Machine Learning (ML) has been increasingly used in medicine to predict outcomes and identify risk factors, particularly for total hip and knee replacements. However, researchers warn that these predictions are currently based on limited data, which can affect their accuracy.
Dr Reza Hashemi from Flinders University’s College of Science and Engineering explains that while machine learning has great potential for processing large amounts of data, it is not without its issues. “The accuracy of predictive models is dependent on the quality of the data sources, and predictions may be significantly affected by the amount of data and the number of variables included,” he says.
Current predictive models for total hip and knee replacements are based mainly on patient-reported factors and imaging variables. As a result, the output of ML models in this area needs to be interpreted carefully.
Assessing the effectiveness of machine learning
To study the application of supervised ML in predictive modelling for post-operative outcomes of total hip and knee replacements, researchers from Flinders University, the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR), Royal Adelaide Hospital and UniSA assessed the most widely used ML techniques, data sources, domains, limitations of predictive analytics and the quality of predictions.
Dr Khashayar Ghadirinejad, a research co-author from Flinders University, explains that the most widely used ML approach in medical sciences is ‘supervised learning,’ which estimates the mapping function for new input data to predict categorised, real values, or time-to-event outputs.
Conventional statistical methods of risk predictions rely on predetermined assumptions and mathematical equations to formalise relations between variables, whereas ML techniques use large amounts of available data to recognise these relationships.
Limitations and future directions
In assessing the effectiveness of ML to assist with total hip and knee replacement procedures, the researchers note that care should be taken by the medical profession when dealing with limited data on specific subjects. Dr Ghadirinejad suggests that ML models should now be assessed and evaluated using randomised cohort studies and controlled trials in real-world settings, rather than just assessing data.
Despite the current limitations of ML, the researchers recognise that there is still a need for models that can predict various outcomes, such as the early identification of prostheses outliers based on the available big data from national joint registries around the world.
Joint registries aim to reduce the revision rates of arthroplasty surgeries by early detection of outlier joint arthroplasty devices. They provide population-based data on the comparative outcome of prostheses within the community. The authors suggest that a future direction for ML in the domain of joint arthroplasty could be to develop decision-making support systems focused on pre-surgical predictions that enable surgeons to determine what is best for their patients individually.
Ritchelle is a Content Producer for Healthcare Channel, Australia’s premier resource of information for healthcare.
