Who holds the keys to the vault?

As healthcare spending spirals, and the system staggers under the weight of burgeoning stressors, there is an abundance of analysis regarding different but invariably 'broken' elements demanding urgent overhaul.

Interestingly, amongst the cacophony of shrill voices, less attention is devoted to decision-making along the clinical trajectory, and its direct, continual impact on resource utilization.

This is all the more pressing given Canada's unique position. A publicly funded system can't rely on market forces to allocate resources. We've committed, in principle, to a moral economy. But in practice we don't have a coherent framework for how clinical decisions should distribute finite resources. We've rejected the market without replacing it with anything else.

That analytical gap is a glaring oversight. The clinical process, and specifically the algorithmic web that weaves together the care pathway, is where resources actually get allocated. These clinical decisions are cellular in scope against the much larger healthcare ecosystem, and ultimately shape how money is spent. Along the way, myriad hyper-complex factors interact to the tune of billions of dollars.

But rarely is the master of the clinical decision thrust into the light. Why not?

I fondly remember some advice provided to me when I was a resident. The attending in a smaller ER was heading out for the evening, and leaving the department to me. Her parting wisdom: not to be afraid to call in the tech if I needed imaging. "No one gets in trouble in the real world from ordering investigations. Get that bit of medical training out of your head."

She nailed it. And it has massive implications for both bedside judgment and health spending.

Given the high stakes, it's fair to ask: should doctors be making clinical decisions based on best scientific evidence while simultaneously acting as comptrollers of finite resources? This was always a legitimate question, but with the onwards march of AI into all social systems, this debate becomes that much more urgent and provocative.

Times have changed. Clinical medicine is far more complex than even a generation back. And the economics (never a strong suit of medical education) have likewise mushroomed into a super-organism, complete with its own multi-layered membranes folded and embedded throughout. These membranes represent a remarkably resilient, self-serving, self-perpetuating, and eminently wasteful bureaucracy that has added no value to the system.

Amidst it all, the sea change that is AI increasingly outperforms in spaces where physicians once reigned supreme.

In this setting, archaic resource distribution mechanisms are no longer effective and cannot serve a sustainable, equitable system in the long term.

The archetypal Canadian patient

Consider this scenario:

A highly deconditioned elderly man, with multiple sub-optimally controlled co-morbidities—the modern-day package of Type 2 diabetes, hypertension, hyperlipidemia, chronic kidney disease, coronary artery disease with compensated heart failure—argues, in response to the suggestion that he walk 30 minutes daily, that he cannot start moving his body because his knee hurts.

Somewhere along the way, in a fit of college-induced moral zealotry, the family doc transitioned him from Tylenol No. 3s to regular NSAIDs, which subsequently cooked his already fragile sugar-soaked over-pressurized kidneys, and also wrecked his mood. It also further compromised his compliance; now he only takes his medications on Mondays, Wednesdays, and Thursdays because he, in his words, is "not that into pills." He says this with the conviction of a man who is doing someone a favour by participating in the charade in the first place.

An x-ray confirms the obvious: osteoarthritis of the knee. In his wisdom, the family doc refers to the orthopod.

Years later, when pressed in heaven, the family physician acknowledges that the man was a poor surgical candidate. But, he tells God, that's for the surgeon to decide. The referral was for an opinion on the knee, to determine whether the man should receive knee replacement to manage the osteoarthritis. A consultation, in medical parlance.

(Down the heavenly queue, the orthopod waits her turn impatiently, cursing all the unvetted referrals piled up on her desk. Her 11-year-old kid, she mutters bitterly, could do a better job managing those referrals).

In any event, the man doesn't do well. His sugary blood ensures that a post-op wound infection takes hold almost as soon as he is wheeled out of the OR. The following week, failing to mobilize and with a worsening infection, they go back in to wash him out.

