Cardiovascular diseases (CVDs) are caused by abnormalities in the heart and blood vessels. The two most important CVDs are coronary heart disease (CHD) and stroke. It has been estimated that over the past three decades, CHD mortality has decreased by more than 50% in many developed countries such as Finland, England, the United States, Canada, Australia, New Zealand, France, and Japan [1‐4]; however, the situation remains challenging in developing nations, which account for more than 60% of the global disease burden [5]. In China, CVD is the leading cause of death and disability among adults [6, 7]. In addition to aging and population growth over the coming 20 years, projected unfavorable trends in blood pressure, total cholesterol, diabetes, and body mass index may accelerate the epidemic [8]. Therefore, CVD should be actively and effectively addressed by future chronic disease prevention and control measures in China.

Many studies have revealed that the epidemiological characteristics of CVD show obvious geographical disparities, and its epidemiological pattern has changed significantly along with the rapid economic development in China [9‐15]. Recent studies have focused primarily on temporal variations in epidemiological characteristics (incidence, fatality, and mortality) or its variations at the regional, provincial, or municipal levels [9, 13, 16‐20]. Counties are the basic administrative unit in China. To the best of our knowledge, no previous studies have examined the spatiotemporal variations in CVD mortality across China at the county level. This lacuna makes it difficult to recognize the areas of high CVD mortality and those with great increases in CVD mortality, which in turn restricts the timely adjustment of intervention strategies for this epidemic in these areas.

The aims of this study sought to describe the spatial and temporal variations in the CHD and stroke mortality rates of all the counties in China in 1991–2009 using a geospatial tool. This study might help to understand the epidemiological characteristics of CHD and stroke across different regions as well as provide a reference for evaluating the effectiveness of CVD prevention control measures to develop future local intervention plans.

The spatial estimation of CVD death rates (ASDRs) at the county level in China achieved satisfactory results according to the R-square and relative RMSE values shown in Fig. 2. Moreover, because of higher R-square and lower relative RMSE values, the estimation of the CHD death rates from 1991 to 2000 was much more satisfactory than that from 2004 to 2005 and 2006–2009. The performance of the stroke estimation changed in the reverse direction during the same period. In comparison, the estimation of the CHD and stroke death rates displayed the same temporal differences by gender as they did in the total population.

We classified the estimated death rates of CHD into four grades: 0–25, 25–50, 50–75, and 75 – (per 10⁵). The estimated death rates of stroke were also classified into four grades: 0–100, 100–150, 150–200, and 200 – (per 10⁵). All of the counties were divided into four groups so that the populations (taken from the Provincial or Municipal Statistical Yearbook) in these grouped counties were summed across different periods. Accordingly, the proportion of the populations in these grouped counties to the national population was calculated to indicate the temporal trend of the CVD death rates in China from 1991 to 2009 (Fig. 3).

In terms of these proportions, CHD and stroke presented different temporal trends, espeically at the fourth grade (CHD: 75 per 10⁵~; stroke: 200 per 10⁵~). As Fig. 3 – a illustrated, the proportion of the population in counties with the fourth grade of CHD death rates (75 per 10⁵~) dramatically increased from 15% to approximately 50% from 1991 to 2005, whereas those of the first and second grades clearly declined. In contrast, the proportion of the population with the fourth grade of stroke death rates (above 200 per 10⁵) was retained at a level of 20% before an obvious decrease to approximately 4% from 2006 to 2009 (Fig. 3b). These results indicate that the period from 2004 to 2005 was the temporal inflection for both CHD and stroke death rates in China during 1991–2009.

In addition to differentiated temporal trends, the spatial patterns of the county-level death rates (estimated ASDRs of CHD and/or stroke) and their changes tended to displayed different features. During the baseline period (1991–1995), the counties with higher CHD death rates than 75 per 10⁵ were mostly distributed in three clustered areas (Fig. 4a). After this period, these counties showed an obvious spatial expansion around these three clusters greater from 1996 to 2000 (Fig. 4b). In particular, these clustered areas were then gradually jointed from 2004 to 2005 and 2006–2009, resulting in an inverted triangle region across this nation (Figs. 4c and d).

In comparison, the spatial variations of the estimated stroke death rates differed from that of the CHD at the county level. As Fig. 5a illustrated, the counties with higher stroke death rates than 150 per 10⁵spatially formed a wide belt from the Southwest to the Northeast across China from 1991 to 1995. From 1996 to 2000 (Fig. 5b), this belt narrowed in its central portion (i.e., the Qinghai, Sichuan, and Gansu provinces), became wider in the West and Southeast from 2004 to 2005 (Fig. 5c), and then decreased and split into two regions from 2006 to 2009 (Fig. 5d). One of these regions was a slim belt from the West to East. The other was the Northeast region (North Heilongjiang and Northeast Inner Mongolia) where stroke mortality was much greater than 200 per 10⁵. In addition, there were some counties in Northeast China continuously possessing higher death rates of both CHD (Fig. 4) and stroke (Fig. 5) during the four calendar periods. However, many more counties in the rest regions didn’t display a similar characterization yet.

