2021 Healthcare Industry Challenges - Using Specialised Datasets

WA Health Hackathon Week

Using Specialised Datasets

Participants will have the opportunity to explore specialised datasets provided by each Challenge Owner while working on solutions for the below challenges. These datasets have been provided solely for use during the WA Health Hackathon 2021. Participants who select and are subsequently assigned to work on these challenges will receive access to the specific dataset(s) provided by the Challenge Owner on Monday 16 August 2021, until Sunday 22 August 2021. After this period, access to the dataset(s) will be revoked.

Participants who work on one of the following challenges will be required to sign a version of this Participation Agreement prior to commencing the Hackathon.

Collaboration Agreement – Specialised Datasets Provided

Iron Woman: A Risk Prediction Model

Iron deficiency is common and the leading cause of anaemia worldwide. It is particularly common in women but often not recognised and undertreated. This can lead to symptoms of fatigue, tiredness, reduced cognition and depression as well as physical symptoms of muscle aches, palpitations, and headache.

Can you develop a risk prediction model that incorporates diet and specific symptoms (fatigue, tiredness, reduced cognition, depression, muscles aches, palpitations, headache, and heavy periods)?

Domain Expert: Prof. Toby Richards, Director at the Iron Clinic & Lawrence-Brown Professor of Vascular Surgery at the University of Western Australia



The Heat Map of Brain Injury

Clinicians recognise typical patterns of burn injury from the end of the bed – a child who walks through hot coals, a young man whose BBQ explodes on his hands, a frail elder who scalds their thighs with a dropped cup of tea. We recognise these will need different treatment but expect different outcomes, but these patterns are lost in our current predictive algorithms that focus on age and size of injury. Burns care is very resource intensive requiring multi-disciplinary input from surgical and theatre staff, nursing, physiotherapy, occupational therapy, psychology, social work, and others with both inpatient and long-term outpatient care components. Improving predictive algorithms for key outcomes (length of stay, time in care, intensity of care, balance of intensity early and late) will guide appropriate allocation of resources at both individual patient and service levels during both routine care and disaster management.

Can market research tools such as clustering, and segmentation identify these clinical phenotypes to improve predictive algorithms?

This challenge was provided by the Burns Service of WA.

Domain Expert: Prof. Fiona Wood, Director of the Burns Service of WA

Children's Diabetes Portal

There are currently more than 1200 Western Australian children diagnosed with diabetes and if not well-controlled, many will go on to develop costly long-term complications. Children attend clinic four times per year but every other day of the year, families are managing the condition on their own, making an estimated 180 health-related decisions per day. Good self-management is vital, and this can only be achieved if patients are informed, motivated, and educated.

Can you create a product to enable people living with diabetes to access their own clinical data, upcoming appointment details, personalised learning opportunities and have a direct communication channel with healthcare professionals?

This challenge was provided by Children and Adolescent Health Services (CAHS).

Domain Expert: Helen Clapin, Project Manager for the Australiasian Diabetes Data Network (ADDN)

i-Rehab Data Challenge

Data is currently collected by clinicians on paper forms during clinic, often transcribed by one clinician while another is examining the patient or talking to the family/carer, so there is a risk to accuracy and quality of data collection. Data collection is not standardised across all clinical areas, so gaps exist. Data collection is not standardised for data entry, so inconsistencies exist (for example data for the same diagnosis may be described in a variety of ways). Existing processes and forms have reduced but not resolved this problem, and inefficiencies remain embedded in the process. Delays in completion of forms, or on paper forms that go missing, means data is not always available in a timely manner for clinical care. Without this pipeline of accurate, timely, reliable data the next steps in developing a learning health care system are compromised.

Can you improve data collection volume, quality, timeliness, accuracy, and standardisation for the Paediatric Rehabilitation Information System (PRIS)?

This challenge was provided by Children and Adolescent Health Services (CAHS).

Domain Expert: Sue-Anne Davidson, Manager of Kids Rehab WA

PCH Outpatient Clinic Scheduling Challenge

Outpatient (OP) bookings at PCH are scheduled and allocated to physical spaces manually. This process relies on departmental clerical staff to administer the routine OP bookings via spreadsheets and the manual creation of MS Outlook calendar entries for staff and bookable resources. This effort is replicated across all PCH departments. OP Clinic bookings include specialties such as Surgical, Medical, Allied Health and Clinical Research.

Can you create a product to reduce manual handling for outpatient clinic scheduling?

This challenge was provided by Children and Adolescent Health Services (CAHS).