Several days later, he becomes confused. Before someone brightly turns off the hydromorph drip—for the patient, the drip was the lone bright spot throughout their grim hospital odyssey—he gets a septic work-up. Somewhere along the line, a chest CT is ordered in response to a dyspneic episode caused by opiate-induced constipation and attendant straining while on the commode. There's no pulmonary embolism, no consolidation, and minimal failure—the lungs are surprisingly fine—but incidentally an adrenal lesion is noted. Two consults and several further investigations ensue.

A year out, the man walks less than before. His GFR is perpetually 20 points lower, now in the mid-teens. His BMI is up and muscle mass down. His med list is longer. Blood pressure remains poorly controlled. He still doesn't take his medication regularly and the hospital stay confirmed his previously-held suspicions: the healthcare system is in shambles and populated by idiots. At least he got a chronic hydromorphone prescription out of it all, much to the chagrin of the tired family doctor, who puts him on the three-year waitlist for that most Sisyphean of specialties, the pain clinic. Fortunately for the patient, before he arrives at that purgatorial waystation to receive the perfunctory pregabalin-duloxetine cocktail, he manages to fall and break a hip.

And so begins the final act in this individual's terrestrial career. The setting for his swan song? The discordant tertiary care facility, an all-too-familiar geriatric warehouse in 21st century Canada. To the detriment of public coffers and personal well-being alike. There are no winners. He dies one evening, alone, amidst alarms and cursing care aides. He had refused to sign a DNR and so the nurses enact a slow code as he fades off.

As absurd as this sounds, I don't know a person practicing in primary care for any length of time who doesn't have a similar story—whether they share it or not. I certainly have a few tucked away in the cabinet labelled "Could have done better."

Our current systemic state

In an environment of self-regulated resource utilization no one caps (nor monitors) a clinician's expenditure of public funds. Coupled with a culture that exalts the 'diagnosis'—while equating resource stewardship with a fool's errand—there is a laissez-faire frontier-like mentality when it comes to spending what's not yours.

The professional make-up greatly exacerbates this Wild West feel. Medical education and corresponding professional socialization abhor the type 2 error. The dreaded false negative. This near-pathological aversion, developed through years of accentuating 'interesting' cases—a euphemism for statistical outliers—morphs effortlessly into a veritable root mass of potentially catastrophic outcomes jockeying for position on the differential.

The result: over-investigation. The system becomes flooded. The insatiable appetite for further testing, the more elaborate the sexier. And the patient—by virtue of human nature—craves the same. Google searches never suggest doing nothing.

Herein lies an inherent contradiction in the physician role as the patient-centred practitioner and the steward of publicly-pooled and finite resources. In our conservative ethos—for the sake of patient best interests, fear of medicolegal repercussions (as my ER mentor observed, hard to get sued for ordering too many tests or making too many referrals), lack of clinical courage, or whatever—we, in healthcare, all err on the side of more.

Meanwhile, indisputably necessary interventions requiring timely execution for optimal results are too often logjammed back in the queue.

To add some embarrassment to the picture, the profession isn't even consistent in its rabid consumption. Inter-operator variability in resource utilization—related to the variability in clinical approach—is rampant, which further compounds the snarl. Family doctors, vaunted in the mainstream narrative as a lynchpin operator in the system's machinations, have wildly varying practices in terms of test requisitioning, referral patterns, and treatment decisions, for instance. The examples are countless, and often mortifying when audited with even cursory scrutiny.

In a similar vein, specialists hide behind guidelines developed without sufficient attention paid to resource scarcity. After all, they are charged as the knowledge keepers of their defined realm, tasked with definitively identifying and managing all problems under their purview. Incidentally, whether the human being's overall wellbeing is negatively or positively impacted by the process is extraneous—an annoying distraction, if considered at all.

There is zero incentive to order less, to intervene less, to prescribe less. Remember resource stewardship, embodied by the Choosing Wisely ideology trotted out in medical school? It was an intelligent, articulate argument. And bound to fail from the start, though few would have thought it could do so in such spectacular fashion. Think of the quaint Ottawa Ankle Rules. On the far distant shore of nostalgia, near where there be dragons, Norman Rockwell tips his hat approvingly.