According to the spatial and temporal trends of the county-level ASDRs of CVD across China, the counties with outstanding increase or decline were discriminated between the first (1991–1995) and the last periods (2006–2009). In total, there were respectively 2600 and 1300 counties with increasing death rates of CHD and stroke, approximately 91 and 44% of the total counties (Table 1). The spatial distribution of the counties that showed temporal changes of estimated CHD (Fig. 6 b and c) and stroke death rates by gender (Fig. 6 e and f) was similar to that of the total population (Fig. 6 a and d). About 800 counties showed a greater than 100% increase of CHD, mainly locating in the Northeast and the South Central regions of China (Fig. 6 a). Meanwhile, 150 counties, sparsely distributed across West Xinjiang, Northeast Inner Mongolia, Southeast Liaoning, Central Henan, West Fujian, and the junction of Yunnan-Guizhou-Guangxi (Fig. 6 d), displaying a greater of 100% increase of stroke. In contrast, about 1600 counties in the Northeast, Beijing–Tianjin–Heibei, and West regions, approximately 55% of the total counties, presented obvious decrease of stroke death rates (Fig. 6d), which made a great contribution to the reduction of the regions with high stroke death rates (Fig. 5). In addition, there were also the counties with simultaneously increasing death rates of both CHD and stroke across China (Fig. 6). However, more counties tended to present opposite trends of these two diseases.

This study was the first to reveal that the epidemiological characteristics of the CHD and stroke death rates in China tended to be spatially and temporally characterized at the county level. During 1991–2009, the epidemic of CHD in China became increasingly serious because of the remarkable spatial expansion of the counties with high CHD death rates, whereas that of stroke obviously declined due to the notable spatial shrinking of the counties with high stroke death rates, especially across 2004–2005. Overall, the outcome assessment of CVD-related interventions in China could benefit reasonably from the temporal and spatial clues derived from the current study. It might provide an important temporal clue for public health authorities evaluating the outcome of CVD-related interventions at different stages in China over the past decades.

Many studies have shown that the ASDR of CHD presented an outstanding ascending trend while that of stroke obviously decreased in China over the past decades [9‐14, 20, 23]. It was also observed in our study. The increasingly serious status of CHD death rates was mainly attributed to the ascending prevalence of the various risk factors in China [31‐34], including hypertension, smoking, dyslipidemia, diabetes, obesity, etc. Meanwhile, high rates of out-of-hospital CHD deaths (e.g., more than 70% in Beijing [35]) was another important cause of the increasing CHD status in this county. Accordingly, more and more counties/districts displayed high ASDR of CHD across China during 1991–2009. Moreover, increasingly higher ASDR of CHD was observed in some areas, especially in the Northeast and Central South regions in this study, which was also occurred in other East Asia countries [36‐38].

On the contrary, the declining mortality trend for stroke in China did not match the related levels of prevalence of these similar risk factors, which may be strongly affected by the increasingly improvements of medical service system by China’s great economic development [39, 40]. In particular, the county-level hospitals in the rural regions were gradually able to carry out thrombolysis therapy and minimally invasive puncture for cerebral hemorrhage [40]. As a result, a decreasing trend of stroke death rates was observed in about 1600 counties that distributed sparsely across this country. Even so, there were still obviously ascending ASDR of stroke observed in 150 counties, which may be related to their unenlightened medical service level in some undeveloped areas (e.g., the boundary between Xinjiang and Tibet, the western part of Guangxi and Fujian).

In addition, a province-level characterization outlined a surrounding belt of nine provinces with high stroke death rates in Northern China despite that its basic data did not cover the entire country [14]. Many more studies have suggested that stroke mortality tends to show northeast–southwest gradient variations [12, 13, 41]. Interestingly, all of the regions with high stroke death rates previously reported (except Tibet) were covered by those described in our study in the 1990s, which indicates that the high stroke mortality belt in China is constantly changing. Compared with the regional, provincial, and municipal variations of CVD death rates characterized in the previous investigations [18, 20, 23, 42‐48], the specific regions with higher, lower, increasing, or decreasing CVD death rates were clearly revealed at the county level in China over the past decades. These results indicated that the geographic distribution of the death rates of CVD (CHD and stroke) can be well characterized on a more detailed scale.

Certain limitations of this study warrant mentioning. First, spatial interpolation using the OK method tends to increase the lowest values and decrease the highest values, especially in regions with few and sparse surveillance points. This disadvantage might be reduced through the densification of surveillance points and an improvement of the DSP system. Second, the data obtained from a retrospective survey of the causes of death from 2004 to 2005 might be somewhat higher than the DSP data and lead to an overestimation of the causes of death such as the large area of high stroke mortality in West China during this period, which should be confirmed by future investigations. Although we have formed an overall understanding on the spatial and temporal changes of CVD mortality in China over the past two decades, certain risk factors, including blood pressure, total cholesterol, diabetes, overweight, climate change, and air pollution, should be obtained in the next step and integrated into some spatiotemporal models for further exploring the potential causes of the spatiotemporal variations in the county-level CVD mortality. However, some important clues regarding the assessment of the present CVD-related interventions were uncovered in this study.

In summary, the epidemiology of CHD and stroke mortality from 1991 to 2009 in China was spatially and temporally featured at the county level. The study was the first to present critical spatial and temporal clues regarding the assessment and adjustment of CVD-related interventions in this country, and it provides useful support for exploring the spatial and temporal variations of the CVD risk factors in China.