Knowing the patterns and trends in causes of death by age and sex in a population is critical to understanding how to target interventions to maximize population health. Approximately 44% of the world’s population has their births and deaths recorded in a vital registration system, leaving a huge portion of the population for which cause of death is unknown.
Many of the populations we know least about reside in resource-poor settings where known interventions could help prevent certain causes of death. In addition, causes of death are frequently inaccurately or inconsistently recorded. Because enormous investments of time, infrastructure, and money would be required to provide complete vital registration worldwide, reliable and timely alternative methods for estimating causes of death must be developed.
The Causes of Death research team is creating new analytic tools that produce consistent trends across versions of the International Classification of Diseases (ICD), redistribute deaths mistakenly assigned to nonfatal diseases or symptoms that are not the actual underlying cause of death, and help fill in the gaps where data are missing. In many instances, for a given population in which there is a lack of data, estimates for causes of death may be made through comparison to similar populations. Pooling all available data on causes of death – including vital registration, household surveys, verbal autopsies, and hospital records – has enhanced the research team’s ability to create high-performing methods.
The research team aims to provide others with a suite of methods and tools to utilize the data available for a given population, no matter how incomplete, to produce a fuller understanding of the causes of death at the population level and give policymakers the information they need to ultimately improve population health.
Key Activities
- Produce cause of death estimates by country, age, and sex for 39 clusters of causes
The research team is producing country-level estimates of 39 clusters of causes of death by age and sex from 1980 to 2010, drawing from vital registration data, disease registries, verbal autopsies, and centrally-collected government data. This effort starts with a massive amount of data from myriad sources. The database now contains more than 4,200 country-years of data on causes of death for 154 countries from 1901 to 2009, encompassing 720 million deaths in ICD versions 1 to 10, as well as country-specific cause lists.
Verbal autopsy data, culled from literature reviews and studies, have also been added. In order to analyze the data, a comprehensive translation map has been created to link analogous codes among the various versions of the ICD and the country-specific causes lists. In addition, significant methodological development has gone into reducing the number of misclassified deaths, redistributing those deaths which have been attributed to nonfatal outcomes, in particular, to likely causes based upon symptomology and analysis of population trends.
There is a trade-off between accurately providing estimates for the largest number of causes of death and providing estimates for the most useful population subgroup. In general, the more causes one wants to estimate, the more researchers must rely upon aggregations of data from across multiple countries, because of the significant gaps in cause of death data for many countries.
Decision-making about health interventions that might prevent particular causes of death is most commonly made at the national or subnational level, rather than the global or regional one. Therefore, from a public policy perspective, estimates at the country level are the most valuable. By grouping causes of death into related clusters, the research team can provide accurate country-level results. Through analysis of the pooled data and consultation with medical doctors about the relatedness of different cases, the research team arrived at 39 clusters of causes for which it is producing estimates for each country by age and sex.
Tuberculosis | Esophageal cancer | Cirrhosis of the liver |
HIV/AIDS | Stomach cancer | Other digestive diseases |
Sexually transmitted diseases (excluding HIV) | Other malignant and benign neoplasms | Genitourinary, musculoskeletal, and skin diseases |
Intestinal infectious diseases | Lung/pulmonary cancers | Congenital anomalies |
Selected vaccine-preventable childhood diseases | Breast cancer | Transport injures |
Malaria | Cervical and uterine cancers | Other unintentional injuries |
Parasitic and vector diseases | Diabetes mellitus | Falls |
Meningitis and encephalitis | Endocrine, nutritional, blood, and immune disorders | Accidental drowning and submersion |
Other infectious diseases | Mental/behavioral/neurological/sense organ diseases | Smoke/fire/flames/heat |
Respiratory infections | Other circulatory diseases | Accidental poisoning, noxious substances |
Maternal conditions | Ischemic heart disease | Self-inflicted injuries |
Perinatal and infant causes | Cerebrovascular disease | Interpersonal violence |
Nutritional deficiencies | Respiratory diseases | War and civil conflict |
The research team continues to hone the methods and is packaging them into software for use by other researchers, most notably CoDMod III for cause of death modeling. The researchers will expand the number of clusters as necessary to provide information that is as focused as possible to decision-makers.
- Produce cause of death estimates for the Global Burden of Diseases, Injuries, and Risk Factors Study 2010
The Global Burden of Diseases, Injuries, and Risk Factors Study 2010 will include estimates of causes of death by age and sex. Results will be generated for both 1990 and 2010, aggregated into 21 distinct regions that comprehensively cover the globe. Approximately 200 causes of death will be estimated by adapting the methods used for the country-specific estimates and drawing from the same database. The research team is adapting these methods and producing the estimates as part of the overall GBD Study 2010.
- Refine hospital methods using data from multiple settings
IHME has developed a method to use hospital data to calculate population-level causes of death. Because data on hospital discharges are collected in facilities in every country, this method is one of the only ways to produce estimates of the distribution of causes of death in countries where there are no vital registration data available. This method is currently being tested and improved by applying it to data from multiple countries, including Australia, Iran, the Philippines, Papua New Guinea, and Bangladesh.
Impact
The methods developed as part of the Causes of Death research team’s work will assist researchers with analytic approaches to draw value from whatever data they have available. The estimates produced using these methods will provide policymakers, in particular, with the most systematically-generated set of numbers available and will encompass the largest known collected set of cause of death data.
