The main objective of this project is to predict clinical outcomes using EHR Data and modern ML/AI
To build the Docker image, first create a config.yml using the file in the docker-resources folder called config.example.yml as a guide. Then build the Docker image:
docker build --tag ampath-iit-prediction-model-<model_version>:<version> .
And run it locally:
docker run --rm -p 8000:8000 -v "$(pwd)/docker-resources/config.yml:/app/config.yml:ro" ampath-iit-prediction-model-<model_version>:<version>
The Docker container produced here provides an API to run the models. The API endpoint is paired with a corresponding SQL script that describes how to extract the predictors for the model from the live database.
For interruption in treatment, we treat patients as belonging to week-long cohorts, running from Monday to Sunday (for whatever reason, there are a small number of appointments scheduled on Saturdays and Sundays). Since appoinment date is not a factor in our prediction models, these cohorts are primarily logical groups. Note that cohorts are labelled with something like "2024-W03",
The idea of these cohorts is that the bulk of patients for a given week are gathered and scored on the Sunday the week before the week containing their appointment. For example, a patient with an appointment on June 26th, 2023 should have their results score on June 18th, 2023. Since patient appointments can be added on a rolling basis, we continue to update the week's appointments every day.
Encounters that have already had predictions generated are not re-scored, as the overall risk score should not change without an intervening appointment. This also means that we do not drop patients out of cohorts if their appointment date is moved to a different week. This should be discussed.