Data Analysis - Hospital Data

Please note that the pages on this website do not reflect the methods that were used to generate the final set of results for the GBD-2010 study, which was published in December 2012.

-- If you are interested in the methods or results of GBD-2010 and subsequent revisions, please visit the official website of IHME's GBD project (http://www.healthdata.org/gbd). The materials described on the website are now obsolete but these webpages have been retained as an internet archive of the work of the group.

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

Last Updated: March 27 2009

Hospital Data Analysis Team (and how to join): Kavi Bhalla is leading the work on developing countries; Maria Segui Gomez is leading the work on the European regions (East, West and Central); Jed Blore is leading the work on all other high income countries from which data is readily available (to start with US, Australia and Singapore). The work of this team is likely to be analysis-intensive and so the current collaborators are members who can volunteer both data sets as well as substantial amount of people hours for analysis. Please do consider joining the analysis sub team if you think you can make such a contribution yourself. Otherwise, you may want to be included as a "lurking member" in these discussions by contacting kavi_bhalla@harvard.edu – lurking members are copied on email conversations so that they can make suggestions to help shape the work.

If you can contribute hospital data for use in this project, please visit the call for data section of this website.

Current work:

    • Role of hospital data in the project
      • Please visit the analysis overview page to see the broad role of hospital data envisioned in this project.
      • There are two key analytical components for the current work with hospital data.
        • Build mappings of external causes (e.g. pedestrian crash, burns, poisonings, etc) to nature of injury sequelae (TBI, skull fracture, etc). These mappings will be applied to predict disability from injuries in regions where only external cause information is available.
        • Estimate incidence of non-fatal injuries. These will be used to develop injury pyramids (deaths: severe injuries: minor injuries) for estimating the distribution of non-fatal events in regions where only mortality data is available.
    • Data extraction scripts: We are currently sending data requests to many researchers at national agencies who may be able to provide hospital data. This requires data extraction scripts, i.e. scripts that translate unit record data into tabulations (age, sex, external cause, nature of injury) using the GBD groupings. GBD groupings can be found here and a worked example of data extraction using Mexican hospital data is described here.
    • May 2009 deadline: The central team has requested for first set of incidence estimates by May 2009. However it should be noted that it is likely that data collected for several months after this deadline will continue to be incorporated into the project.

Related Analytical issues:

    • Multiple injuries:
      • Issue: Generating the data extraction scripts requires choosing a method for handling multiple injuries. We need a solution by Sept 21st 2008.
      • What are we doing: As a first step, we are checking the fraction of injury cases that get assigned to multiple sequelae to assess the extent to which this will be an issue.
      • Relevant discussion topic: Topic 5: Dealing with multiple injuries, lead by Belinda Gabbe.
    • Multiple admissions for same injury:
      • Issue: Repeat hospital visits for the same injury may result in incorrect estimates of incidence
      • What are we doing: The current thought is to request data separately for ER visits that do not result in admissions, first hospital admissions, and hospital deaths
      • Relevant discussion topic: Topic 9 Recurrent Injury, lead by Caroline Finch.
    • Estimating incidence from hospital data: [needs to be resolved over the next few months]
      • Issue: Ultimately hospital data needs to be translated into incidence rates for non-fatal injuries. However, denominator population is rarely known for hospital admissions.
      • What are we doing: At present, we are planning on estimating incidence envelopes for non-fatal injuries from national health surveys. Hospital admission data would be used to estimate cause specific fractions for external causes.
      • Relevant discussion topic: None existing. Advice is being sought from Colin Cryer.
    • Dealing with unspecifieds:
      • Issue: What is the best strategy for reassigning unspecifieds
      • What are we doing: The default method is age-sex proportional redistribution unless a better method is provided
      • Relevant discussion topic: Topic 7: Dealing with unspecified and poorly specified categories in case data sets (lead by Kavi Bhalla and James Harrison)