Healthspan-lifespan gap differs in magnitude and disease contribution across world regions

Abstract

Background

Longevity gains have not been matched by equivalent advances in healthy longevity, giving rise to the healthspan-lifespan gap. This study maps, by world region, the healthspan-lifespan gap; identifies gap-associated demographic, economic, and health indicators; and deciphers disease burden patterns contributing to gap profiles.

Methods

World Health Organization (WHO) Global Health Observatory, United Nations World Population Prospects and Global Health Expenditure Database were interrogated. The healthspan-lifespan gap was quantified from estimates of life expectancy and health-adjusted life expectancy. Regression analysis evaluated healthspan-lifespan gap correlates with a spatial error model used to adjust for confounders arising from geographic proximity. Dimensionality reduction by principal component analysis and clustering by machine learning discriminated disease burden patterns linked to healthspan-lifespan gap identity. Supervised machine learning enabled validation of disease burden pattern distinctness.

Results

Charted for six WHO-designated regions, comprising 183 member states, the healthspan-lifespan gap differs in size across regions. Life expectancy, gross domestic product, and noncommunicable disease burden most consistently correlate with the healthspan-lifespan gap. Unsupervised machine learning identifies three clusters delineating global morbidity patterns. Cluster-informed stratification discerns inter- and intra-regional gap heterogeneity. Africa, although exhibiting the narrowest healthspan-lifespan gap, is overrepresented in countries with larger than predicted healthspan-lifespan gaps and shows the greatest gap expansion and disease burden pattern restructuring. In contrast, Europe is overrepresented in countries with healthspan-lifespan gaps smaller than anticipated. Projections into 2100 forecast continuous widening of the healthspan-lifespan gap across regions.

Conclusions

The healthspan-lifespan gap is universal yet differs in magnitude and disease contribution among world regions. Gap identities imposed by distinct disease burden patterns caution against global generalization, necessitating region-informed solutions to maximize equitable healthy longevity.

Plain language summary

Life expectancy has increased, yet people do not necessarily live longer in good health. In fact, there has been an increase in the period of life lived with disease. The difference between the number of years lived (lifespan) and the number of years lived in good health (healthspan) is known as the healthspan-lifespan gap. In this study, we mapped the healthspan-lifespan gap by world regions encompassing 183 countries. We identified demographic, economic, and health contributors to the healthspan-lifespan gap. Our study shows that healthspan-lifespan gaps are universal, yet differ in size and disease contribution among regions. Africa exhibited the greatest gap expansion and the most rapid changes in disease burden. We applied artificial intelligence, which recognized three distinct clusters of disease burden underpinning healthspan-lifespan gaps around the globe. Our projections forecast a widening of the healthspan-lifespan gap into the next century. Our work highlights areas that could be targeted to reduce the healthspan-lifespan gap across and within different world populations.

Garmany and Terzic measure, map, and decipher the healthspan lifespan gap across world regions. They trisect the world into three disease burden patterns, which underpin the distinct health span-lifespan profiles across geographies.

Introduction

Human life expectancy has continued to increase, surpassing presumed longevity ceilings^1–4. While extended life expectancy has arisen from declining mortality rates⁵, it has not been matched with equivalent gains in healthy longevity^6–10. The ensuing healthspan-lifespan gap underscores years lived with disease that jeopardize the longevity dividend^6,11. Charting the healthspan-lifespan gap is especially relevant in the context of geographic, socioeconomic, and health inequity^12,13. The widening of the healthspan-lifespan gap globally¹⁴ urges deep interrogation of gap pathogeography.

The present study maps the healthspan-lifespan gap by world regions, identifies gap indicators per region, and delineates region-contributing disease burden patterns. The study follows for the past two decades, and projects to the end of the century, the healthspan-lifespan gap of World Health Organization (WHO) member states across the WHO designated regions of Africa, Americas, Eastern Mediterranean, Europe, South-East Asia, and Western Pacific. WHO assembles world regions by geographic proximity, culture, and epidemiologic similarity, enabling collection, aggregation, and standardized reporting of publicly accessible data sources¹⁵.

This study finds that the healthspan-lifespan gap is universal but not uniform, differing in size and disease contribution across world regions. Life expectancy, gross domestic product, and noncommunicable disease burden most consistently correlate with the healthspan-lifespan gap. Unsupervised machine learning identifies here three clusters of global morbidity patterns that describe gap differences. Africa, although exhibiting the narrowest healthspan-lifespan gap, shows the greatest gap expansion and disease burden restructuring. The study projects widening of the healthspan-lifespan gap into the next century. The growing number of years burdened by distinct regional disease contributors shapes the healthspan-lifespan gap identities impacting healthy longevity among global populations and necessitating region-informed solutions.

