This cause-specific mortality study was carried out in the Bavi district epidemiological field laboratory (FilaBavi), a DSS in which the original sample was selected randomly with probability proportional to population and which covers a range of geographical regions in the district. The sampling unit was hamlet or village subdivision, or cluster; it included 67 clusters with an estimated 11,300 households and a total population of about 51,000. The methods of this study have been published elsewhere (9).

All cases of death among people aged 20 years and older that occurred from January 1, 1999 through December 31, 2003 (n = 1067) were captured for analysis by death registration obtained during quarterly household visits. A verbal autopsy questionnaire, based on a World Health Organization (WHO) instrument and adapted for Vietnam, was administered by trained interviewers. Interviewers assigned the most likely cause of death by obtaining information from a close relative or caretaker of a deceased person about circumstances, signs, and symptoms during the illness (10).  When the verbal autopsy questionnaire had been completed, it was forwarded to two physicians who undertook the interpretation and diagnosis procedure for the data. They each made independent diagnoses based on the reported circumstances, signs, and symptoms. In cases in which the physicians differed, they held discussions to reach consensus. A Kappa test was used to assess degree of agreement between the two physicians. NCD deaths were classified as CVD, cancer, or other NCDs.

Person-time generated from the dynamic cohort of people under surveillance during the same period were calculated based on the date of events (i.e., in-migration, out-migration, births, and death) for age- and sex-specific groups and used as denominators (8). Age was first categorized into the following three groups: 20 to 49 years (young adults), 50 to 74 years (mid-adults), and 75 years or older (elderly). To provide more significant results, age was reclassified into the following two groups: 20 to 49 years and 50 years or older.

Individual socioeconomic status was estimated by assessing education level and household economic status prospectively at the time of entering the cohort in the DSS. Education level was classified as either no formal education (including illiteracy) or formal education (completion of any level of schooling). Economic condition of households was described as either poor (average monthly income per person less than VND 45,000 or U.S. $3.30) or nonpoor (average monthly income per person more than or equal to VND 45,000 or U.S. $3.30) (People Committee of Bavi: Decision number 59, Ministry of Labor, Invalids and Society, adapted for Bavi district, 1999).

Analyses were carried out using Stata software, version 8 (Stata Corp, College Station, Tex). CVD mortality rates per 1000 person-years were calculated. Crude and adjusted Mantel-Haenszel rate ratios and corresponding 95% confidence intervals were used to compare mortality rates by sex. Multivariate Cox proportional hazards models were used to examine associations between sex, age, socioeconomic status, and CVD mortality. The effect of each independent variable was assessed by examining relative risk while controlling for other independent variables in the model. To take into account the nonrandom sampling nature of verbal autopsy, cluster analysis was introduced on the calculation and modeling.

The surveyed population is described in Table 1 by sex, age, education, and economic status. There were 14,289 men and 16,713 women aged 20 years and older at the beginning of the study period. The proportion of people without formal education was higher among women (18%) than men (5%). About 12% of men and 14% of women were classified as poor.

We have shown that a substantial proportion of deaths are attributable to CVD, which was the leading cause of death among adults. The substantial effects of CVD mortality can be partly explained by the aging of the population in the setting. The proportion of people aged 50 years and older rose from 16.1% (13.4% of men and 18.7% of women) in 1999 to 16.9% (14.2% of men and 19.5% of women) in 2001 and to 17.9% (15.0% of men and 20.5% of women) in 2003. 

Epidemiological transition pertaining to CVD epidemics has classically been described in four stages (11). In stage one, the predominant circulatory diseases are rheumatic heart disease, those attributable to other infections, and nutritional deficiency-related disorders of the heart muscle. In the second stage, as infectious disease rates decline and nutrition condition improves, diseases related to hypertension, such as hemorrhagic stroke and hypertensive heart disease, become more common. In the third stage, which includes the highest CVD mortality rates, atherosclerotic processes lead to a high incidence of ischemic heart disease and atherothrombotic stroke, especially at ages older than 50 years. During the fourth stage, increased efforts to prevent, diagnose, and treat ischemic heart disease and stroke typically delay these diseases until older ages.

Our results suggest that rural Vietnam, as represented by the Bavi district, is in the second stage of the CVD epidemic, as are China and some other Asian countries. India is already progressing to stage three of the CVD epidemic, whereas developed countries are typically in stage four (11).

Of particular interest in this article are differences in CVD mortality by demographic and socioeconomic conditions because, according to transition theory (12), subsets of populations may be at different stages of the CVD epidemic.

