Describing health inequalities without distortion: Simple-Means MAIHDA vs Random-Effects MAIHDA

In the study of public health, researchers have long sought to understand why some groups of people are healthier than others. Traditionally, this has been done by looking at averages: calculating the average health of a city, a neighborhood, or a specific demographic group. However, averages can hide the truth. Just as a river's average depth might suggest it is safe to wade, while hiding a deep, dangerous hole in the center, an average health statistic can mask the severe struggles of specific, smaller groups living within that same area. To see the full picture of inequality, scientists need to look beyond the average and examine how the environment itself shapes the lives of individuals. This involves understanding not just who is sick, but how the places they live and their social positions interact to create health patterns. The challenge lies in measuring these patterns accurately without smoothing over the very differences that need to be seen.

A team of researchers in Sweden has developed a new way to map these health inequalities that refuses to hide the details. They focused on two very different health outcomes in the city of Malmö: the use of medication for mental health conditions and the choice to see a private doctor instead of a public one. By analyzing data from over 43,000 people, they compared a new, straightforward method of counting against the more complex statistical models that are currently the standard in the field. Their work reveals that while standard models agree closely on broad summary measures, they can smooth out the sharpest edges of inequality for specific small groups, making them appear less distinct than they truly are. The researchers found that a simple, direct count of what is actually happening in the data provides a clearer, more honest picture of health disparities than the complex models that try to predict what should be happening.

The researchers began by dividing the population of Malmö into 300 distinct groups. Each group was defined by a combination of where a person lived and their social characteristics, such as their age, sex, and income level. This created a detailed mosaic of the city, where every neighborhood contained twelve different social profiles. They then looked at two specific health behaviors. The first was the use of psychotropic medication, which is prescribed for mental health issues. The second was the choice to see a private general practitioner rather than a public one. These two outcomes were chosen because they represent opposite ends of the spectrum: one is driven almost entirely by personal circumstances like income and age, while the other is heavily influenced by geography and where one lives.

When the researchers applied their new method, which they call Simple-Means MAIHDA, they simply counted the number of people in each of the 300 groups who used the medication or chose a private doctor. They did not try to adjust these numbers based on statistical assumptions or mathematical formulas. Instead, they reported exactly what the data showed, including the uncertainty that comes with smaller groups. For the use of mental health medication, the results showed a clear pattern driven by money and age. People with lower incomes were significantly more likely to use these medications than those with higher incomes, regardless of which neighborhood they lived in. The differences between rich and poor neighborhoods were very small, and the environment itself played almost no role in shaping these health outcomes; the inequality was purely a matter of individual social position.

In stark contrast, the choice to see a private doctor was shaped powerfully by geography, but not by replacing personal income. The researchers found that high-income individuals were consistently more likely to choose private doctors than low-income individuals across all areas. However, geography acted as a powerful amplifier: the gap between rich and poor residents was significantly wider in wealthy neighborhoods than in poor ones. The environment didn't just shift the overall numbers; it actually changed the shape of the inequality. In wealthy areas, the difference between rich and poor residents was even wider than in poor areas. The geography itself amplified the social divide, making the choice of healthcare provider a story of both personal status and location.

The paper argues that the standard statistical models used by most researchers, known as Random-Effects models, fail to capture these sharp realities for specific groups, even though they agree on the overall picture. These models work by "shrinking" the estimates for small groups toward the overall average. The logic is that if a group is small, its data might be unreliable, so it is safer to assume it is closer to the city-wide average. The researchers demonstrate that this process acts like a filter that blurs the edges of the map. For the mental health medication, the standard model did not significantly change the finding that geography was negligible, as the geographical effect was already absent. However, for the private doctor choice, the standard model pulled the extreme values of the wealthy and poor neighborhoods closer together, making the geographical divide look less severe than it actually was for specific small strata.

The researchers emphasize that this "shrinking" is not just a mathematical quirk; it has real consequences for how we understand inequality. When a model smooths out the data, it can make a small, marginalized group appear to have a health risk that is closer to the average, when in reality, they might be facing a much more severe situation. By reporting the raw, observed numbers, the new method preserves the true magnitude of these inequalities. It shows that for some health issues, the place you live is a powerful force that reshapes your life chances, while for others, your personal circumstances are the dominant factor. The study does not claim that the complex models are useless; they are good for making predictions when data is sparse. But for describing the reality of health inequality as it exists right now, the researchers argue that the simplest approach—counting what is there—is the most honest.

The study concludes that to truly understand and address health disparities, scientists and policymakers must look at the data without smoothing it over. The new method provides a tool to do this, allowing researchers to see exactly how much of an inequality is due to where people live and how much is due to who they are. In the case of Malmö, the data revealed that while mental health struggles are largely a story of individual poverty, the choice of healthcare provider is a story of geography and segregation that amplifies social differences. By refusing to hide the small, sharp differences in the data, the researchers offer a clearer path toward understanding the true structure of health in society. The work suggests that if we want to fix health inequalities, we must first see them exactly as they are, not as a statistical model thinks they should be.