Is the Family Doctor Heading for Extinction? AI in Healthcare and the Future of General Practice in New Zealand
Healthcare in crisis: Artificial Intelligence (AI) offers a solution
In March of this year, I sustained serious spinal injuries while in Chile. After a month of recovery, I was well enough to return to Wellington, New Zealand. I scheduled an appointment with a medical specialist, but the wait was lengthy, and it wasn't until May 18th that I saw him. During the examination, I mentioned that the delay between the accident and the consultation meant that most of the opportunity for medical intervention had passed, with 90% of the healing already completed. I was politely expressing my frustration. He agreed that nothing more could be done and praised the comprehensive rehabilitation plan I had devised for myself. We then shook hands and parted ways.
New Zealand's healthcare system is presently in crisis. Once efficient, it now struggles with frequent shortages of doctors and nurses and increasing waiting lists. Artificial intelligence (AI) is being touted as the answer to this problem. But is it really? In this article, I present my personal exploration of artificial intelligence in healthcare.
AI in healthcare is here to stay
Artificial intelligence in healthcare is progressing much more rapidly than most people realize. Currently, much of this progress seems relatively harmless and, in many instances, highly beneficial, though there are significant caveats. A general practitioner can now utilize AI to document and summarize consultations, prepare clinical notes, search medical literature, check medications and interactions, review guidelines, and assist with decisions regarding investigations and referrals. In New Zealand, the Royal New Zealand College of General Practitioners is already piloting AI-supported simulated consultations in GP training. The College aims to determine where AI truly aids while maintaining clinical judgment and human connection at the core of general practice.
I am not entirely opposed to these developments. I use artificial intelligence myself, and I can see potential for it to improve healthcare. What interests me, and concerns me, is where all of this may eventually lead. We should not assume that because the first applications of AI are designed to assist the doctor, the technology will always remain in that subordinate role.
My concern is supported, in part, by an article by Gloria J. Maloney, Miracle Machines or Medical Mirage? The Dangerous Hype of AI Healthcare. Maloney examines the growing enthusiasm for AI-enabled medicine while raising questions about evidence, safety, regulation and the possibility that technological excitement may run ahead of what has actually been demonstrated to work safely. I think the questions she raises deserve consideration.
From the doctor's assistant to the doctor's replacement
Consider what happens during an ordinary visit to the family doctor. The doctor listens to your story, asks questions, examines you where appropriate, considers possible explanations, reviews your history, decides whether investigations are required and then recommends treatment or refers you elsewhere. Much of that process involves collecting information, recognising patterns and making decisions based upon accumulated medical knowledge.
These are precisely the areas in which artificial intelligence is developing rapidly. An AI system can review a lifetime of medical records in seconds, compare symptoms with enormous bodies of medical literature, identify possible drug interactions, follow clinical guidelines, recognise patterns in laboratory results and imaging, and suggest further investigations. None of this means that present-day AI can safely replace a doctor. It does mean, however, that an increasing proportion of the intellectual and administrative work traditionally undertaken in general practice is becoming technically capable of being assisted, and perhaps eventually performed, by machines.
The next phase is easy to foresee, as elements of it are already emerging around us. Watches and other wearable gadgets can constantly gather physiological data. Home monitoring devices can send readings remotely. In the future, we can anticipate more advanced sensors being integrated into clothing, household items, and implanted technologies like pacemakers and even brain implants. This could lead to a constant flow of physiological data instead of the occasional snapshot collected during a doctor's visit. Some of these advancements make me uneasy, and we need thorough discussions and guidelines instead of freely allowing their complete and immediate implementation.
The World Health Organization is already describing AI as a technology with applications in diagnosis, clinical care, disease surveillance and health-system management, while simultaneously warning that governance, regulation and safety mechanisms need to keep pace.
Combine these advancements, and your AI health assistant could access your lab history, medications, past illnesses, injuries, family history, scans, specialist reports, exercise habits, sleep patterns, heart rate, and other physiological data. Instead of waiting for you to notice something is wrong, it might eventually detect changes before you are even aware of them. I partially support this approach as it aligns with my strategy of staying 20 years ahead of diseases throughout a person's life. However, we must be careful not to medicalize every human process and life stage. I am concerned that the latter is occurring, while the strategy of staying 20 years ahead is not being adequately prioritized.
