Corresponding Author Tekebash Araya, School of Public Health, College of Health Sciences, Addis Ababa University, Addis Ababa, Ethiopia. Tel.: +251 911234655; Fax.: +251 115513099; E-mail: Tekebash@gmail.com
Objective To determine the level of HIV-related mortality reduction after the introduction of large-scale antiretroviral therapy (ART) using a burial surveillance system coupled with verbal autopsy (VA) in Addis Ababa, Ethiopia.
Methods Prospective burial surveillance was established in 2001 at cemeteries in Addis Ababa. VA interviews were periodically conducted on a random sample of adult burials registered between 2001 and 2009. Independent physicians reviewed the completed VA questionnaires and assigned underlying causes of death. The period before 2005 was defined as pre-ART and that since 2005 as the ART era. HIV-specific mortality fractions were calculated by age, sex and year of burial to examine the mortality trends before and during the ART era.
Results Of the 4239 VA physician diagnoses, 1087 (25.6%) were ascribed to HIV-related deaths. HIV-related deaths in 2009 were 33% fewer than in 2001. The proportion of HIV-related deaths was reduced from 44.0% in the pre-ART period to 20.0% in the ART era. Mortality in women (36.7%) declined more than in men (30%). A marked reduction in HIV-specific mortality was observed in the age group 30–39 years (from 69.1% pre-ART to 46.8% during ART era) compared to 20–29 (from 60.5% pre-ART to 41.0% during ART) and 40–49 year olds (49.7%) pre-ART to 34.4% during ART provision).
Conclusion Burial surveillance combined with VA demonstrated a significant reduction in HIV-related deaths during the provision of free ART. Replication of burial surveillance is recommended in similar settings, where a vital registration system is non-existent, to track large-scale population-level interventions.
Objectif: Déterminer le taux de réduction de la mortalité liée au VIH après l’introduction de l’ART à grande échelle, en utilisant un système de surveillance des enterrements coupléà l’autopsie verbale (AV) à Addis-Abeba, en Ethiopie.
Méthodes: La surveillance prospective des enterrements a étéétablie en 2001 dans les cimetières à Addis-Abeba. Des interviews d’AV ont été régulièrement menées sur un échantillon aléatoire des enterrements d’adultes enregistrés entre 2001 et 2009. Des médecins indépendants ont passé en revue les questionnaires remplis d’AV et assigné les causes de décès. La période avant 2005 a été définie comme pré-ART et celle depuis 2005, étant l’ère ARV. Les fractions de mortalité spécifique au VIH ont été calculées par âge, sexe et année de l’enterrement afin d’examiner les tendances de la mortalité avant et pendant l’ère ARV.
Résultats: Sur les 4239 AV/diagnostics de médecin, 1087 (25,6%) ont été attribués à des décès liés au VIH. Les décès liés au VIH en 2009 étaient de 33% de moins qu’en 2001. La proportion de décès liés au VIH est passée de 44,0% dans la période pré-ART à 20,0% dans l’ère ARV. La mortalité chez les femmes (36,7%) a reculé davantage que chez les hommes (30%). Une réduction marquée de la mortalité spécifique au VIH a été observée dans le groupe d’âge 30–39 ans (passant de 69,1% pré-ART à 46,8% pendant l’ère ART), comparée au groupe d’âge 20–29 ans (passant de 60,5% pré-ART à 41,0% pendant l’ère ART) et au groupe d’âge 40-49 ans (passant de 49,7% pré-ART à 34,4% pendant l’ère ART).
Conclusion: La surveillance des enterrements combinée à l’AV a démontré une réduction significative des décès liés au VIH pendant la période de l’offre gratuite de l’ART. La réplication de la surveillance des enterrements est recommandée dans des contextes similaires, où un système d’état civil est inexistant, pour permettre le suivi des interventions de grande envergure basées sur les populations.
Objetivo: Determinar el nivel de mortalidad relacionada con el VIH, tras la introducción del TAR a gran escala, utilizando un sistema de vigilancia de entierros junto con autopsias verbales (AV) en Addis Ababa, Etiopía.
Métodos: En el 2001 se estableció un estudio prospectivo de los entierros en cementerios de Addis Ababa. Se realizaron autopsias verbales de forma periódica en una muestra aleatoria de entierros de adultos registrados entre el 2001 y 2009. Una vez completados los cuestionarios de las AV, estos fueron revisados por médicos independientes quienes asignaron las causas de muerte subyacentes. El periodo anterior al 2005 se definió como pre-TAR y desde el 2005 como la era del TAR. Las fracciones de mortalidad VIH específicas se calcularon por edad, sexo y año del entierro, para examinar las tendencias en mortalidad antes y durante la era del TAR.
