Limiting to an age range efficiently

[robinsonha]

Every attempt our group has made to limit to an age range fails to circumvent a design flaw that Synthea will still generate patients outside of that age range and just discard them as it goes along. We need to fulfil a computationally large (and therefore expensive) generation task monthly, has anyone found a way to coerce patient generation within an age range at the point of generation? We have tried removing columns from demographics.csv, editing the synthea.properties file and editing generator.html but nothing has resolved this so far. All suggestions very welcome.

[jawalonoski]

"Design flaw" is in the eye of the beholder.

You can try using -a and --generate.only_alive_patients=true together in combination (see below), however it won't entirely eliminate discarded patients (i.e., those who are born in the appropriate years, but die in the simulation before the present day), but at least they won't be exported in your final data set.

./run_synthea -a 40-50 --generate.only_alive_patients=true -p 1000000

Sometimes the simulated patients die, due to the stochastic elements of the models. You call this a design flaw, but it is a central part of the system. If we just forced everyone to survive and live through everything, then it wouldn't be a very good model of disease.

[robinsonha]

Thank you, I do not mean this in a critical way necessarily but not having control over age at the generation level does limit efficiency in creation of a cohort with a defined age range. We do not wish to bias our dataset by including only patients that are alive, would this be a consequence of filtering in the manner you suggest? I would expect the majority of patients under age 65 to still be living, therefore they would continue to be generated before being discarded from the final cohort.

[robinsonha]

Could what we are aiming for (a lower age limit of 65+yrs with no upper limit) be achieved using a birth date limitation in an altered base code file given that a date range can be specified for records?

[jawalonoski]

If you omit only_alive_patients you will get a mix of alive and dead patients. With only a specified age range over 65 (e.g., -a 65-110), you should see a distribution something like this (for the chart, I rounded all the ages down to the nearest decade):

Yes, there will be some dead patients under 65 that you'll have to discard (less than 15% of the total). You should be able to filter them out by post-processing the data based on the birth date and whether or not they are alive.

We could provide a method to strictly filter based on provided age ranges, but that is not currently supported, unless you want to specify that criteria in a keep module.

Apologies if this doesn't meet your requirements.

[robinsonha]

Hi Jason, Thanks for the suggestion, this may work, assuming the alive filter filters at the point of the defined start date rather than the end date. We are modelling claims data, so death and the surrounding costs is still relevant and needs capturing. I'd like to work on this more when I have more time over the next few weeks. There are probably a few ways to insert a workaround for this, I'd like to try and resolve it and will share if I come up with anything useful. Best wishes, Heather

…

On Wed, Oct 15, 2025 at 9:52 PM Jason Walonoski ***@***.***> wrote: If you omit only_alive_patients you will get a mix of alive and dead patients. With only a specified age range over 65 (e.g., -a 65-110), you should see a distribution something like this (for the chart, I rounded all the ages down to the nearest decade): ages.png (view on web) <https://github.com/user-attachments/assets/71fa7211-1725-41f2-9336-74dc76cfe86d> Yes, there will be some dead patients under 65 that you'll have to discard (less than 15% of the total). You should be able to filter them out by post-processing the data based on the birth date and whether or not they are alive. We could provide a method to strictly filter based on provided age ranges, but that is not currently supported, unless you want to specify that criteria in a keep module <https://github.com/synthetichealth/synthea/wiki/Keep-Patients-Module>. Apologies if this doesn't meet your requirements. — Reply to this email directly, view it on GitHub <#1624 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A6CZ7GVE4KIKWHCZSHITOD33X2XYRAVCNFSM6AAAAACJI3AG72VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTINRZGE2DQMY> . You are receiving this because you authored the thread.Message ID: ***@***.*** com>