Call for data

Please note that the GBD-2010 Injury expert group no longer exists.

- Please visit the website www.globalburdenofinjuries.org to find out more about other closely related collaborations of our group members.

- If you are interested in the ongoing work of IHME's GBD project, please visit their official website http://www.healthdata.org/gbd.

Introduction

Please see our recent call for contribution in the January 2009 issue of PLoS Medicine. We are requesting individuals and organizations that can provide us with data for estimating the Global Burden of Injuries to please contact Kavi Bhalla (kavi_bhalla@harvard.edu) and/or James Harrison (james.harrison@flinders.edu.au). We can provide scripts for extracting and tabulating data from most databases (see example below). We will work with you to provide the most efficient way of translating your data sets into inputs for this project.

What are relevant data sources and variables?

1. Data sources for estimating deaths: High quality cause of death registration data is the gold standard. In its absence, we can estimate from incomplete or sub-national death registration, mortuary records, funeral/cemetery records, and police reports.
2. Data sources for estimating burden of non-fatal events:
   1. Health surveys: Population representative health surveys that include questions on injuries are important sources for estimating incidence of non-fatal injuries. Surveys which disaggregate by type of care are particularly useful.
   2. Hospital registries: Hospital records provide ICD coded injury diagnosis, which is needed for estimating public health burden. We are particularly interested in registries that include both external causes and injury diagnosis. Registers of admitted patients and registers of patients attending emergency departments are both valuable.
3. Variables of interest: This project is primarily interested in the injured person’s age, sex, external cause of injury, and injury diagnosis. We are also interested in variables that enable best use of the data to estimate population incidence of injury (e.g. variables that allow control of double counting of cases).
4. Time period of interest: GBD 2005 will generate estimates for the years 1990 and 2005. However when data from these years is not available, the project will analyze information from the closest time period for which data is available.
1. Characteristics of good data sources:
   1. For estimating incidence: High coverage/completeness or population representativeness; few cases assigned to unspecified categories.
   2. For estimating burden of non-fatal injuries: medically certified diagnosis (e.g. ICD coded hospital records); inclusion of both external causes and injury outcomes; follow up studies to estimate long term disability consequences.

How can you contribute?

You can contribute by helping us get access to data sources in an appropriate format. Data can be shared with our project in any of the following ways:

1. Provide us with de-identified case level records. Availability of unit level records (microdata) allows us to use advanced analytical tools to correct for data quality (e.g. reassignment of ill-defined cause categories). However we are aware that privacy concerns can often make sharing of case level records difficult.
2. Run our data extraction scripts and send us tabulations: We can provide scripts for most commonly used statistical packages that would produce tabulations of appropriately grouped of the relevant variables. These scripts involve tabulating data into GBD age-sex groups, GBD external cause groups, and GBD nature of injury groups. See examples below.
3. Provide us with any reports or publications that tabulate the relevant variables and we will use algebraic mappings to the GBD definitions.

We request permission to share data sets with other researchers but will respect all data access restrictions requested by the owners of the information.

What data sources does the project already have access to?

The table below lists all data sources currently being used by the project. If a country or data source is not listed below, please assume that we do not currently have access to it.

Note about year: Please note that the GBD project is making estimates for the years 1990 and 2005. The year listed in the table above refers to the most recent year for which data is available to the project. For many countries, we also have data for 1990 already available via the WHO mortality database. These countries are listed in the mortality data analysis section of our website.

Examples of data contributions

The following examples take case level data as input and produce tabulations grouped according to GBD age-sex, external cause and nature of injury categories.

Cautionary note: It is important that the data translation process be as similar as possible to the examples described below. Whenever this is not possible, it is essential that detailed documentation of the methods used be provided along with the data set.

Example 1: Death Registration Data

Death registration data needs to be translated into tabulations of deaths within GBD groupings for age, sex, and external cause. These GBD groupings (and ICD 9 & 10 code mappings) are described on the mappings page. These mappings were used to extract data from the WHO mortality database as shown here.

Example 2: Hospital Registration Data

The following describes the process by which hospital data is transformed into non-fatal nature of injury/external cause-linked tabulations for injuries by age and sex, according to Global Burden of Disease Project external cause and nature of injury sequelae groupings. Mexican hospital data is used as an example.

Process steps:

(1) Identify variables of interest: age, sex, external cause, first three diagnosis codes

(2) Apply inclusion criteria:

(a) Keep only non-fatal cases

(b) Keep only cases presenting at first point of contact with the healthcare system

(c) Keep only cases with S&T codes (ICD10 - nature of injury) within first three diagnoses

(3) Map data (age and ICD codes) to GBD groupings

(a) Recode age to GBD age groupings

(b) use ICD =>GBD mappings for external cause groups

(c) use ICD =>GBD mappings for nature of injury groups

(4) Generate separate tabulations for admitted care (inpatient) and ambulatory care (outpatient) for external cause by nature of injury, age, and sex.

Click here for worked example using Mexican hospital data

Example 3: Survey Data

Surveys vary substantially in their coverage and the depth of detail in which injuries are characterized. The excel file (Survey data request.xls) provides a templates for reporting data to the GBD project with the understanding that the templates be filled at the maximum level of detail possible.

If the instructions listed above are confusing, please do not hesitate to email kavi_bhalla@harvard.edu. We would be very happy to work with you closely to generate these results.

Last Updated: March 20th 2009