How Six Nations Diverge on the Road From Health to Stroke

Stroke rarely arrives without warning. For most people, the path to a cerebrovascular catastrophe is paved over decades, passing through recognizable waypoints: first hypertension, sometimes diabetes, then the dangerous combination of both. A new analysis published in BMC Public Health has mapped those waypoints with unprecedented geographic scope, tracking nearly a quarter of a million middle-aged and older adults across six countries to ask a deceptively simple question: does the journey from health to stroke look the same everywhere? The answer, the researchers report, is a qualified no, and the differences they uncovered carry significant implications for how nations should design their prevention strategies.

The study, led by Lianlu Jiang of the First Affiliated Hospital of Hunan College of TCM, together with Xiaodan Ren and Feng Liu, drew on harmonized longitudinal data from six of the world’s major aging surveys: CHARLS in China, ELSA in England, HRS in the United States, KLoSA in South Korea, MHAS in Mexico, and SHARE, which spans continental Europe. In total, the team followed 247,863 participants aged 45 to 100 years, accumulating 599,161 observed transitions between health states, including 40,911 transitions to death. By pooling these cohorts through the Gateway to Global Aging Data harmonization project, the researchers were able to compare disease trajectories that are usually studied only within single national borders.

Methodologically, the paper stands out for its use of a multi-state modeling framework, a statistical approach that treats disease progression not as a single event but as a sequence of movements between defined states. Here the authors defined six states: healthy, hypertension only, diabetes only, the combination of hypertension and diabetes, stroke, and death, with death serving as an absorbing state from which no further transitions occur. Because participants could die before ever developing stroke, the investigators modeled death as a competing risk rather than simply censoring those individuals, a decision that guards against the well-known bias of overestimating disease incidence when mortality is ignored. They fitted cause-specific hazard models with complementary log-log links, estimating the association of age, sex, education, and smoking with each of 17 observed transition types, while adjusting for within-person correlation arising from repeated measurements.

The baseline picture alone revealed striking international variation. The proportion of participants classified as healthy at the start of the study ranged from 45.5 percent in Mexico to 74.5 percent in China, a near twofold difference that reflects divergent lifetime exposures to dietary patterns, physical activity, healthcare access, and diagnostic intensity. Conditional on leaving the healthy state, the most common destination was hypertension only in every single country, accounting for between 42.2 and 66.9 percent of exits, followed by death, which ranged from 9.0 to 33.7 percent. In other words, wherever the researchers looked, elevated blood pressure was the dominant first stop on the road away from cardiovascular health, but how quickly people reached that stop, and how often they instead died without any prior cardiometabolic diagnosis, differed substantially from nation to nation.

Age emerged as the most consistent driver of progression. Each additional year of life was associated with an increased hazard of progressing to stroke from the healthy state, with hazard ratios ranging from 1.032 to 1.084 across the six cohorts, meaning the annual risk climbed between roughly 3 and 8 percent per year of aging. The age gradient persisted through intermediate states as well: moving from hypertension to stroke carried hazard ratios of 1.022 to 1.050 per year, and from combined hypertension and diabetes to stroke, 1.013 to 1.032. Death itself showed the steepest and most uniform age association, with hazard ratios of 1.107 to 1.125 per year. Yet even this seemingly universal biological constant bore the fingerprints of geography. Formal interaction tests using country-by-covariate terms identified statistically significant cross-national differences in age effects for all major transitions, with Wald tests yielding p values below 0.05, indicating that the steepness of the age gradient itself varies by country.

Beyond age, the study identified modifiable and socioeconomic determinants whose influence also refused to conform to a single global template. Smoking was associated with progression to the combined hypertension-plus-diabetes state, a finding consistent with tobacco’s established role in accelerating vascular and metabolic damage. Education showed protective associations in several cohorts, echoing a large literature linking socioeconomic advantage to earlier detection, better treatment adherence, and healthier environments. The fact that these associations appeared in some populations but not uniformly across all six underscores a central message of the paper: the determinants of cardiometabolic progression are not simply present or absent worldwide, but differ in strength depending on context.

What might explain such heterogeneity? The authors point to plausible biological and structural candidates. National differences in diet, sodium intake, obesity prevalence, physical labor patterns, and air pollution could genuinely alter the speed at which vascular damage accumulates. Healthcare systems matter too: hypertension and diabetes are diagnoses, not just physiological states, and countries with more aggressive screening will detect and record these conditions earlier, shifting where their populations appear to sit within the state model. Mexico’s relatively low healthy proportion, for example, may reflect both a true high burden of cardiometabolic risk and the epidemiological transition the country has undergone in recent decades. Conversely, China’s high healthy fraction could partly reflect younger cohort composition or underdiagnosis in rural areas. The study’s design cannot fully disentangle these mechanisms, and the authors are explicit that design and measurement differences across the cohorts should be considered competing explanations for the observed patterns.

The technical sophistication of the competing-risk framework deserves emphasis, because it changes how the results should be read. In a conventional survival analysis, participants who die before experiencing stroke are removed from the risk set, implicitly assuming they could still have experienced the event. Here, the researchers kept individuals at risk until death or censoring and estimated cause-specific hazards for every possible transition, allowing death to compete directly with cardiometabolic progression. With more than 40,000 deaths among the observed transitions, this was not a statistical formality: in populations with high background mortality, ignoring competing risks would have materially distorted the estimated probabilities of moving from hypertension to stroke. The complementary log-log link, meanwhile, provides a proportional hazards formulation well suited to interval-censored panel data collected in waves every two years or so, the characteristic structure of aging surveys.

For public health policymakers, the practical takeaway is that one-size-fits-all prevention programs are unlikely to be optimal. Because progression patterns and their determinants differ statistically across countries, interventions should be tailored to the dominant local pathways. In countries where the transition from health directly to stroke is relatively rapid, aggressive early blood pressure screening may yield the greatest benefit. Where the hypertension-to-diabetes cascade dominates, integrated cardiometabolic care that manages both conditions simultaneously becomes the priority. The consistent age gradients confirm that aging itself remains the irreducible backdrop against which all prevention operates, but the heterogeneous magnitudes suggest that national context, from food systems to clinical practice, modulates how quickly the biological clock translates into disease.

The study also demonstrates the power of harmonized international data infrastructure. None of the individual aging surveys was designed to answer cross-national questions about multi-state disease progression, yet the Gateway to Global Aging harmonization made it possible to align variable definitions, follow-up structures, and health-state classifications across six continents’ worth of evidence. As aging populations grow worldwide and stroke continues to rank among the leading causes of death and disability, this kind of comparative architecture will become essential for identifying which countries are beating the odds, and why. The findings, published as an open-access article with a permanent DOI, arrive with appropriate statistical caution, but their central message is hard to escape: the road from health to stroke is universal in its waypoints, yet every nation travels it at a different speed, and prevention must follow the local map rather than a single global itinerary.

Subject of Research: Cross-national progression patterns from healthy state through hypertension and diabetes to stroke in aging populations

Article Title: Cross-national patterns of cardiometabolic disease progression from healthy state to stroke: a cause-specific Cox transition analysis of six international cohorts

Article References: Jiang, L., Ren, X., & Liu, F. (2026). Cross-national patterns of cardiometabolic disease progression from healthy state to stroke: a cause-specific Cox transition analysis of six international cohorts. BMC Public Health. https://doi.org/10.1186/s12889-026-29756-1

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29756-1

Keywords: stroke, hypertension, diabetes, cardiometabolic disease, multi-state model, cause-specific hazard, cross-national comparison, aging, CHARLS, ELSA, HRS, SHARE