NHS AI Lab Skunkworks project: Long Stayer Risk Stratification

A pilot project for the NHS AI Lab Skunkworks team, Long Stayer Risk Stratification uses historical data from Gloucestershire Hospitals NHS Foundation Trust to predict how long a patient will stay in hospital upon admission.

As the successful candidate from a Dragons’ Den-style project pitch, Long Stayer Risk Stratification was first picked as a pilot project for the AI (Artificial Intelligence) Skunkworks team in April 2021.

Background

Hospital long stayers, those with a length of stay (LoS) of 21 days or longer, have significantly worse medical and social outcomes than other patients. Long-stayers are often medically optimised (fit for discharge) many days before their actual discharge. Moreover, there are a complex mixture of medical, cultural and socioeconomic factors which contribute to the causes of unnecessary long stays.

This repository contains a proof-of-concept demonstrator, developed as part of a research project - a collaboration between Polygeist, Gloucestershire Hospitals NHS Foundation Trust, NHSX, and the Home Office’s Accelerated Capability Environment (ACE). The project aimed to achieve two core objectives:
firstly, to determine if an experimental artificial intelligence (AI) approach to predicting hospital long-stayers was possible; secondly, if so, to produce a proof-of-concept (PoC) risk stratification tool.

Stratification Tool

The tool displays the LTSS for a patient record, between Level 1 and 5; with 5 being the most severe risk of the patient becoming a long stayer. The tool allows exploration of various factors, and enables the user to edit those entries to produce refined or hypothetical estimates of the patient's risk.

The tool has shown good risk stratification for real data, with Level 1 consisted of 99% short stayers, and minor cases, with less than 1% of long-stayers being classified as very low risk. Moreover, 66% of all long-stayers were classified as Risk Category 4 and 5, with proportions steadily increasing through the categories. Risk Category 5 also stratified those patients with long and serious hospital stays under the long-stay threshold (serious and lengthy stays).

Documentation:

Docs	Description
REST API	API Endpoint descriptions and usage examples
LTSS Flask App API	Package documentation for the `ltss` Python package and incorporated submodules
Deployment Instructions	Build and run instruction for development or production deployments
WebUI Overview	Description of UI components and application structure
Configuration Files	Overview of provided configuration files
Production Build Configuration Files	Overview of the configuration files provided for production build Docker containers
Training	Description of the training process for the models used in the LTSS API