Our results show that CVD mortality rates were significantly higher for men than women, and the differences in mortality by sex were larger for CVD than for other NCD causes. The greater risk for men of dying from CVD was also stated by the American Heart Association (13) and can be explained by the differences in risk factor profiles between men and women. In our setting, men smoked more and had a higher prevalence of elevated blood pressure (14,15). In this study, age was proven to be more strongly associated with CVD mortality than other NCD causes.

We found that CVD mortality rates decreased considerably among educated people compared with those without formal education, even after adjusting for other independent variables such as sex, age, and economic status. This finding is similar to the findings of numerous studies that showed an inverse socioeconomic gradient in CVD mortality in developed (15-18) and developing countries (19).

Likewise, education was an important factor for health among men and among women, particularly in rural areas, because we found that education is usually associated with increased knowledge about health matters and consequent reduction in risky health behaviors. Explanations for the differences in education levels we observed among those who died from CVD include differences in risk factors such as blood pressure, blood cholesterol, smoking, and obesity. Indeed, available evidence from other studies in the Bavi district support the observation that people with lower education levels smoke more (20) and have more hypertension (14). 

In this study, CVD mortality was found to have stronger association with education than other NCD mortality in both sexes. This finding is similar to findings from studies in England (21), Israel (22), and Korea (23).  

A strong correlation between old age (50 years and older) and education was also found in relation to CVD mortality. Even though the risk of dying from CVD associated with lower education level was similar for men and women, the effects of women's education are particularly noteworthy. Given the results, together with the relatively higher proportion of women aged 50 years and older (30% of women aged 20 years and older), of whom 55% had no formal education, better education for women may substantially reduce CVD mortality. Better education likely leads to healthier lifestyles (e.g., lower rates of smoking, drinking, physical inactivity) and is likely to improve access to health care.

Economic status was not significantly associated with CVD mortality in the multivariate regression model. This model showed a possibly rising burden of CVD mortality among the poor, which demonstrates the shift from stage one to stage two of a CVD epidemic.

Because we were working in a setting without medical death certification, our findings depend on the quality of the verbal autopsies. There are numerous factors that may affect the validity of verbal autopsies, including cause of death, characteristics of the deceased individuals, classification of causes of death, the design and content of the questionnaire, and interviewing procedures. The verbal autopsy process is difficult to validate rigorously, particularly for detailed differentiation between different types of CVD. Based on experiences with verbal autopsy in 1999 (24), we chose interviewers with more appropriate qualifications (e.g., medical background, field experience), carefully organized the training, designed a detailed questionnaire, and involved one local clinical physician as an assessor of the questionnaire instrument. Because of the lack of availability of medical documents and records stored either by the family or in health facilities, reliable standards such as health facility or hospital diagnoses and postmortems were not available. However, in settings with poor vital registration and weak health systems, especially at the community level, the verbal autopsy method is the only viable option for determining cause of death (24). We are currently working on the development of a probabilistic model for interpreting verbal autopsy, which may lead to a more objective assessment of cause of death (25).

In this study, education was used as a main indicator of socioeconomic position, which has several advantages. It occurs causally before occupation and income and is stable throughout life after young adulthood. Unlike occupational class, education allows classification of individuals who do not work, including, for instance, most of the elderly individuals in our study. Educational status is an individual measure of socioeconomic position, which may be a better indicator than household measures, such as household income, which are difficult to measure in our setting (26).

Economic status is a household-level indicator, can change over time, and is difficult to measure correctly in our context (26). It may, however, provide another dimension of the relationship between socioeconomic status and CVD mortality.

In conclusion, this study demonstrated a heavy burden of CVD mortality among the adult population in a rural community of Vietnam, during a period of epidemiological transition. It also suggests that education may be an important factor in preventing deaths from NCD and CVD, especially in reducing CVD mortality among women. This study also demonstrates the potential of the DSS as a means of characterizing a major community-based epidemic in a setting where routine data collection is insufficient for effective health policy making and planning.

There is an urgent need to develop and implement effective CVD policies and interventions in Vietnam. Interventions should include both primary prevention, (e.g., educating people about health hazards, promoting healthy lifestyles) and secondary approaches (e.g., improving geriatric services). Whether adult education programs can offer any prevention against CVD remains an open question. Further studies over longer periods are required to give more insight into the CVD epidemic. The link between CVD risk factors and mortality in transitional settings also needs further investigation to prioritize health promotion messages.