Resultados: De los 4239 diagnósticos de AV realizados por médicos, 1087 (25.6%) fueron asignados a muertes relacionadas con el VIH. Las muertes relacionadas con el VIH en el 2009 fueron un 33% menos que en el 2001. La proporción de muertes relacionadas con el VIH se redujo de un 44.0% en el periodo pre-TAR a un 20.0% en la era TAR. La mortalidad en mujeres (36.7%) disminuyó más que en hombres (30%). Se observó una reducción marcada de la mortalidad VIH específica en el grupo de edad de los 30–39 años (de 69.1% pre-TAR a 46.8% durante la era TAR) comparada con el grupo de 20–29 años (de 60.5% pre-TAR a 41.0% durante el TAR) y de 40–49 años (49.7%) pre-TAR a 34.4% durante la era TAR).
Conclusión: La vigilancia de entierros combinada con las autopsias verbales demostró una reducción significativa en las muertes relacionadas con el VIH durante la entrega gratuita del TAR. Se recomienda hacer réplicas de la vigilancia de entierros en emplazamientos similares, en donde no exista un sistema de registros vitales, para mantener un registro de intervenciones a gran escala en la población.
Consistent and reliable cause-specific mortality data are essential for monitoring the success of antiretroviral therapy (ART). However, the chances of a death being registered and the causes of death being documented strongly depend on the socioeconomic status of the community and nation in which it occurs (Byass 2007). In rich countries, mortality statistics are important part of vital registration systems. However, most sub-Sahara African countries lack systems to fully capture deaths and distribution of causes of death (Mathers et al. 2005). In these countries, data on cause-specific mortality often rely on records from medical facilities that are often incomplete, inaccurate and biased (Bennett et al. 2006) and hospital statistics are often not representative of the population because not all communities have equal access (Cooper et al. 1998). Health services utilisation is often low and selective and most deaths occur at home, making it difficult to establish the cause of death (Murray et al. 2007). For AIDS, in particular, there is evidence that some families withdraw HIV-diagnosed relatives from hospitals (Arthur et al. 2000).
There is a particular need to develop methods to evaluate the impact of interventions in countries with widespread HIV/AIDS epidemics. The ultimate goal of the scale-up of ART is to reduce AIDS mortality (Porter & Zaba 2004), and monitoring the success of such programmes relies on accurate evaluation of HIV-related mortality at population level (Diaz et al. 2005). In African settings, studies on cohort of patients on ART in demographic surveillance sites (DSS) have indicated population-level decline of AIDS mortality after the introduction of ART (Jahn et al. 2008; Herbst et al. 2009). Similarly, a study using evidence from registers, coffin sales and funerals reported mortality reduction associated with ART (Mwagomba et al. 2010).
Similar to most sub-Saharan Africa, HIV/AIDS has been a great challenge to the Ethiopian health system and remains among the major causes of adult deaths over the past two decades (Seyoum et al. 2009). Recent HIV-prevalence figures indicate a generalised epidemic probably stabilizing or even declining in major urban areas (Berhane et al. 2008). In Ethiopia, a co-pay scheme ART was introduced in 2003 and free ART roll-out was declared in January 2005 (Berhane et al. 2008). A hospital-based cohort of HAART patients reported improved survival and decreased tuberculosis incidence (Jerene et al. 2006).
In Addis Ababa, HIV prevalence among women aged 15–24 years has shown a decline of 35% between 1996 and 2005, falling from 20.7% to 13.5% in 2005 (MOH/HAPCO 2006). A study using the burial surveillance method in same population indicated a decline in adult AIDS mortality by 38.2% and 42.9% for men and women, respectively, between 2005 and 2007 (Reniers et al.2009).
The DSS in 20 countries, the Sample Registration System (SRS) in India and the Disease Surveillance Points (DSP) in China regularly use verbal autopsy (VA) on a large scale, to assess the causes-of-death structure of a defined population (Soleman et al. 2006). It has also been practiced in rural (Chandramohan et al. 1998; Lulu & Berhane 2005; Fantahun et al. 2006) and urban (Araya et al. 2004; Reniers et al. 2009) parts of Ethiopia to identify all causes of death including AIDS mortality. These studies clearly demonstrate the value of the VA method in resource-poor settings to support clinical observations and advance epidemiological studies.
