Article Text

Download PDFPDF

Burden Calculator: a simple and open analytical tool for estimating the population burden of injuries
  1. Kavi Bhalla1,
  2. James E Harrison2
  1. 1Johns Hopkins International Injury Research Unit, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  2. 2Research Centre for Injury Studies, Flinders University, Adelaide, South Australia, Australia
  1. Correspondence to Dr Kavi Bhalla, Johns Hopkins International Injury Research Unit, Johns Hopkins Bloomberg School of Public Health, 615 N Wolfe Street E8138, Baltimore, MD 21205, USA; kavibhalla{at}


Background Burden of disease and injury methods can be used to summarise and compare the effects of conditions in terms of disability-adjusted life years (DALYs). Burden estimation methods are not inherently complex. However, as commonly implemented, the methods include complex modelling and estimation.

Objectives To provide a simple and open-source software tool that allows estimation of incidence-DALYs due to injury, given data on incidence of deaths and non-fatal injuries. The tool includes a default set of estimation parameters, which can be replaced by users.

Development of the software tool The tool was written in Microsoft Excel. All calculations and values can be seen and altered by users. The parameter sets currently used in the tool are based on published sources.

Using the software tool The tool is available without charge online at To use the tool with the supplied parameter sets, users need to only paste a table of population and injury case data organised by age, sex and external cause of injury into a specified location in the tool. Estimated DALYs can be read or copied from tables and figures in another part of the tool.

Conclusions In some contexts, a simple and user-modifiable burden calculator may be preferable to undertaking a more complex study to estimate the burden of disease. The tool and the parameter sets required for its use can be improved by user innovation, by studies comparing DALYs estimates calculated in this way and in other ways, and by shared experience of its use.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The 2010 Global Burden of Disease (GBD-2010) project is a major technological advance for measuring the population burden of injuries.1 GBD-2013 is bringing further developments and the rolling programme of releases foreshadowed by the Institute of Health Metrics and Evaluation can be expected to implement further changes. GBD-2010 and its subsequent revisions include a wide range of new types of injury data sources and methods for processing these data. Broader developments, such as new disability weights based on rating short descriptions of health states2 and the development of new analytical tools have also improved global estimates of the burden of injuries. Some of the tools, such as Cause of Death Ensemble model (CODEm)3 and the latest version of the generic disease modelling system, DISMOD-MR,4 are computationally intensive and require a supercomputing cluster, which makes them unavailable to most researchers and complicates replication.

While these new tools serve an important purpose in global disease modelling, they may not be necessary for researchers interested in developing burden of injury estimates in specific populations (eg, a country or province). For instance, while CODEm is useful for estimating mortality from poor-quality or sparse cause-of-death data,3 researchers focusing on a particular population often have estimates of mortality from injuries available from reliable sources. Similarly, while DISMOD-MR is particularly useful in generating consistent estimates from data on incidence, prevalence and duration,4 most researchers seeking to generate burden of injury estimates only have access to case data on the incidence of injuries and should not require DISMOD.

We developed a spreadsheet-based analytical tool that can be used to estimate the burden of injuries in a population using data on the incidence of deaths and non-fatal injuries, and estimates of mean case duration and GBD disability weights. The tool simplifies the GBD-2010 methodology while retaining the features most important for evidence-based estimation of the burden of injuries. The tool is freely available on the project website:


This project aims to develop a simple and open-source software tool to allow estimation of incidence-DALYs due to injury from data on the incidence of deaths and non-fatal injuries. By default, the tool provides a full set of parameter values, which can be replaced by users.

Development of the software tool

Our model, Burden Calculator, is conceptually similar to the 1996 Australian Burden of Disease tool and the 2001 WHO manual ‘National Burden of Disease Studies: a Practical Guide’5 but incorporates several new developments from later GBD projects. The analysis is implemented using Microsoft Excel.

The project website provides detailed documentation of methods, and the sources of parameter values included by default in the tool. These are summarised below.

Estimating premature mortality

Burden Calculator estimates years of life lost (YLLs) using the equation Embedded Image

where D is number of deaths and L is the standard life expectancy for people surviving to that age. The user is expected to supply the incidence of deaths (disaggregated by age, sex and external cause). The standard reference table for life expectancy provided by default in the tool is the global standard life table developed by GBD-2010.6 However, users can modify these to other values, such as those recommended by WHO,7 which are provided in the documentation accompanying Burden Calculator.

