Not how many abnormalities we can identify, or how accurately we can identify them. But how much healthier people become, and how much longer they live, because we found something.
Thinking about detection in terms of additional years of healthy life changes the way you approach the problem. It forces you to look beyond the technology and consider the entire process of finding patients, getting them into care, and ultimately improving their outcomes.
There is a rather interesting idea in health economics called the Quality Adjusted Life Year, or QALY. It attempts to measure something medicine has traditionally had difficulty putting a number on. Not simply how long people live, but how well they live.
The concept is straightforward. One year of life in perfect health is worth one QALY. A year lived with significant illness or disability is worth some fraction of that. An intervention that allows someone to live longer, or spend more of their remaining years in better health, generates QALYs.
It is an imperfect measure, of course. Reducing the human experience to a number between zero and one has obvious philosophical problems. But it gives us a way to compare very different medical interventions. A cancer treatment, a cardiac procedure and a diagnostic technology can all be evaluated in terms of the additional years of healthy life they produce.
And importantly, what it costs to produce them.
This is where things get interesting for diagnostics.
We tend to think of detecting disease as an inherently good thing. Find it early, treat it early, and everyone is better off. But that is not necessarily true. Detection itself does not improve anyone’s health. It simply provides information. The value comes from what happens next.
Consider aortic stenosis, a common form of structural heart disease in which the aortic valve progressively narrows. The disease can develop over many years, often without obvious symptoms. Eventually the heart begins to struggle against the obstruction. Once severe symptoms develop, the prognosis without valve replacement can be quite poor.
Fortunately, we have effective treatments. Surgical valve replacement has been available for decades, and transcatheter aortic valve replacement (TAVR) has transformed treatment options, particularly for older patients.
The problem is getting patients into the appropriate treatment pathway at the right time.
A patient whose disease is recognized and monitored can be evaluated for intervention when the clinical evidence supports it. Another patient with essentially the same disease may remain undiagnosed until symptoms become severe, cardiac function deteriorates, or an acute event brings them into hospital.
Both patients may eventually receive treatment. But they have not necessarily had the same opportunity to benefit from it.
This distinction matters when calculating QALYs. The value of earlier detection is not simply that a diagnosis was made sooner. It is that the diagnosis may change the patient’s clinical trajectory. It may prevent deterioration, preserve cardiac function, avoid a dangerous presentation, or make effective treatment possible during a more favorable window.
Those changes can translate into additional years of life and better quality of life.
There are two populations worth thinking about here.
The first is the undiagnosed population. These are people living with clinically important disease that has not yet been recognized.
ome have no symptoms. Others have symptoms attributed to aging, poor fitness or an unrelated condition. They are effectively invisible to the healthcare system, at least as far as their underlying cardiac disease is concerned.
The second is the care-gap population. These patients may already have a diagnosis, but are not receiving the surveillance, specialist assessment or treatment their condition warrants. Being diagnosed and being effectively managed are not the same thing.
Both populations are substantial.
Population studies of valvular heart disease have repeatedly found important disease in people previously unaware of it. And the problem extends beyond the valves. Heart failure, atrial fibrillation and other cardiovascular conditions all have patients whose disease is unrecognized, incompletely evaluated or inadequately managed.
Across cardiovascular medicine, we are talking about millions of people. Not all need immediate treatment, and not every additional diagnosis will improve an outcome. But the scale of the opportunity is difficult to ignore.
It also exposes a peculiar weakness in modern medicine.
We have invested extraordinary resources in developing sophisticated treatments for cardiovascular disease. We can replace heart valves through catheters, restore blood flow to coronary arteries, and manage conditions that would once have been fatal. Yet our ability to systematically identify the people who could benefit remains surprisingly uneven.
In many cases, the treatment technology is considerably more advanced than the process used to find the patient.
So why not simply screen everyone?
Because healthcare systems have finite resources, and detecting more abnormalities is not the same as finding more patients who will benefit from treatment.
Every positive screening result creates work. Someone must evaluate it, order tests, interpret findings, arrange follow-up and decide what happens next. In structural heart disease, this often means echocardiography, specialist consultations and ongoing surveillance.
Some of that activity is precisely what we want. It represents patients entering appropriate care who might otherwise have been missed. But poorly designed detection programs can generate enormous volumes of unnecessary testing, incidental findings and referrals that accomplish very little.
A detection strategy that overwhelms diagnostic capacity may actually make the system less effective, increasing waiting times for the very patients who need timely evaluation.
Designing a better detection strategy
The interesting question, then, is not simply whether we can detect more cardiovascular disease. It is how we should go about doing it.
Where should we look? Which patients should we examine? How frequently? Who should perform the examination? How accurate does it need to be? And what happens when we find something?
These questions are closely related, and the answers can produce very different clinical and economic outcomes.
Start with the population. Aortic stenosis is predominantly a disease of older adults. Screening 30-year-olds would produce very few meaningful findings, regardless of how good the technology might be. Examining people over 65, particularly those with relevant symptoms or risk factors, is a different proposition.
Disease prevalence determines how many examinations are needed to find a patient who might benefit. It also affects how many positive results turn out to be false alarms.
Then there is the question of where to look.
Primary care seems an obvious opportunity. Patients are already there, and the examination could become part of routine care. But preoperative assessments, urgent care visits and encounters with patients being evaluated for other cardiovascular conditions may also offer opportunities to identify unrecognized disease.
