Can we please stop with this due date nonsense already?
Yes, the due date is not a precise date, only an estimate. And yes, most women will deliver within 2 weeks of their so called “due date”.
But the due date has bigger problems.
Actual birth data, from real pregnant women, shows that the due date is not only an imprecise delivery estimate–as any single day would be–it is 1-4 days too early. Statistically-speaking, it is biased. And bias is bad.
If you knew nothing about due dates, you might reasonably suppose that the due date is the average date of delivery.
And you’d be wrong. The due date is NOT the average delivery day of a large, modern reference sample of pregnant women. The due date is also NOT the day on which pregnant women most commonly deliver (the mode). It is NOT the day by which half of pregnant women have typically delivered, and half have not (the median). No, the average, the mode, and the median delivery days are all 1-4 days after the “due date“.
So why care about the due date at all? Well, modern medicine has given us one very concrete reason for concern: Due dates act as a medical deadline. Go a week past yours, and you will be urged to have labor induced. Go two weeks past and an induction is all but assured. If you have severe gestational diabetes, and your due date will likely be treated as a hard stop, and your doctor will insist on an induction at the 40 week mark.
So what IS a due date?
The “due date” comes from a calculation known as Naegele’s rule, in which 280 days are added to the first day of a woman’s last menstrual period.
Naegele’s rule has proved remarkably enduring. Naegele, a director at a German hospital, developed his rule in the early 1800s, long before ovulation predictor kits, over-the-counter pregnancy tests, prenatal ultrasounds, fetal nonstress tests, pitocin–in short, before all of modern obstetrics. His calculation of pregnancy length relied not only on the duration of actual, observable pregnancies, but was also greatly influenced by biblical scholarship indicating that the gestation of Christ lasted for ten lunar months.
Not surprisingly, Naegele’s rule has some problems. The first, imprecision, is pretty obvious to people with ovaries: His rule assumes menstrual cycles are consistent in length, exactly 28 days long, with ovulation occurring precisely on day 14. Ovulation is less predictable than that. And many, many women’s cycles are shorter or longer than 28 days. Have a longer cycle, and the due date will likely underestimate the real delivery date. Have a shorter cycle, and the due date may overestimate it.
The second problem, bias, is far more concerning: The average actual delivery date is later than the one predicted by Naegele’s rule. This is especially true for first-time mothers. In a study of 114 Caucasian women, women with 1 or more deliveries under their belts delivered an average of 3 days after their due dates; first time mothers delivered an average of 8 days after their due dates. Another study of 1,514 women with reliable conceptions dates found that first-time mothers went an average of 4 days “late;” women with a previous birth went 2 days “late.”
Other large studies back up these findings: Naegele’s rule consistently underestimates the average delivery date by 2-4 days (see here).
Note, though, that estimates apply best to Caucasian women. For unknown reasons, women of Asian or African ancestry tend to give birth about a full week earlier than do Caucasians.
Is there a more accurate way to calculate a “due date”?
Yes. One way is to shift Naegele’s rule. Adding 282 days to the first day of the last menstrual period, instead of 280, provides a more accurate estimate. But, if available, an estimate based on a first trimester ultrasound is the best option. Using ultrasound-based due dates, 92% of pregnancies delivered within the normal range of 37-42 weeks; using due dates, this percentage fell to 87%.
One day off. Two days off. Who really cares? Actually, this bias matters a lot. An estimate off by even a couple of days significantly raises the percentage of women regarded as post-term (past 42 weeks), who are then urged to have inductions.
Consider the results of a UK study of over 24,000 women. The researchers studied inductions for post-term (past 42 week) pregnancies. When evaluated by ultrasound dates alone, they found the majority, 71.5% percent, of these pregnancies were not actually post-term.
So when are pregnant women actually most likely to deliver?
In her book Expecting Better, Emily Oster used the 2006 annual U.S. data for singleton births, from the Centers for Disease Control, to try to answer this question. She noted that in 2006 the most common week to deliver in was the 39th; about 30% of births occurred then. The data from 2012 look similar to those from 2006, but hint: there’s a problem with her approach.
What’s the problem? These data are fine for figuring out when a pregnant woman is most likely to deliver, assuming she is agnostic about how she delivers. But they are not particularly useful for figuring out when spontaneous labor is most likely. This is because the CDC birth data includes all deliveries. The CDC combines spontaneous births with scheduled inductions and c-sections, and these are typically scheduled for the 39th week of pregnancy.
Are there really enough scheduled inductions and c-sections to affect the distribution of births for the whole U.S.? Absolutely. Since 1990, the rate of medical inductions has more than doubled. In 2012, more than 1 in 5 pregnant women were induced, compared to just under 1 in 10 in 1990. In 2012, 32.9% of women had c-sections, a 60% increase from 1996, and about half of these c-sections were scheduled.
To see how big the effect is, compare the CDC’s singleton birth data for 2012 with the same data from 1990:
Scheduled inductions and c-sections shift the bulk of the distribution to the left. In 1990, slightly more women delivered in their 40th week of pregnancy than in their 39th week, and about 10% made it to their 42nd week. In 2012, most women delivered in their 39th week, and only 5% delivered in their 42nd week.
This comparison is not perfect. In 1990, scheduled inductions and c-sections were commonplace, albeit less frequent than in 2012. Still, the 1990 data are closer to what would occur without medical intervention and are consistent with other historical data. A large Swedish study of births found that 10% of women go past 42 weeks. The distribution of births in Canada in 1972 and 1986 shows a similar pattern:
Now that I am full term, how likely is it that I will deliver in the next week?
As a full-term pregnant woman, the above graph did not address my central question: when was someone like me most likely go into labor? Predicting this using the average delivery day, while a superior approach to using the due date, is still suboptimal. This is because the distribution of births is skewed. In 2012, preterm births accounted for just under 12% of all U.S. births. These preterm births create a long tail that pulls the average delivery date to the left.
Moreover, by 37 weeks, I knew I was not going to deliver in my 35th week. It would be nice to update my delivery estimate with that knowledge. Boiled down to a simple question, what I wanted to know was: How likely am I to go into labor within the next week, considering that I have not yet given birth? Statisticians would call this the conditional probability of giving birth in a given week.
For the sake of comparison, the table below shows the conditional probabilities for 1990 and 2012.
*Note: the CDC data combine week 37-38 deliveries. The probabilities shown for week 37 are therefore for delivery over the next two weeks.
Until past the 40-week mark, the odds of labor in the next week are less than 50%.
Because it is standard medical practice to induce labor at 42 completed weeks, the chance of delivery by 42 weeks appears to be 100%. Of course, many of these births are not spontaneous, but induced. From historical data, it appears that without medical intervention 5-10% of women would deliver after 42 weeks.
Like this post? Don’t forget, sharing is sweet!