The birthday problem

A friend with a baby who was born a couple weeks late commented that he would’ve liked a website that gave an estimated due date that was something more accurate than the usual “last menstrual period + 40 weeks.” I did a quick google and found this and this. Based on their descriptions of what information they use, the first site looks like it might be good, and the second site looks iffy. But I don’t really know.

8 thoughts on “The birthday problem

    • Thanks for posting this link; I was thinking of the same article. It helps drive home the more important point for expectant parents — natural variation in pregnancy length is large. Trying to get a more precise date is an interesting statistical exercise, but I’m skeptical that the models reduce the uncertainty very much. Understanding the uncertainty is what’s important for the parents to be.

      Length of human pregnancy and contributors to its natural variation
      Jukic, et al. Human Reproduction. 2013.
      http://humrep.oxfordjournals.org/content/early/2013/08/06/humrep.det297.full

  1. Pingback: Featured on Andrew Gelman’s website | babybirthday

  2. Ok, the barometric pressure is measured. Is there any hard data, in which way that should influence birth date? Is there anywhere on that page real data mentioned, according to which the modell is set up?
    A study is mentioned on blog post May 24, 2011 without simplest data like study size, let alone study protocoll. There is a link to more details but it is dead for me. In short, I can see no reason to put any belief in the babybirthday app whatsoever.

    • We applied the babybirthday app to a sample group of 30 women at around 37-38 weeks, and found its prediction to be about 11x better than the due date itself. Also, while small typical variations in barometric pressure are irrelevant, larger dips (of order ones that might occur 10 times per year) do impact rupture of membranes and delivery dates for both term and preterm gestations. The app also takes into account many other small factors like day-of-the-week and labor unit staffing which does impact delivery date (since delivery is about more than just labor and has to include scheduled inductions and planned cesareans also).
      Interestingly I sometimes see other calculations taking into account the phase of the moon, but although enticing, delivery data has not shown any true relationship.
      Thank you for your interest!

  3. I did try to get data in France, from social security, hospitals, etc. I have never been able to get data to study accurately pregnency length (even if data do exist). A colleague (statistician) told me that there might be surprises in Paris, where it is difficult to ‘book’ a bed in hospitals: some women do lie about the date to get a bed, get it, then – accidentally – deliver a few weeks in advance.
    I’d really love to get proper datasets (e.g. on the age of the mother, the rank of the baby, etc).

  4. An early ultrasound (well before 20 weeks) is pretty accurate (though you can detect variation in size even when the embryos are pretty small). Note the mode of gestational age is closer to 39 weeks using ultrasound dating, but it’s probably better to have a later date in mind anyway as waiting is difficult!

Comments are closed.