Methods

Quantifying healthspan-lifespan gap

In accordance with Human Health Services regulations for the protection of human subjects in research (45 CFR §46), the present study was exempt from institutional review board approval and informed consent was not required as aggregate publicly available datasets were used. The World Health Organization (WHO) Global Health Observatory was used to extract standardized life expectancy and health-adjusted life expectancy datasets for 183 member states and six administrative regions between 2000 and 2019^16,17 with publicly available data accessed and analyzed between January 2024 and November 2024. Regions and affiliated states were defined in accord with the WHO classification scheme. Healthspan-lifespan gap was calculated as: $\mathrm{LE}-\mathrm{HALE}$, where LE represents life expectancy at birth for both sexes and HALE represents health-adjusted life expectancy at birth for both sexes. Life expectancy, health-adjusted life expectancy, and lifespan-healthspan maps were constructed using the ggplot2 package in R (v4.4.4). The life expectancy-adjusted healthspan-lifespan gap (LEA-GAP), representative of the percentage of lifespan lived with disease, was calculated as follows: $\frac{Healthspan-LifespanGap}{LifeExpectancy}.$

Measuring disease burden

Morbidity burden was computed from WHO Global Health Observatory estimates of years lived with disability (YLDs) calculated per 100,000 persons¹⁷. WHO hierarchical classification of noncommunicable diseases; communicable, maternal, perinatal, and nutritional conditions (CMPN); and injuries was applied. The relative contribution to total disease burden was calculated as: $\frac{{YLD}_x}{YLDt}$, where YLD_x represents YLD attributed to disease x and YLD_t represents total YLD. Mortality burden was assessed as years of life lost calculated per 100,000 persons.

Resolving healthspan-lifespan gap associations

Demographic, economic, and health indicators were retrieved from the publicly available United Nations World Population Prospects¹⁸, Global Health Expenditure Database¹⁹, and WHO Global Health Observatory¹⁷. Associations with the healthspan-lifespan gap were probed by linear regression. Tested indicators were z-score normalized for comparable measure of respective effect sizes. A multivariate model for predicting the healthspan-lifespan gap from demographic, economic, and health indicators was derived using a forward selection approach to maximize the adjusted R² value. Ninety-five percent confidence intervals for slopes and estimates from linear regression are provided. To assess member state performance in each region, deviation of the observed healthspan-lifespan gap from the multivariate regression estimate was quantified for member states with all parameters quantified. A healthspan-lifespan gap above or below the 99% confidence interval from the multivariate regression projection was classified as gap deviation. Overrepresentation was determined by comparing the proportion of member states from a particular region with healthspan-lifespan gaps larger or smaller than predicted using a Chi-square test of independence. To adjust for potential confounders that may be influenced by geographical distribution, a spatial error model was constructed. In brief, the rnaturalearth package²⁰ was used to extract spatial data for WHO member states with the spdep²¹ package used to create a neighborhood list and row-standardized weights based on polygon continuity.

Defining disease burden patterns

Disease burden patterns were derived from relative YLD for 22 WHO disease categories annotated based on the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10). Sudden infant death syndrome was excluded as it had no morbidity burden. The heatmap of the relative contribution of each disease category was plotted with the pheatmap package in R with hierarchical agglomerative clustering based on Euclidean distances. Principal component analysis was conducted for dimension reduction in R (v4.4.1) following scaling and centering. K-means clustering identified the number of unique disease pattern clusters. The elbow method determined the ideal cluster number. The ellipse package in R delineated 95% confidence intervals for each cluster.

Supervised classification and feature selection

Random forest classification, a supervised machine learning algorithm, was conducted with the RandomForest²² R package for classification of regions based on relative morbidity contribution of disease subclasses and to validate distinctness of clusters identified by K-means clustering, an unsupervised machine learning algorithm. Accuracy was calculated using internal cross-validation. Feature selection was conducted with the Boruta algorithm (Boruta R package)²³, a wrapper method based off random forest, to validate loading plot feature selection.

Projecting healthspan-lifespan gaps

To project the healthspan-lifespan gap from life expectancy, 183 WHO member state-specific regression models were derived from actual values over the last two decades. Life expectancy estimates from the United Nations World Population Prospects¹⁸ served as input to project the healthspan-lifespan gap up to the year 2100. Individual member state projections were aggregated to derive mean healthspan-lifespan gap values and confidence intervals for each region, namely $X_{dy}\pm 1.96\times {SEM}_{dy}$, where X_dy represents the mean projected healthspan-lifespan gap for region d on year y and SEM the standard error of the mean.

