deaths/live births to tragedies/pregancies, we get this:

Home
Numerator:       47 + 1 = 48
Denominator: 13,250 + 1 = 13,251
Ratio: .362%

Hospital
Numerator:       6,571 + 3,035 =     9,606
Denominator: 2,740,845 + 3,035 = 2,743,880
Ratio: .350%

362/350 = 1.03

On the surface, hospitals still appear to do marginally better (3% better, in this case), however I'll pose the same question as before: How many make it to full term? If we do the same searches and take out the full term requirement, we see an awful truth:

Home
Filter: $cnt++ if substr($_, 41, 1) eq "4" && substr($_, 409, 1) =~ /3|4/}{print $cnt
Fetal deaths  (A):      1
Infant deaths (B):     58
Live births   (C): 14,694
Tragedies / Pregnancies ((A + B)/(A + C)): 59 / 14,742 = .400%

Hospital
Filter: $cnt++ if substr($_, 41, 1) eq "1" && substr($_, 409, 1) eq "1"}{print $cnt
Fetal deaths  (A):    47,646
Infant deaths (B):    25,261
Live births   (C): 3,594,203
Tragedies / Pregnancies ((A + B)/(A + C)): 72,907 / 3,641,849 = 2.002%

Twice as many fetal deaths as infant deaths! Tragic outcomes of 2% vs. 0.4%. Unless my understanding of these files is totally wrong, then with standard medicalized pregnancy, you are 5 times as likely to either lose the baby, or have it die in infancy, than if you receive prenatal care from a midwife.

Forget Apgar scores. Forget the assumption that cherrypicking the data for full-term babies means that you are doing an apples to apples comparison. You aren't. The babies with worse Apgar scores under midwife care? They would probably have been miscarried under the care of an obstetrician. Why? The emotional stress of being told all the bad things that they need to test for, and the tendency to try not to love a baby you believe you'll lose. Convincing moms they need amniocentesis because x,y, or z proteins came up on a blood test. Ultrasounds... watch your baby when you get an ultrasound. Does he try to run away from the probe? He should, it's heating up the amniotic fluid, damaging his cells, and causing growth to slow.

Conversely, what does a midwife do? She says "You can do it. Go out and get some sunshine, breathe some fresh air and get some exercise. Make sure you eat lots of protein and calcium, and drink some nettle tea for nausea. Let me use my tape measure and hands to check on the baby. You look great. I'm proud of you. You're going to be a great mom!" And this standard of care resulted in one stillbirth out of 14 thousand pregnancies.

I think we're both pretty cross-eyed from numbers and charts at this point, so I'll save my commentary on the outcomes of African-American moms for a later post. But suffice to say this: Midwives are also color-blind, and hospitals are not, and the numbers unambiguously show it.

Thanks for reading this far, and I hope you'll consider at least talking to a local midwife group about their style of prenatal care, what home births are like, and the history of the group's birth outcomes. If you're anything like me, a couple visits and you'll be very likely to swear off doctors.

If you live in Columbus Ohio, here are two good groups to start with:
The Center for Humane Options in Childbirth Experiences - CHOICE
BORN Community Midwives