In Addis Ababa, where the dead are buried and not cremated, registration of burials provides mortality data with background characteristics and address of a deceased. A convenient feature of the burial surveillance is that it taps into an existing infrastructure of burial sites (Sanders et al. 2003). While several scientific reports describe the surveillance of burials, the purpose of this paper is to describe the application of burial surveillance method complemented by VA technique as a timely source of mortality data to determine the level of HIV-related mortality following the introduction of large-scale ART in Addis Ababa.
Description of study area
Burial surveillance has been conducted since 2001 in Addis Ababa, the capital of Ethiopia. In 2001, the projected population of Addis Ababa based on the 1994 census was estimated at 2.6 m (CSA 1995). In the 2007 census, the population was estimated at 2.7 m, comprised of 1.3 m (47.6%) men and 1.4 m (52.4%) women; two-thirds (1.8 m, 66%) in the age group 15–45 years with 53% women. By religion, 75% were Orthodox Christian, 16% Muslim and 8% Protestant, and the rest were Catholics or followers of traditional religions (CSA 2010).
A governmental co-pay ART program was introduced in 2003. Patients possessing a poverty certificate were given free ART. From January 2005, free ART was available to everyone and the number of patients benefiting from the programme increased quickly (Berhane et al. 2008; Reniers et al. 2009). By the end of January 2009, a total of 50 705 patients had ever started ART (MOH/HAPCO 2009). As would be expected, the rapid increases in enrolment rates were accompanied by important reductions in AIDS mortality (Reniers et al. 2009).
The burial surveillance data registered at all cemeteries (N = 88) in Addis Ababa were the primary source of data. On the funeral day, information about the deceased including name, age, sex, address and other identifiers was collected by trained cemetery clerks who interview the deceased’s relatives while performing prerequisites for funeral. The methodology of burial surveillance is described elsewhere (Sanders et al.2003; Araya et al. 2004; Reniers et al.2009). Bodies that are unidentifiable are sent by police or hospitals to a municipality-based cemetery named ‘Baytewar’ for burial. This unique cemetery comprises ≥15% of all burials lacking addresses and more than 90% have no name, age or other identifiers (Sanders et al.2003). Lay cause of death within burial surveillance has been validated as a useful instrument by which to measure the impact of ART on HIV-related mortality at population level (Reniers et al. 2009).
The second source of data used in this study was VA, a method of identifying the cause of death based on an interview with next of kin or other caregivers regarding symptoms, signs and circumstances preceding death (Soleman et al. 2006). It has been used by other investigators to assess HIV-related deaths (Kamali et al. 1996). The World Health Organization (WHO) has standardised the VA method, and it is the only practical option in developing countries for ascertaining cause of death (Baiden et al. 2007).
Verbal autopsies were conducted on a one-in-ten random sample of 60 036 adult burials registered in 3 months of 2001, in 8 months of 2003, in 4 months of 2006, all of 2007 and 2008 and for 7 months of 2009. A pair (men and women) of trained interviewers conducted the interview by tracing the home address of the deceased. The interviews were undertaken between one and 6 months (mean = 3.2 months) from the date of death. Each interview lasted between 30 and 45 min.
Estimating cause of death
In this study, two physicians independently reviewed the completed questionnaires and assigned underlying causes of death and, where possible, contributing and immediate causes of death. The physicians were given a copy of the ICD10 (WHO 1993) and forms on which to record diagnosis. A non-physician research assistant checked for consistency of diagnosis between the two physicians and assigned final diagnosis and ICD10 code. A diagnosis was considered consistent when both physicians agreed on the underlying cause of death. First time consistent diagnoses were 70% of interviewed VA questionnaires. If physicians disagreed about the cause of death, the questionnaire was reviewed by a third independent physician. If all three physicians failed to reach a consistent diagnosis, the questionnaire was reviewed by panel and, where possible, a diagnosis was assigned by consensus. A cause of death for which physicians could not assign a specific diagnosis was considered undetermined (5%).
Data processing and analysis
For quality assurance, clearly documented data cleaning syntaxes in Access and STATA software were used. Cleaned data were prepared for analysis by stripping identifiers from the master database. As the VA procedure requires the complete address of the deceased, all ‘Baytewar’ burials were excluded from VA procedures. This particular study focuses on adult deaths (age ≥12 years) (Table 1). Among the households selected for VA interview, on average 70% were successfully interviewed, while the rest were not interviewed for various reasons (20% household not found, 5% refusal, 5% caregiver unavailable). This study used only adult VA interviews with physician diagnosis for the aforementioned study years.