Statistics and reproducibility

For multiple comparisons, one-way ANOVA or Kruskal–Wallis test was applied after normality assessment with Wilk–Shapiro tests. Post-hoc testing used Dunn or Tukey’s Honest Significant Difference (HSD) tests with Benjamini–Hochberg correction (BH-p) to adjust for multiple comparisons. Statistics were conducted in R (v4.4.1).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Healthspan trails lifespan across world regions

Median life expectancy of 73.7 years, with a range from 50.8 to 84.3 years, characterizes WHO member states (Fig. S1). Per region, the median lifespan (minimum, maximum) is (in years) lowest in Africa at 64.1 (50.8, 77.1) years, compared to 70.4 (59.4, 84.3), 72.6 (69.1, 79.6), 73.9 (56.5, 81.0), 75.9 (64.0, 82.2), and 78.6 (69.5, 83.4) for Western Pacific (BH-p = 2.5e-4), South-East Asia (BH-p = 2.9e-3), Eastern Mediterranean (BH-p = 3.3e-4), Americas (BH-p = 1.4e-9), and Europe (BH-p = 7.2e-20), respectively (Fig. 1a). Median health-adjusted life expectancy, which reflects years lived free from disease, lags at 64.5 years (Fig. S1; [Range: 44.2–74.1 years], p < 2.2e-16). Median healthspan (minimum, maximum) is (in years) lowest in Africa at 55.6 (44.2, 66.4) compared to 62.1 (52.6, 74.1), 63.4 (60.3, 70.0), 64.0 (49.7, 70.1), 65.8 (55.8, 71.3), and 68.8 (62.0, 72.5) for Western Pacific (BH-p = 1.2e-4), South-East Asia (BH-p = 2.6e-3), Eastern Mediterranean (BH-p = 6.6e-4), Americas (BH-p = 2.7e-9), and Europe (BH-p = 8.6e-22), respectively (Fig. 1a). Across the globe, life expectancy—a longevity measure—exceeds corresponding health-adjusted life expectancy that quantifies healthy longevity (Fig. 1a). Lifespan and healthspan are notably distinct among world regions, with Africa exhibiting the shortest lifespan and healthspan.

Fig. 1. Healthspan lags lifespan across world regions resulting in healthspan-lifespan gaps.

a Healthspan versus lifespan for World Health Organization member states presented by regions and visualized by kernel density estimates. b World map of the healthspan-lifespan gap across the 183 World Health Organization member states colored based on gap size. c The healthspan-lifespan gap of member states (data points) stratified by region. *denotes maximum Benjamini–Hochberg corrected p-values versus South-East Asia, Eastern Mediterranean, Americas, and Europe. Boxes represent 25th, 50th, and 75th percentile. Whiskers extend to extreme values within 1.5 times interquartile range.

Healthspan-lifespan gap: universal but non-uniform

Healthspan-lifespan gaps around the globe reflect health-adjusted life expectancy trailing life expectancy in every WHO member state (Fig. 1b). Globally, the median gap is 9.1 years and ranges from 6.5 in Lesotho to 12.4 years in the United States of America (Fig. 1b). Mapping the gap documents regional variance (Fig. 1c, p = 2.5e-11). In Africa, the median gap (minimum, maximum) is (in years) 8.3 (6.5, 10.7), equivalent to 8.4 (6.8, 12.1) years in Western Pacific (BH-p = 0.30) but lower than 9.6 (7.6, 10.5), 9.8 (6.8, 11.1), 9.6 (8.2, 12.4), and 9.9 (7.5, 11.3) in South East Asia (BH-p = 0.02), Eastern Mediterranean (BH-p = 1.4e-5), Americas (BH-p = 1.0e-7), and Europe (BH-p = 3.6e-9), respectively (Fig. 1c). Africa and Western Pacific have distinctly smaller healthspan-lifespan gaps compared to Europe, Americas, Eastern Mediterranean, and South-East Asia which exhibited respectively larger gaps. Beyond the number of years burdened by disease, the percentage of lifespan compromised by disease, defined as the life expectancy-adjusted healthspan-lifespan gap (LEA-GAP), was next assessed. The median LEA-GAP for the 183 WHO member states is 12.7% with a minimum of 10.5% in Democratic People’s Republic of Korea and a maximum of 15.8% in the United States of America (Supplementary Data 1 and Fig. S2a). Assessed per region (Fig. S2b), the median (minimum, maximum) LEA-GAP for Europe is 12.4% (10.8%, 13.9%), lower than the 12.9% (11.9%, 15.8%), 12.9% (12.3%, 14.5%), 13.3% (11.9%, 14.7%) observed in the Americas (BH-p = 0.03), Africa (BH-p = 7.7e-4), and Eastern Mediterranean (BH-p = 1.0e-4), respectively; but larger than the 11.8% (10.8%, 14.6%) for Western Pacific (BH-p = 0.049). Healthspan-lifespan gaps are thus omnipresent, but non-uniform across world regions. Notably, although exhibiting a smaller healthspan-lifespan gap compared to other regions, Africa has a substantial proportion of lifespan burdened by disease.

