Improving the speed and accuracy of clinical coding is top of the ‘to do’ list for one research team at Western Sydney University (WSU), a Digital Health CRC participant. The team is also providing the IT grunt for a ground-breaking research project on ageing, and it’s considering future research into the better regulation of mobile health apps.
Senior Lecturer in Health Informatics at WSU, Dr Anupama Ginige (pictured), leads the team.
She says its work on clinical coding has ‘real world’ implications for healthcare and health research in Australia and internationally – and digital solutions could play a key role.
“There are two classification systems used in Australian hospitals – the International Classification of Diseases (ICD) Australian Modification, and the Australian Classification for Health Intervention (ACHI)” she says.
“When you go to hospital, you get treated and go home – and that’s the end of the story as far as you are concerned. But that is the start of the story for your data.”
“Clinical coders are specially trained to read through your clinical documents and assign specific codes for your illness, diagnosis and treatment.”
“These codes are provided to the Australian Institute of Health and Welfare – so it can determine, for example, how many cases of cardiovascular disease there might be in Australia each year.”
“This, in turn, can underpin healthcare policy.”
“The data is also used by the World Health Organisation (WHO) and other international health agencies.”
“Different countries are often interested in different conditions – for example, health policy-makers in Switzerland might be more interested in skiing-related injuries, whereas in Australia we want to know what kind of a spider bite someone received or the prevalence of skin cancer.”
“But regardless of the injury or illness, coding is critical the world over.”
“And given there are hundreds of codes tracking each different type of illness and injury, it is incredibly important that we ensure these codes are accurate – not only between hospitals in Australia, but also between countries.”
Calibrating the codes
The International Classification of Diseases (ICD) Australian Modification is updated every two years. In 2019, the contract to undertake this work was given to WSU and the University of Sydney. Dr Ginige is the IT lead for the consortium.
She and her team are also involved in work to calibrate Australia’s coding classifications with international classifications.
“Because all countries code slightly differently, we occasionally need to align them all” she says.
“For example, in Australia we might code a colon growth as 20.11 with bleeding and 20.12 without bleeding, but another country might code it the other way around – so from time to time we need to calibrate our coding with other countries to ensure the data and statistics remain consistent.”
“If they aren’t, it can impact on global epidemic research, international health reporting and even cures to diseases – so it is critical we get it right.”
Dr Ginige is a member of a small international committee – the WHO Implementation of Terminology Committee – that is looking at these issues.
“Through this committee, we have brought together health information managers from across the world to better assess the differences in coding between countries” she says.
“There is discussion around using Machine Learning (ML) and Natural Language Processing (NLP) techniques to do this mapping, because there are thousands of codes involved, and to do this manually will take many years.”
No free lunches!
Back in Australia, the WSU team is also undertaking research in how ML and NLP might be used to boost the accuracy and efficiency of clinical coding.
“At the moment, clinical coding is a manual process” Dr Ginige says.
“Clinical coders have to read through the patient diagnosis and treatment summaries, and then input the codes. This is time-consuming and has significant potential for errors.”
“We are researching whether we could use ML and NLP to autocode patient records and make the coding process quicker and more accurate.”
“Ultimately, the aim is to get to the point where a coder or even a clinician could dictate a patient’s history and treatment into a computer, and it will code everything accordingly.”
But Dr Ginige cautions that the No Free Lunch Theorem is just as much at play in the clinical world as it is in the legal world – individual algorithms do not always produce the same result across different datasets.
“In case law, an algorithm can still make an incorrect finding of guilt or innocence against an accused person” she says.
“It’s the same when it comes to clinical coding – the algorithm doesn’t always get it right.”
“We are looking at why this is, and how we can improve it so we get the 95-98% accuracy that we need.”
“Just like judges might have different views on a legal case, a clinical coder at one hospital might have a slightly different understanding of an illness or operation from a coder at another hospital, so they might input different codes.”
“We need to look at different solutions to this, but one option could include giving higher ratings to a particular dataset because it has been coded by a more experienced coder.”
Maintain Your Brain
Dr Ginige and her team have also led the IT development of a major online Australian clinical trial, Maintain Your Brain – the largest trial in the world to attempt to prevent cognitive decline, and potentially dementia, through an online intervention program.
The project has been funded by the NHMRC, and is being undertaken by a consortium of universities led by the Centre for Healthy Brain Ageing at UNSW.
“The trial has been about intervening in older peoples’ lives for one year, to get them to make lifestyle changes in diet, exercise, stress alleviation and cognitive training” Dr Ginige says.
“The aim has been to ascertain whether the year of intervention has delayed the onset of dementia until after 75 years of age.”
“Our team at WSU has provided the digital platform to enable this, and we have got it to the point now where participants can go online to report their data from home – rather than having to go into a clinic.”
“This has helped us build our participant numbers to over 6000, making it better for research purposes.”
“We have also provided IT platforms for other studies, including one tracking the journey of bariatric patients in Blacktown Hospital; and one comparing older twins and their health outcomes, particularly around brain health.”
The rise (and rise) of mobile apps
The need to better evaluate mobile health apps is another area in which Dr Ginige has a strong interest.
“There are currently over 380,000 mobile health apps, and it’s growing every day” she says.
“A few years ago, we had some students evaluate mobile apps for mental health, diabetes, and a range of other conditions – we quickly found that you can evaluate 100 apps today and within five months there will be 10 more.”
“We are considering researching how we could automatically review apps using Machine Learning.”
“For example, if you have an app for weight control, we can use the data from the app to assess its success (or otherwise) in whether you lost weight.”
“But we’d need to ensure that such a tool recognises all the contributing factors – for example, whether the app user has truly followed the app’s instructions.”
Dr Ginige believes there is a need for Australia to develop frameworks around approving mobile apps in Australia.
“Currently, mobile apps don’t need to be approved by the TGA. If they are classified as a medical device they do, but developers can often argue their app is not a therapeutic product.”
“In the US, they have started classifying mobile apps as therapeutic products, and as of last year more than 280 apps were already listed as FDA-approved products.”
“I think we need to develop similar frameworks here.”
“The big issue is that, while drugs can be controlled because they are physically imported into the country, how do we control an app that is available to everybody over the internet?”
“There are other issues to consider as well – can GPs prescribe approved apps, is there a healthcare funding model around it, and can insurance companies pay for it?”
“At WSU, we are doing some initial research into how different countries are handling this issue – and I know regulators here are looking at these things too.”
“We believe we could contribute to policy development in this area, including around the privacy and security of data, and the collection of usage data from apps – much like adverse drug outcomes are reported to the TGA at the moment.”
“If we want to seriously use digital health to improve population health we cannot ignore this space.”
For more information on WSU’s work in the digital health space, visit https://ehrg.scem.westernsydney.edu.au
For more information on Maintain Your Brain, visit www.maintainyourbrain.org
Email Dr Anupama Ginige at [email protected]
Article by Patrick Daley