One advantage of AI that should not be dismissed
There is another frustration with general practice that receives less attention than waiting lists. You may have had the same family doctor for 10, 20 or even 30 years. Over that time, your doctor comes to know far more about you than can be found in a list of diagnoses and blood results. They know your temperament, family circumstances, how you tend to respond to illness, what you worry about, what you are inclined to ignore and sometimes even what you mean when you say, "I just don't feel right."
Then one day your doctor retires, moves away or leaves the practice. You find yourself sitting opposite somebody new and, in many respects, starting again. Your medical records may have transferred perfectly, but the relationship has not. Trust and understanding have to be rebuilt, while the new doctor needs time to discover the nuances the previous doctor accumulated through years of conversations and observation.
Paradoxically, this is one area where an AI health agent could have an enormous advantage. Properly designed, and subject to appropriate consent, privacy and data safeguards, it need never retire or change practices. It could retain a longitudinal account of your health extending over decades: illnesses, injuries, medications, investigations, laboratory results, scans, responses to treatment, exercise, sleep and nutrition, increasingly supplemented by information collected from personal monitoring devices. It might even retain seemingly insignificant observations that become important many years later.
This is one reason I am not entirely opposed to artificial intelligence entering healthcare. Used intelligently, I think it could become one of the most useful developments in medicine during my lifetime. It could give doctors access to a depth and continuity of information that no human being could reasonably be expected to remember while relieving them of repetitive administrative work that consumes valuable clinical time.
“Continuity of information and continuity of human relationship are not the same thing.”
AI in healthcare could transform New Zealand general practice
New Zealand general practice is already under considerable workforce pressure. The Royal New Zealand College of General Practitioners reports that 35 per cent of GP respondents to its workforce survey intend to retire within five years. It also cites projections that New Zealand could be short about 300 GPs within a decade, with the number of GPs per 100,000 people projected to fall.
Anyone who struggles to get a GP appointment, let alone one with the hospital, understands the attraction of another solution. If people must wait days, weeks, or even months for an appointment, cannot enrol with a practice, repeatedly encounter different doctors or spend a substantial amount for a brief consultation, an AI health service available 24 hours a day becomes increasingly attractive. I do not need to assume that anybody has deliberately created this shortage in order to introduce AI medicine. The pressure already exists. But history tells us that when a population becomes sufficiently frustrated with an existing system, resistance to an alternative can fall very quickly. A technological solution that might once have appeared impersonal or unacceptable can suddenly become attractive when the alternative is waiting three weeks to see a doctor.
This is where economics will exert enormous pressure. If an AI health agent can deal with thousands of routine enquiries for a fraction of the cost of providing thousands of face-to-face consultations, governments, insurers and healthcare organisations will inevitably be interested. The question will gradually change from, "Can AI assist the doctor?" to, "Which parts of the doctor's work still require the doctor?" "Do we even need a doctor at all?" "Will robots do the surgery?"
I still want a human being in the room
This is where I draw an important distinction. I welcome AI as an extraordinarily powerful medical tool. I would nevertheless be very sad to see the day when the general practitioner disappears altogether and is replaced by an AI health agent that receives our symptoms and physiological data, orders investigations, interprets the results and decides whether we should be referred to a physiotherapist, podiatrist, psychologist, surgeon, gastroenterologist, cardiologist or another specialist.
A good GP does something that is difficult to reduce to an algorithm. They observe the person walking through the door. They notice how somebody moves, speaks, breathes and responds to questions. They can examine an abdomen, look into an ear, feel a pulse, notice a tremor, examine a swollen joint and recognise when the person sitting before them does not quite fit what the laboratory results or computer screen are saying. There is also something less tangible but equally important: the accumulation of clinical wisdom that tells an experienced practitioner that something is not quite right.
This principle has guided much of my own work. The test provides clues. The person provides the context. A laboratory result, scan, computer algorithm or hair tissue mineral analysis can be extraordinarily useful, but none of these should become more important than the human being sitting in front of us.
“The test provides clues. The person provides the context.”