Distinct regional healthspan-lifespan gap performance

The healthspan-lifespan gap was assessed across world regions as a function of demographic (i.e., life expectancy, median age, birth rate, death rate, rate of natural change, population size, and population density), economic (i.e., gross domestic product and healthcare expenditure per capita), and health (i.e., total morbidity burden; communicable, maternal, perinatal and nutritional conditions; injuries; and noncommunicable diseases) indicators (SI Extended Results; Figs. S3–S14). Life expectancy, gross domestic product, and noncommunicable disease burden most consistently correlate with the healthspan-lifespan gap. Accordingly, the performance of individual WHO member states was determined by comparing actual healthspan-lifespan gaps per country versus gaps predicted in considering life expectancy, gross domestic product, and noncommunicable disease burden. Sixty-one member states have a healthspan-lifespan gap larger than anticipated (Fig. 2a) with 41%, 23%, 15%, 11%, 7%, and 3% from Africa, Americas, Eastern Mediterranean, Europe, South-East Asia, and Western Pacific, respectively (Fig. 2a inset). Africa is overrepresented in countries with larger than anticipated healthspan-lifespan gaps (Fig. 2a). Overrepresentation persists following adjustment with a spatial error model to account for potential confounders arising from geographic proximity (Fig. S15). Conversely, 58 states have lower gaps than anticipated (Fig. 2b) with 45%, 31%, 16%, 5%, 3%, and 0% from Europe, Western Pacific, Americas, South-East Asia, Eastern Mediterranean, and Africa, respectively (Fig. 2b inset). After adjusting for potential confounders that may be shared by geography, Europe is overrepresented in countries with smaller than anticipated healthspan-lifespan gaps (Fig. S16). In considering demographic, economic, and health indicators, regional disparity in healthspan-lifespan gaps is thus observed, with Africa overrepresented in countries with higher than projected gaps and Europe overrepresented in countries with lower than anticipated gaps.

Fig. 2. Healthspan-lifespan gap deviation for World Health Organization member states.

a Sixty-one states exhibited a healthspan-lifespan gap larger than predicted from demographic, economic, and health profiles with overrepresentation from Africa. b Fifty-eight states exhibited a healthspan-lifespan gap smaller than predicted from demographic, economic, and health profiles with overrepresentation from Europe. Insets show relative proportion of states from each region with larger (in a) and smaller (in b) gaps than anticipated.

Regional disparity in attributable disease burden

Healthspan-lifespan gap profiles were resolved from disease class contribution, measured as relative YLD. Globally, noncommunicable diseases are responsible for 56%–90% of a WHO member state’s total disease burden. Communicable, maternal, perinatal, and nutritional conditions (CMPN) account for 3%–37%, and injuries between 4% and 18%. Contributions of disease classes shows regional differences. The lowest contribution from noncommunicable diseases is in Africa (Fig. S17a) with a median contribution of 68% compared to 78%, 80%, 82%, 83%, and 84% in South-East Asia (BH-p = 0.003), Western Pacific (BH-p = 1.4e-5), Europe (BH-p = 3.1e-15), Eastern Mediterranean (BH-p = 1.6e-6), and Americas (BH-p = 1.6e-12), respectively. The highest contribution by CMPN is in Africa with a median of 27% compared to the 14%, 13%, 10%, 9%, and 5% in South-East Asia (BH-p < 0.02), Western Pacific (BH-p = 1.3e-5), Eastern Mediterranean (BH-p = 3.8e-5), Americas (BH-p = 1.9e-9), and Europe (BH-p = 5.5e-26), respectively (Fig. S17b). The percentage of total disease contributed by injuries also differed by region. Africa has the smallest contribution attributed to injuries (5%) compared to 6%, 7%, 7%, 7%, and 11% in Western Pacific (BH-p = 4.4e-3), Eastern Mediterranean (BH-p = 1.3e-3), Americas (BH-p = 2.2e-5), South-East Asia (BH-p = 0.04), and Europe (BH-p = 1.6e-20), respectively (Fig. S17c). Thus, regional healthspan-lifespan gap profiles originate from different morbidity contributions, with Africa demonstrating the most distinctive disease burden make-up.

