By Jill MacCorkle
Issue 117, March/April 2003
Media reports of the latest study on homebirth were short and to the point: Infants born at home had twice the risk of death, and both mothers and infants had higher risks of other complications. The study, "Outcomes of Planned Home Births in Washington State: 1989-1996," appeared in the August 2002 issue of Obstetrics and Gynecology, the official journal of the American College of Obstetricians and Gynecologists (ACOG).1 On the day of publication, ACOG, which represents the interests of 40,000 obstetricians and gynecologists in the US, issued a press release titled "Homebirths Double Risk of Newborn Death."2 The lead author of the study, Jenny W. Y. Pang, said in media reports, "It's still a small risk, but women should know there is an added risk with homebirth."3
In contrast to the headlines, the study itself states, "The results... suggest that planned homebirths are associated with an increased risk of adverse neonatal and maternal outcomes, particularly among nulliparous women. Nonetheless, more light needs to be shed on this controversial topic before practitioners and expectant parents can be fairly counseled about the safety of planned homebirths."4 This presents a tale quite different from the one spun for the media.
Knowing that most consumers and reporters will not read the actual study, ACOG can be reasonably confident that, if its statements simplify or overstate the conclusions, few will realize it. But midwives and homebirth experts were immediately skeptical, because the study appears to contradict a large body of research on homebirth that demonstrates that planned, attended homebirth for low-risk women is as safe as, or safer than, hospital birth. The list of studies that confirm that idea is impressive in its length, depth, and breadth. What, then, to make of the Washington State data?
Findings of the Study The stated objective of the Washington study was to evaluate the risk of neonatal (i.e., less than 28 days of age) death for planned home deliveries with professional providers compared with that of intended hospital deliveries. In addition, the authors analyzed a short list of other complications. They used linked data from Washington State birth and death certificates as their data source. Unfortunately for the accuracy of their research, "Washington State birth certificates do not identify which homebirths are planned."5 The researchers acknowledged the importance of determining that the homebirth group contains only planned births; unplanned births and unattended births are associated with much higher risks for mothers and babies.6
The homebirth group, after correcting for gestational age of 34 weeks or more and for certain pregnancy complications, included 6,133 singleton births, 279 of which were classified as planned homebirths that were transferred to the hospital during labor. The hospital birth cohort included 10,593 singleton births. These formed the main groups of the study. In further analysis, the authors compared only those births of at least 37 weeks and babies with birthweights of more than 2,500 grams (5 lbs. 8 oz.).
Demographic data for the two groups showed that the homebirth mothers were more likely to be white, married, older, nonsmokers, and parous (having had a previous birth). Curiously, the researchers failed to match the two groups for risk factors, instead matching for birth year only. Since the stated objective was to isolate the effect of birth environment on birth outcomes, the researchers should have made the two groups as alike as possible. They did not do so. Although they later adjust for factors such as age and parity, they report the results of these adjustments inconsistently in their data.
There were 20 neonatal deaths noted in the homebirth group, for a rate of 3.3 per thousand, compared to 18 deaths (1.7 per thousand) in the hospital group. The homebirth group also had higher risks of very low (0-3) five-minute APGAR scores and slightly higher rates of assisted ventilation of more than 30 minutes. The risk of assisted ventilation was statistically significant only for babies born to first-time mothers. The researchers also report that mothers in the homebirth group had slightly higher rates of prolonged labor and postpartum bleeding; again, this was statistically significant only for women having a first birth.
The five authors of this study are all physicians. One works in pediatric hematology and oncology, one is an obstetrician-gynecologist, and three are epidemiologists (one of these is also a professor of orthodontics). None of the five has direct experience with homebirth, and no midwives were included in the research group. The assumptions they made, as well as the outcomes and analysis they omitted from the study, illustrate their lack of experience in homebirth and their firm roots in the medical model of childbirth. This was confirmed for me when I spoke at length with Jenny Pang.
Faulty Assumptions in the Study Design Because Washington State birth certificates do not indicate whether a homebirth was planned, the authors attempted to isolate planned births by limiting the group to births of more than 34 weeks' gestation. According to Pang, this cutoff point was chosen because a baby born at 34 weeks or older would not necessarily need medical backup or transfer to a tertiary care center. However, midwives, as a rule, do not attend births of less than 37 weeks gestation; using 34 weeks as an initial cutoff gives the unfortunate impression that midwives are attending preterm births at home. The researchers' later analysis, which used cutoffs of 37 weeks and a birthweight of greater than 2,500 grams, may have eliminated some additional unplanned homebirths, but may also have skewed the results in favor of the hospital group, as I explain below.
