SynPath - Toy Model

Data Science and Applied AI Team, Data & Analytics (D&A)

This project is in development, and should not be relied upon. Please do not use any code without consultation of the Data Science and Applied AI team (D&A).
Report Bug Request Feature

About the Project

This repository holds python code for a toy simulation of patient pathways to be optimised for both patient outcomes and system cost.

Figure 1: Schematic of the simulation showing a patient with attributes interacting with multiple randomly assigned pathways which each have randomly assigned overlapping actions

Note: Only public or fake data are shared in this repository.

Built With

• Core Python packages (e.g. numpy, pandas)
• Core python visulisation packages (e.g. matplotlib, networkx, seaborn)
• heapq for the priority queue implementation

Data

As this is a toy simulation, fake data is generated within the code using random numbers. It does not reflect any real world data or insights.

Usage

This code has been created to demonstrate some of the features required when optimising patient pathways. It is to be used as a demonstration only.

The code structure contained in 'project' includes:

• 'main.py'
• 'config.py' Config (default set to 10 patients, 10 pathways, 10 actions, 30 time steps)
  • At 10 patients this takes less than 1 second to run. At 1,000 patients this takes 1hr30.
  • 'patient.py' Patient Class (incl. age, sex, diseases, comorbidities, clinical values, sickness, outcomes)
  • The progress_disease method simulates disease occurence over time
  • The clinical_decay method simulates the patient getting sicker over time
  • The apply_action methods simulates the patient getting better due to an activity
  • The score_outcomes method calculates the queue and clinical penalities for an individual
• 'action.py' Action Class (incl. capacity, effects of the action on the patient clincial values, cost, duration, queue)
  • The update_capacity allows some dynamic changes in supply
  • The assign method deals with how the individuals are placed in a queue for the activity using heapq
  • The execute method
• 'pathway.py' Pathway Class (incl. valid transitions and thresholds)
  • The next_action method chooses from the valid transitions using q-learning to decide for each pathway which action to take dependent on the patient age-group and sickness and optimise over cost, benefit to the patient, patient queue time, and system overall queue length.
• 'build.py' Simulation Build (Creates a set of random actions randomly connected with transistion and threshold matrices to define possible links)
• 'run.py' Simulation Run
  • Runs for two major time steps to compare raw versus learnt systems
  • Runs for a user set number of steps
  • For each patient look at each pathway if the patient is on this pathway choose a next action for them to be added to the queue for. Calculate the outcomes and log the activity.
• 'vis.py' Visualisations of outcomes

The 'Experiments' Folder contains a full and minimium working example of the code as notebooks.

Outputs

Visualisations of the system to be generated as well as figures showing usage, queues and outcomes - saved on outputs/. The random seeds can be uncommented in the first cell to create reproducible results.

Contributing

Contributions are what make the open source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

See CONTRIBUTING.md for detailed guidance.

Contact

datascience@nhs.net

License

Unless stated otherwise, the codebase is released under the MIT Licence. This covers both the codebase and any sample code in the documentation.

See LICENSE for more information.

The documentation is © Crown copyright and available under the terms of the Open Government 3.0 licence.