Evolving disease burden

Regional disease class contributions redistribute over time. In the last two decades, noncommunicable diseases, communicable diseases, and injuries exhibit an average change of 3%, −3%, and −0.4% in participating in the total global disease burden, respectively. Redistribution varies by region. The greatest increase in noncommunicable disease burden contribution is in Africa which shows a 5% median increase, equivalent to the 6% in South-East Asia (Fig. S17d; BH-p = 0.35) and greater than the 3%, 3%, 2%, and 2% observed in Americas (BH-p = 1.0e-3), Western Pacific (BH-p = 4.0e-3), Eastern Mediterranean (BH-p = 8.8e-4), and Europe (BH-p = 1.1e-10). Concurrently, Africa exhibits the greatest reduction in the contribution by communicable diseases with a median decrease of 5%, equivalent to the 6% decline in South-East Asia (Fig. S17e; BH-p = 0.22) and greater than the median decrease of 3%, 2%, 2%, 0.7% observed in Western Pacific (BH-p = 0.06), Americas (BH-p = 2.3e-3), Eastern Mediterranean (BH-p = 0.02), and Europe (BH-p = 3.7e-16). Changes in the contribution from injuries are more modest, yet regional differences are observed (Fig. S17f; p = 0.0002). Africa exhibits a 0.3% increase, higher than the 0.3% and 0.7% decreases in Americas (BH-p = 0.04) and Europe (BH-p = 6.8e-4). Over the last two decades, Africa thus exhibits the greatest reshaping of its disease landscape, afflicted by a rising burden of noncommunicable diseases.

Disease burden patterns distinguish regions

To inspect disease profiles at higher specificity, the relative morbidity burden per disease subclasses—across 22 WHO disease categories – was assessed. Common to all regions, mental and substance use disorders are overrepresented (Fig. 3a). Hierarchical agglomerative clustering distinguishes the disease burden profile of Africa characterized by high contributions from nutritional deficiencies, and infectious and parasitic diseases (Fig. 3a). Principal component analysis segregates Africa from Europe, documented by non-overlapping 95% confidence intervals. A lesser separation is detected for Americas, South-East Asia, Eastern Mediterranean, and Western Pacific reflecting mixed disease profiles in these regions (Fig. 3b). Distinctness of disease burden contributions to the healthspan-lifespan gap among regions was assessed with random forest. Informed by the relative morbidity burden imposed by disease subclasses, this supervised machine learning algorithm classifies with 84% accuracy WHO member states into corresponding regions. Classification accuracy is region dependent (Table S1), with greater accuracy achieved for Europe (98%), Africa (91%), and Americas (88%), and lower accuracy attained for Eastern Mediterranean (80%), Western Pacific (57%), and South-East Asia (36%). Classification accuracy parallels principal component analysis segregation, with the most pronounced distinction between Europe and Africa. Disease subclasses contributing to segregation on principal component analysis were revealed by assessing loading values, which represent the correlation coefficient between each disease subclass and a given principal component (Fig. 3c). Malignant neoplasms, unintentional injuries, oral conditions, musculoskeletal diseases, neurological conditions, other neoplasms, and cardiovascular diseases positively associate with the Europe pole (Fig. 3c). By contrast, respiratory infections, nutritional deficiency, congenital anomalies, maternal conditions, infectious and parasitic diseases, and neonatal conditions directly associate with the Africa pole (Fig. 3c). Boruta analysis, a feature selection wrapper based off random forest, determines that queried disease subclasses contribute to classification. Neurological conditions, unintentional injuries, infectious and parasitic diseases, digestive disease, other neoplasms, musculoskeletal disease, and cardiovascular diseases are deemed most important (Table S2). Regional differences in the proportion of total years lived with disabilities attributed to disease categories contributes to regional segregation (Figs. S18 and S19). Dimensionality reduction thus extracts regional disease burden patterns and sets Africa and Europe apart based on morbidity profiles with distinctness confirmed by supervised machine learning classification.