There is another consideration that becomes increasingly important as machines assume greater responsibility: who is accountable when something goes wrong? The World Health Organization has warned that AI adoption in healthcare is moving faster than legal and ethical safeguards in many jurisdictions, raising questions about patient safety, privacy, bias and responsibility when an AI system makes an error. More recent WHO work has continued to highlight transparency, fairness, accountability, privacy and possible harms from rapid deployment as areas requiring careful oversight.
When my doctor gives me advice, there is a person standing behind that advice. If an algorithm gives me the advice, who ultimately carries the responsibility?
Information is not the same as wisdom
Artificial intelligence will almost certainly become capable of knowing far more medical information than any individual doctor. In some areas, it already can process quantities of information no human being could possibly absorb. But knowledge and wisdom are not identical.
Medicine is full of uncertainty. People frequently have several health problems at once. They may be taking multiple medications, dealing with family pressures, financial problems, grief, poor sleep or loneliness. They may describe their symptoms badly. They may omit the very piece of information that turns out to be important. Sometimes the wise clinical decision is not to do more but to wait, observe and reassess.
Modern medicine is extraordinary when it is used appropriately. Trauma care, surgery, emergency medicine, infection management and many specialist procedures have transformed our lives. My concern has never been with medicine itself. It is with allowing tests, technologies, drugs and protocols to replace thinking about the whole person.
An AI system may eventually become extremely good at navigating uncertainty. Perhaps it will outperform us in ways we have not yet imagined. But that is a proposition to be demonstrated carefully, not something we should assume merely because the technology is impressive.
Are we medicalising ordinary human life?
For many years, I've been troubled by a broader issue: the growing trend of integrating ordinary stages of human life into the medical system. Life phases such as preconception, pregnancy, childbirth, childhood, adolescence, adulthood, menopause, aging, and the later years can all become occasions for testing, monitoring, medication, and intervention. A recent documentary on menopause, shown on Television One, highlighted this concern. It highlighted, for me, the tendency to quickly medicalize a natural life stage without first considering and addressing the underlying root causes of the transition being excessively uncomfprtable.
Medicine plays a crucial role when issues arise at any of these stages. Nevertheless, the distinction I want us to consider is between medicine being accessible when biology requires help and medicine taking over the responsibilities of biology itself.
Artificial intelligence could accelerate that process enormously. When we continuously monitor ourselves, normal biological variation can begin to look like pathology. Every fluctuation in heart rate, sleep, glucose, hormones or some future biomarker could potentially generate an alert. Every alert can lead to another investigation. Every investigation can uncover another abnormality. Will it produce a nation of overmedicalized, overdrugged hypochondriacs?
Commercial incentives deserve scrutiny. Healthcare is an enormous global industry, and technologies that create new forms of monitoring, diagnosis and treatment can create equally large commercial opportunities. That does not make the technology bad. It does mean we should continue asking an old-fashioned question: does this intervention actually make people healthier? So often, the answer is, "No!" I will be writing more on this topic.
Spending more on medicine does not guarantee better health
The United States provides an interesting warning against assuming that more healthcare expenditure automatically produces a healthier population. OECD figures show that the United States spends about US$14,885 per person on healthcare, compared with an OECD average of US$5,967, and devotes 17.2 per cent of GDP to health compared with an OECD average of 9.3 per cent. Yet American life expectancy is 78.4 years, 2.7 years below the OECD average, while preventable and treatable mortality are also higher than OECD averages. When it comes to health trends New Zealand follows America. Keep that in mind.
We must exercise caution in interpreting these findings. In OECD countries, there is generally a positive link between health spending and life expectancy, so it would be incorrect to assert that increased medical spending leads to poorer population health. The United States, however, stands out as an exception: its exceptionally high spending has not resulted in equally impressive population health outcomes. The main reason for the disproportionate spending and unsatisfactory results in the United States and New Zealand could be the allowing of direct-to-consumer pharmaceutical advertising.
New Zealand and the United States are the only countries in the OECD that permit direct-to-consumer advertising of pharmaceutical drugs. This issue warrants a separate discussion.
For me, the lesson is that healthcare expenditure and health are not interchangeable. Health is also built through nourishing food, physical activity, sleep, strong families, meaningful relationships, healthy environments, purposeful lives and personal responsibility. Medical technology can assist enormously, but it cannot manufacture health on its own.
What happens to the family doctor?