In contrast to the Pang study, another study of homebirth in Washington State, by Janssen, Holt, and Myers, included a licensed midwife among the authors; the expertise brought to the study design by previous knowledge of homebirth is obvious to even a casual reader.7 The Janssen study concluded that licensed non-nurse midwives' outcomes in out-of-hospital settings were as safe as outcomes for physicians in hospitals or CNMs in or out of hospitals. [See sidebar for additional studies that show the safety of homebirth.] The study used a list of licensed midwives provided by the Department of Health to cross-check data from attendant codes on birth certificates, allowing them to state with confidence that no attendant in their study was misclassified. Pang and her colleagues took no such step, merely selecting records where the birth certificate listed a physician, nurse, or midwife as the attendant. They did not account for the possibility that anyone, regardless of training, experience, or licensing, could self-identify as a midwife.
The authors further restricted their study group by eliminating women who had one or more pregnancy complications listed on the birth certificate, assuming that women with those complications would be considered high-risk and therefore would not plan a homebirth. This is another questionable assumption, given that some of the complications listed (such as a previous large baby, previous preterm or low-birthweight baby, or herpes) do not always automatically exclude women from homebirth care. Midwives usually consider risk case by case, taking into account the history and entire health of a woman rather than simply relying on risk labels. Thus, some well-planned births may have been thrown out of the study group.
Pang told me that they were aware of this possibility, but because diagnosis and estimation of the risk of these complications are subjective, they chose to be conservative and exclude all complications. This seems fair, but may not be. More than 24 percent of pregnancies in the hospital birth group were excluded for one or more of these complications, while only 18 percent of homebirths were. Several of these complications might be overdiagnosed in the hospital group, or even caused by medical management; excluding them may obscure the negative effects of hospital birth.
Another telltale sign that the authors did not have a complete picture of homebirths lies in the rate they found of transfer from home to hospital. According to their data, the transfer rate was 4.8 percent. However, studies of midwifery practice indicate that a typical transfer rate for planned homebirth ranges from 8 to 16 percent.8 In a state where homebirth is well established and more prevalent than the national average, it is unlikely that the transfer rate for experienced, knowledgeable midwives such as one finds in Washington State would be so far below normal. In fact, another homebirth study found a high probability in Washington State of appropriate, selective transfer from home to hospital.9 Pang and her colleagues either may have missed a substantial number of homebirth transfers or may have assigned an inflated number of births to the planned homebirth group. If they had been familiar with homebirth, they would have recognized that the transfer rate they found was abnormally low, indicating a problem with their population.
Taking into account all of the potential ways in which the data may have been incomplete or misclassified, it is clear that the authors cannot state with confidence that they adequately controlled the makeup of their study groups. They are well aware of this fact, stating that "misclassification of any unplanned births as planned homebirths in this study would result in inflated risk estimates of neonatal mortality and other outcomes for planned homebirths."10 If the data going into the analysis are not accurate, the results will not be either.
Questionable Choice of Outcomes The choice of outcomes measured in this study is noticeably incomplete, ignoring the negative effects of hospital birth and medical management of low-risk labors. Pang et al. chose to include in their list of outcomes only neonatal death, postneonatal death (the period up to one year of age), very low Apgar score, assisted ventilation of more than 30 minutes, postpartum bleeding, and prolonged labor. They failed to report on any measures of other serious maternal and infant morbidity, such as cesarean section, induction of labor, forceps or vacuum delivery, maternal infection, perineal lacerations, total maternal blood loss, hysterectomy, use of pain medications, postpartum rehospitalization, meconium aspiration, neonatal infection, cord prolapse, placental abruption, uterine rupture, fetal distress, jaundice, or failure to breastfeed. Many of these outcomes are readily available on birth certificates, and other studies of homebirth have documented increased risk of maternal and infant mortality for hospital births in these categories.
