If we manipulate the data enough, we can make it look like homebirth has a lower risk of postpartum hemorrhage.

Does homebirth decrease the risk of postpartum hemorrhage?

Now, a group of statisticians, Andrea Nove, Ann Berrington and Zoë Matthews, has published a paper in BMC Open (Comparing the odds of postpartum haemorrhage in planned home birth against planned hospital birth: results of an observational study of over 500,000 maternities in the UK) purporting to show that homebirth decreases the risk of postpartum hemorrhage. Too bad the paper violates many standards of statistical analysis in an effort to reach a desired conclusion.

Before looking at the paper, let’s consider what you would need to investigate if you wanted to find out whether homebirth has a lower risk of postpartum hemorrhage.

The first rule of statistical analysis is that you must compare like with like. So in comparing women who gave birth at home to those who gave birth in the hospital, you MUST:

  • Use contemporary data. Out of date data leads to out of date conclusions.
  • Use a standard definition of postpartum hemorrhage.
  • Use intention to treat at the start of prenatal care. Otherwise, homebirth mothers who become high risk during pregnancy will end up in the hospital group and lower the risk of homebirth group.
  • Include a comparison of the characteristics in both groups to make sure they are as similar as possible. For example, if you find out that women who choose homebirth have fewer pre-existing medical conditions, or are less likely to smoke, you must correct for those difference in the final analysis.
  • Assure that there is no difference in the way that postpartum hemorrhage is measured in the different settings.

The authors violated every one of these requirements.

The authors used data that was 12-24 years old.

Instead of looking at postpartum hemorrhage (defined as blood loss greater than 500cc) the authors chose to look at severe postpartum hemorrhage (blood loss greater than 1000cc).

According to the Royal College of Obstetricians and Gynaecologists (RCOG), although an estimated blood loss of at least 500ml counts as a PPH, in the UK a case should be considered an “emergency” only when the blood loss exceeds 1000ml. For this reason, the definition of PPH adopted for this analysis was the loss of at least 1000ml of blood.

The authors used intention to treat at the start of labor. Therefore, any women who developed high risk conditions during pregnancy were excluded from the homebirth group and added to the hospital group.

Women were classed as having intended a hospital birth if: (a) a hospital birth was intended at booking and the baby was delivered in hospital, or (b) a home
birth was intended at booking but the baby was delivered in hospital, and SMMIS recorded the change in intention as having taken place before labour commenced.

The authors assumed that midwives working alone at home accurately recorded blood loss. Although clinicians often try to downplay complications, in the hospital setting there are additional observers, raising the possibility that the blood loss estimates at homebirth were minimized compared to blood loss estimates at hospital birth.

It is notoriously difficult to estimate accurately the amount of blood lost during labour and delivery, and the normal method used (visual estimation) has been found to be inaccurate. However, there is no reason to suppose that the estimates in hospital were systematically higher than the estimates at home, so this inherent inaccuracy is unlikely to have biased the relative risk estimates when comparing home and hospital births.

Finally, the authors did not know which women had a history of postpartum hemorrhage. One of the biggest risk factors for postpartum hemorrhage is previous postpartum hemorrhage. A history postpartum hemorrhage is a contraindication to homebirth. Therefore, women with a history of postpartum hemorrhage would deliver in the hospital and thereby increase risk level of the hospital group.

So the authors did not compare the incidence of postpartum hemorrhage in women who intended to deliver at home with women who intended to deliver in the hospital. They compared the risk of severe postpartum hemorrhage in women who intended to deliver at home AND developed no complications in pregnancy (and probably had no history of PPH) with women who intended to deliver in the hospital, had a variety of complications PLUS women who intended to deliver at home but developed complications during pregnancy. That is NOT comparing like with like.

The authors claim to have corrected for some of these problems in their analysis. A characteristic of any good scientific paper is that it includes enough data so that the reader could perform the same analysis if desired. This paper does not include the most critical data: the incidence of risk factors in each group. For example, although women having a first baby have a much higher risk of postpartum hemorrhage than women having a second or subsequent baby, the authors fail to tell us just how many primips and multips are in each group.

The authors conclude:

This study aimed to compare the risk of PPH between those who intended a home birth at the end of pregnancy (whether or not they went on to experience a home birth) and those who had a planned hospital birth. It found significantly higher odds of PPH among those who had a planned hospital birth than among those who intended a home birth. This raises questions about the safety of hospital birth from the perspective of the mother‟s wellbeing.

Really?

Let’s look at some other variables that were highly statistically significant in this analysis. According to Table 3, other variables that resulted in highly statistically significant differences (P< .001) in the risk of PPH include giving birth to a girl, having more than one ultrasound in pregnancy and giving birth in 1999 or 2000. Moreover, giving birth between 4-8 PM was also a risk factor for PPH (P< .01).

Though the study demonstrated statistically significant differences in these factors, no one, least of all the authors, believes that means that gender, ultrasound or giving birth in the late afternoon causes PPH or that women should be counseled about the purportedly increased risk of PPH in association with these factors.

Why did the authors fail to use standard definitions, fail to use intention to treat at the start of prenatal care, fail to provide standard data on the risk profile of each group and conclude that place of birth was a risk factor while ignoring other variables that were also statistically significant? I suspect that it was because doing the correct analysis, using the correct data would not have shown homebirth to have an advantage. In addition, pointing out that their analysis also led to the conclusion that gender, ultrasound, and late afternoon birth increased the risk of PPH would have undermined the validity of the analysis.

The authors claim:

The results will provide further evidence to help pregnant women, their partners and maternity care providers to make a more informed choice about place of birth than has been possible with previously available evidence.

But that is completely untrue for four important reasons.

First, the authors never compared the risk of postpartum hemorrhage between women intending home birth vs. those intending hospital birth. They compared the risk of severe postpartum hemorrhage in low risk women attempting homebirth, with higher risk women intending hospital birth.

Second, the history of previous postpartum hemorrhage was not known and this almost certainly affected the difference in postpartum hemorrhage between the two groups.

Third,the fact that gender, ultrasounds and late afternoon birth appeared to increase the risk of PPH suggests that the analysis is not particularly useful.

Fourth, the authors did not demonstrate causation, as the authors themselves have acknowledged.

This result highlights a statistical association between intended place of birth and PPH; it does not prove a causal relationship, nor does it explain why the association exists.

Therefore, this paper is interesting as an observation, but tells us NOTHING about whether homebirth decreases the risk of postpartum hemorrhage. It will not help providers make a more informed decision about place of birth because it does NOT tell us whether place of birth affects the risk of postpartum hemorrhage.