Spotlight Series: When Neurology Meets the Limits of What We Can Measure

Dr. Yun Hwang on Progressive Disease, the EMR Journey, and Why AI Validation Matters More Than AI Excitement

Part 2 of the Medora Advisors Spotlight Series

----------------------------------------------------------------------

Most healthcare technology is designed around conditions that can be measured.

Blood pressure. Glucose. Tumour markers. Infection markers. The logic is clean: capture the number, track the trend, adjust the treatment.

But what happens when the condition you're managing doesn't produce reliable markers? When the gap between what a patient is experiencing and what clinical tools can detect is wide — and growing?

Dr. Yun Hwang works in that gap every day.

A Neurologist and a PhD on biomarkers in neurodegenerative disorders, Dr. Hwang specialises in some of the most complex and progressive conditions in medicine: Parkinson's disease, cognitive impairment, and dementia. He has co-developed an AI-powered Parkinson's monitoring application with DXC Technology, published research in The Medical Journal of Australia, and led work on integrated metropolitan neurology clinic models of care.

He has also watched the Australian health system transition from paper to digital — from DOS-based blood result terminals and faxed request forms to EMR, cloud computing, and AI-assisted diagnostics.

We asked him four questions about complex care, technology transitions, implementation advice, and innovation. His answers reveal something the digital health industry needs to hear more clearly: the process matters as much as the output — and in medicine, getting lucky with the right answer through the wrong process is not good enough.

-----------------------------------------------------------------------------------------------

Q1: What aspect of managing ongoing care for patients with complex or progressive neurological conditions appears straightforward on paper, but is far more nuanced in real-world practice?

Dr. Hwang:

There are two.

The first is monitoring disease progression and capturing fluctuations — especially when there is a discrepancy between what symptoms the patient is experiencing and the markers we can use to assess disease progression. Often these markers do not exist in neurology. I may diagnose a patient with Parkinson's Disease, Alzheimer's Disease, or Chronic Inflammatory Demyelinating Polyneuropathy, and there is a pathway or criteria to diagnose these conditions. But the way the condition progresses in each patient is different, and a lot of the time we end up taking a very reactive approach to managing the condition long term: "Do you have any new symptoms?" "Do you feel you are getting worse, better, or about the same?"

Clinical encounters are very brief, and we are all more liable to recency bias. If a patient had a good week in the lead up to the appointment, they are more likely to say "I'm doing great" — while they may have been struggling for an entire month before that, or vice versa.

The second challenge, which is more pertinent to many progressive neurological conditions that do not have a cure, is how we can improve the quality of lives of these patients. They have symptoms which are limiting their lives in some way — balance problems, weakness, speech difficulties — and the way we enable and support them to live their daily lives has not changed significantly this century.

I am hoping that we will be able to develop more tools, harnessing the significant increase in our ability to capture biometric data using wearable devices or in-home cameras — to assess strength or gait on a frequent basis, for instance — to provide more information for the treating team on symptom progression and fluctuation. This would allow us to identify problems earlier and tailor a plan to the specific individual and their situation. Access to cloud computing, reliable wireless connectivity, and the ability to connect many devices add to their potential. The flip side is that these monitoring devices need to undergo rigorous validation to demonstrate that they produce reliable, reproducible, and actionable data.

Q2: Can you share an example where a technology change, system transition, or digital tool either meaningfully improved — or disrupted — care delivery for complex patients?

Dr. Hwang:

I started medicine in Australia at the cusp of the transition from paper-based systems to EMR in NSW. When I was a medical student, there were very rudimentary PACS systems in hospitals, and different teams would often congregate at a handful of terminals in the radiology department to review scans. Private providers would give films directly to patients to take to their doctor. Blood results were on a DOS-based window — you could only ever see a particular grouping of results at a time, and you would have to copy them manually into the notes. Request forms were written on paper, then faxed, or dropped off in a box.

I did an elective placement in Boston in 2007 and saw doctors ordering tests on their iPhones. I remember thinking: what kind of magic is this?

Having a system where we can access blood results, radiology, and clinical documentation in one place, and request further tests from the same interface, is phenomenally useful and efficient. I am looking forward to the adoption of a single medical record and EPIC in NSW Health, which I understand has even greater capabilities.

Q3: If you could offer one piece of advice to a health system preparing to implement new technology or workflows in a complex care or neurology setting, what would it be?

