Reliable direct measurement of causes of death in low- and middle-income countries

Abstract

Background: Most of the 48 million annual deaths in low- and middle-income countries (LMICs) occur without medical attention at the time of death so that the causes of death (COD) are largely unknown. A review of low-cost methods of obtaining nationally representative COD data is timely.

Discussion: Despite clear historic evidence of their usefulness, most LMICs lack reliable nationally representative COD data. Indirect methods to estimate COD for most countries are inadequate, mainly because they currently rely on an average ratio of 1 nationally representative COD to every 850 estimated deaths in order to measure the cause of 25 million deaths across 110 LMICs. Direct measurement of COD is far more reliable and relevant for country priorities. Five feasible methods to expand COD data are: sample registration systems (which form the basis for the ongoing Million Death Study in India; MDS); strengthening the INDEPTH network of 42 demographic surveillance sites; adding retrospective COD surveys to the demographic household and health surveys in 90 countries; post-census retrospective mortality surveys; and for smaller countries, systematic assembly of health records. Lessons learned from the MDS, especially on low-cost, high-quality methods of verbal autopsy, paired with emerging use of electronic data capture and other innovations, can make COD systems low-cost and relevant for a wide range of childhood and adult conditions.

Summary: Low-cost systems to obtain and report CODs are possible. If implemented widely, COD systems could identify disease control priorities, help detect emerging epidemics, enable evaluation of disease control programs, advance indirect methods, and improve the accountability for expenditures of disease control programs.

Figure 1

Quality of the cause of death statistics from low- and middle-income countries (LMICs) reported to WHO. Only five countries outside China, India and Latin America have national COD surveys: Afghanistan, Mozambique, Thailand, Vietnam and Zambia. Zambia's survey covers 4 of 10 provinces. The remaining countries are nationally representative. Source: Mathers et al. [3], updated based on data from Murray et al. [27].

Figure 2

Simplified schema of death and disability estimates in indirect methods. Modified from World Bank data [25].

Figure 3

Cause of death (COD) data in the Global Burden of Disease (GBD) study for selected countries. The GBD measured the major COD types in 110 low- and middle-income countries (LMICs) based on data available from national surveys with verbal autopsy (VA) in five countries; final results were based on a ratio of estimated to actual COD data that was less than 0.12% (or about 1 in 850). Most health measurements rely on disability estimates as opposed to mortality, and measurement error often exceeds the desired change in health outcomes following an intervention. For example, in seeking a 10% improvement in health outcomes in children under 5 years of age, it is not possible to accurately assess the outcome of an intervention if the measurement error exceeds 10%. By contrast, as death is a discernible, objective outcome, restricting analyses to mortality significantly reduces measurement error [1,28].

Figure 4

Low- and middle-income countries with Demographic and Health Surveys (DHS), INDEPTH network Demographic Surveillance Sites, or Sample Registration Systems (China and India). Source: INDEPTH [41] and DHS [43], reproduced with permission [1].