Of all of the complications that the study's authors could and should have considered, cesareans have the greatest impact on immediate and long-term maternal and infant health. Existing homebirth studies already inform us that cesarean rates for women planning to birth at home are drastically lower than for women planning to birth at the hospital.11 The failure to account for the difference in women being unnecessarily subjected to major surgery suggests the authors' bias toward the belief that hospital birth is the standard of safety. The medical model of birth views cesarean section as an inherent risk of birth itself, rather than as an increased risk of hospital birth. But unnecessary cesareans are a serious problem in the hospital setting. In addition to the immediate higher risks of having a surgical birth, such as blood loss, infection, and death, cesareans confer lifelong higher risks to reproductive and general health. Cesareans also leave the next baby at increased risk of complications, including death.12
In our interview, Pang stated that the researchers considered reporting the differences in cesarean section rates but felt they would be unable to fairly compare the two because ascertainment of cesarean rates would be less reliable in the homebirth group. However, they did choose to consider the differences in rates of postpartum bleeding and prolonged labor. Complications such as cesarean delivery, postpartum bleeding, and prolonged labor are indicated by means of check boxes on the birth certificate form. The odds that the person filling out the certificate would accurately report subjective findings of minor morbidity but inaccurately report objective occurrences of major abdominal surgery is difficult to believe. At the very least, we know that the rate of transfer to the hospital was 4.8 percent; since cesareans are not done at home, we know that the average cesarean rate for the homebirth group could not have been higher than that, and was probably lower. In contrast, the overall national cesarean rate for the years of this study ranged from 20.7 to 22.8 percent.13
Out of the wealth of data available, the authors chose two outcomes that are not in and of themselves serious complications, but that seem to cast homebirth in a negative light. Women giving birth at home may have had longer labors (although the study provided no definition of "prolonged" labor), but women in the hospital were likely not allowed to have labors that exceeded "normal" length. A longer labor in the hospital is usually "treated" with Pitocin augmentation, instrumental delivery, fundal pressure, episiotomy, or cesarean section. Women in the homebirth group whose labors were longer than those in the hospital group probably had less morbidity as a result, certainly less iatrogenic (doctor-caused) morbidity. When asked what the significance of the finding was, Pang maintained that prolonged labor could have serious consequences in the presence of a positive culture for Group Beta Streptococcus (GBS). However, the researchers did not provide data on the incidence of GBS, nor did they report any deaths due to GBS. Two babies from the homebirth group and three from the hospital group died of unspecified infection/sepsis, but the difference in rates of death from infection was effectively the same. We can therefore assume that the prolonged labors in the homebirth group did not result in a higher rate of neonatal death from GBS or other infection.
Pang et al. also offer no information about the consequences of cases of postpartum bleeding; the information that women at home had higher rates of bleeding does not mean anything by itself. Although, Pang said, the authors would have liked to examine rates of postpartum transfusion instead, those data were not reported on birth certificates. But if postpartum bleeding does not result in complications like severe anemia, hysterectomy, or transfusion, it is just one more variation of normal. And, as in the case of prolonged labor, women in the hospital are routinely given injections of Pitocin in the third stage, making their rate of postpartum bleeding artificially low. Since the authors acknowledge that maternal complications of labor and delivery are reported with questionable accuracy on birth certificates, and since there was no reported clinical significance of the two complications they reported, their inclusion of these outcomes is questionable.
Deficient Analysis When the authors restricted their analysis to pregnancies of 37 weeks or more gestation and birthweights of 2,500 grams or more, they did not report on the resulting differences between the groups. By comparing the number of babies excluded from each group for low birthweight and/or preterm birth, we find that 81 (1.3%) of babies in the homebirth group were born between 34 and 37 weeks, or mildly premature. In the hospital group, 392 babies were born between 34 and 37 weeks. This preterm birth rate of 3.7 percent is almost three times the rate for planned homebirth babies. These births were excluded after restricting the group for complications of pregnancy; they occurred in an otherwise low-risk population. This apparent increased risk of prematurity for the hospital group was not noted or explained in the study. It may be due to chance, or could reflect babies forced out of their mothers' wombs early by induction or elective cesarean. In any case, Pang et al. may have unknowingly found an unstated risk for hospital birth of having a mildly premature baby. Mild prematurity has consequences: A recent study by Kramer et al. demonstrated that rates of infant death (from birth to one year) are increased threefold in the US for babies born between 34 and 36 weeks gestation.14 Excluding these babies from analysis means it is impossible to say whether the increased risk of mild prematurity and potentially higher death rates would have changed the results of the study.
Pang and her colleagues include information on the causes of death for each group. One cause that accounts for five of the deaths in the homebirth group is congenital heart disease, with approximately 1.7 times as many babies in the homebirth group dying from this complication. However, the authors fail to provide data on the overall rate of congenital heart disease in either group, including babies who did not die. Without the denominatorâ€”that is, without knowing how many total babies in the homebirth group had congenital heart problems to begin with the comparison to hospital birth is meaningless.
Another serious deficiency in the data is the distribution of deaths in the homebirth group. The Pang study does not state which deaths from the homebirth group actually occurred at home rather than after transfer to the hospital. In fact, according to Pang, the researchers did not check this, and thus could not report it. This is important because one of their contentions is that death from congenital heart disease and respiratory distress might be "expected to be amenable to prevention in the hospital setting."15 From that statement it appears that they assumed all deaths in the homebirth group happened at home. However, in another study from Washington State, which used some data from the same years as the Pang study, L. Cawthon found that, depending on the cause of death, 85 to 100 percent of the deaths in the Pang study homebirth group happened after transfer to the hospital.16 If this is true also for the Pang population, it would negate their implication that the deaths from congenital heart disease and respiratory distress in the homebirth group might have been prevented had the babies been in a hospital. Cawthon's data demonstrate that there is a good chance these babies probably were in the hospital when they died.