Fig. 3. Disease profiles segregate states.

a Heatmap display of 22 disease categories (rows) in 183 World Health Organization member states (columns). Dendrograms represent hierarchical agglomerative clustering based on Euclidian distances. Color scale represents the relative contribution of each disease category to the member state’s total morbidity burden. b Principal component analysis extracts disease profiles for 183 World Health Organization member states. Colored ellipses represent 95% confidence intervals for Europe (green), Americas (blue), and Africa (tan) clusters. c Loading values revealed disease categories contributing to segregation of Africa and Europe. Dashed lines denote cut-off loading values|>0.2|. The tan shading represents disease categories positively contributing to segregation on the Africa pole and the green shading represents disease categories positively contributing to segregation on the Europe pole.

Machine learning-resolved disease pattern clusters stratify healthspan-lifespan gaps

Unsupervised machine learning recognizes three disease pattern clusters encapsulating global morbidity (Fig. 4a). Each cluster exhibits a distinct regional composition. Cluster 1, characterized by a predominance of respiratory infections, nutritional deficiency, congenital anomalies, maternal conditions, infectious and parasitic diseases, and neonatal conditions (Fig. 3c), encompasses 91%, 36%, 33%, 30%, 0%, and 0% of states in Africa, South-East Asia, Western Pacific, Eastern Mediterranean, Americas, and Europe, respectively (Fig. 4b). Cluster 2, featuring prevalent sense organ disease, diabetes mellitus, and genitourinary disease (Fig. 3c), contains 88%, 70%, 64%, 38%, 18%, and 9% of states in Americas, Eastern Mediterranean, South-East Asia, Western Pacific, Europe, and Africa, respectively (Fig. 4b). Cluster 3, hallmarked by an overrepresentation of malignant neoplasms, unintentional injuries, oral conditions, musculoskeletal diseases, neurological conditions, other neoplasms, and cardiovascular diseases (Fig. 3c), comprises 82%, 29%, 12%, 0%, 0%, and 0% of states in Europe, Western Pacific, Americas, South-East Asia, Eastern Mediterranean, and Africa, respectively (Fig. 4b). Thus, nearly 75% of states assigned to Cluster 1 are from Africa, whereas 75% of Cluster 3 assigned states are from Europe (Fig. 4c). Cluster 2 encompasses countries across regions with each region representing 6% to 41% of the states with this pattern (Fig. 4c). Notably, healthspan-lifespan gap sizes are distinct among disease pattern clusters with a median of 8.3, 9.4, and 10.3 years for states in Cluster 1, Cluster 2, and Cluster 3, respectively (Fig. 4d; p < 2.2e-16). Distinct contribution of disease burden subclasses to the healthspan-lifespan gap for member states of each cluster were assessed by the supervised machine learning algorithm random forest. Notably, 95% classification accuracy is achieved when considering the relative contribution of disease subclasses to overall morbidity burden with classification accuracies of 95.0%, 95.8%, and 96.1% for Cluster 1, Cluster 2, and Cluster 3, respectively (Table S3). Among queried disease subclasses, 21 contribute to cluster classification with infectious and parasitic diseases, maternal conditions, malignant neoplasms, neonatal conditions, nutritional deficiencies, and other neoplasms deemed most important (Table S4). Beyond distinguishing interregional gap differences, intraregional disease pattern clusters within Europe, Americas, Eastern Mediterranean, Western Pacific, and Africa stratify states by gap size (Fig. 4e, p < 0.01). Validated by supervised classification, unbiased clustering of disease burden patterns thus groups the globe into three distinct clusters which correlate with and inform the healthspan-lifespan gap size across and within regions.

Fig. 4. Artificial intelligence clusters disease burden patterns stratifying healthspan-lifespan gaps.

a K-means clustering of disease profiles from 183 World Health Organization member states, following dimensionality reduction by principal component analysis, identified three disease burden pattern clusters. Colored ellipses represent 95% confidence intervals for each cluster. b Proportion of member states from each region assigned to a disease burden pattern cluster. c Composition of disease burden pattern clusters by region. d Violin plots of healthspan-lifespan gaps for 183 World Health Organization member states classified according to disease burden pattern clusters. *denotes maximum Benjamini-Hochberg corrected p-values versus all other clusters. e The healthspan-lifespan gap of disease burden pattern clusters within each world region. *denotes maximum Benjamini–Hochberg corrected p-value comparing disease burden pattern clusters within a region. In (d, e), Boxes represent 25th, 50th, and 75th percentile. Whiskers extend to most extreme values within 1.5 times interquartile range.

