Thursday, March 27, 2008

New study showing high homebirth death rate is poorly done

I often complain about the fact that there is a lot of poorly done research in medical journals. That's why it is important to read the study, to determine if the conclusions are justified by the data. Here is a new study that purports to show that homebirth has a rate of intrapartum death that is high and rising. It is certainly possible that this is the case, but the study is based on so many unproven assumptions as to be essentially worthless.

The study is An estimation of intrapartum-related perinatal mortality rates for booked home births in England and Wales between 1994 and 2003 by Mori et al. in the April issue of the British Journal of Obstetrics and Gynecology. According to the authors:
Results: The overall IPPM rate for England and Wales improved between 1994and 2003. However, data to obtain a precise estimate of IPPM rate for booked home birth were not available. The average IPPM rate for all births in the study period was 0.79 per 1000 births (95% CI 0.77–0.81), and the estimated IPPM rate for booked home births was 1.28 or 0.74 per 1000 births, depending on the method of calculation (range 0.49–1.47). The IPPM rates for the completed home birth group appeared to be lower throughout the study period compared with the unintended home birth groups. Those women who had booked for a home birth, but later needed to transfer their care for a hospital birth, appeared to have the highest risk of IPPM in the study period.

Conclusions: The results of this study need to be interpreted with caution due to inconsistencies occurring in the recorded data. However, the data do highlight two important features. First, they suggest that IPPM rates for home births do not appear to have improved over the study period examined, even though rates did so overall. Second, although the women who booked for home births and had their babies at home seemed to have a generally low IPPM rate, those who required their care to be transferred to hospital did not. Women who book for home births should be offered comprehensive evidence-based information about the potential benefits, risks and uncertainties associated with their choice of birthplace by the healthcare professional responsible for supporting their decision. It is of considerable concern that the data recorded nationally in England and Wales do not provide accurate information about when and why a transfer from home to hospital booking occurs and about their outcomes.
These conclusions could be true, but the paper does not show that they are true. Indeed, the paper does not show much of anything.

In reading the paper, you should keep in mind that the authors use a new statistic, intrapartum related mortality.
IPPM rate is defined as deaths from intrapartum 'asphyxia', 'anoxia' or 'trauma', derived from the extended Wigglesworth classification 3, which is used by CEMACH. This includes stillbirths and death in the first week. The denominator was all births (live births and stillbirths).
What they appear to mean is perinatal mortality caused by intrapartum asphyxia. I have not seen this statistic used before in any other homebirth studies.

The initial comparison is straightforward: "A total of 4991 intrapartum perinatal deaths occurred in England and Wales between 1994 and 2003 among 6 314 315 births" for an IPPM rate of (0.79 per 1000 births [95% CI 0.77–0.81].In contrast, "[t]here were 125 intrapartum-related deaths among the 130 700 home births in England and Wales in this period" for an IPPM rate of 0.96 per 1000 births (95% CI 0.79–1.03).

Now comes a gigantic problem, or rather two gigantic problems. First, the homebirth group includes both planned and UNPLANNED homebirths (rather like the Pang study). Second, the homebirth group does NOT include women who planned to deliver at home but transferred to the hospital. Therefore, you cannot draw any valid conclusions from a direct comparison.

Here's where the authors get into trouble. As an editorial accompanying the paper explains what they did:
Mori et al. attempt to compensate for this by using figures for the proportion of deaths associated with planned and unplanned home births collected since 1994 by the Confidential Enquiry into Maternal and Child Health. Furthermore, they impute from previous smaller studies in various parts of England and Wales the proportion of home births planned to be at home but transferred to hospital because problems had developed in labour (this group also has a poor outcome). This enabled the authors to make an estimate, with confidence intervals, of the likely perinatal mortality associated with an intention to give birth at home. Although this should be a selected low-risk group, their perinatal mortality between 1994 and 1997 was no lower than that of hospital births and subsequently was consistently about 80–250% higher (significantly so in 1998–99 and 2002 03).
In other words, in order to arrive at their calculated intrapartum related mortality rates, the authors estimated BOTH the denominator (the number of intended homebirths) and the numerator (the number of deaths that purportedly occured in the estimated intended homebirth group).

When the authors claim that the purported intrapartum related mortality rate in the intended homebirth group is 0.96/1000, they are really saying this:

125 deaths + estimated transfers deaths - estimated unplanned homebirth deaths
______________________________________________________________ = 0.96
130,700 births + estimated transfers - estimated unplanned homebirths

That could be true, but since four out of the six variables are only crude estimates based on past data, it is just as like to be false. It could be lower, or it could be higher; there is absolutely no way to know. Moreover, since the numerator is so small, even tiny errors in estimation will have a big impact on the result.

The only part of the paper that we can be certain is accurate is this:
It is vital that data are collected prospectively so that an accurate picture can be established of both intended and unintended home birth rates, together with a clear indication about when and why a transfer to hospital care occurred. The fact that reliable data are not currently available to inform a key health debate is a matter of great concern.
Fortunately, a study is underway to correct this deficiency. The National Perinatal Epidemiology Unit in the UK is currently collecting data for the Birthplace Study, which is "designed to compare outcomes of births planned at home, in different types of midwifery units, and in hospital units with obstetric services." The results should be available next year.

Labels: ,

0 Old Comments: