For 1,500 years, epidemics of communicable diseases such as plague, cholera, smallpox, typhus, and dysentery were documented as major causes of death in Ethiopia. In the early 20th century, one-fifth of the Addis Ababa population died of influenza (1). In the late 20th century, meningitis, malaria, cholera, and AIDS ravaged the country (2). Analogous data for noncommunicable diseases, however, are scarce (3). Thus, it is difficult to assess the double mortality burden, that is, the increasing threat of noncommunicable diseases as a cause of death while deaths from infectious diseases remain highly prevalent. The health system continues to rely heavily on the conventional infectious disease paradigm and is unresponsive to the emerging epidemiological shift.

Evidence on the causes and patterns of death is required to understand the overall epidemiological profile of diseases in Ethiopia and to help planners and decision makers prioritize the public health agenda. The objective of this study was to examine the proportion of deaths in Addis Ababa caused by communicable diseases, noncommunicable diseases, and injuries using verbal autopsy, a method of interviewing relatives or caregivers of the deceased about their signs, symptoms, lifestyle behaviors, and other characteristics before death and the circumstances surrounding their death.

We conducted verbal autopsies of a sample of deceased Addis Ababa residents for whom data were collected by the Addis Ababa Mortality Surveillance Program from September 2006 through December 2009. Our sampling frame was all burials recorded by burial surveillance, which has been conducted since 2001 in all cemeteries within the city limits. Cemetery clerks submitted forms for all deaths (approximately 20,000 per year). In principle, burial surveillance records data for all deceased residents of Addis Ababa, although biases exist because residents may die or be buried outside the capital, just as nonresidents may be buried inside city limits. Some of these biases are identified and corrected, but others inevitably go unnoticed (4-6). However, the 2007 Ethiopian census showed that Addis Ababa had 18,686 deaths of adults (defined for our study as aged ≥15) in the year before the census, a death rate of 9.2 per 1,000 (7). This is similar to the number identified by burial surveillance, 18,013, an age-specific death rate of 8.9 per 1,000. Similarly, adult deaths identified in 2008 and 2009 were 17,984 and 18,154, respectively.

The study population was all adults buried in Addis Ababa cemeteries. Addis Ababa has 89 cemeteries; 64 are Ethiopian Orthodox Church-based, 10 are municipal, 9 are mosque-based, and the rest are associated with the Catholic Church, Greek Orthodox Church, and synagogues. Because cremation is not practiced in Addis Ababa, virtually all deaths result in burials conducted at religious or municipal sites (8).

The burial surveillance identified 58,010 deaths during the study period. Of these, 49,309 (85%) deaths were eligible for verbal autopsy; the rest were not because they had been buried without close relatives or friends who could provide information for verbal autopsy interviews. The verbal autopsy questionnaire was adapted from a standardized verbal autopsy questionnaire from the World Health Organization (WHO) and International Network for the Demographic Evaluation of Populations and Their Health in Developing Countries Network. Three pairs of trained data collectors administered the surveys. They underwent frequent refresher training with strict supervision.

We randomly selected 10% (4,931) of deaths from the eligible burial surveillance sample for verbal autopsy; we used a sample for financial and logistical reasons. Of these, 91% (4,494) were aged 15 or older. We completed verbal autopsies for 3,709 (83%) deaths. All underwent physician review, and underlying causes of death were assigned. Verbal autopsy was not conducted for the rest, 785 (17%) of the sample because some caregivers were unwilling to participate and others were not available, despite repeated visits to their residences.

The age group 15 or older was selected for this study because the prevalence of noncommunicable diseases is rapidly increasing for this age category in developing countries and because the age group 15 to 64 is considered the most economically productive. In addition, adult death rates are an indicator for overall death rates (9). Furthermore, burial surveillance tends to underreport infant deaths (10,11).

We adapted the 2006 Global Burden of Diseases and Risk Factors to classify causes of death in our study as follows: Group I (communicable diseases, maternal conditions, and nutritional deficiencies), Group II (noncommunicable causes), and Group III (injuries) (12).

We calculated percentages and proportional death rates using Stata software (StataCorp LP, College Station, Texas). We used a binary logistic regression model to assess associations and significant differences; adjusted odds ratios and 95% confidence intervals were calculated to assess strength of associations.

Of the 3,709 adults for whom verbal autopsies were completed, approximately half were women and 60% were aged 15 to 64 (Table 1). The median age at death was 55 (range, 15-105).

Overall, 51% of deaths were attributed to noncommunicable diseases, 42% to communicable diseases (of which a negligible percentage were due to maternal and nutritional conditions), and 6% to injuries. The cause of death was ill-defined for 7% of deaths (Table 2). These percentages exceeded 100% because multiple causes could be recorded for 1 death. Proportions of male and female deaths due to noncommunicable and communicable diseases were similar, but more women than men died of both. In contrast, injuries were responsible for 10% of male and 3% of female deaths.