Regional disparity in healthspan-lifespan gap trends

Over the last two decades, the global median healthspan-lifespan gap grew from 8.4 to 9.1 years with a median rate increase of 0.05 years/year (Fig. 5a). Healthspan-lifespan growth rates of 0.07 (0.01, 0.15), 0.06 (−0.05, 0.07), 0.05 (0.01, 0.07), 0.05 (0.01, 0.1), 0.03 (0, 0.07), 0.03 (−0.01, 0.09) are registered for Africa, South-East Asia, Eastern Mediterranean, Europe, Western Pacific, and Americas, respectively (Fig. 5a). Africa exhibits the most rapid growth in healthspan-lifespan gap comparable to South-East Asia (BH-p = 0.07), and significantly greater than Americas (BH-p < 0.0001), Western Pacific (BH-p < 0.0001), Europe (BH-p < 0.001), and Eastern Mediterranean (BH-p < 0.01). Thus, while the gap accrual is disparate among regions, Africa shows the fastest healthspan-lifespan gap expansion.

Fig. 5. Healthspan-lifespan worldwide gap trends and projections.

a Mean change per year, over the last two decades, in the healthspan-lifespan gap for 183 World Health Organization members states grouped by region. *denotes maximum Benjamini–Hochberg corrected p-value versus Europe, Eastern Mediterranean, Western Pacific, and Americas. Boxes represent 25th, 50th, and 75th percentile. Whiskers extend to most extreme values within 1.5 times interquartile range. b Projected healthspan-lifespan gap for each world region through 2100. Solid line represents the average projection for World Health Organization members states within each region, and colored ribbons represent 95% confidence intervals for each region.

Healthspan-lifespan gap projections

The evolution of the healthspan-lifespan gap was projected based on WHO member states longevity trends. By the year 2100, the median healthspan-lifespan gap is predicted to rise globally by 22% (Fig. 5b), ranging from 7.3 to 18.8 years (Supplementary Data 2). Across all WHO regions, the healthspan-lifespan gap is anticipated to widen throughout this century (Fig. 5b). By the turn of the next century, the healthspan-lifespan gap (in years) is projected to reach (mean ± SEM) 12.1 ± 0.3, 12.1 ± 0.4, 11.7 ± 0.2, 11.0 ± 0.3, 10.5 ± 0.5, and 10.1 ± 0.2 in Americas, Eastern Mediterranean, Europe, Western Pacific, South-East Asia, and Africa, respectively (Fig. 5b). These projections forecast an ongoing widening of the healthspan-lifespan gap across world regions.

Discussion

This study maps the healthspan-lifespan gap across the globe identifying regional heterogeneity in size and underlying disease contributors. The study highlights the number of years burdened by disease in world populations in the context of demographic, economic and health indicators. Life expectancy, gross domestic product, and noncommunicable disease burden correlate consistently with the healthspan-lifespan gap across and within regions. Africa is overrepresented in the number of member states that have a larger healthspan-lifespan gap than predicted from demographic, economic and health correlates. Europe has the most countries with a lower than anticipated gap. Regional differences in disease burden patterns, derived through dimensionality reduction, are identified with the most distinctive profiles distinguishing Africa and Europe. Artificial intelligence assigns 183 WHO member states into three clusters based on disease burden patterns. Cluster designation stratifies states by healthspan-lifespan gap size globally and regionally. Africa shows the fastest rate of gap widening associated with a restructuring of the disease burden landscape towards noncommunicable disease dominance. Further gap expansion is projected throughout this century for all regions. The pathogeography of the healthspan-lifespan gap, mapped multidimensionally herein, threatens the quality of human longevity imposing an escalating global health risk.

Disparities in longevity, healthy longevity, and disease burden have been reported^7,24–27, yet regional characteristics of the healthspan-lifespan gap remained unexamined. The present study addresses this through analysis of all WHO designated world regions over the last two decades. Identifying regional differences helps recognize inequalities in disease impact, relevant to the United Nations sustainable development goals^28,29 and consistent with multipolar, regional efforts^30,31.

Africa exhibits a shorter lifespan and healthspan compared to other regions, translating into a smaller healthspan-lifespan gap in line with longevity profiles³². Importantly, however, Africa presents with a substantial percentage of lifespan compromised by disease compared to other regions and displays the fastest widening of the healthspan-lifespan gap. This trend points to regional vulnerability and need for action as Africa undergoes demographic transitions with population growth, increased longevity³², and as documented here, in disease burden restructuring. Increasingly, Africa is under a dual threat from a persistently large burden imposed by communicable, maternal, perinatal, and nutritional conditions aggravated by a mounting epidemic of noncommunicable diseases^33,34. In this regard, health-positive actions with multistakeholder engagement need to incorporate region-led healthy longevity initiatives³² to ensure equitable region-benefiting outcomes.

