I have a set of data with information about women and their expected delivery dates for childbirth.

I have more columns in the table but for simplicity let's just focus on the below and assume that my table is as follows with hundreds/thousands of rows like the example:

Patient ID Expected Delivery Delivery Risk
1 11/11/2022 Medium

I would like to predict going forward the number of expected deliveries on a given day and also the risk involved in those deliveries.

I feel like there must have been some studies into distributions of deliveries around the due date, maybe it is normally distributed around the due date? Or perhaps it is different and women are more likely to give birth after due date for example.

I am very much in the investigation stage at the moment but wondered if anyone may have some suggestions in for the following questions

  1. How does the distribution look like if it has been investigated/modelled?

  2. How might one apply such a model to the dataset I have? I could implement in Python for instance? Or maybe it doesn't need to be so advanced and it could be done in excel.

  3. How might the output look for such a model - ideally I would like to be able to point to a specific day and say things like - look based on the model we are expecting more deliveries today or more higher risk deliveries, or today looks like it will be less deliveries than yesterday and they will be lower risk.

Any advice or suggestions would be welcome at this stage, thank you.

  • $\begingroup$ Does your data contain both the expected date and actual date? (I assume it's past data). If not, there's not much you can do because you don't have the true value. If so the only thing you could is to count the number predicted any particular day, $\endgroup$
    – Erwan
    Nov 4, 2022 at 15:41
  • $\begingroup$ yes I have actual date too $\endgroup$
    – Tom
    Nov 4, 2022 at 15:47
  • $\begingroup$ Then you at least can easily calculate the diff $|expected - true|$ date, and observe the distribution of this value in your data. $\endgroup$
    – Erwan
    Nov 4, 2022 at 18:34
  • $\begingroup$ Yes but how would I go about then using this distribution to predict the future? $\endgroup$
    – Tom
    Nov 7, 2022 at 13:45
  • $\begingroup$ Your question 1 was about the distribution, hence my comment. For prediction you need features which represent information likely to help knowing the value of the target. Apparently you have 'risk', but do you have any other feature? Anyway even an expert wouldn't be able to predict the date of an individual delivery very well, a lot depends on chance. However you can probably predict the approximate number of deliveries by day, this is a standard time series problem: there might be a global trend and a seasonality. $\endgroup$
    – Erwan
    Nov 7, 2022 at 15:07


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.