Currently, treatment decisions for elective surgery are currently based on clinicians’ assessments of the benefits the surgery will bring to the patient. This assessment is derived from the clinicians’ personal experience and knowledge of published evidence of effectiveness.

In New Zealand, eligibility for surgery is based on a prioritisation score derived from this clinical assessment of the patient. The scoring usually takes little account of how various patient-specific factors might affect the net benefit the patient receives from surgery, and the cut-off remains somewhat arbitrary. Relevant patient-specific factors include co-morbidities (e.g. obesity in the case of hip replacement), sex, age and the patient’s life expectancy. The result is that health outcomes and the total health gain from these procedures is highly variable, as is their cost-effectiveness. Patients and payers of health services would therefore value a more accurate prediction of the outcomes and cost-effectiveness of a treatment.

The Health Outcomes Prediction Engine (HOPE) project produced a prototype electronic clinical decision support system to make precise health outcome predictions tailored to the specific circumstances of individual patients.

This involved combining the existing knowledge base of the treatment effectiveness and cost-effectiveness of one (or more) elective interventions, with a supervised machine learning algorithm that iteratively increases the predictive accuracy of the initial model.

There were three research streams within the HOPE project that lead to prototypes:

• HOPE for Stroke Outcomes Prediction
• HOPE for AAA (Abdominal Aortic Aneurysm) Risk Prediction
• HOPE for Patient Reported Outcome Measures

HOPE for Stroke encompassed a predictive model and clinical decision support system for stroke patient outcome prediction within Waitemata District Health Board. Stroke recovery can vary significantly, so the tool is designed to predict the outcome of a stroke, giving patients and their families the opportunity to adjust their expectations and work with care providers to prepare for recovery. The innovative new tool is ready to be tested with stroke patients, as the machine learning predictions have proven to be very accurate. The next stage is for clinicians to find the most useful way to use this tool in their systems.

HOPE for AAA involved creation of a prototype for predicting a person’s risk of abdominal aortic aneurysms within a primary care setting. In a pilot study, the precision screening algorithm was proven to detect the condition and put patients with AAA on the path to a positive outcome.

This project’s predictive models were presented at the 2017 HiNZ conference. If you’d like to learn more, further information can be found here for Stroke and here for AAA.