Europe is among the regions with the largest healthspan-lifespan gap yet is overrepresented in countries with a healthspan-lifespan gap smaller than anticipated from demographic, economic, and health indicators. As observed here the morbidity burden of Europe is characterized by a large contribution of musculoskeletal diseases, unintentional injuries, neoplasms, and other noncommunicable diseases³⁵. While Europe presents a large contribution from mental disorders³⁵, the burden of mental and substance use disorders is relatively conserved across regions and does not contribute to segregation of disease burden patterns³⁶. South-East Asia, Eastern Mediterranean, Americas, and Western Pacific exhibited heterogenous disease burden patterns that did not discern these regions with neither supervised machine learning nor dimensionality reduction. Accordingly, unsupervised clustering was used to extract disease burden pattern similarities segregating 183 WHO member states into three clusters. Highlighting commonality of disease contributors, disease pattern cluster assignment informs the magnitude of the healthspan-lifespan gap. Mapping the healthspan-lifespan gap for each of the six world regions and deciphering disease burden patterns contributing to the gap, the present study reconfigures the globe from traditional geographic, demographic, and economic groupings into a tripartite cluster scheme underscoring shared disease patterns and comparable gap profiles. Identifying disparities in disease burden contribution helps guide public health policy efforts³⁷. Notably, understanding influencers of the healthspan-lifespan gap is relevant in this age of accelerated healthspan-lifespan gap widening. Quantifying healthspan-lifespan gap by region and identifying disease pattern clusters segregating states by gap magnitude offers a new dimension for population health audit enhanced by quantitative and qualitative measures impacting healthy longevity.

The healthspan-lifespan gap metric relies on estimates of health-adjusted life expectancy which weigh years of life lived with disease based on perceived quality of life³⁸ derived from diverse populations³⁷. Perceived quality of life varies by geography and expected disease impact may be under- or over-estimated^39,40. In addition, while distinct disease burden patterns were associated with a larger healthspan-lifespan gap, they also present distinct regional distribution. Regions also differ in economic, social, and health indicators precluding direct implication of a specific disease burden profile to an exaggerated healthspan-lifespan gap. Additionally, the healthspan-lifespan gap quantified herein does not specify the period of life impacted by disease. Lifespan analysis has been enhanced by measures of entropy assessing the distribution of mortality across age groups^41–43. Similarly, age strata-dependent dissection of morbidity would help deepen understanding of the healthspan-lifespan gap.

Distinct disease pattern clusters – introduced here – provide insight into contributors to the healthspan-lifespan gap across and within world regions. These findings draw attention to healthy longevity threats, stress the need for expedited transition to health-centric systems of care, and invite disease burden-informed solutions for world populations to help narrow the widening healthspan-lifespan gap globally^44,45.

Author contributions

Conceptualization: A.G. and A.T.; Data curation: A.G.; Formal analysis: A.G.; Funding acquisition: A.T.; Investigation: A.G. and A.T; Methodology: A.G. and A.T.; Project administration: A.T.; Resources: A.T.; Software: A.G.; Supervision: A.T.; Validation: A.G. and A.T.; Visualization: A.G. and A.T.; Writing—original draft: A.G. and A.T.; Writing—review and editing: A.G. and A.T.

Peer review

Peer review information

Communications Medicine thanks Andrea Nigri and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Data availability

All original datasets are publicly available at World Health Organization Global Health Observatory (https://apps.who.int/gho/athena/data/GHO/WHOSIS_000001,WHOSIS_000015,WHOSIS_000002,WHOSIS_000007?filter=COUNTRY:*&format=xml&profile=excel) and United Nations (https://population.un.org/wpp/Download/Standard/MostUsed/) sites. The data used in Figs. 1, 4, and S1–S14 are found in Supplementary Data 3 and 4. The data used in Figs. 2, S15, and S16 are found in Supplementary Data 5 and  6. The data used in Fig. S17 are found in Supplementary Data 7. The data used in Figs. 3A, S18, and S19 are found in Supplementary Data 8. The data used in Figs. 3B, C and 4 are found in Supplementary Data 9, 10, and  11. The data used in Fig. 5 are found in Supplementary Data 12 and 13.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-025-01111-2.