Abandonment of the seminal gatekeeping role on the front line has been equally detrimental. In the opening vignette, the diagnosis could likely have been reasonably established in one part of one visit, without any imaging. Our 80-year-old could have been told that his knee hurts because he has osteoarthritis and that there are several possible interventions, first-line being painful physical rehabilitation with or without the guidance of a physiotherapist—a move which would serendipitously improve other parameters of his chronic disease picture.

One intervention not on the table should have been a trip to the orthopod and ultimately her OR.

An elder physician once told me, Don't send them to a surgeon unless you want them cut. Similar maxims could just as easily apply to other habitual primary care moments. Don't send them for imaging unless you can do something about what you find. Don't intubate unless you're convinced you will be able to extubate. Don't prescribe unless it will make things better, etc.

A key role of the now-defunct general practitioner was to protect patients from specialists, useless (or even harmful) prescriptions, meaningless labels, and the ilk. Medicine can be harnessed for good, to be sure, but also with potential to cause harm when incorrectly deployed, like any form of magic.

But, it takes courage to stand firm on what is viable and what is not. Far easier, and safer from both patient and medicolegal perspectives, is for a physician to kick the can down the road.

And so, with the demise of generalism throughout the sector—and the corresponding inability to function outside a narrow sliver of specific conditions—the faculty to effectively gatekeep has gone the way of the dodo bird.

Who guards the citadel?

Under the current miasma, if physicians are not best positioned to simultaneously manage both clinical care and resource distribution, who then should take the reins? It's a contentious question, especially as technology reshapes the sector. The much-maligned bean counters within healthcare? Government purse holders? Ethicists (certainly the healthcare equivalent of that kid who intentionally positions themselves to minimize risk of touching the ball)?

Some intersection between those with the funds (government) and those whose decisions determine how those funds are spent (clinicians, hopefully with some reluctant ethicists in tow) can likely be found near the mathematic fundamentals of probability and risk shared by both medicine and accounting.

Medicine is ultimately about probabilities. Until the day when it is possible to rule out everything—a logical impossibility under the inductive method—and yet have precise enough tests to rule in only one thing (in other words, the ability to answer a question without incidental findings arising through the process of investigation), we are stuck with probabilities.

The crux then becomes, what is a tolerable risk of not pursuing a diagnosis? Flipped around, what is the cost of chasing increasingly obscure diagnoses in the quest for a concrete answer, often without any clinically significant improvement in outcome—except for a House-like diagnostic triumph? The process results in increasingly diminished returns at the margins.

A poor investment.

AI at the gate

With the ongoing emergence of AI, complexity can be mitigated through the greatly enhanced capacity for calculation. Multiple variables can be inputted and a numerical result—so coveted by clinicians—can be generated.

Family physicians have been previously imagined as natural overseers of the patient journey. And perhaps at one time, they were. But in 21st century clinical Canada, that position has increasingly morphed into a Fata Morgana, an anachronistic mirage reflecting a bygone era. The family doc refers onwards and washes their hands of future decision-making. Got diabetes? See endocrine. Heartburn? Request a scope from GI. No wonder waitlists for cancer diagnoses remain untenable.

It is a sad pronouncement, but one that is glaringly obvious: in far too many cases, an AI doctor could do as good a job as a family doctor—and often better. The field needs to confront its looming obsolescence—or get out of the way.

Similarly, the specialist on the receiving end functions from the assumption that they are being asked to offer what is in their toolkit. Their creed is neither to vet referrals, nor to consider the whole person when elucidating what is on offer. And regardless, that algorithm is cleaner when followed by the machine.

In these unsettled seas, there is no captain, no steward of finite resources. The vessel floats through the waters rudderless. To be clear, not everyone operates this way. But there are incentives to act with no accountability paid to distribution. And, conversely, no incentive to deviate from the status quo.

