Accelerating Burden of Hypertensive Heart Disease in Low‑ and Lower‑Middle‑SDI Countries, 1990–2023: The Contribution of Cold Temperature Exposure

Data source

The latest 2023 GBD database has comprehensively evaluated diseases, injuries, and risk factors across populations of different ages, regions, and genders. The GBD study provides a comprehensive framework for estimating disease burden and risk factors worldwide, including 371 diseases and injuries and 88 risk factors across 204 countries and territories. The methodological framework was based on previously published GBD 2021 studies, while the latest estimates analysed in this study were extracted from the GBD 2023 database, including mortality, DALYs, ASRs, temperature‑attributable burden, and socio‑demographic index (SDI) values. All estimates underwent standard GBD modelling and uncertainty assessment procedures [9, 13]. Data related to deaths and DALYs for HHD can be viewed and downloaded from the VizHub—GBD Results (IHME) website (https://vizhub.healthdata.org/gbd-results/).

Data for HHD (ICD‑10 codes I11–I11.9) were extracted exclusively for the 79 low‑ and lower‑middle‑SDI countries using the GBD Results Tool. Selected metrics included: (1) number and ASRs of deaths and DALYs (per 100,000 population) with 95% uncertainty intervals (UIs); (2) low‑ and high‑temperature‑attributable deaths and DALYs (Level 4 risk factors in the GBD comparative risk assessment framework); (3) country‑level SDI values from 1990 to 2023.

Definition

Although the GBD 2023 HHD data include all age groups (0–100+ years), this study focused on individuals aged 15 years and older, consistent with the age stratification presented in the results. HHD incidence and mortality are extremely low in children and adolescents, and age‑specific estimates in these younger groups are less reliable. Age‑standardised rates and decomposition analyses were therefore conducted for the ≥15‑year‑old population. The selected age range is consistent with the specific age range specified in the GBD 2023 database.

Disability weights (DWs) represent the degree of poor health associated with a specific disease or disability, relative to a healthy state. In the GBD, DWs are primarily used to calculate years lived with disability (YLD) as part of the DALY metric, where the YLD formula is: YLD = disease prevalence × disability weight. This allows YLD to be combined with years of life lost (YLL) due to premature death to form the total DALY [14].

The SDI is a comprehensive index formed by standardising indicators such as education level, fertility rate, and per capita income, and is used to reflect the development level of a country. This study used the most recently available SDI data (reference year: 2023) as the classification standard, and countries were assigned to the corresponding SDI groups (low and low–middle) according to their SDI values for that year. The inclusion criteria were as follows:(1) SDI data availability: countries were required to have available SDI values in the GBD 2023 dataset. (2) Completeness of health burden data: countries had to report HHD deaths and DALYs in GBD 2023 sufficient for calculating ASRs and conducting trend analyses. (3) Acceptable data quality: countries were excluded if data were missing, estimates were unstable, or key indicators lacked sufficiently reliable UIs, to ensure analytical robustness. Based on these criteria and the GBD 2023 SDI values, a total of 79 eligible low‑SDI and low‑middle‑SDI countries and territories were included. These countries were considered to represent the full set of low socio‑demographic development settings with complete HHD data available, and were therefore used for cross‑national comparisons and temporal analyses. The SDI value ranges from 0 to 1, with a higher value indicating a higher level of development. Based on these criteria, the SDI values of 204 countries and regions are divided into five groups. The latest grouping is as follows: low SDI (<0.45), lower‑middle SDI (0.45 ∼ < 0.61), middle SDI (0.61 ∼ < 0.69), upper‑middle SDI (0.69 ∼ < 0.80), and high SDI (≥0.80) [15].

The types of HHD studied were extracted using the International Classification of Diseases (ICD‑10) codes: HHD(I11–I11.9) [16].

Statistical analyses

The DALYs and death indicators of the studied HHD were quantified using the ASRs. This method eliminates the impact of differences in age composition, ensuring the comparability of the research indicators [17]. In the GBD database, all data on ASRs and their 95% UI estimates are expressed per 100,000 population. The UI is defined as the 2.5th and 97.5th percentiles obtained from 1000 ordered samplings. The basic calculation method of ASR is as follows:

where $a_i$ denotes the age $i^{th}$ age group, $w_i$ is the number of people in the standardised population within each age group, A represents the number of age groups, and 95% UI is defined as the 2.5th and 97.5th values of the ordered 1000th extraction.

Decomposition analysis was performed to quantify the contributions of population growth, population ageing, and changes in age‑specific mortality rates to changes in HHD deaths over time. Demographic decomposition was performed using the Das Gupta method to partition changes in HHD deaths between 1990 and 2023 into three components: population growth, population ageing, and changes in age‑specific HHD death rates. Decomposition was conducted for each SDI stratum separately, with results aggregated across countries within each stratum. The interaction term was distributed proportionally between demographic components following standard practice.

