Challenge 2: Improving end-of-life care

Challenge Statement: How can the health system intervene to improve the quality and choice of end-of-life care?

Supporting Mentors

Dr Andy Papa-Adams – Chief Medical Officer and GM Member Outcomes HBF

Ian McCrum – Manager Hospital Network HBF

Vidhatri Lakkim Setti – Head of Customer Communications Management HBF

Honzik Jurza – Data Scientist HBF

Billy Martin – GM Architecture & Innovation HBF

Potential Focus Questions 

This is a complex Challenge and you may find it useful to consider the following questions when building out your prototype:

  • How can we better predict those who are at end of life (less than one year to live)?
  • Can we predict what services certain conditions will necessitate for a particular patient at end of life?
  • How can we identify care that makes no difference to outcome; eg ICU stays, last line chemotherapy chemotherapy?
  • Is there any evidence that completion of an Advance Health Directive could or does lead to higher quality care at end of life?
  • What is the cost of healthcare in the last year of life and what kinds of care contribute to that cost?
  • Who and what benefits can be achieved with improved end-of-life care?
  • Is better end of life care provided by palliative care physicians compared to other specialists?

Critical Concepts 

 

Of the 94,800 palliative care-related hospitalisations, almost 3 in 5 (56%) ended in the patient dying in hospital in 2021–22. 80% of people who have expressed a need to die at home, 20% of them had this need meet. (Cancer Council Australia in a study on advance cancer patient).

Identifying palliative care (including end-of-life care) in existing data collections and health settings remain a key issue, particularly for care delivered in community, primary care, and residential aged care settings. For example, limited national data are currently available on community-based palliative care services and Medicare Benefits Schedule (MBS)-subsidised services provided by general practitioners and non-palliative care medical specialists.

You may want to talk to your Challenge Mentors about the following ideas as you developing you prototype:

  • How patients at end of life are currently managed.
  • What the key drivers are of poor experience at end of life.
  • How we currently try to predict when someone is at end of life.
  • Whether end of life care is managed more efficiently and efficaciously by certain doctor groups.
  • Whether we can build a machine learning model to better predict end of life care needs and when a patient is likely to die.
  • Model validation methodology.
  • Outcomes of the machine learning model to be potentially applied in clinical care.

 

Supporting Data Sets 

The HBF Health Hospital Episodes data contains data both patient demographic and diagnostic classification data for individuals who experienced a hospital episode and claimed for said episode through their HBF Health Insurance policy and includes data from July 2014 to August 2024. This data is real world data and as such certain sensitive values, such as customer and hospital identifiers, have been hashed to ensure the privacy of our members. If any of you are unfamiliar with hashing, this website is a great introduction. The data set contains data covering a wide range of hospital admissions and not just the episodes relating directly to the question your team will be tackling, this is by design.

Other variables include;

  • Financial data on the hospital stay:
    • Amount charged by Hospital
    • Amount Charged by Specialist
    • Amount paid by HBF
    • Number of services received
    • Allied health components
  • Product Information:
    • The Hospital product the patient held at admission
    • The Extras product the patient held at admission
  • Patient data:
    • Age at admission
    • Postcode at admission
    • Sex

 

Other data

Potential Solution Pathways 

You are free to resolve this Challenge by developing your prototype in whatever means you may like. Our mentors, partners and organising team have thought of the following techniques as being viable methods to resolve the Challenge:

  • Predictive models to provide a forecast for palliative care demand.
  • Simulation models of patient journey to end of life.
  • Create personalised care plans by analysing patient histories, preferences, and outcomes
  • Predictive models can identify patients who may benefit from palliative care earlier
  • Provide decision support to healthcare providers, helping them make informed choices about treatments and interventions based on the latest evidence and patient data