Why Diabetes Prevention Works Better When You Find the Right People First

Effective diabetes prevention is less about a perfect intervention than about aiming it well. Risk stratification is how you aim. It means sorting a population by how likely each person is to develop the disease over a defined window of time, so the most intensive help goes to the people who stand to gain most, and people at low risk are spared effort that would do little for them. Treating everyone identically feels fair, but it is usually the least effective and least equitable option. This is educational, not medical advice; to know where you personally sit, talk with your own clinician.

I have spent my research life on the production side of risk. My doctoral work at the Lund University Diabetes Centre is on the genetics of type 2 diabetes, and a meta-analysis I co-authored in Diabetes Care looked at how a basic feature of glucose biology differs across populations. Both point at the same lesson: risk is not spread evenly, and pretending it is wastes the resources prevention depends on.

What does risk stratification actually mean?

Here is the short, quotable version. Risk stratification is the practice of dividing a population into groups by their estimated probability of a future outcome, then matching the intensity of an intervention to each group's risk. In diabetes prevention, the outcome is usually progression from a borderline glucose state to type 2 diabetes within several years.

The idea rests on a fact that sounds dull until you sit with it. Most people in any large group will not develop the disease over any given span, even when it is common. Offer an intensive lifestyle program to a few thousand average adults, and the great majority were never going to progress in that window anyway. The benefit concentrates in a minority whose trajectory was genuinely bendable. Find that minority and the program looks powerful. Spread it flat across everyone and it looks weak and expensive. Stratification is not a way of withholding care. It is how you find the people for whom care will actually move the number.

Why treating everyone the same is the expensive option

Public health has long balanced two approaches: a small intervention aimed at the whole population to nudge the average, and a strong intervention aimed at the people at highest risk. Good prevention uses both, and the trap is assuming that uniform effort equals uniform benefit. A flat program spends finite things, including clinician time, screening visits, and the patience of people asked to change their lives for a threat that was statistically distant for most of them. Spend that on people at low absolute risk and you buy very little while burning the budget and goodwill the high-risk group needed. There is a quieter cost too. Tell a large low-risk group they are "pre" something and you generate anxiety and follow-up testing that chases a risk which was never going to materialize. Stratification protects low-risk people from that overreach as surely as it directs the high-risk toward real help.

How do you sort people by risk without getting it wrong?

The mechanics are not exotic. You combine signals you already have: age, family history, body size and where the weight sits, blood pressure, a fasting glucose or an HbA1c, increasingly a genetic risk component. A model weighs these and returns an estimated probability. Validated questionnaires can do a surprising amount before a single lab test is ordered, which matters when screening capacity is the bottleneck.

But a risk model is only as good as the population it learned from, and a model tuned on one group can be quietly miscalibrated for another. The Diabetes Care meta-analysis I worked on examined how the relationship between insulin sensitivity and the body's insulin response shifts across ethnic groups. The practical upshot is that the same glucose number, or the same body-mass index, can sit at a different distance from disease depending on a person's background. Someone may carry meaningful risk at a body size that a model calibrated elsewhere would wave through as safe. A stratification tool that ignores this will systematically under-refer the very people who needed referring, and do it invisibly, with the false confidence of a number.

That is the heart of precision prevention. It is a refusal to let one threshold, calibrated on one population, decide who gets help. Measure the relevant signal, account for context, and stop treating a single cutoff as if it meant the same everywhere.

Where this meets the clinic and the software

A clinician in a short appointment cannot hold all of this at once: the guideline, the contraindication, the patient whose background means the usual threshold sits in the wrong place. That gap, between what the evidence knows and what fits inside a fifteen-minute visit, is the real bottleneck in prevention.

It is the problem I worked on directly. I co-developed EASY Diabetes, an AI clinical decision-support system for type 2 diabetes built with patients and clinicians. A tool like that does not replace judgment. It carries the nuance, surfacing the risk a flat rule would miss and flagging the patient who is not behaving like the average. Its registered randomized controlled trial (EASY-1, NCT03258268), a multi-clinic study, compared it against standard care. The efficiency half is not a footnote. Stratification that takes too long to apply does not get applied, and the patient at the edge slips back into the undifferentiated middle.

Such tools also have to know their own uncertainty. A system that confidently assigns the average risk to an edge-case patient has reproduced the original error in software with a cleaner interface. The harder task is building something that can say, in effect, look closer, this one does not fit.

The honest limits

Risk stratification is a sorting tool, not a verdict. A high-risk label is a reason to act, not a sentence, and the models are probabilistic; they inherit the gaps in the data they learned from. Used carelessly, the same machinery that targets help can entrench old inequities by encoding a narrow population as the norm.

So the discipline is twofold. Aim the strong help at the people who can use it, and keep checking that your aim is fair across the groups your tool was never built around. Do both, and prevention stops being a flat tax on everyone's attention and becomes the right help, reaching the right person, early enough to matter.