AI use in healthcare: opportunity or threat?

It is undoubtedly the topic of the year! Artificial Intelligence has captivated imaginations and generated headlines more than ever, heralding the dawn of the 4^th Industrial Revolution. Even in a part of the world still going through the agrarian revolution, we are realising a diversity of AI for Health applications, from Cape to Cairo.

As the technology rapidly transitions from sci-fi to real life, concerns about its safety have been echoed across the world. Particularly, in an industry where the ethical foundation of “do no harm” is sacrosanct, we must evaluate whether AI for health is ultimately an opportunity or a threat.

AI offers immense potential to improve access to care and help Africa achieve the daunting UN SDG 3 target of universal health coverage. AI can augment the ability of doctors to provide accurate diagnoses and optimize treatment plans.  

AI is being used by governments to manage disease outbreaks, inform the decisions of health policy-makers and allocate resources within health systems. Across Africa, AI is being used across all clinical disciplines of medicine from paediatrics to neurology, from radiology to cardiology and even in mental health.

Notable examples

Notable examples include hearX, a patented smartphone application for the early detection of hearing loss that is addressing the shortage of audiology lab services. Another is Babyl in Rwanda which provides an AI-supported teleconsultation service accessible via feature phones.

There is Ubenwa in Nigeria, which is a mobile app that uses machine learning to analyse a baby’s cry to detect early signs of anomalies from asphyxia or brain injury.  

There is also Diagnosify, which uses AI to diagnose skin diseases from uploaded images. This could be very helpful given that Zimbabwe has only 2 registered dermatologists! Radiology is the field with the most applications of machine learning AI as algorithms are being built to interpret medical imaging.

Dr CADx is a local start-up that succeeded in developing a system with high accuracy for diagnosing pneumothorax – a life-threatening complication, often missed on a chest X-ray. These are just a few examples on an ever-growing list of AI for Health applications in Africa. 

Side effects

Every medicine, regardless of how effective it is, has side effects or adverse reactions. As we appreciate the growth of AI use, we must also consider the possible threats it poses to healthcare.

One of the biggest problems with AI is obviously inaccuracy. An AI solution is as good as its algorithm which in turn, is based on the data used to formulate it. So it can easily be junk in, junk out! Algorithms based on health data from America or Europe will not work in our African context.

I once evaluated a US-based symptom checker that failed to recognise that in our setting, diarrhoea is most usually from an infection rather than from an inflammatory condition or malignancy. Infections such as malaria, bilharzia and cholera though common in our setting, are regarded as ancient diseases in the developed world.  Therefore building algorithms solely based on western data sets results in inaccurate solutions that have encoded bias.

The role of regulators such as the Medical and Dental Practitioners Council of Zimbabwe (MDPCZ) is to primarily protect the public from unsafe practitioners. The widespread use of AI poses a threat to this regulatory control. How can the MDPCZ regulate an AI-based app that diagnoses and issues prescriptions, whose HQ is in Europe that is available as a free download to the public?

If there is a case of malpractice, who should the Council invite to their offices for tea and biscuits? Should it be the software engineers or the researchers who supplied the biased data, or should it possibly be the app market managers for allowing such an app on their platform? Should it be all of them? With AI, it is yet to be determined who carries the actual liability. 

Will AI replace human doctors?

This is a question often asked – the answer to which is no, not in the foreseeable future. However, AI will certainly affect the way doctors are practising medicine in the foreseeable future by altering the dynamics of the patient-doctor relationship.

Already, Dr Google is posing a challenge. Patients who google their symptoms prior to consultation often question or disagree with their doctor’s assessment. This could be amplified by AI applications. If the doctor’s opinion differs from that of the AI, who then could be wrong in such a scenario?

Is it the licensed doctor with years of study and experience or could it be the algorithm based on terabytes of health data collected from thousands of doctors all over the world? This breeds mistrust that can lead to patient non-adherence to treatment or unnecessary spending of money looking for alternative opinions.

With the growth of AI, we may become heavily reliant on it, to the extent of losing human autonomy. Many of us can recall the frustration of going to the bank and failing to transact because “the system is down” Imagine if this could be the same story when you visit your doctor! We could end up so reliant on AI that doctors may fail to function when there’s a power cut and the computers are down! Such loss of autonomy is what leads to a Skynet dystopia.

With so much investment being driven to AI, there’s a lot of unethical data mining and harvesting going on. Without our control, rogue entities can amass our demographic, financial, genetic, medical and biometric data to use it against us. The emergence of COVID-19 witnessed several incidences of technologies being used to invasively capture personal data for supposed disease surveillance. Such data in the wrong hands could lead to wider discrimination, political persecution or even genocide. 

Is AI an opportunity or a threat?

So, is AI use in healthcare an opportunity or a threat? It is both! AI’s opportunities and challenges are inextricably linked. Therefore, we must navigate the space in a way that maximizes on the opportunities whilst mitigating the risks. In line with this, The WHO assembled a group of experts to issue guidance on the ethics and governance of Artificial Intelligence for Health. They recommended six core principles:

(1) Protect autonomy;

(2) Promote human well-being, human safety, and the public interest; 

(3) Ensure transparency, explainability, and intelligibility; 

(4) Foster responsibility and accountability; 

(5) Ensure inclusiveness and equity; 

(6) Promote AI that is responsive and sustainable. 

Ensuring transparency, explainability, and intelligibility is the key factor. To many, AI is like magic – they do not actually understand what AI is and how it works despite using it every day. As a society of digital natives, we innately trust and depend on technology, assuming all is safe, lacking the curiosity to want to know what’s under the hood. This is dangerous!

The level of transparency, explainability, and intelligibility needed in AI should be similar to the rigorous process new medicines go through. There’s no such thing as a miracle drug. The active molecules inside the drug are declared, their pharmacological actions are explained, the details of the clinical studies it went through to prove efficacy are published and pharmacovigilance is exercised to report any new side effects of the drug. This should be extended to the algorithms behind AI solutions. Greater transparency also allays misconceptions that often stand in the way of good solutions.

As a country, as a continent and as a people, we should avoid being mere spectators or passive participants of the 4^th Industrial Revolution. We need to vie for equity in distribution. There is opportunity for us to build AI systems for health that can understand what nyon’o is, systems that understand the sensitivities of the relationship between tezvara nemuroora and systems functional in our diverse African languages because they will be built off our local data. We need to vie for equity in ownership. The AI systems we build will have their HQs in Africa, where they pay taxes and offer us an opportunity to trade in their stocks in local currency. Their CEOs would also be summonable to hearings in our halls of parliament! Only this way, will we harness the opportunities AI presents whilst also mitigating the threats it poses.

Also read:

Zimbabwe is not ready to take full advantage of what AI has to offer

A specialised AI to help find every word spoken by any politician on any topic

Using AI to tailor pricing to individuals, devise advertising strategies and more