Primary supervisor
Jackie RongCo-supervisors
- Lachlan Dalli
- Monique Kilkenny
- Eric Kuo
Description:
Acute patient length of stay is the leading contributor to hospital costs in Australia. The Victorian Auditor General has reported that ≈145,000 extra bed-days could be made available if all hospitals managed length of stay more efficiently. Such optimisation of length of stay could translate into $125 million in annual savings for Victorian taxpayers.
In the context of stroke, length of stay is thought to be largely influenced by disease-specific factors (e.g. stroke severity, post-stroke complications, and need for intensive care). However, length of stay varies substantially between hospitals and new tools are needed to support more effective allocation of hospital beds for patients with stroke.
Background:
The accurate estimation of length of stay in hospital for acute stroke is crucial for understanding medical costs and planning patient care. Prolonged length of stay not only impacts healthcare expenditure, but also impacts accurate allocation of healthcare services, particularly in optimising complex stroke treatments for all patients. Delaying the discharge of patients from hospital with stroke can also adversely affect patient experience and perceptions of healthcare. There is a strong impetus to more effectively manage length of stay within hospitals to ensure more efficient and responsible allocation of healthcare resources.
According to the Australian Stroke Clinical Registry, the median length of stay for patients with acute stroke is 5 days (interquartile range: 2-9 days). However, variation in length of stay is observed by type of stroke, hospital location, and whether specialist stroke unit care is provided. The emergence of Artificial Intelligence (AI) now provides new methods to more reliably and comprehensively predict length of stay for stroke. However, research in this area to date has been largely limited to single hospital studies, with small sample sizes. By accurately predicting length of stay for large and diverse populations, we can potentially reduce hospital costs and facilitate more personalised in-patient planning tailored to individual patients.
Student cohort
Aim/outline
This project aims to leverage AI to develop a new clinical risk prediction tool to support hospital clinicians to more accurately estimate length of stay for patients with stroke based on patient, clinical, and system factors.
Project activities for Masters placement:
The Masters student will work on one of the projects within the Big Data, Epidemiology and Prevention Division of the Stroke and Ageing Group and will be expected to undertake some of the following activities as part of their placement:
- Undertake data processing, verification and management processes to ensure the integrity of data for analysis
- Performing statistical analyses as instructed
- Data entry and data cleaning, as required
- Undertaking literature reviews, assisting with the preparation of grants and ethics submissions, as required
- Contribute to the production of conference and seminar papers, and publications, deriving information and input from research undertaken
- Attend and participate in staff meetings and seminars
- Perform other research activities as directed by the research fellows and chief investigators
What type of activities will the student undertake?
- Undertake a literature review on AI and length of stay using Covidence and EndNote.
- Enhance your statistical analysis package (e.g. Python, R, and Stata) to write do files for data analysis and reporting.
Potential project outputs:
- Report and presentation
- Potential publication
URLs/references
Required knowledge
- Proficient in Python (preferred) or R-studio programming
- Solid understanding of machine learning and data analysis modelling
- Basic understanding of deep learning
- Aptitude for learning about the Australian Health System and health conditions such as stroke, pneumonia