The leading cause of death was cardiovascular disease (CVD) (24%); proportions for hypertension (12%) and stroke (11%) were similar and constituted most of the CVD deaths. The second and third most common causes of death were HIV/AIDS (19%) and tuberculosis (12%). Malignant neoplasm (10%) was the fourth leading cause of death; stomach cancer (2%) and other neoplasms (4%) were the most common neoplasms. Digestive tract diseases caused 9% of all deaths; chronic liver disease accounted for 4%. These 5 leading causes of death accounted for three-fourths of all deaths. The other main causes of death were injury (6%) and type 1 and type 2 diabetes (5%) (Table 2).

Most of the deaths in this study were among the young and middle age groups (aged 30-64), reflecting a typical mortality burden for low- and middle-income countries (12). Noncommunicable diseases were the leading cause of death by a considerable margin. Few mortality studies are available for Africa, but this finding is in line with the 2001 estimate for low- and middle-income countries by the Global Burden of Diseases analysis, in which 36% of the deaths were due to Group I, 54% to Group II, and 10% to Group III (12). This finding also supports World Bank estimates for Madagascar in 2006, where the percentage of deaths among adults aged 15 to 44 due to Group I and Group II diseases was equivalent (40%) but for adults aged 45 or older was 18% and 63%, respectively (13). Our findings also support those of a study from Banjul, Gambia, where noncommunicable diseases were reported as the leading causes of death among adults (14). These similarities may be due to similar lifestyle and sociodemographic characteristics of urban settings in Africa.

On the other hand, our finding is different from a 2004 WHO estimate for deaths of African adults aged 15 to 59, in which 62% of the deaths were due to Group I diseases, 25% to Group II, and 13% to Group III (15). The World Bank estimate for Kenya in 2006 is also different; 75% of deaths among people aged 15 to 44 were due to Group I, and 11% were due to Group II; 57% of deaths among people aged 45 or older were from Group I, and 27% were from Group II (13).

Moreover, our findings are different from the estimates by the 2002 WHO analysis for Ethiopia (16). In that analysis, noncommunicable diseases accounted for 23% of deaths, and the corresponding estimate for communicable, maternal, perinatal, and nutritional causes was 71%. The possible explanation for these differences could be that cause-of-death analysis by WHO used cause-of-death models because of lack of information on death rates at the country level (17,18). Our results are not directly comparable to a study conducted in rural Ethiopia (19), where leading causes of death are Group I. Our study is from an urban setting dominated by a sedentary, Western lifestyle (eg, cigarette smoking, obesity, stress, and consumption of refined foods).

In our analysis, cardiovascular diseases, HIV/AIDS, tuberculosis, malignant neoplasm, and digestive tract diseases were the leading causes of death and accounted for three-fourths of all adult deaths. This finding supports the indication that an epidemiological transition is occurring, especially in the urban population, while people are also hard hit by HIV/AIDS and tuberculosis (20). An earlier community-based study in Addis Ababa also showed that high blood pressure and physical inactivity were highly prevalent among adults (21). In addition, a study from Zimbabwe, an urban blood pressure follow-up survey, and data from a central registry confirmed high prevalence of hypertension (22).

Our finding that men were more affected by injury concurs with that of a cause-of-death analysis of sub-Saharan Africa (13). Even though our study found that both men and women were almost equally affected by CVD, HIV/AIDS and malignant neoplasm caused more deaths among women than men. Another study found that more women than men die of HIV/AIDS in Addis Ababa (23). The higher proportion of deaths from malignancies among women requires further epidemiological study.

Among young and middle-aged adults, HIV/AIDS, tuberculosis, and injuries were leading causes of death in our study, while CVD, malignancies, and digestive tract diseases were the leading causes among adults aged 45 or older. In general, the proportion of deaths from communicable diseases and injuries decreased with age, while that from noncommunicable diseases increased. This implies that the premature death rate from communicable diseases is higher than that from noncommunicable diseases, a finding that is in line with those of the World Bank's estimates for causes of death in sub-Saharan Africa that reported young adults (aged 15-44) had a high prevalence of HIV/AIDS and high incidence of injuries (13). The pattern of cause of death by age is also in line with the study from South Africa that reported that deaths from communicable diseases decreased with increasing age (24). The contribution of noncommunicable diseases to premature death increases with age starting at age 35. These premature deaths from noncommunicable diseases were mainly due to selected cancers, diabetes, hypertension and stroke.

This study had several limitations. Burial surveillance may not record data for all deceased residents of Addis Ababa, some of whom died outside the city. Verbal autopsies may be less accurate than physical autopsies, and causes of death were ill-defined in some cases.

In conclusion, this study closes some gaps in the estimates of causes of death for Addis Ababa residents that may be relevant to health policy. Noncommunicable diseases are the leading cause of death among adults in Addis Ababa. Together with the existing burden of communicable diseases, this double morality burden requires the attention of policy makers and planners. Leading causes of deaths such as CVD, HIV/AIDS, tuberculosis, malignancies, digestive tract diseases, injuries, and diabetes should be prioritized for attention.

Finally, the observed pattern of double mortality burden is unlikely to be unique to Addis Ababa and provides new insight into the progression of the epidemiological transition in urban Ethiopia. This study may therefore serve as a baseline for further assessments. Similar studies or surveys should be conducted with large cities and rural areas to examine the pattern of epidemiological transition and the factors associated with it.