With child mortality rapidly declining, an increasing proportion of all deaths in low and middle-income countries (LMICs) will occur at adolescent and adult ages in the next decades. The forthcoming Sustainable Development Goals (SDGs) set out by the United Nations reflect this shift: over the next 15 years (2015-2030), several SDG targets will focus on reducing deaths from causes that affect primarily those age groups (e.g., maternal mortality, road traffic accidents, suicides). Investments in adolescent and adult health from US federal institutions, international organizations and governments of LMICs are thus expected to increase significantly. The mortality impact of such investments may however remain unknown because few LMICs have vital registration systems that allow measuring mortality precisely. Despite current efforts to improve such systems, household-based surveys (e.g., the Demographic and Health Surveys, DHS) will remain the primary data source on adolescent an adult mortality in LMICs. These surveys entail asking a random sample of respondents to report their siblings' survival history (SSH), i.e., whether each of their siblings 1) is alive, 2) how old s/he is, and 3) if deceased, how old s/he was at the time of death and the time since death. SSH permit directly estimating mortality rates at ages 15 and older, and also include basic assessments of the causes of siblings' deaths. They are however still affected by pervasive reporting errors. For example, respondents may omit some deaths, whereas other deaths may be added, or displaced in time. The net effects of these reporting errors on mortality indicators are large and hard to predict, but existing analytical corrections are based on very limited validation studies and do not capture this complexity. They risk accentuating bias and/or misrepresenting uncertainty in mortality estimates. In this project, we propose to improve the accuracy of survey-based estimates of adolescent and adult mortality through a) innovative data collection techniques (e.g., event history calendars, recall cues) and b) integrated Bayesian methods that account for sampling and non-sampling errors. Results from this study will help develop and target adolescent and adult health interventions in low and middle-income countries, and evaluate the effectiveness of global health initiatives.
In low-income countries with limited vital registration systems, the trends and causes of adolescent and adult mortality are measured retrospectively during surveys, but these surveys often yield inaccurate data. We propose to improve the accuracy of survey-based estimates of adolescent and adult mortality through a) innovative data collection techniques (e.g., event history calendars, recall cues) and b) integrated Bayesian methods that account for sampling and non-sampling errors. Results from this study will help develop and target adolescent and adult health interventions in low-income countries, and evaluate the effectiveness of global health initiatives focused on preventing premature deaths in those age groups (e.g, PEPFAR).
