The labor window is the shortest span of calendar days that contains at least half of the remaining probability that labor begins on its own. Suppose the labor calculator shows a window of October 3 to October 11. The model assigns about 50% probability to labor beginning within those nine days and about 50% to labor beginning before or after them.
A 50% window is not a prediction that labor will begin inside it. The probability inside the window and the probability outside it are approximately equal.
Why the window uses 50%
A wider interval, such as 80% or 90%, would contain most outcomes but would be too long to plan around. Because spontaneous labor timing is widely spread, the span between the 10% and 90% dates calculated at 36 weeks covers more than three weeks in the calculator's baseline model.
A 50% window is short enough to be useful. In the baseline model for a first pregnancy, it covers about 12 days when calculated at 36 weeks. It narrows to about 11 days at 39 weeks and about 8 days at 40 weeks, as earlier days are removed from the forecast. The calculator also reports the 10% date, the day by which there is a one-in-ten chance that labor has already begun, for users who want an early planning date.
How the probability curve is built
The window is calculated from a probability distribution with two components.
The first component covers labor before 37 weeks and holds 5% of the total probability. This value is rounded from the 4.5% rate of spontaneous preterm delivery among U.S. singleton births reported by Gyamfi-Bannerman and Ananth (2014) for 2012. Answers about a previous preterm birth, IVF, smoking and other factors change the share held by this component.
The second component covers 37 weeks onward. It is a normal distribution positioned so that the median of the complete distribution falls on day 284 for a first pregnancy and day 282 for a later one. Those are the medians Smith (2001) reported for non-elective deliveries. The width of the distribution was set so that it also reproduces a late-pregnancy benchmark. In the cohort studied by Marquette and colleagues (2014), 67.6% of 15,253 uncomplicated pregnancies still ongoing at 41 weeks began labor spontaneously within the following week.
The window is then located by scanning the distribution for the narrowest span containing half of its area. The displayed dates are rounded outward, so the calendar range contains at least 50%.
Why the window changes from day to day
The distribution describes the whole pregnancy, but only the portion after today is relevant to the user. Each time the calculator runs, days that have already passed are removed and the remaining probability is rescaled to 100%.
Two effects follow. The window moves later as the due date approaches, because early days are removed from the calculation. A window of September 28 to October 6 at 37 weeks may become October 2 to October 9 one week later with no change in the answers entered. The window also narrows, because the remaining distribution is shorter and half of its area fits within fewer days.
Neither change indicates that the earlier forecast was incorrect. The forecast answers a conditional question: given that labor has not yet begun, when is it likely to begin?
What the model does not include
The model uses a small number of inputs. They are the due date or last menstrual period, whether the user has given birth before and when that labor began, age, BMI, smoking status, IVF, and miscarriage history. It does not use cervical examination findings, ultrasound measurements, current symptoms, or any information held by the user's care team. Those factors can change an individual forecast by more than any input on the form.
The model covers only spontaneous onset. Scheduled inductions and planned cesareans are outside it. If an induction is planned at 41 weeks, the model's probability for later days does not describe an event that will occur.
The spread in the forecast reflects biological variation, not only model limitations. Jukic and colleagues (2013) measured ovulation directly in 125 naturally conceived pregnancies and found a 37-day range in gestational length after excluding preterm births and related conditions. A calculator cannot narrow that range. It can only indicate where within the range the probability is concentrated.
Using the window
The window is the block of days to prepare for, and the 10% date is the day to be prepared by. The single most likely date is the peak of the distribution. In the baseline model the probability assigned to that one day is about 4% when calculated at 36 weeks and about 6% at 40 weeks.
For a fixed date such as a trip or a family arrival, the weekly probabilities on the result page give the chance that labor has begun by that date. That figure is more informative than whether the date falls inside the window. The methodology page describes the model in full, including each adjustment and its source.