Cardiac intensive care: Machine learning to improve patient flow

ISRCTN ISRCTN10085738
DOI https://doi.org/10.1186/ISRCTN10085738
Secondary identifying numbers 1
Submission date
17/11/2017
Registration date
08/12/2017
Last edited
05/01/2018
Recruitment status
No longer recruiting
Overall study status
Completed
Condition category
Circulatory System
Prospectively registered
Protocol
Statistical analysis plan
Results
Individual participant data
Record updated in last year

Plain English summary of protocol

Background and study aims
Patient flow describes the movement of patients throughout the ward. Entering the ward, having surgery, being moved back to the CICU, and recovering over a period of around 5 to 10 days. Self-learning machines refer to machines that are capable of taking feedback into account. If a self-learning machine predicts that a patient will take five days to recover and then is informed that it was correct, it will strengthen its prediction algorithm. If it predicts a similar patient to take five days to recover but they instead take eight, it will investigate the differences between the patients more closely in order to determine the cause of its failure and change its algorithm in order to take this into effect. As this machine is trained on 35,000 patients, this will eventually lead to accurate predictions for many different types of individuals. The study aims to improve patient flow through the cardiac ICU (CICU) via analysis of patient recovery times. Self-learning machines will be developed to adjust to patients and predict accurate recovery times, allowing inefficient planning methods to be revised and fine-tuned in order to provide accurate bed, pharmaceutical (medication), and staff management.

Who can participate?
Patients in the CICU

What does the study involve?
This study uses a NHS database to access routinely collected data about those who are in the CICU and have had heart surgery. The data from 2009 until present is collected about patient flow in the CICU in order to train the self-learning machines.

What are the possible benefits and risks of participating?
There are no benefits or risks of participating

Where is the study run from?
Bristol Royal Infirmary Cardiac Intensive Care Unit (UK)

When is the study starting and how long is it expected to run for?
January 2017 to September 2020

Who is funding the study?
National Institute for Health Research (UK)

Who is the main contact?
Mr Duncan Shillan
ds17453@bristol.ac.uk

Contact information

Mr Duncan Shillan
Public

Bristol University
Senate House
Tyndall Avenue
Bristol
BS8 1TH
United Kingdom

Phone +44 (0)117 928 9000
Email ds17453@bristol.ac.uk

Study information

Study designObservational cross-sectional study
Primary study designObservational
Secondary study designCross sectional study
Study setting(s)Hospital
Study typeDiagnostic
Participant information sheet Not available in web format, please use the contact details to request a patient information sheet
Scientific titleApplication of machine learning to improve patient flow through the cardiac intensive care unit
Study objectivesThe study aims to improve patient flow through the cardiac ICU via analysis of patient recovery times. Self-learning machines will be developed to adjust to patients and predict accurate recovery times, allowing inefficient planning methods to be revised and fine-tuned in order to provide accurate bed, pharamceutical, and staff management.
Ethics approval(s)Not provided at time of registration
Health condition(s) or problem(s) studiedPatient flow
InterventionThere are no interventional components to this study. Machine learning systems are being developed in order to show hypothetical increases in patient flow throughout the ward (meaning that beds are at near full occupancy with some left for emergencies etc).

This study uses routinely collected observational data. These patients are all from the cardiac intensive care ward and will have had heart surgery. As this study only uses routinely collected observational data, absolutely nothing happens to the patient as part of this trial. The data is collected from gaining access to an NHS database. The data from the original state of patient flow from 2009-present. Data is taken to see if machines can learn to analyse data from the Cardiac Intensive Care Unit and that could improve patient flow management.
Intervention typeOther
Primary outcome measureBed occupancy in the Cardiac ICU (CICU). This is to be kept close to full, with room for emergencies, and with beds neither empty nor double-booked due to bad estimations of patient recovery times.
Secondary outcome measuresPrediction of potential complications in patients with preventative measures recommended to hospital staff.
Overall study start date01/01/2017
Completion date01/09/2020

Eligibility

Participant type(s)Patient
Age groupAll
SexBoth
Target number of participants35000, data from 2009 - present
Key inclusion criteriaPatients of the CICU
Key exclusion criteriaThere are no participant exclusion criteria
Date of first enrolment01/08/2017
Date of final enrolment01/08/2018

Locations

Countries of recruitment

  • England
  • United Kingdom

Study participating centre

Bristol Royal Infirmary Cardiac Intensive Care Unit
Bristol
BS2 8HW
United Kingdom

Sponsor information

University of Bristol
University/education

University of Bristol
Senate House
Tyndall Avenue
Bristol
BS8 1TH
England
United Kingdom

Phone +44 (0)117 928 9000
Email ds17453@bristol.ac.uk
ROR logo "ROR" https://ror.org/0524sp257

Funders

Funder type

Government

National Institute for Health Research
Government organisation / National government
Alternative name(s)
National Institute for Health Research, NIHR Research, NIHRresearch, NIHR - National Institute for Health Research, NIHR (The National Institute for Health and Care Research), NIHR
Location
United Kingdom

Results and Publications

Intention to publish date31/12/2020
Individual participant data (IPD) Intention to shareYes
IPD sharing plan summaryAvailable on request
Publication and dissemination planPlanned publication in a high-impact peer reviewed journal around the end of 2020.
IPD sharing planThe datasets generated during and/or analysed during the current study are/will be available upon request from Duncan Shillan, ds17453@bristol.ac.uk.

Editorial Notes

05/01/2018: Internal review.