Boston 2009 Meeting

Title: Estimating the burden of injuries in Africa
When: Oct 7 2009
Where: Harvard University, Cambridge, MA, USA
Agenda: Agenda.doc
Attended by:  Kunuz Abdela, Jerry Abraham, Kidist Bartolomeos, Kavi Bhalla, Nicole DeSantis, Vanessa Fawcett, Lois Fingerhut, James Harrison, Michael Lipnick, Richard Matzopoulous, Robert Mtonga, Megan Prinsloo, Fred Rivara, Saeid Shahraz, Maria Valenti, Pon-Hsiu Yeh, Diego Zavala

Summary:  The group at Harvard University has initiated a new project to generate regional estimates of the burden of injuries in Sub-Saharan Africa with funding from the World Bank Global Road Safety Facility. The project will last for one year starting October 2009. The work will provide input to the ongoing GBD-2005 study. The purpose of this meeting in Boston was to connect with key collaborators, discuss existing information sources, and identify the most efficient approaches for generating country level estimates in Africa. As a starting point, the meeting took a regional focus on Eastern and Southern Africa with participants with ongoing related research in Ethiopia, Zambia, South Africa, Uganda and Mozambique. 

These meetings were one of a cluster of meetings associated with the ongoing GBD-2005 Injury work that were held in Boston from Oct 7-11 2009. These included a a joint meeting of the GBD-Injury Expert group and the International Collaborative Effort (ICE) on Injury Statistics and a meeting of the ICD-11 revision group for external causes of injuries. Details of the other meetings can be found on the GBD-Injury Expert Group website

Meeting Notes:
The meetings began with an introduction to the GBD-2005 project and the Injury Expert Group by James Harrison.  Following this, Kavi Bhalla introduced the new project on estimating the burden of injuries in Sub-Saharan Africa within the GBD-2005 framework. He summarized the ongoing work at Harvard University for estimating national level road injuries in low-income countries, where a general methodology has been developed to estimate national incidence of road injuries by triangulating from multiple national data sources. This was illustrated by an example from Iran where national death registration data was used for estimating mortality and a national health survey was coupled with a sample of hospital data to estimate the incidence of non-fatal injuries. This methodology relies heavily on national death registration data, hospital discharge data, and national health surveys. However during the course of this project it has become clear that such data sources are rare in Africa and an alternate strategy is needed. This is the primary reason for the new work being undertaken in Africa. 

Next, Jerry Abraham described findings from an environmental scan of the data sources that could potentially inform estimates in Africa. These data sources are broadly classified into verbal autopsy data for causes of death (usually from rural Demographic Surveillace Sites), mortuaries (usually located in major urban centers), hospital-based injury surveillance projects, and community/national health surveys. Jerry described a pilot study of how such data sources could be used to build national estimates in Mozambique. He illustrated the process of extrapolating from urban mortality rates estimated in Maputo city from mortuary data and rural mortality rates estimated from the Manhica DSS site to national mortality estimates. And, he described the use of a national health survey (DHS) and hospital data to estimate the incidence of non-fatal injuries. 

Following the morning tea-break, Kidist Bartolomeos from the WHO led a discussion of the long term benefits of developing capacity in Africa for better injury surveillance. She described ongoing work being done by the WHO and regional partners. It was emphasized that while the current push for better current estimates from the existing data is necessary, it is important that this be part of a long-term agenda for building data infrastructure and analytical capacity in the region.

The post lunch session focused on country-specific plans and strategies.  
Mike Lipnick described the architecture of data sources from Uganda that could be brought to such an analysis. These include six months of retrospective data gathered from a mortuary in Kampala, the Mulago hospital mortuary, two rural DSS sites that collect data on causes of death using verbal autopsy, the ICCU hospital-based injury surveillance data, and a household survey (ongoing). 

Richard Matazopoulos and Megan Prinsloo described ongoing work in South Africa. In particular, Megan described preliminary estimates of the burden of injuries in the Western Cape. They have used corrected all-cause mortality estimates coupled with the injury profile from the city of Cape Town and two rural districts. Richard described ongoing efforts to establish a hospital-based non-fatal injury surveillance system. 