Estimates were calculated via a generalised linear regression model to evaluate the EAPCs in ASRs, which relates the natural logarithm (ln) of the ASRs to time, revealing the evolution of the ASRs over time [17, 18]. EAPCs and their 95% CIs were calculated using EAPC = 100 × [exp(β) − 1]. The basic calculation of EAPCs is as follows:

where ln (ASR), α denotes the intercept, X denotes the calendar year, ℇ denotes the error term, and β corresponds to the linear positive or negative trend of ASR. EAPC was considered statistically significant if the 95% CI excluded zero.

Socioeconomic inequalities across the 79 countries were quantified using the SII for absolute differences and the CI for relative differences, both based on SDI ranking. A negative SII value indicates that the burden is disproportionately higher in low‑SDI countries, whereas a positive value would indicate higher burden in high‑SDI countries [19]. Global and high‑SDI data are included in Table 1 as reference benchmarks to contextualise the divergent trends observed in low‑resource settings.

To quantify the contribution of ambient temperature to HHD burden, we extracted estimates from the GBD 2023 Comparative Risk Assessment framework. In the GBD framework, temperature‑attributable burden represents population‑level burden estimated from modelled temperature exposure distributions, exposure–response relationships, and the theoretical minimum risk exposure level (TMREL). The population attributable fraction is calculated based on these parameters and subsequently applied to total HHD deaths and DALYs to estimate the burden associated with non‑optimal temperature exposure. Therefore, temperature‑attributable estimates should be interpreted as risk attribution at the population level rather than evidence that temperature exposure independently determines temporal changes in HHD burden. Detailed descriptions of temperature exposure assessment, TMREL estimation, lag structures, and exposure–response modelling (including distributed lag non‑linear models) have been reported in previous GBD publications and methodological documentation. In this study, we did not independently perform temperature exposure modelling or risk estimation; instead, all attributable burden estimates were obtained directly from the GBD database. Uncertainty intervals (95% UIs) were derived from 1000 posterior draws provided by the GBD study.

To project future burden while avoiding implausible extrapolation of ASRs under rapid demographic change, we used Bayesian age–period–cohort (BAPC) models with default settings to estimate multiplicative effects of age, period, and cohort. BAPC models were used to project future age‑standardised HHD rates by sex, incorporating age, period, and cohort effects. Absolute numbers of deaths and DALYs were subsequently derived by applying projected age‑specific rates to World Health Organization (WHO) medium variant population forecasts. Age, period, and cohort effects were modelled using second‑order random walk priors (RW2) with non‑informative hyperpriors on precision. Model convergence was assessed using Potential Scale Reduction Factor (PSRF < 1.1) and Deviance Information Criterion (DIC). Back‑testing was performed by fitting the model to data up to 2018 and predicting 2019–2021, which confirmed that the model adequately captured observed trends. Posterior distributions were summarised to obtain median estimates and 50%, 80%, and 95% credible intervals. Although specific numerical point estimates for ASRs are not reported in tables, the posterior median and UIs are visualised in Figure 7 to illustrate projected trends. Models were fitted to observed 1990–2023 data and extrapolated using the WHO medium variant population forecasts for the 79 countries [20, 21]. Projections were limited to absolute numbers of deaths and DALYs. No specific point estimates for ASRs by sex in 2050 were reported. All analyses and visualisations were performed in R version 4.4.2 using the BAPC and INLA packages.

HHD burden at the global, regional, and national levels in 2023

In 2023, substantial geographical heterogeneity was observed in the ASRs of HHD, with the highest burdens concentrated in Central Sub‑Saharan Africa and other low‑SDI regions (Figure 1).

Figure 1

Compared with 1990, the low‑SDI region showed the most significant growth in the number of deaths and DALYs in 2023 (221%). The number of cases and ASRs of HHD deaths and DALYs in 1990 and 2023 are detailed in Table 1 and Figure 2.

Figure 2

By 2023, however, the region with the highest ASR shifted to Central Sub‑Saharan Africa, with an age‑standardised mortality rate (ASMR) of 44.60 (24.89, 70.65) and an age‑standardised DALY rate (ASDR) of 900.36 (512.99, 1406.13) (Table 1; Figure 2). The burden increased progressively with age across all SDI regions, reaching the highest levels in those aged 95 years and older. Among SDI regions, males and females in the low‑SDI region had the highest disease burden in terms of crude deaths and DALYs (Table 1; Figure 3), reflecting the combined effects of high prevalence of untreated hypertension and advanced population ageing in these settings.

Figure 3

Trends in HHD burden from 1990 to 2023

Globally, the absolute burden of HHD rose substantially from 1990 to 2023 due to population growth and ageing, but ASRs generally declined, reflecting improvements in hypertension management in many settings (Table 1; Figure 4).