Dr. Hwang:

Build for openness and flexibility — and create a clear mechanism for feedback and trialling changes as suggested by users and as circumstances evolve.

Things are changing so quickly that by the time a full-stack tool has been developed, parts of it will already be outdated or superseded. Can those parts be removed, replaced, or updated easily? There may be new requirements. A new diagnostic tool may become available and its output needs to be entered. A new wearable may be adopted — does it produce output that is compatible with the existing EMR? Can that information be integrated directly into patient records?

The end users on the coalface are using these systems most often. What are their pain points? How can these be improved? Do users have a mechanism to make suggestions — to add or alter aspects of the system?

One example: a colleague was receiving blood test results from two separate pathology companies and needed to track the trends of tumour markers. But results from Company A and results from Company B could not be displayed on the same graph. He ended up creating a widget using Claude Code to work around the issue — but this became another programme running in the background that he needed to open separately, and he would sometimes need to export the graph as an image to insert into the notes. Ideally, he could have contacted his EMR provider and said: can you make this happen within the system? That feedback loop needs to exist — and it needs to be responsive.

Q4: What innovation in complex care or neurological medicine are you most excited about today — and what operational realities must be addressed for it to succeed at scale?

Dr. Hwang:

I'm excited by two things: the rise of wearable devices, and the many different efforts to deploy AI effectively in clinical practice — whether in imaging analysis software, fundus imaging, or other AI tools designed to assist and improve on what we are doing in medicine.

My concern, however, is that many of these tools are being deployed without adequate testing and validation for accuracy and consistency, and without sufficient understanding of how they work and how they reach the conclusions they do.

No system is perfect. No human is perfect. I don't think we can expect AI to get everything right every time — and in some ways this is a function of limited and incomplete information provided to the AI. But when there are errors, people's lives can be affected in a serious way.

There is an established mechanism to review decision-making when humans are involved — Morbidity and Mortality meetings, for instance. Hopefully we learn something from the incident, changes are made to the system and to individual practice, and we don't make the same mistake again. But how AI reaches a particular output is often opaque. The output may not be consistent. The process is not always reviewable.

And in medicine, the process matters as much as the output. One can get lucky and reach the right output — but if the process is not correct, errors and mistakes will invariably occur.

--------------------------------------------------------------------------

Why This Matters

Dr. Hwang's insights expose a tension at the heart of digital health innovation: the gap between what we can build and what we can validate.

Wearables can capture gait. AI can analyse fundus images. Algorithms can flag tumour marker trends. But if the data is unreliable, the process is opaque, and the feedback loop between clinicians and system developers is broken — the technology creates new risk rather than reducing existing risk.

The operational wisdom here is not anti-innovation. It is pro-rigour.

Build flexible systems that can be updated as clinical practice evolves. Create genuine feedback mechanisms so that the people using the system can shape it. Validate before you deploy. And remember that in medicine, the right answer reached through the wrong process is not a success — it is a near-miss.

These are not abstract principles. They are the lessons of a clinician who has watched the Australian health system transition from paper to digital, from DOS terminals to cloud-hosted EMR, and who is now watching the next transition unfold — from EMR to AI-assisted care — with both genuine excitement and clear-eyed caution.

 -----------------------------------------------------------------------------------------------

About Dr. Yun Hwang

Dr. Yun Hwang is a Neurologist trying to balance a mix of private practice, on call roster participation at the local public hospital and family life while dabbling in medical education, research, and innovation. He qualified as a neurologist in 2015 and completed a PhD on biomarkers in neurodegenerative disorders at University of Sydney in 2023. His current academic interest is effectively utilising and deploying the latest development in technology to improve the diagnosis and management of neurological disorders to improve the quality of life of patients with these illnesses. His first (of many, he hopes) foray in this direction has resulted in an AI-powered Parkinson's Disease monitoring application co-developed with DXC Technology, which is currently in the process of validation. 

Dr. Hwang trained in medicine in Australia during the transition from paper-based systems to electronic medical records, and has practised across NSW Health, but also in ACT Health in Australia and the NHS in England during his fellowship. He brings a perspective that is simultaneously clinical, technical, and deeply grounded in the operational realities of complex and progressive disease management.

The Medora Advisors Spotlight Series features healthcare executives and clinicians who've actually led operations — not just consulted on them. If you've sat in the seat and want to share your story, connect with us at medoraadvisors.com.