Pang et al. also failed to examine the hospital charts of the 38 babies who died in both groups. Had they done so, they could have excluded any babies who could not have lived no matter where they were born or what kind of care they received. Deaths unrelated to birth environment should not be included in an analysis of home vs. hospital birth. They might also have been able to find more information about the course of labor and delivery among the babies who died, shedding more light on what effect the birth environment had, if any. With such a small number of records to check, the task would not have been overly difficult. When asked, Pang stated that they lacked the resources to do so and that it would have introduced a new level of complexity to their study. Nevertheless, this information would have added a great deal of credibility.
Clearly, the potential for error in this research is great. In addition to making assumptions out of ignorance of homebirth practices, the authors made no attempts to cross-check their data. They freely admit that the study "has several limitations that are related to the reliance on birth certificate data. These include the potential for misclassifying unplanned home births as planned home births and for misclassifying various outcomes and covariates."17 In a study where the differences between two groups are measured in single digits and tiny percentages, the proper classification and interpretation of data are crucial. A few misclassified records could lead to very different conclusions. Moreover, when researchers omit clearly important data, even if by oversight or lack of resources, the results cannot be considered reliable.
Asking the Wrong Question Despite this flawed study, the existing research demonstrates as well as we can ever expect that homebirth is a safe and valid choice for mothers and babies. What needs to happen now, in both the world of research and in practice, is to accept what is known about the safety of homebirth and move on to determine what changes could make home - and hospital - births even safer. Currently, the controversy over the safety of homebirth actually makes it more dangerous. In states where physician groups have managed to make homebirth midwifery illegal, or where obstetricians refuse to provide reliable backup care to midwives, homebirth is not as safe as it could be. A lack of midwives in some areas makes it more difficult for mothers to find a well-trained, experienced homebirth attendant. Increasing the number of practicing midwives and ensuring a coordinated system of transfer for hospital care when necessary will add to the proven safety of homebirth.
We know, too, that hospital birth can be made safer by adopting the midwifery model of care, which has been shown to result in lower rates of intervention and better outcomes, regardless of setting. We already have a comprehensive blueprint for how to achieve better hospital birth: the Mother-Friendly Childbirth Initiative from the Coalition for Improving Maternity Services (www.motherfriendly.org).
1. J. W. Y. Pang et al., "Outcomes of Planned Home Births in Washington State: 1989-1996," Obstetrics and Gynecology 100, no. 2 (2002): 253-259.
2. American College of Obstetricians and Gynecologists, "Home Births Double Risk of Newborn Death," news release, July 31, 2002. Available online at www.acog.org/from_home/publications/press_releases/nr07-31-02-3.cfm
3. Reuters Health, "Home Births Linked to More Infant Deaths," Health eLine, August 5, 2002; available to subscribers at www.reutershealth.com/en/index.html 4. See Note 1.
5. See Note 1.
6. C. A. Burnett et al., "Home Delivery and Neonatal Mortality in North Carolina," Journal of the American Medical Association 244, no. 24 (1980): 2741-2745.
7. P. A. Janssen et al., "Licensed Midwife-Attended, Out-of-Hospital Births in Washington State: Are They Safe?," Birth 21, no. 3 (1994): 141-148.
8. P. A. Murphy and J. Fullerton, "Outcomes of Intended Home-births in Nurse-Midwifery Practice," Obstetrics and Gynecology 92, no. 3 (1998): 461-470.
9. L. Cawthon, Planned Home Births: Outcomes Among Medicaid Women in Washington State, Report 7.93 (Olympia: Office of Research and Data Analysis, Washington State Department of Social and Health Services, 1996).
10. See Note 1.
11. O. Olson, "Meta-analysis of the Safety of Home Birth," Birth 24, no. 1 (1997): 4-13
12. J. MacCorkle, "Fighting VBAC-lash: Critiquing Current Research," Mothering 110 (2002): 58-66.
13. S. Ventura et al., "Births: Final Data for 1998," National Vital Statistics report 48, no. 3 (2000); available online at www.cdc.gov/nchs/births.htm
14. M. S. Kramer et al., "The Contribution of Mild and Moderate Preterm Birth to Infant Mortality," Journal of the American Medical Association 284, no. 7 (2000): 843-849.
15. See Note 1.
16. See Note 9.
17. See Note 1.
Unless otherwise cited, all statements by Dr. Jenny Pang are from a telephone interview with the author.
Jill MacCorkle, BA, ME-PD, is a freelance writer and editor of The Clarion, the newsletter of the International Cesarean Awareness Network. She lives in Paris, France, with her husband, Sean, and their children: Griffen (5), born in a hospital, and Carlin (2), born at home.