AI in Healthcare: Democratising or Dystopian?

Gary Kantor, a Clinical Consultant at Insight Actuaries and Consultants, brings to light a topic of urgent global significance – the intersection of AI and healthcare.

Gary Kantor, a Clinical Consultant at Insight Actuaries and Consultants, brings to light a topic of urgent global significance – the intersection of AI and healthcare.

Healthcare systems around the world are facing tremendous pressure due to ageing populations, the burden of new illnesses like COVID, increased rates of non-communicable disease (NCD), along with rising care expectations. While demand is growing, resources remain limited.

Kantor proposes that the advent of AI could be a solution or a trigger for additional stress – the outcome is ours to determine.

But what does this mean exactly?

Let’s consider AI technologies that have been in use for a while and are now gathering momentum. They augment the capabilities of healthcare professionals in certain diagnostic and treatment tasks. For example, in operating theatres, AI applications are being utilised to predict low blood pressure and to guide needle placement in common invasive procedures like nerve blocks.

The average or mean blood pressure (BP) is a reflection of the amount of blood flow going to vital organs. A value of 65 or less is the number below which organ injury  – to the brain, heart or kidneys – becomes more likely. Data in electronic medical records from thousands of anaesthetics reveals that brief periods of low BP – a minute or less – are really common, but maybe not so important. More worryingly, much longer periods – 10 to 15 minutes – still happen a lot – 30 percent of the time. A new commercial monitor based on machine learning algorithms gives a real-time index between 0 and 100, which predicts the onset of low blood pressure in the next few minutes.

Consider this case: the area into which anaesthetists insert a needle for a brachial plexus nerve block is full of blood vessels and close to the lungs. Placing the needle correctly requires extensive experience to master the interpretation of fuzzy grey ultrasound images. Now, AI-based real-time colour coding and annotation of the image could be of substantial assistance.

Patients who have battled severe diabetes for an extended period may eventually lose their sight, a condition known as diabetic retinopathy. Typically, only experts are trained to diagnose it. However, AI now possesses the capability to autonomously identify this condition with the same level of accuracy as an expert. Commercial versions of this AI-based examination are billed in the US at $50 per instance, yet an international charity is piloting it globally for free. It should be noted that a rather costly camera is required to capture the necessary images.

One more example. A non-radiologist can utilise a portable machine like this one to capture ultrasound images of breast lumps. With the assistance of AI, one can attain near-expert levels of accuracy in diagnosing cancer, comparable to those achieved with images from a conventional, larger ultrasound machine interpreted by a radiologist.

The most revolutionary development may be the emergence of “generative AI” based on ‘Large Language Models’, such as ChatGPT, which passed the US Medical Licensing Exam half a year ago. While this doesn’t imply that ChatGPT will replace your doctor, it does open up a realm of possibilities for healthcare advancements.

The impact of AI will be comprehensive and far-reaching. It could enable earlier and more precise diagnoses, along with more effective treatments that are personalised to individual patients. It could expedite the drug discovery process and enhance the functionality of prosthetics and robotic devices used in surgery. It has the potential to aid remote patient monitoring and virtual consultations. AI could also detect mental health conditions, propose suitable treatments, and even offer therapy. Furthermore, it could predict disease outbreaks and epidemics and bolster the security of health data, detecting and pre-empting cyber threats.

However, based on past experience, the integration of AI into healthcare might not go as smoothly as we might hope. For instance, the introduction of electronic health records (EHR) in America led to doctors spending nearly half their time on EHR and desk work rather than patient interaction. AI could provide a solution by transcribing doctor-patient conversations into medical notes, potentially reducing documentation time by half and decreasing burnout. Yet, such advancements are not without drawbacks – privacy concerns, missing or false details, human review requirements, and significant costs.

Despite AI’s potential, it’s essential not to overlook the necessity for high-quality evidence proving that these systems work safely and effectively. The predictive capabilities of AI, can sometimes fall short, as shown in a recent clinical study  of the low blood pressure prediction index and ChatGPT’s performance in a licensing exam for gastroenterologists.

Kantor emphasises that the solution lies in a deeper understanding of healthcare systems and the manner in which technology is introduced into them. This suggests the necessity for AI regulation and certification, as highlighted by Sam Altman, the CEO of Open AI, the company behind ChatGPT, in his recent testimony to the US Congress.

Moving forward, AI in healthcare necessitates a nuanced approach. Peter Lee, head of Microsoft’s Healthcare Division, advocates for local optimisation rather than disruption or transformation. He says we need to understand the work of healthcare professionals, connect with their aims, and improve their day-to-day effectiveness and satisfaction.

Those most experienced with AI in healthcare maintain that humans must remain in the loop. The aim isn’t to replace humans but to augment human professional capability, and automate certain non-clinical but necessary administrative or clerical tasks. This approach is especially relevant for under-resourced settings such as most of the public health sector in SA where skilled professionals are in shortest supply.

In conclusion, techno-optimism hasn’t always produced the goods, but AI-powered new technology that’s well-aimed, well-regulated and based on sound evidence of safety and efficacy could greatly enhance care and administrative processes. The focus should be on fostering collaboration between humans and AI, improving ethical standards, safety, and regulation, and bringing more care and compassion into the healthcare system. Whether AI in healthcare is democratising or dystopian is a decision we make based on our actions in the coming months and years.



Get an email whenever we publish a new thought piece


By signing up you consent to our terms and conditions

More from Gary Kantor