Table 1. Number and percentages distribution of registered burials by cemetery type, age and sex as reported by caregivers of the deceased (2001–2009), Addis Ababa, Ethiopia
Total number of registered burials
159 774 (100.00)
All Baytewar burials
25 219 (15.80)
All non-Baytewar burials
134 555 (84.20)
Child non-Baytewar (age <12 years)
15 949 (11.85)
Adult non-Baytewar (age ≥12 years)
118 606 (88.15)
58 616 (49.43)
59 976 (50.57)
Subtotal with sex record
118 592 (99.99)
Missing sex record
We defined HIV-related death as a death resulting from HIV/AIDS assigned by panel of physicians. In this study, HIV-specific mortality fraction is the proportion of HIV-related deaths divided by the total number of VAs reviewed by physicians. All VA physician diagnoses were categorised into major causes of death and analysed for patterns of mortality compared to HIV-related deaths in the years VAs were conducted. Mortality owing to HIV/AIDS by age was computed to identify the most severely affected age group. HIV-specific mortality fractions were calculated by sex and year of burial to evaluate the mortality trends. The study period was dichotomized into ‘pre-ART era’ (before 2005) and ‘ART era’ (since 2005).
Ethical clearance was secured from the institutional review board (IRB), Faculty of Medicine of Addis Ababa University with biannual renewal and at national level, from the National Health Research Ethics Review Committee of the Ministry of Ethiopian Science and Technology with annual renewal. Official permission was obtained from religious leaders, municipal officials and cemetery authorities before initiation of burial registration. Before obtaining the necessary information, cemetery clerks explained the purpose of the surveillance to close relatives or friends of the deceased. Before VA interviewers conducted interviews, consent was obtained from the closest adult caregiver of the deceased.
A total of 4239 adult VA diagnoses were analysed: 200 from 2001, 790 from 2003 and 3249 from 2006–2009. Slightly more than half (2123, 50.1%) of the diagnoses were for men and 2116 (49.9%) for women. Deaths attributing to communicable diseases excluding HIV vary between 15% in 2003 and 16% in 2009 and owing to injuries between 6.45% in 2003 and 7.14% in 2009 (Fig. 1). The overall proportion of HIV-related deaths was 25.6% (n = 1087) comprising 22.9% (n = 485) men and 28.5% (n = 602) women. The trend of HIV-related deaths among men declined from 40.6% in 2001 to 10.6% in 2009. The same pattern of decline was also observed among women (from 50.6% in 2001 to 13.9% in 2009) (Fig. 2). This equates to a decline of about 30% for men and 37% for women from the pre-ART era to 3 years after the introduction of free ART. Mortality as a result of HIV-related declined significantly from 44.0% in the pre-ART era to 20.1% in the free ART era (Fig. 3).
Before the provision of free ART, HIV-related deaths were highest in the 30–39 years age group (69.1%), followed by 60.5% and 49.7% in the 20–29 year and 40–49 year groups, respectively. The same age groups were also highly likely to die of HIV-related after the introduction of free ART with 46.8%, 41.0% and 34.4% among 30–39, 20–29 and 40–49 year olds, respectively (Fig. 4). A more pronounced reduction in HIV-specific mortality was observed in the 30–39 year olds when compared with the 20–29 and 40–49 year olds. Deaths ascribable to HIV-related in the 30–39 year olds decreased from 69.1% pre-ART to 46.8% during the ART era, while for the 20–29 and 40–49 year olds, it decreased from 60.5% pre-ART to 41.0% after free ART and from 49.7% pre-ART to 34.4% after free ART provision, respectively (Fig. 4).
While there are abundant intervention programmes offering prevention, treatment and care for AIDS patients, documentation of their effectiveness at the population level is limited (Assefa et al. 2009). In Ethiopia, data on the effectiveness of HIV treatment at population level are scarce and appropriate systems to evaluate their interventions within the health care system are also lacking (Reniers et al. 2009). While there are several reports indicating significant decline of HIV prevalence (MOH/HAPCO 2007) and adult deaths within the years immediately after introduction of ART (Reniers et al. 2009), studies documenting the level of AIDS-attributable mortality after accelerated access to ART are scarce. Hence, we believe the burial surveillance approach in combination with VA data can bridge that gap.