My prediction is that general practice may be approaching a profound restructuring. Initially AI assists with notes. Then it assists with research, medications, investigations and interpretation. It begins integrating information from wearable devices and home monitoring. It follows people between consultations, flags changes and manages routine follow-up. Eventually the patient begins interacting directly with the AI rather than accessing it through the doctor.
At this juncture, a clear economic question emerges: why is a doctor necessary for every routine interaction? Could AI and robots handle the entire process or consultation? How dystopian would that scenario be, and could we be just a few years away from it becoming reality?
Even though I might not prefer it, I can envision a scenario where the AI health agent serves as the initial point of contact, with human doctors overseeing larger groups and stepping in when the system detects complexity, uncertainty, or danger. Specialist and procedural medicine might be less immediately affected. Until robots can perform surgeries, we will still require surgeons to operate, ophthalmologists for procedures, and skilled clinicians for physical interventions that cannot be conducted via a computer screen. However, the role of the traditional family doctor may undergo significant changes, and mosty at risk of redundancy, as much of general practice involves collecting information, interpreting it, and determining the next steps.
This is my forecast, not a set path for New Zealand healthcare. Significant technical, clinical, ethical, legal, and social challenges lie between our current state and such a system. I am worried that technology is advancing faster than the discussions on ethics and safety, as well as humanity's role in this context.
The dangers should not blind us to the benefits
It would be equally foolish to reject this technology simply because it presents risks. AI could give people in remote communities access to medical expertise that is currently difficult to obtain. It could identify dangerous medication interactions, detect subtle changes in health, reduce paperwork, monitor chronic illness and help an exhausted doctor recognise something they might otherwise miss.
For an elderly individual with a complex medical history, an AI that can store and analyze 40 years of health data would be incredibly beneficial. For a general practitioner seeing a new patient, this system could offer a clear summary of the patient's history, rather than requiring the doctor to sift through disjointed records in a brief 15-minute appointment. Even the most brilliant physicians have a limited capacity for retaining information. This is where AI steps in as the assistant, and the doctor's personal librarian.
This is why I do not see the desirable future as doctor versus artificial intelligence. I see the possibility of a doctor equipped with an extraordinarily capable AI companion. The AI remembers everything, searches everything, watches for patterns and does not become tired late on a Friday afternoon. The doctor brings physical examination, judgement, accountability, experience and, above all, the human relationship.
“I do not see the desirable future as doctor versus artificial intelligence. I see a doctor equipped with an extraordinarily capable AI companion.”
That combination could be considerably better than the general practice medicine we have today.
Progress should serve the person
The philosophy behind FreeRangers includes a simple idea: progress should serve humanity rather than replace it. I think that principle is particularly appropriate as we decide how artificial intelligence should enter healthcare.
I do not want to preserve an outdated model of general practice merely because it is familiar. If AI can remove paperwork, improve diagnosis, remember my complete medical history, identify dangerous patterns, save money, and give my doctor more time to actually practise medicine, then I welcome it. What I do not want is for efficiency to become the only measure of success.
There should, in my view, remain a human being within the transaction, particularly at the coalface of healthcare. The family doctor has traditionally occupied that position. The GP knows that the patient is not merely a collection of laboratory values, diagnostic codes and physiological signals. The patient is a person with a history, a family, fears, hopes, responsibilities and circumstances that may never fit neatly into an algorithm.
“Artificial intelligence may eventually provide better continuity of information than the traditional family doctor while still being unable to provide continuity of human relationship.”
We should try to preserve the best of both.
I would be disheartened by the idea of no longer seeing a family doctor and only dealing with an AI health agent that relies on a microchip implanted in my body to decide my next actions. While such a future might be technologically possible and economically attractive, it would lack humanity.
Whether it represents better medicine is a question we should be asking now, while we still have the opportunity to decide what kind of healthcare system we want.
“Progress should serve humanity rather than replace it.”
Further reading: Gloria J. Maloney, Miracle Machines or Medical Mirage? The Dangerous Hype of AI Healthcare.
Medical Disclaimer
This information is provided for educational purposes only and is not intended as personal medical advice, diagnosis or treatment. Always consult an appropriately qualified healthcare professional regarding your individual circumstances, medications and health conditions.
Comments