Domain Expert: Jo Fleming, Clinical Nurse Manager


Predictive Modelling for Patient Treatment

Medicinal cannabis is unregistered and there are no best practice care models in dosing or decision making, or in determining patient outcomes. Emyria has been prescribing medicinal cannabis for nearly 3 years, and during that time has collected regular validated data in a clinical trials database and now has millions of data points from thousands of patients treated in their clinics, including patient demographics, patient and clinician reported outcomes, vital signs, adverse events, and concomitant medicines, among others. This data can be used to generate digital biomarkers and can be used predictively in determining better models of care for individual patients or patient cohorts. By finding digital biomarkers within the available data, we believe that we can provide better individual patient care, ensuring better outcomes for all patients and helping our doctors make smarter and more informed treatment decisions and setting the gold standard for clinical care in this field.

Can you use existing data to general digital biomarkers as part of a predictive model that could be used to determine better models of care for individual patients or patient cohorts?

This challenge was provided by Emyria.

Domain Expert: Sebastian Roth, Behavioural Economist at Emyria

Patients Insights Dashboard

Emyria’s doctors have been prescribing medicinal cannabis to patients for the past 2.5 years. During this time, Emyria has collected standardised and clinically validated patient and clinician outcome measures in order to track patient response to treatment over time. Our patients come to us because we care about monitoring outcomes. We are not simply prescribing medicinal cannabis like many other clinics out there; we are tracking patient outcomes and adverse events from baseline through their treatment, and it would be amazing to show that data back to the patient. Patients will gain a greater understanding of their own outcomes but will also learn more about how well medicinal cannabis works for their condition in general.

Can you create a Patient Insights Dashboard using existing data to allow patients to tracked progress against their own baseline?

This challenge was provided by Emyria.

Domain Expert: Sebastian Roth, Behavioural Economist at Emyria

Tracking Patients of Concern

After hours identification & monitoring of patients at risk of deterioration is often a slow, innacurate and manual process. This creates a number of challenges, including reducing the ability of clinicians to share data on patients, and reducing the hospital’s overall ability to track activity or trends to gain insights.

Can you use open source data to create a model that would identify patients potentially at risk of deterioration? Then can you develop a user-friendly interface to give clinicians insights into patients at potential risk of deterioration?

Domain Expert: Dr Tim Bowles, Head of Department at Health in a Virtual Environment (HIVE), Intesive Care Unit Consultant at Royal Perth Hospital (RPH) and Head of Department – SAFE Team

Critical Care Referral Outcome Tracking

Hospitals admits significant numbers of patients per year to the Intensive Care Unit (ICU), at significant cost. Currently limited feedback exists for clinicians as to the long term outcomes of these patients, to inform decisions around timing and suitability for admission. By linking 6 and 12 month outcomes, in terms of mortality and frequency of hospital readmission, the scarce ICU resource can be better allocated. Intensive Care is an expensive and highly limited resource. At the moment, patients are admitted to the intensive care unit based on clinician assessment, which is informed by experience, clinical assessment, and a global assessment of the patient’s ability to benefit. However, we know that a significant number of patients suffer long term disability as a result of ICU admission, which may affect their decision to receive this treatment. Equally, there may be people who would benefit from ICU, who may be excluded on the basis of perceived poor outcomes.

Can you link 6 and 12 month outcomes, in terms of mortality and frequency of hospital re-admission, to better inform decisions about admissions to ICU?

Domain Expert: Dr Tim Bowles, Head of Department at Health in a Virtual Environment (HIVE), Intesive Care Unit Consultant at Royal Perth Hospital (RPH) and Head of Department – SAFE Team

End of Life Prognosis

Due to difficulties in identifying which patients are in the last 6-12 months of their life, many of these patients do not receive well-coordinated, high quality palliative care. Palliative care has been increasingly shown to improve patient outcomes in symptom management, quality of life and patient and family satisfaction. The World Health Organisation recognises palliative care as a crucial part of integrated, people-centred health services. They acknowledge that relieving serious health-related suffering, be it physical, psychological, social, or spiritual, is a global ethical responsibility. Involvement of palliative care is long proven to have a direct effect on patient outcomes improving symptom management, quality of life and the coping abilities of care providers. In addition to the improved symptom control and quality of life, early involvement of Palliative Care empowers patient to make decisions and have discussions about what is important to them in the lead up to their death. Having an Advanced Health Care Directive or an Advanced Care Plan lessens the emotional and psychological toll on the patient, their carers and families as the patient’s wishes are clearly defined. In addition, a recent Journal of the American Medical Association article showed that palliative care intervention led to improvements in caregivers coping and depression symptoms.

Gold Standards Framework Prognostic Indicator Guidance (GSF PIG) is a tool for clinicians to support earlier recognition of patients nearing the end of life. It describes a number of factors both general and disease specific which suggest patients may be nearing the end of their life.

Can you apply a set of clinical indicators (GSF PIG) to a patient mortality dataset to create a predictive model for patients who are in the last 12 months of life? These use this to provide insights to facilitate consistent and early involvement of Palliative Care Services?