As a substitute for hospital data, VAs are used to monitor cause-specific mortality and VA physician diagnosis has been validated among adult deaths in African countries (Chandramohan et al. 1998) including Ethiopia (Fantahun et al. 2006). Mortality surveillance systems using VA procedures represent a cost-effective and sustainable medium-term solution to this problem (Soleman et al. 2006). In this paper, we document the impact of large-scale ART intervention on AIDS mortality using a combination of burial surveillance and VA methods. The results demonstrate significant reduction in mortality attributed to AIDS during the study years.
Scattered studies have demonstrated the impact of ART on AIDS mortality in different African settings. A decline in population-level mortality was observed shortly after the introduction of ART (Jahn et al. 2008; Herbst et al. 2009). These studies mainly focus on rural-based DSS with relatively delineated smaller populations (Herbst et al. 2009) or ART cohorts (Jahn et al. 2008). Our earlier study of the same population and using similar methods documented the decline of AIDS deaths using lay-reported cause of death from burial surveillance. The study also suggested that periodic re-anchoring of lay reports may be necessary in the absence of continuous VA (Reniers et al. 2009). The availability of continuous VA will further complement previous studies without any underlying assumptions.
Our results depict a relatively constant trend in causes of death other than HIV-related death and non-communicable diseases. Excluding AIDS, the proportion of communicable disease is less than 20% with a stable trend over the study years. However, HIV-related mortality steadily declined by 50.0% and 54.0% in the male and female populations, respectively, in 2007. By the end of the study period, the proportion of HIV-related deaths had almost halved (41.0%vs. 26.0% for men and 51.0%vs. 27.0% for women).
The findings of our study are consistent with reports in Addis Ababa that showed reduction in HIV-related deaths after years of ART scale-up (WHO/UNAIDS/UNICEF 2008). Studies conducted in similar African settings, such as Botswana and Malawi, were highly suggestive of an association between AIDS mortality and ART scale-up programmes (Stover et al. 2008; Mwagomba et al. 2010). Likewise, the immediate reduction in HIV-related mortality observed in this study just after accelerated free ART is highly indicative of such a relationship. In this regard, using the burial surveillance method with VA technique to measure such outcome is an alternative technique of documenting population-level cause of death and evaluating large-scale interventions, like ART.
We argue that burial surveillance is feasible and less costly for poor countries like Ethiopia, where cremation is not practiced and burials are conducted at dedicated religious or municipal burial sites. More importantly, in a traditional African community like Ethiopia, where religious institutions are influential, information generated using this system may have more social and cultural acceptance, in turn influencing policy decisions. The inclusion of VA further strengthens the estimate of cause of death. The completed VA interviews included in this study varied by year owing to several reasons such as number of deaths registered, eligible burial records, budgetary limitation and completeness of interviewed VAs. Since June 2009, we have included reporter’s address to improve identification of deceased households and this has improved completeness of VA interviews. The fluctuation in the numbers of deaths at burial sites is mainly attributed to establishment of new cemeteries, closing of cemetery at some sites and body repatriation. Hence, documenting the completeness of death registration is essential to strengthen the estimate of the population-level cause of death.
For determining the level of HIV-related outcomes, mortality data are the most basic health outcome indicator to measure population health status and effectiveness of interventions (Mathers et al. 2005). Thus, burial surveillance offers opportunities to evaluate HIV-related mortality and to provide a solid platform for policies and decisions.
In an urban setting like Addis Ababa, where vital registration is absent or incomplete, hospital deaths and clinical records are scarce and cremation is not practiced, burial surveillance is a cost-effective method of capturing deaths at population level. The use of VA is vital in documenting patterns of cause of death and estimating HIV-related deaths in poor settings during the scale-up of ART. We recommend assisting the establishment of municipal infrastructure and controlled cemeteries to strengthen burial surveillance in resource-poor settings.
The Addis Ababa Mortality Surveillance Program has been made possible with major financial support from the AIDS Foundation of Amsterdam, the Netherlands (grant no. 7022); WHO (Second Generation Surveillance on HIV/AIDS, contract no. SANTE/2004/089-735); Centers for Disease Control and Prevention (EPHA-CDC cooperative agreement no.U22/CCU022179); Mellon Foundation (pilot project grant to the Population Studies Center of the University of Pennsylvania); and Hewlett Foundation grant to the University of Colorado at Boulder for the African Population Studies Research and Training Program. The project receives institutional support from the Faculty of Medicine of Addis Ababa University and the Ethiopian Public Health Association. Religious Leaders and Addis Ababa Labour and Social Affairs Bureau facilitated our access to cemeteries. Physicians who reviewed the VA questionnaires are grateful.