Figure 4

EAPCs of global ASMR and ASDR for HHD calculated from 1990 to 2023 were −0.67% (95% CI: −0.84, −0.50) and −0.79% (95% CI: −0.96, −0.62), respectively, both showing a downward trend. In contrast, both indicators were positive in low‑SDI countries, whereas in lower‑middle‑SDI countries only the ASMR showed a slight increase. The combined number of absolute deaths in low‑ and lower‑middle‑SDI countries increased by approximately 192% (from 157,950 to 461,110), with DALYs rising by approximately 171% (Table 1; Figure 4).

These divergent trends highlight the persistent challenge of hypertension control in low‑resource settings and underscore the growing contribution of low‑ and lower‑middle‑SDI countries to the global HHD burden.

Decomposition analysis of HHD burden

From 1990 to 2023, ageing and population growth led to an increase in the death burden, while epidemiological changes resulted in a decrease in the death burden. In the gender‑specific analysis, the decomposition results for males were consistent with the overall global trend (for both genders). However, for females, epidemiological changes, ageing, and population growth all contributed to the increase in the death burden.

Contrary to the global trend, age‑specific mortality rates in low‑SDI countries exhibited an upward trajectory over the study period, thereby contributing positively to the overall increase in HHD‑related deaths. Decomposition analysis within these regions indicated that population ageing accounted for approximately 35% of the rise, population growth contributed roughly 25%, and alterations in age‑specific HHD mortality rates explained nearly 40% of the total increase in deaths between 1990 and 2023. In low‑middle SDI regions, the corresponding contributions were approximately 30% from ageing, 20% from population growth, and 50% from changes in age‑specific mortality rates. These estimates, derived from decomposition analysis, reflect the relative contributions of demographic and epidemiological factors to changes in HHD mortality (Figure 5). Notably, East Asia, which lies outside the primary focus on low‑ and lower‑middle SDI regions, showed a disproportionately large absolute contribution (∼40 × 10 deaths) in the global decomposition figure. This is largely attributable to its large population size and rapid demographic ageing rather than higher per‑capita HHD risk. Therefore, this regional contribution should be interpreted cautiously and not as a focal finding of the present study.

Across all SDI regions, epidemiological changes generally reduced burden in line with the global pattern. Notably, in low‑ and lower‑middle‑SDI regions, epidemiological improvements had a weaker mitigating effect, resulting in positive contributions from all drivers in some strata and underscoring the limited impact of preventive advances in resource‑constrained environments (Figure 5).

Figure 5

Health inequality analysis of HHD burden

In 1990 and 2023, the analysis of absolute inequality in HHD across countries worldwide showed that the SII (from −18.93 to −20.66) for HHD in 2023 was more negative than that in 1990. This indicates that, over this period, the inequality in ASMR burden, driven by differences in SDI levels, increased. As a result, a higher HHD burden is now more prevalent in countries with lower SDI rankings.

Relative inequality, measured by the CI, remained stable over the period, indicating no reduction in the proportional disadvantage faced by lower‑SDI countries (Figure 6).

Figure 6

Projected HHD burden from 2024 to 2050

The BAPC projections indicate that absolute numbers of HHD deaths and DALYs in low‑ and lower‑middle‑SDI countries will increase substantially by 2050 under continuation of current trends (Figure 7). This rise is driven predominantly by population ageing and a growing proportion of elderly women, who face higher vulnerability due to longer life expectancy, lower historical rates of hypertension screening, and reduced postmenopausal cardiovascular protection. These patterns highlight the urgent need for targeted, gender‑responsive prevention strategies in resource‑constrained settings to avert a widening burden of preventable heart failure.

Figure 7

HHD burden attributable to non‑optimal temperature

To explore the changes in the impact of environmental factors on HHD burden across different regions, we analysed the trends in risk factors related to HHD from 1990 to 2023. In this study, in low‑SDI and lower‑middle‑SDI regions, cold temperature consistently accounted for a greater temperature‑attributable HHD burden than high temperature throughout the study period, indicating a substantial contribution of cold exposure to temperature‑related cardiovascular burden, with only minor fluctuations between 1990 and 2023. In contrast, the burden attributable to hot temperature gradually increased over the same period, contributing to the overall upward trend. In high‑middle SDI regions, cold temperature‑attributable HHD burden exhibited a sharp decline around the early 2000s, with ASMRs dropping from ∼4 per 100,000 to ∼1–2 per 100,000, before stabilising thereafter. This abrupt decline likely reflects improvements in hypertension management and healthcare access, as well as potential artefacts from sparser historical data. In high SDI regions, both cold‑ and hot temperature‑attributable burdens remained relatively stable throughout the study period (Figure 8).

Figure 8