New US$100,000 Prize
The Roux Prize rewards bold action to improve population health through disease burden evidence. Nominations close March 31, 2014.
What can we conclude from death registration? A new method for evaluating completeness
Published in PLoS Medicine, April 2010
Novel techniques can make better use of incomplete vital registration systems for population health studies, according to new research. The study, What can we conclude from death registration? Improved methods for evaluating completeness, describes an approach to check the completeness and accuracy of databases that compile information from death certificates. The work was done in collaboration with scientists at the University of Queensland and Harvard University.
The study findings describe three death distribution methods (DDMs) that provide the best mortality estimates. However, even these improved methods yielded uncertainty intervals of roughly ±one-quarter of the estimate.
Researchers systematically evaluated the performance of 234 DDMs in three different validation environments. These methods use the relative proportion of deaths in each age group for recorded deaths and the age distribution of the general population to estimate the number of deaths that have not been officially counted. The World Health Organization uses DDMs to monitor adult mortality in nearly 100 countries. These methods, however, have limitations, including application of methods with little scientific literature to guide their selection, a lack of extensive validation in real population conditions, and the fact that DDMs do not generate uncertainty intervals. Prior to this study, researchers had not previously compared the performance of the various types of DDMs. From this analysis, the researchers identified the three DDMs that yielded the best estimates and demonstrated the application of the optimal variants in eight countries.
Policymakers rely on accurate information about mortality patterns to set public health priorities. Most high-income countries generate this information through death registration systems that capture nearly every death in the country. However, in low-resource settings, including most African countries, less than 25% of deaths are officially recorded. IHME is working to improve the analytical methods that can be applied to incomplete death registration data and produce more accurate estimates of death rates.
Recommendations for future work
Citation: Murray CJL, Rajaratnam JK, Marcus J, Laakso T, Lopez AD. What can we conclude from death registration? Improved methods for evaluating completeness. PLoS Medicine. 2010 April 13; 7(4):e1000262.
Data and Methods
One of the fundamental building blocks for determining the burden of disease in populations is to reliably measure the level and pattern of mortality by age and sex. Where well-functioning registration systems exist, this task is relatively straightforward. Results from many civil registration systems, however, remain uncertain because of a lack of confidence in the completeness of death registration.
Incomplete registration systems mean not all deaths are counted, and resulting estimates of death rates for the population are then underestimated. Death distribution methods (DDMs) are a suite of demographic methods that attempt to estimate the fraction of deaths that are registered and counted by the civil registration system. IHME has completed a systematic analysis of 234 variants of the DDMs.
Here, we provide the STATA code to apply the Generalized Growth Balance method, the Synthetic Extinct Generations method, and the Adjusted Synthetic Extinct Generations method. The code automatically calculates 78 variants of each method by varying the age groups included in the analysis. It also calculates diagnostics for each variant.
The code can produce the diagnostics and the completeness estimates for the 234 variants of the DDMs in a short amount of time. For instance, it takes about 3 minutes on a laptop to produce results for 10,000 populations.
Methods and data for download. Death distribution methods code and simulation sample data (9.8MB zip)