Finding a way forward

This needs to change. To be relevant in an AI universe, the clinician needs to be reinvented. A substantial part of this reinvention is how resources are doled out.

The entrance of AI has understandably brought terror to those who guard algorithms jealously. There is a Beowulfian unease as Grendel's fangs glisten somewhere outside the gate. But unlike Beowulf, we cannot fight this Grendel head on. That doesn't mean, however, that we cannot harness its power.

Because it's fairly clear there is no real contest: the computer does it better. Much better. Chess' grandmasters held out hope in the 80s and 90s of besting the machine. Now, on bent knee, their main preoccupation is to keep the machine out of their game—because the machine is too good. There's no contest.

Clinical medicine is at a similar crossroads. It has been for some time, but too many have convinced themselves that instead of crossing the Rubicon, they can somehow continue to float down it. An example par excellence—medical schools still force students to memorize. Memorization in the era of nanotechnology? Think about it. We might as well use carrier pigeons to send in prescriptions. It would be cute if lives weren't at stake. Or billions of dollars.

This antiquated approach to the craft—a nod to the days when the doctor personified the body of knowledge contained in books too heavy to carry—is as futile as a grandmaster trying to beat a computer. Both are a throwback to an era when the well-dressed physician attended a house call, listened to the chest and gravely prescribed bed rest for 10 days, not an hour sooner.

For better or for worse, that world is gone. The ship has sailed.

To overhaul the healthcare edifice this needs to be acknowledged. It is critical to rethink how decisions are made—or generated, if you prefer—regarding the way that resources are doled out. The algorithm excels at this. It also holds potential, if done correctly—a critical 'if', one should note—to remove prejudices that plague healthcare, including those rooted in race, gender, class, and disability, to name but a few defining axes.

As with all advances, transformation is imperative. But, the human should not be removed. The algorithmic calculation trends towards human obsolescence, to be sure, but the translational component—that space between us and the machines—remains sacrosanct.

In the first instance, algorithms demand guardrails. Inclusion of externalities—the so-called social determinants of health are the obvious example—into treatment models requires direction. So do the many cases that are too nuanced to leave to machine-based decision-trees. These cruxes provide an opportunity for an expanded, evolved role for the contemporary clinician.

Just as importantly—if one accepts the public value placed on the centrality of the physician-patient bond—the cold calculus of algorithm must be offset by translation into a discourse that is both intelligible and palatable to the human condition. AI can outcompete Magnus Carlsen at nearly every turn. But the chess world wants Magnus. There is a relationship between the player and the audience, and it extends to intangibles off the board and beyond the match. Without getting into the weeds of the AI-versus-human morass, it is safe to say that—for now at least—people still crave human-human relationships. We would do well to celebrate, perhaps even nurture, that sanctity.

The engagement between clinician and patient—that relational aspect that may be all that is left of medicine as the glaring reckoning sets in—is intrinsically biased. Assuming we all want the best for the patient, and assuming that, by and large, people want some certainty as to their path forward, that bias is an accepted—even desirable—part of the compact.

But it also flies in the face of effective population-wide dispersal of finite resources. Again, that is not a bad thing in and of itself. People want to trust that their clinician has their best interests in mind and will do 'everything' to ensure that those interests are met. The clinician takes the contract seriously as well—and, likewise, confronts the added pressure of the ingrained terror of that type 2 error lurking in the shadows, ever ready to pounce on the festering inadequacy that is a primordial element of the professional psyche.

There is nothing wrong with wanting the highest of standards for those we serve. On the contrary, this impetus should be celebrated. And used to ensure market forces do not supplant moral economy principles when it comes to allocating resources in an equitable fashion. Nevertheless, that part of the process desperately requires systematization. We cannot simply allow for a Black Friday-like stampede for limited loot. The advent of AI—regardless of one's perspective—has forced the issue.

Therein lies the opportunity amidst the tumult. And none too soon. Because, in its current state, the system is not meeting the healthcare needs of the people.

Just like food supply and hunger, it is not a question of resource quantity. It's a question of distribution.