Estimating health loss from non-fatal injuries

In some settings, case data are available that include information on both injuries (traumatic brain injury, hip fracture, etc) and their external causes (RTC, fall, etc). Well-coded data of this type are most often available for admitted patients, though sometimes are also available for patients who attended emergency departments. In most settings globally, however, data sources that can provide information on the incidence of injuries due to at least some external causes are more common than sources that provide information characterising the injury conditions sustained. However, estimates of the incidence of injury conditions are needed in order to estimate the health loss attributable to injury (‘disability’ in GBD terminology). In GBD-2010, we responded to this by developing mappings from external causes to injury ‘sequelae’. These mappings are estimates of the probability that a health event (ie, a hospital-admitted incident injury case) due to a particular external cause (such as a fall) will result in a particular injury (such as a hip fracture): Embedded Imagewhere N is the incidence of sequelae, E is the incidence of external cause of injuries and p is the probability of the occurrence of each sequela.

We developed such mappings based on large hospital administrative and surveillance databases from 28 countries from South-East Asia; Central, Eastern and Western Europe; Central, Southern and Tropical Latin America; North Africa/Middle East; North America High Income and East Sub-Saharan Africa. The databases included dual coding of discharges by external cause and sequelae based on ICD9 or ICD10 groupings. In most instances, hospital records included multiple injury diagnoses for each patient. Therefore, a heuristic was used to identify the most disabling injury, and each patient was assigned one GBD-2010 external cause and sequelae.8 In the resulting database, a relatively high proportion of sequelae are coded to ‘Other & Unspecified’, for which disability weights and duration estimates are not available. We reattributed these cases to the specified sequelae pro rata within age, sex and external cause groups. Although we aimed to produce separate mappings for each age–sex group, this was not always possible due to small case counts in the source data. When there were fewer than 1000 cases available for a particular external cause, we combined age–sex groups. Therefore, the external cause to sequelae mappings are available separately for the 36 age–sex categories for some external causes (such as RTCs and falls) but have not been disaggregated by sex, or use lumped age groups (<5, 5–14, 15–74 and 75+ years), or use a single age–sex category depending on the external cause.

The illustrative example shown by default in Burden Calculator uses external cause to sequelae mappings for RTCs but the user can replace these with the mappings appropriate for other external causes, as needed. The mappings for other sequelae are provided as a separate Excel file on the project website.

The model estimates years lived with disability (YLD) asEmbedded Imagewhere N is the number of incident cases of each sequelae, pperm is the proportion of cases that will have permanent disability; DW is the disability weight that reflects the severity of the health decrement on a scale from 0 (perfect health) to 1 (dead); DWst is the short-term disability weight; DWperm is the permanent disability weight; Dst is the short-term duration of the disabling event; and Lp is the period life expectancy of the population, reduced (if necessary data are available) to allow for elevation of all-causes mortality among people living with permanent sequelae.

The burden estimates obtained per incident case of injury can be expected to be similar to those obtained by other burden of injury studies that incorporate the same model and assumptions.

GBD-2010 estimates of duration of injuries have not been published. Therefore, the default values for these parameters (percentage of incident cases that develop persisting disability, and the duration of disability of cases that have short-term disability) provided in Burden Calculator are based on estimates of these parameters that were used by GBD prior to the 2010 revision. The user can replace these parameter values with more appropriate data from GBD as they become available, or from another suitable source.

Disability weights used in our model are based on the values used in GBD-2013.9 Alternative sets of disability weights, such as from previous GBD studies, have been included in the project documentation for researchers interested in studying the sensitivity of burden estimates to disability weights.

Estimating net health loss: DALYs

Finally, our model estimates overall health loss in terms of disability-adjusted life years (DALYs) lost as the sum of weighted time lived with disability and time lost to premature mortality:Embedded Image

Note that these burden estimates are neither discounted nor age-weighted because such adjustments were discontinued by GBD-2010 after a consideration of their ethical implications and other arguments.6 Our model estimates incidence-DALYs, that is, it estimates the stream of future health loss connected with events that occur in a particular period, which is the approach that has been previously used by GBD.10 It should be noted, however, that GBD-2010 and subsequent revisions use a hybrid approach that combines incidence-YLLs and prevalence-YLDs. For injuries, incidence-DALYs may be preferable for the following reasons. First, from the perspective of injury prevention policy, it is often important to ascribe the entire burden to the point in time at which the injury occurred so that the net benefits of prevention can be ascertained. Second, unlike the situation that exists for many disease conditions, much of the data available on injuries and injury outcomes are incidence-based, making incidence-DALYs easier to compute and understand.

Using the software tool

The tool (Excel file) and its associated documentation can be downloaded from the project website: The tool is organised in a series of tabs, including one tab on which users can input data on the incidence of injuries, several tabs that include parameter values and steps in burden of disease analysis and several others that tabulate and illustrate the estimated burden of injuries.

