Undesired Variability

2026

Why does the same patient, with the same diagnosis, receive different care depending on the day, the department, or the professional? The best healthcare systems in the world are not just those with the best typical patient. They are those with the smallest gap between the best and worst patients they treat.

There is a comfortable idea that persists in the collective imagination. Medicine is a science, and science tends towards homogeneity. Identical patients should receive identical treatments. Clinical guidelines exist precisely to ensure that this happens.

The reality of the data tells a different story.

In almost every area where variability was measured, it was found. These are not small fluctuations, but differences ranging from two to ten times in the probability of receiving a particular treatment, depending on the location, service, or doctor making the decision. And this variation is not explained by the patient’s severity or their preferences.

We call this difference, unwanted variability . It is perhaps the greatest silent cost of modern healthcare systems.

What do international data say?

In 2014, the OECD published a landmark study on geographical variation in healthcare, analyzing 13 countries. Some of its conclusions deserve close attention.

The cesarean section rate in Portugal is about 50% higher than in Finland, even after adjusting for the mother’s age. In Italy, the internal variation between provinces can be as high as six times. Cesarean sections are more frequent in private settings and among women with higher socioeconomic status, which hardly corresponds to a different clinical need.

The rate of knee replacement surgery varies by up to four times between OECD countries, and by up to five times between regions in Canada. Cardiac procedures vary by more than three times between countries, and in more than half of OECD countries they represent the area with the greatest internal variation.

Source: OECD, Geographic Variations in Health Care, 2014

In the UK, the NHS Atlas of Variation revealed differences that seem improbable in a single system. There is a variation of more than ten times in the percentage of patients receiving reference treatment after a transient ischemic attack (TIA). Amputation rates in type 2 diabetics are almost double in the southwest compared to the southeast, and it is known that about 80% of these amputations are potentially preventable.

Source: NHS Atlas of Variation in Healthcare, Public Health England

In a more recent analysis, surgery for stress urinary incontinence in England varies between 20 and 106 procedures per 100,000 women per year, depending on the region. Even after adjusting for age, ethnicity, chronic disease, and socioeconomic deprivation, more than 60% of the variation remains without demographic explanation.

Source: BMJ Open, 2019, England national cohort study

The Portuguese case

In Portugal, the evidence is less systematic, but equally revealing. Cesarean section rates consistently place the country among the highest in Europe, despite the existence of guidelines from the Directorate-General of Health on the subject. The Hospital Referral Network for Cardiac Surgery, under public consultation since 2023, explicitly acknowledges significant regional asymmetries in surgical activity between the North, Center, and South, with very different waiting times for the same type of patient.

But the most common example isn’t in central hospitals. It’s in small prescribing decisions. The same patient with the same diagnosis of hypertension might leave a consultation with three different medications, depending on the clinician’s preference, rather than the patient’s clinical response. The same headache might result in a CT scan or a prescription for paracetamol, depending on the emergency department where the patient presents.

None of these decisions is necessarily wrong. But their variability cannot be explained solely by differences between patients.

Because it happens

The literature on the subject, consolidated by the Dartmouth Institute since the 1970s, distinguishes three categories of care where variability appears.

Supply-sensitive care, where the volume of care primarily reflects installed capacity. There are more surgeries where there are more surgeons, not where there is more disease.

Patient-sensitive care, where there is more than one clinically valid option and the choice should reflect the patient. In practice, it primarily reflects the physician’s style.

Effective care, where there is clear evidence of the best option and yet it is not followed uniformly. Here, variation represents substandard care.

Source: Sutherland & Levesque, J Eval Clin Pract, 2020, review of 836 OECD studies.

The invisible cost

Variability has a double cost.

The first is financial. The British Atlas showed variations of up to 50 times in spending on certain surgeries between regions, with no corresponding difference in results. That’s money that could buy better healthcare if it were applied consistently.

The second is ethical. Variation means that a patient receives or does not receive quality care depending on the geography or the schedule of the clinician attending them. It is a silent, undeclared inequality that coexists with systems that define themselves as universal.

The right question

Most healthcare systems have invested decades in improving average quality. The next question is different. What is the dispersion around that average? And I would even add, what is the advantage of using the average as a reference?

Reducing unwanted variability is not about eliminating clinical autonomy. It’s about ensuring that this autonomy is exercised within a defensible range of evidence. It’s about measuring what you do, comparing it with peers, and being willing to discuss the results.

If we want to move towards the concept of Value in Healthcare, how will this be possible given the strong heterogeneity of care pathways? How can we ensure outcomes that are generally similar, with care pathways exhibiting significant and undesirable variability? Are we focusing on clinical outcomes and forgetting the robustness of the process and its monitoring?

The best healthcare systems in the world are not just those with the best typical patient. They are those with the smallest gap between the best and worst patients they treat.

Next article: Efficient steps do not make an efficient process, about the dead time between steps where good processes are lost.