These settings are not interchangeable. They serve different populations, have different disease prevalence, and offer different opportunities for follow-up. A finding during an annual examination may lead to orderly evaluation and surveillance. A finding during a preoperative assessment may uncover an important condition immediately before a planned procedure. Both can be valuable, but for different reasons.
Who performs the examination matters as well.
If every screening examination requires a cardiologist, the program is unlikely to scale very far. If it can be performed reliably by a primary care physician, nurse or another healthcare professional during an existing encounter, the economics change considerably.
Convenience is not trivial. A test requiring a separate appointment, specialized equipment and an hour of someone’s time will reach fewer people and consume more resources than one completed in a minute or two during an existing visit.
Accuracy is equally important, although it is more complicated than simply asking whether a test is good.
Sensitivity tells us how many patients with the target disease the test correctly identifies. Specificity tells us how many patients without it the test correctly excludes. Both matter, but their relative importance depends on the population and the consequences of getting the answer wrong.
Consider a hypothetical screening program examining 10,000 people, of whom 2% have clinically important disease. A test with 90% sensitivity and 90% specificity would identify 180 of the 200 affected patients. But it would also produce 980 false-positive results.
That is 1,160 positive examinations, of which only 180 represent true disease.
If every positive result leads to echocardiography, the health system must perform more than six echocardiograms for every affected patient identified. And 20 patients with disease will still be missed.
Now apply that same test in a population where disease prevalence is 10%. With identical sensitivity and specificity, 900 of 1,800 positive examinations would represent true disease. The diagnostic yield has changed dramatically, even though the test has not improved at all.
These are illustrative numbers, but they expose an important principle. The performance of a detection strategy depends as much on where and how it is used as on the accuracy of the technology itself.
There is another consideration that is easy to overlook. Finding disease is only useful if the patient subsequently receives appropriate care.
A positive examination that never leads to confirmatory testing accomplishes little. A confirmed diagnosis that is never followed up may accomplish little more. And identifying someone already receiving appropriate management may add no incremental health benefit.
The entire pathway matters, from the first examination through diagnostic confirmation, surveillance and, where appropriate, treatment.
This is where QALYs become particularly useful. They force us to follow the consequences of detection rather than simply count diagnoses.
For example, detecting moderate aortic stenosis may not lead to immediate intervention. But it may establish surveillance that allows the patient to receive valve replacement at the appropriate time, rather than presenting years later with advanced disease. The benefit arises from what happens over those intervening years, not from the diagnosis itself.
A good detection strategy must balance several objectives. It must reach enough people to matter, concentrate where meaningful disease is likely to be found, identify patients accurately, minimize unnecessary downstream work, and connect findings to effective management.
It must also fit into the way healthcare is delivered. An excellent screening program requiring an entirely new clinical infrastructure may be less attractive than a simpler program deployed across millions of existing encounters.
There is no reason to assume a single strategy will be optimal everywhere. Primary care, preoperative assessment and urgent care may each warrant a different approach, even when the underlying technology is identical.
Measuring what matters
One useful way to think about this is to ask how many additional years of healthy life a detection strategy produces for every 10,000 people examined.
Consider a hypothetical strategy applied during 10,000 routine healthcare encounters. Suppose 700 patients receive findings that meaningfully change their care, and the resulting improvements in treatment and outcomes produce 250 additional quality-adjusted years of life.
That is a quarter of a millennium of additional healthy human life, potentially generated from examinations performed in minutes, during visits that were already taking place.
The benefits would not be distributed evenly. Some patients might gain years of life through timely treatment. Others might avoid years of declining cardiac function. Many would experience little or no measurable benefit. But collectively, the impact could be substantial.
The important question is what it costs to produce those 250 years, and whether the strategy can deliver them without placing an unreasonable burden on the healthcare system.
This is a rather different way of measuring success. Instead of celebrating the number of abnormalities detected, we begin asking how many patients actually benefit, how much benefit they receive, and what it costs to deliver it.
And that is ultimately what a QALY-based economic analysis is intended to help us understand.
Beyond the technology
At Sparrow, we develop technology to detect cardiovascular disease from the sounds produced by the heart. It is fascinating work. The ability to extract clinically useful information from a brief acoustic examination using artificial intelligence opens up possibilities that were difficult to imagine even a few years ago.
But the technology is only one part of the problem.
We can build a remarkably accurate detection system and still accomplish very little if it is deployed in the wrong populations, creates too many unnecessary referrals, or fails to connect patients to appropriate care.
Conversely, a well-designed detection strategy can create substantial value by finding the right patients, at the right time, in places where they are already receiving healthcare, and directing them toward treatment that can genuinely change their lives.
This is why we think about detection in terms of QALYs, clinical pathways and healthcare resources, rather than diagnostic accuracy alone. It changes the questions we ask, the way we evaluate opportunities, and ultimately how we design the technology and its use.
The detection technology is the exciting part. It is what people see, and understandably what attracts the most attention. But it is only a small part of the overall design.
Connecting detection to real health benefits requires understanding the entire cascade, from the first examination to the patient’s eventual clinical outcome. And for a company developing diagnostic technology, taking responsibility for understanding that cascade is not someone else’s job.
It is our job.