This study is part of the Addis Ababa Mortality Surveillance Program (AAMSP) that uses the verbal autopsy method on adult mortality. As usable mortality data are lacking in Ethiopia to measure the impact of socioeconomic and behavioral factors on causes of mortality, it has been necessary to apply the verbal autopsy method on data from burial surveillance for such analysis. The AAMSP analyzes data from surveillance of burials in Addis Ababa to capture causes of adult deaths. The burial surveillance has been conducted since 2001 in all cemeteries under the city limit of Addis Ababa. Addis Ababa harbors 89 cemeteries (670 are church based, 9 are mosque based, while 10 are municipality owned cemeteries) . The data used are for the period of September, 2006 to December, 2009.
The burial surveillance is used as a sampling frame for the verbal autopsy procedure . As cremation is not practiced in Addis Ababa, all burials of deaths are conducted at the above mentioned religious or municipality based cemeteries. Thus, in principle, the burial surveillance captures all deceased residents of Addis Ababa, although biases exist because residents may die and/or be buried outside the capital just as non-residents may be buried inside Addis Ababa. Some of these biases are mostly identified and corrected while others inevitably go unnoticed.
The burials are registered by cemetery clerks who are well trained and regularly given annual refresher trainings. They report all deaths (n ≈ 18,000 per year) from cemeteries using structured forms. They collect the information from relatives or close friends during burial ceremonies. The information include: date of burial, age, name, sex, address, marital status, region of birth, ethnicity, religion, and a lay reported cause of death .
Adult deaths aged 15 years and above that were captured by the burial surveillance in 2007 were 18,013, making age specific death rate of 8.9 per 1000 for Addis Ababa. This figure is very close to the results of the census by the Ethiopian Statistical Authority for the same year (18,686 total deaths of aged 15 years with age specific death rate of 9.2 per 1000) . Adult deaths captured by the burial surveillance in 2008 and 2009 were 17,984 and 18,154 respectively. Overall, the burial surveillance has identified 58,010 deaths during September 2006 to December 2009. Among these, 49,309 (85%) deaths were eligible for verbal autopsy procedures while the remaining could not be subject to such procedures because the burials were conducted without close relatives or friends who could provide information for the verbal autopsy interviews.
Verbal Autopsy is interviewing the relatives or caregivers about the signs, symptoms, lifestyle behaviors and other characteristics experienced by the deceased before their death and the circumstances surrounding their death . The verbal autopsy questionnaire was piloted and adapted to local situation from a standardized WHO and International Network of field sites with continuous Demographic Evaluation of Populations and their Health in developing countries (INDEPTH Network) verbal autopsy questionnaires [17, 18]. Some modifications were made on the questionnaire for the purpose of enhancing local comprehension and ensuring cultural acceptability. This helped the program to obtain internationally comparable verbal autopsy data with the burial surveillance approach. The questionnaire consists of identification of respondents and care givers, identification of the deceased, death related information, signs and symptoms during illness and list of possible risk factors.
Three data collectors, who are high school graduates with similar previous experiences, are deployed to visit each of the sampled households on average of two months after the death occurred. Data collectors are given extensive training on the objective of the program, on the questionnaire and on skills of interview. To maintain the quality of the data, annual refresher trainings, monthly supervisions and weekly meetings are conducted.
After data entry and cleaning, 10% (4,931) of the deaths were randomly drawn using the Visual Basic Computer Program for verbal autopsy. Among the drawn sample, 91% (4,494) were adults of age 15 years and above, and 9% (437) were under 15 years of age. Verbal autopsies were completed among 3,709 (83%) of the adult deaths, while care givers were not willing to participate or not available with repeated visits for the rest of the cases.
All completed verbal autopsies also underwent physician review before assigning underlying causes of death. Initially, two physicians reviewed every completed verbal autopsy interview blindly to assign possible cause/s of death. Any inconsistent case within the initial review is referred to a third physician. If the results of the three physicians are inconsistent, it would be referred to a panel discussion. In the few cases where it is difficult to arrive at a probable cause of death even after panel discussion, the cause of death will be labeled as “undetermined”. Finally, the International Classification of Diseases Version 10 (ICD-10) is used for standardization and for comparison with other studies.
Behavioral risk factors
The adult verbal autopsy questionnaire for this study contains similar questions to that of the internationally standard verbal autopsy questionnaire on consumption of tobacco and alcohol . In addition, khat chewing has been included as a locally important behavioral risk factors  with the globally identified risk factors of tobacco and alcohol consumption . Since a system for surveillance of behavioral risk factors is lacking in Ethiopia, the verbal autopsy with burial surveillance can be considered as one alternative approach to examine the prevalence of behavioral risk factors and their associations with mortality. The questionnaire elicited information on the duration and frequency of tobacco and alcohol consumption as well as khat chewing. Alcohol consumption and khat chewing were categorized into: frequently in a week (at least four times a week) and occasionally; while tobacco use was categorized as frequently (at least once per day) and occasionally. Furthermore, there were categories for getting drunk with alcohol as frequently, once in a week and occasionally.
Demographic and socioeconomic factors
The causes of deaths were analyzed in relation to demographic (age, sex and marital status) and socioeconomic (religion, educational and occupation) factors. The age of the deceased (in completed years) is categorized as: 15 to 24, 25 to 44, 45 to 54, 55 to 64, and 75 and older. Educational status is categorized as: no education, primary education, and secondary education, above secondary and others (traditional). Occupation is categorized as: professional (technical/managerial/sales/clerical/self-employed), manual labor (skilled/unskilled), housewives, unemployed, retired, and others (students, farmers).
Data management and analysis
Double data entry has been implemented with Microsoft Access program followed by a thorough data cleaning using STATA driven .do files. For this analysis the age group 15 years and above was selected due to the fact that this group is economically productive and is highly affected by over-consumption of the substances at issue . Using this age group for analysis is also considered appropriate since the burial surveillance method is prone to under-reporting child deaths [14, 15, 21].
Further analysis focused on identified leading causes of diseases in the study area such as cardiovascular diseases, malignancy, chronic liver diseases, HIV/AIDS and tuberculosis and presented meaningful findings . The 2006 Global Burden of Diseases classification was adapted to classify causes of death in this study. The classification categorized diseases into: Group I (communicable diseases, maternal conditions and nutritional deficiencies); Group II (non-communicable causes); and Group III (injuries) .
Statistics such as frequencies and cross tabulations are applied to examine the distribution of each demographic, socio economic and behavioural risk factors. To compare proportions of behavioural risk factors by causes of death, the chi-squared test is used. Further, binary logistic regression was applied to examine the relationship between demographic, socioeconomic and behavioural risk factors with selected causes of death using the STATA software (Stata Corp LP, College Station, Texas).
The protocol for the Addis Ababa Mortality Surveillance Program has been approved by the Institutional Review Board (IRB) of the College of Health Sciences, Addis Ababa University as well as the Ethics Committee of the Ethiopian Ministry of Science and Technology. Permission for conducting the study within the burial sites has also been obtained from the local authorities. Individual interviews are conducted after getting verbal informed consent from the immediate kin and caregivers of the deceased. At the program office, individual information is accessible only to the research team and is kept strictly confidential.