  1. Gold Standard Framework Prognostic Indicator Guidance (GSF PIG) https://www.goldstandardsframework.org.uk/
  2. Ocallaghan, Anne & Laking, George & Frey, Rosemary & Robinson, Jackie & Gott, Merryn. (2014). Can we predict which hospitalised patients are in their last year of life? A prospective cross-sectional study of the Gold Standards Framework Prognostic Indicator Guidance as a screening tool in the acute hospital setting. Palliative medicine. 28. 10.1177/0269216314536089. 

Domain Experts:

  1. Dr Derek Eng, Palliative Medicine Physician at Royal Perth Hospital, Head of Department Palliative Care at St John of God Subiaco Hospital.
  2. Dr Eva Jones, Resident Medical Officer

Visualisation of hospital performance in real time for the Armadale Health Service (AHS)

Timely intervention in hospital logistics that includes access block, patient demand, resource management before ambulance ramping occurs at Armadale Health Service (AHS) is a real challenge. We need early detection of this information in one place that directs the hospital to the areas of concern relating to Western Australian Emergency Access Target scores, ED ramping status, admission and discharge streams from inpatient services as well as the ED, Operating Theatre performance, inpatient bed capacity issues, staffing profiles and deficits and specifically for a general hospital such as AHS, outlier patient inward and outward demand in a simple dashboard utilising datasets already available. We need to integrate and visualise this data with a resource that can provide a visual cue that focuses attention on demand pinch points in real time, giving back time to key decision makers that would otherwise be spent gathering data before being able to decide on interventions. Reducing ambulance ramping and bed block at AHS will assist our patients to be treated in the appropriate timeframe allowing for quicker recovery, treatment and return to normal activities of daily living. By ensuring that our theatres are operating at maximal capacity allows AHS to reduce pain and discomfort on our surgical patients who are on the waitlist. This in turn has a positive flow on effect for the community by ensuring that GPs have more capacity for treating acute illness in the community. Identifying outlier patients will help AHS manage the movement of patients to and from Royal Perth Hospital (RPH) and other tertiary hospital sites so patients are able to relocate and recover in their local hospital at AHS and be closer to friends and family. Creating innovative solutions for this problem will also help increase the efficiency of patient flow and bed management at AHS, releasing capacity for staff and beds which has significant flow-on effects for bed ramping and/or bed block at other East Metropolitan Health Service (EMHS) hospitals including RPH, Bentley Health Service and St John of God Midland Public Hospital. 

Can you create a one-stop-shop visualisation of hospital performance in real time for the Armadale Health Service (AHS) to support health care quality, safety, staff and patient experience?

Domain Experts:

  1. Tim Leen, Project Director at Health in a Virtual Environment (HIVE)
  2. Deb Reid, Senior Registered Nurse – Patient Flow and Bed Management

Improving Sepsis Recognition

NOTE: Access to specialist data pending

Patients continue to die from sepsis, which is partly due to delayed recognition when they present at a healthcare facility. Current sepsis algorithms focus on predicting the outcomes in patients with sepsis. However, delayed recognition and identification remain a significant problem. Current solutions focus on sepsis pathways, but when identification of patients with sepsis is delayed or missed, this means that patients may not be placed on the pathway and miss out on appropriate care.

Can you build on the existing sepsis algorithm to reduce delayed or missed identification of sepsis?


  1. Hou, N., Li, M., He, L. et al. Predicting 30-days mortality for MIMIC-III patients with sepsis-3: a machine learning approach using XGboost. J Transl Med 18, 462 (2020). https://doi.org/10.1186/s12967-020-02620-5
  2. Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of Clinical Criteria for Sepsis: For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315(8):762–774. doi:10.1001/jama.2016.0288

Domain Expert: Dr Matthew Anstey, Intensive Case Specialist

Breaking Down Access Block

NOTE: Access to specialist data pending

Access block refers to a problem whereby ‘patients in the emergency department (ED) requiring inpatient care are unable to gain access to appropriate hospital beds within a reasonable time frame’ (Fatovich DM, Nagree Y, Sprivulis P, 2005). It is well known that Emergency Department access block increases patient mortality, over those without delays to admission. The reasons for this might be multi-factorial – delayed access to inpatient specialty treatments, delayed access to investigations, delayed access to specialty nursing care or lack of ongoing re-assessment by treating clinicians.

Can you find a way to use data science and/or digital technology to reduce access block issues?


  1. Fatovich DM, Nagree Y, Sprivulis P. Access block causes emergency department overcrowding and ambulance diversion in Perth, Western Australia. Emerg Med J. 2005 May;22(5):351-4. doi: 10.1136/emj.2004.018002. Erratum in: Emerg Med J. 2005 Jul;22(7):532. PMID: 15843704; PMCID: PMC1726785. 

Domain Expert: Dr Matthew Anstey, Intensive Case Specialist