Discussion Medical Care

Estimating the burden of injuries that do not receive medical care

Last Update: May 1 2011

Purpose of this webpage: The GBD injury expert group is developing a model to estimate the burden of non-fatal injuries. This requires estimating the incidence and burden of cases that do not receive medical care. The project is conducting a survey of clinical experts to estimate certain key parameters. This page provides background and reference material for the survey.

Who is Involved: Study design is being led by Lisa Knowlton, James Harrison, Kavi Bhalla and Theo Vos. Others involved include Richard Gosselin, Kelly McQueen, Charles Mock, Richard Coughlin, Tom Weisner and Jason Nickerson.

About GBD: The new Global Burden of Diseases, Injuries, and Risk Factors Study (the GBD 2010 Study), which commenced in the spring of 2007, is the first major effort since the original GBD 1990 Study to carry out a complete systematic assessment of the data on all diseases and injuries, and produce comprehensive and comparable estimates of the burden of diseases, injuries and risk factors for two time periods, 1990 and 2005. By Sept 2011 the project will produce a final set of estimates. The GBD 2010 Study brings together a community of experts and leaders in epidemiology and other areas of public health research from around the world to measure current levels and recent trends in all major diseases, injuries, and risk factors, and to produce new and comprehensive sets of estimates and easy-to-use tools for research and teaching. It is led by a consortium including Harvard University, the Institute for Health Metrics and Evaluation at the University of Washington, Johns Hopkins University, the University of Queensland, and the World Health Organization (WHO). More information about the GBD study is available at GBD website.

About the Global Burden of Injuries Model: There are two aspects to the models for assessing the global burden of injuries. The first is a mortality model that uses all available existing data sources (including death registration, verbal autopsy studies, mortuary data sets) to estimate the years of life lost to mortality from injuries in all global regions. The second is the morbidity model that uses existing data sources to estimate the incidence and burden of non-fatal injuries globally. The measurements from this survey will be used to fill key information gaps in empirical evidence being used in the morbidity model.

MORBIDITY MODEL: The morbidity model uses empirical evidence from a wide variety of data sources. These include incidence of injuries from various external causes (road injuries, falls, etc) derived from approx. 120 household surveys, distribution of nature of injury (TBI, fractures, etc) for these external causes derived from approx. 30 national hospital discharge databases, the long term outcomes (% deaths, and disabled) measured in the Victorian Trauma Register and Orthopaedic Outcomes Register (Australia), and estimates of probability of admission based on empirical work being conducted in New Zealand. Note that the two key data sources that are being used to model long-term outcomes of injuries in this project are from a resource-rich setting. Unfortunately, the project was not able to find many reliable data sources to estimate the outcomes of injuries that do not get hospitalized and treated in resource poor settings. This information gap in the available data is being filled by asking clinical experts to estimate certain key parameters for the project.

A brief sketch of the morbidity model is as follows:

• Step 1: Estimate regional incidence of hospitalized external causes (e.g. road injuries) from survey data. The project has collected data from household surveys from most global regions. Typically, these surveys ask respondents if they have been injured recently, the external cause of the injury, and the type of medical care received for the injuries. This incidence data is being used to construct regional estimates of the incidence of hospitalized injuries (by external cause).
• Step 2: Estimate regional incidence of hospitalized sequelae (TBI, fractures, etc) of each external cause (H0). This mapping from external cause to sequelae incidence is done by using hospital discharge data that contains information about both external causes and sequelae.
• Step 3: Estimate regional incidence of sequelae (TBI, fractures, etc) that are not hospitalized because of limited access to care (H1). Note that Step 1 and 2 allow estimating the incidence of hospitalized sequelae only. In resource-poor settings, we expect that there are incident cases that would have been hospitalized in a resource-rich setting but are not hospitalized because of resource constraints. We are attempting to estimate this parameter using survey data but the quality of available information is poor. Thus, we are supplementing this information by eliciting expert opinion about the fraction of all incident cases that are not hospitalized in the most resource poor regions.
• Step 4: Estimate regional incidence of sequelae (TBI, fractures, etc) that are not hospitalized because they do not need to be hospitalized (H2). A probability-of-admission model for each sequelae is being developed using data from New Zealand. In addition, we are supplementing this information by eliciting expert opinion about the fraction of all incident cases that are not hospitalized.
• Step 5: Estimate short-term disability durations and fraction of cases with long-term (persisting) disability for H0, H1 and H2: Empirical data for sequelae that are hospitalized (H0) is available from the Victorian Trauma Register and Orthopaedic Outcomes Register (Australia). The information on durations and the fraction of sequelae with long term disability for H1 is being estimated by eliciting expert opinion.
• Step 6: Estimate DALYs (i.e. burden) by applying empirically measured disability weights for H0, H1 and H2. Note that the current revision of the GBD project is doing field studies to measure disability weights. A brief description of this aspect of the project is available in this publication. (Salomon et al. 2010)

Note that while there are several limitations of this modeling strategy, this method is a substantial advance on previous GBD methods. A key consideration in the development of this modeling strategy was to make optimal use of the imperfect data and theory available to construct GBD estimates. While more sophisticated YLD models for GBD-Injury are possible, we do not believe they are warranted given the quality and quantity of global data available. Thus this model also highlights the enormous gaps in knowledge and provides a basis for framing priorities for future work.

Figure 1 (a) Estimating incidence of sequelae Figure 1 (b) Estimating burden from incidence of sequelae

Additional Resources:

Table 1: Classification of regions by access to medical care

(Note that this classification of regions was done for the purpose of the survey of expert opinion on access to medical care being conducted by the GBD-Injury Expert Group)