Basic use

Researchers have often found it valuable to construct estimates of the burden of injuries in a particular population of interest.11 ,12 Burden Calculator aims to make this task easier by providing the infrastructure to do these calculations without needing to start from scratch. In order to estimate the burden of injuries using Burden Calculator, the user needs simply to supply population counts, disaggregated by age and sex, and the incidence of fatal and non-fatal injuries, disaggregated by age, sex and external cause. In most cases, this will require the user to preprocess their data prior to inclusion in the INPUTs tab of the model. This includes processing data so that it is suitably aligned with the age, sex and external cause groups used in our model. Notably, it also includes ensuring that the data on incidence of injuries has been reasonably adjusted for biases and completeness so that it is a reliable estimate of the true incidence of injuries. For instance, it is common for administrative datasets to classify a substantial proportion of cases to unspecified causes. If only cases with specified causes are used as inputs, the model will underestimate the true health burden of injury. Therefore, researchers should consider reattributing unspecified cases using simple algebraic tools (such as pro rata adjustments13) or more sophisticated statistical tools for handling missing information prior to input into Burden Calculator. We encourage users to share their experience with the Burden Calculator and their research findings based on its use in publications and on the project website.

Modifying and enhancing the tool

Sections of Burden Calculator that the user will not typically modify are protected to prevent accidental modification but instructions are provided on how to remove all protections, giving the user full control over the analysis. There are many reasons for advanced users to edit parameter values and/or make structural changes to the analytical model. For instance, users might have access to parameter values that may be more accurate than those currently used, or more appropriate for a particular application. Such data are likely to arise from ongoing research programmes that are improving estimates of the disability weights for injuries, duration of disability, and injury severity,14–16 among other analysis parameters. More commonly, researchers may have access to local data on non-fatal injuries that is dual coded and allows generating external cause to sequelae mappings that are specific to the population of interest.

Researchers may also be interested in developing the analytical tools further by adding modules to Burden Calculator. For instance, with some programming in Excel Visual Basic for Applications (VBA), it should be possible to conduct simulations that allow studying the sensitivity of burden estimates to alteration of parameters, and characterising the uncertainty in burden estimates. Users may also want to make more elaborate extensions that allow use of data that estimate injuries that received healthcare but were not admitted to a hospital, and to allow direct use of data organised by injury conditions (as distinct from external causes). Finally, users might also transfer the calculator to one of the open-sourced spreadsheet applications that are available. We invite researchers who modify and enhance Burden Calculator to share these developments on the project website for the benefit of other users.


GBD studies have been profoundly influential in shaping the priorities of global health and development actors. Partly driven by the success of the global studies, there has been growing demand for conducting studies of burden at the national and subnational level. Unfortunately, recent efforts at improving global estimates of the burden of disease have hampered the ability of local researchers to construct estimates for particular jurisdictions of interest to them. GBD-2010 introduced new analytical tools, such as CODEm and DISMOD-MR, designed for constructing estimates from data aggregated from multiple sources of varying reliability, which is important in a global context. However, these tools are computationally intensive and inaccessible to the broader research community.

We have developed a simple and open-source tool for assessing the population burden of injuries, in terms of DALYs, from estimates of the incidence of fatal and non-fatal injuries. We developed Burden Calculator in Microsoft Excel because it is widely available and familiar to many people who are not adept at programming. Such a tool has many potential applications. It could allow researchers conducting studies of descriptive epidemiology of injuries to easily report their findings in DALYs, and thus be able to compare their results with other diseases. The tool can also be used easily to study the relative importance of various estimation parameters and thus provide guidance on how to prioritise work to improve these for better measurement. Finally, Burden Calculator can also be used as a teaching tool that can help develop the intuition of researchers about the functional relationships between inputs, analysis parameters and burden estimates.

We envisage that the tool will evolve through contributions and collaborations of the injury statistics community. We encourage users to conduct validation studies, perform cross-country comparisons, study the sensitivity of the models to the inputs parameters and to contribute their results and modifications to the website. We expect that as the tool improves, it will enhance the ability of researchers worldwide to use local injury surveillance data to generate metrics that inform safety policy.

What is already known on the subject

  • Burden of disease and injuries studies provide valuable information for shaping population health and safety priorities.

  • New analytical tools introduced starting in Global Burden of Disease-2010 have hampered the ability of local researchers to construct burden estimates for their local jurisdictions.

What this study adds

  • We have developed a simple and open-source tool, Burden Calculator, for estimating the burden of injuries from estimates of the incidence of fatal and non-fatal injuries. The tool and the associated website are designed to enable contributions and collaborations of the injury statistics community.


We are grateful to Dr Colin Mathers and Dr Lynelle Moon for their thoughts on an early version of this paper. Burden Calculator builds on the work of many members of the GBD-2010 Injury Expert Group, which the authors of this paper jointly led, and the broader burden of disease research community.



  • Contributors Both authors worked together to develop the first draft and subsequent revisions of the manuscript. KB had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

  • Funding The development of Burden Calculator is partially supported by a grant from the World Bank to Johns Hopkins University.

  • Competing interests None declared.

  • Provenance and peer review Commissioned; externally peer reviewed.

  • Data sharing statement The software tool developed in this project is an open-source tool that is available online at