Kunuz Abdela described the relevant data sources from Ethiopia. These included the District Health Information System (hospital based injury surveillance system), the fatal injury surveillance system (mortuary based), a community injury survey, traffic police data, and a rural DSS site. Kunuz also emphasized the need for affordable methods that rely on periodic data collection that provide snapshots of burden estimates.

Diego Zavala and Robert Mtonga described the challenges in doing data collection in their multi-national injury surveillance project in Congo, Kenya, Nigeria, Uganda and Zambia. The surveillance focused on violence related injuries and road traffic injuries. Robert Mtonga also described various other data sources in Zambia, including the national health management information system, insurance companies, police, and mortuary. 

A group discussion followed on the assumptions and limitations of the broad method and the feasibility of replicating the Mozambique model in multiple countries. In general, the broad methodology was considered suitable for replication in other countries. However, the following methodological issues were identified as being of special concern:
  • Violence and conflict: The methods being developed are ill suited for reliable estimates of injuries resulting from violence and conflict. It was noted that though Harvard’s group has a special interest in improving estimates of road injuries, the project seeks to estimate the burden of all injuries. However, various key collaborators and participants in this meeting have experience in working in conflict settings and may be able to help improve estimates.  
  • Uncertainty estimation: The extrapolation methods discussed provide point estimates. Estimates of the uncertainty associated with these methods need more attention.
  • Possible biases in mortuary based data: It is likely that this project will rely heavily on mortuary based data collection. One key issue with using such data for estimating mortality incidence is ensuring that underlying catchment population is accurately estimated and that all injury deaths in this population are brought to the mortuary.
  • Poor nature of injury codes in hospital-based surveillance data. 
  • Other methodological issues: We need to be alert for misclassification (in some countries “Road injury - pedestrian” is a common dump code for homicides; Quality assessment of data sources should consider the magnitude of the less-specific “other” categories; 
In the days following these meetings, several smaller meetings were held to plan the logistics of the collaborative research. 


Introduction (Speaker: James Harrison)

Google Presentation

Introduction (Speaker: Kavi Bhalla)

Google Presentation

Data Sources in Africa (Speaker: Jerry Abraham)

Google Presentation

Case Study : Mozambique (Speaker: Jerry Abraham)

Google Presentation

Google Presentation

Multinational Injury Surveillance Pilot Project
(Speaker: Diego Zavala)

Google Presentation

Ethiopia (Speaker: Kunuz Abdela)

Google Presentation

Estimating the Burden of Injuries for the Western Cape, 2006
(Speaker: Megan Prinsloo)
Authors: Megan Prinsloo, Debbie Bradshaw, Pam Groenewald, Ria Laubscher, Nadine Nannan & Ian Neethling Burden of Disease Research Unit, Medical Research Council of South Africa e-mail: 

Cause of death statistics play an important role in the prioritization of health services. Despite improvements in vital registration, under-registration and mis-classification of causes remain a challenge. Burden of disease estimates were derived nationally and provincially for 2000 after careful examination of data on the levels and causes of mortality. The aim of this paper is to explore the feasibility of applying the provincial method used previously to develop contempory burden of disease estimates for 2006 for the Western Cape.
The demographic and AIDS model, ASSA 2003, was used for estimates of the total number of deaths and the number resulting from AIDS. StatsSA cause of death data for 2006 was used for the profile of other natural causes, after adjusting for mis- classification. The injury profile for the province was obtained from the City of Cape Town and two rural districts. Numbers of deaths, age-specific mortality rates and Years of Life Lost (YLL) were estimated. Patterns of cause specific mortality rates for the Western Cape were compared with selected subregions from the WHO GBD mortality estimates for 2001.
Results indicate that HIV/AIDS is the leading cause of death whilst ischaemic heart disease, diabetes, homicide and stroke were among the top 5 leading causes of death. The YLLs from these preliminary estimates reiterate the need to focus interventions on HIV and TB, as well as the prevention of homicide and road traffic fatalities. Chronic diseases of lifestyle contribute significantly to the burden of disease in the Western Cape, indicating that the province is well into the health transition and should address the risk factors that cause them.
Comparisons with alternative data sources appear to confirm the plausibility of these estimates of the injury burden. Particularly high homicide rates among young males were even higher than the average experienced in Central/Latin America. Additional demographic and epidemiological analysis and modelling are required to enhance these estimates.
Kavi Bhalla,
Oct 26, 2009, 11:38 AM