Generating synthetic health data with an agent based model
Synthea is great if you have/ want really in depth patient records. But, sometimes you don't get stuff in that much detail. This is an attempt to generate synthetic health data for settings where you kinda don't have that much information.
Goal is synthetic health data in a rural community health setting. What I want to use this for is to generate fake patients so i can build tools without using real health data.
using maternal health outcomes as a test case. see dag.png
pregnancy_model.R is the current model. It's a fairly basic and straightforward agent based model. It also throws away a large fraction of its records in order to simulate real life missing-ness of data. The output right now is "one line per visit".
- add in age,
wealth, education,health visits & kids vars - add in danger signs + interaction with age + kids + health visits
- add deaths
- model each ANC visit?
figure out how to get it to give me the state of each individual at each timestep instead of the number of individuals in each state- read up on interoperability and put this into HL7 format lmao
- 2020-03-14 - pregnancies, births, miscarriages are modeled.
- 2021-02-09 - scrapped using individual package; modeling using other means instead for flexibility. I think this is necessary to generate data that looks like it could conceivably be medical record data.
- 2021-02-09 - added in facility births
-
- limitation: paper assumes no interaction btw variables. obvs not true given that wealth + education are usually correlated. but whatever this is the best data i have right now anyway