Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A Bayesian hierarchical model for mortality data from cluster-sampling household surveys in humanitarian crises

A Bayesian hierarchical model for mortality data from cluster-sampling household surveys in... BackgroundThe crude death rate (CDR) is one of the defining indicators of humanitarian emergencies. When data from vital registration systems are not available, it is common practice to estimate the CDR from household surveys with cluster-sampling design. However, sample sizes are often too small to compare mortality estimates to emergency thresholds, at least in a frequentist framework. Several authors have proposed Bayesian methods for health surveys in humanitarian crises. Here, we develop an approach specifically for mortality data and cluster-sampling surveys.MethodsWe describe a Bayesian hierarchical Poisson–Gamma mixture model with generic (weakly informative) priors that could be used as default in absence of any specific prior knowledge, and compare Bayesian and frequentist CDR estimates using five different mortality datasets. We provide an interpretation of the Bayesian estimates in the context of an emergency threshold and demonstrate how to interpret parameters at the cluster level and ways in which informative priors can be introduced.ResultsWith the same set of weakly informative priors, Bayesian CDR estimates are equivalent to frequentist estimates, for all practical purposes. The probability that the CDR surpasses the emergency threshold can be derived directly from the posterior of the mean of the mixing distribution. All observation in the datasets contribute to the estimation of cluster-level estimates, through the hierarchical structure of the model.ConclusionsIn a context of sparse data, Bayesian mortality assessments have advantages over frequentist ones already when using only weakly informative priors. More informative priors offer a formal and transparent way of combining new data with existing data and expert knowledge and can help to improve decision-making in humanitarian crises by complementing frequentist estimates. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Epidemiology Oxford University Press

A Bayesian hierarchical model for mortality data from cluster-sampling household surveys in humanitarian crises

Download PDF
9 pages

Loading next page...
 
/lp/ou_press/a-bayesian-hierarchical-model-for-mortality-data-from-cluster-sampling-0gYdysI0lv
Publisher
Oxford University Press
Copyright
© The Author(s) 2018; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association
ISSN
0300-5771
eISSN
1464-3685
DOI
10.1093/ije/dyy088
Publisher site
See Article on Publisher Site

Abstract

BackgroundThe crude death rate (CDR) is one of the defining indicators of humanitarian emergencies. When data from vital registration systems are not available, it is common practice to estimate the CDR from household surveys with cluster-sampling design. However, sample sizes are often too small to compare mortality estimates to emergency thresholds, at least in a frequentist framework. Several authors have proposed Bayesian methods for health surveys in humanitarian crises. Here, we develop an approach specifically for mortality data and cluster-sampling surveys.MethodsWe describe a Bayesian hierarchical Poisson–Gamma mixture model with generic (weakly informative) priors that could be used as default in absence of any specific prior knowledge, and compare Bayesian and frequentist CDR estimates using five different mortality datasets. We provide an interpretation of the Bayesian estimates in the context of an emergency threshold and demonstrate how to interpret parameters at the cluster level and ways in which informative priors can be introduced.ResultsWith the same set of weakly informative priors, Bayesian CDR estimates are equivalent to frequentist estimates, for all practical purposes. The probability that the CDR surpasses the emergency threshold can be derived directly from the posterior of the mean of the mixing distribution. All observation in the datasets contribute to the estimation of cluster-level estimates, through the hierarchical structure of the model.ConclusionsIn a context of sparse data, Bayesian mortality assessments have advantages over frequentist ones already when using only weakly informative priors. More informative priors offer a formal and transparent way of combining new data with existing data and expert knowledge and can help to improve decision-making in humanitarian crises by complementing frequentist estimates.

Journal

International Journal of Epidemiology – Oxford University Press

Published: Aug 1, 2018

References

  • Improving standards in international humanitarian response: the sphere project and beyond
  • The Sphere Project: the implications of making humanitarian principles and codes work
  • Assessing the impact of humanitarian assistance in the health sector
  • Interpreting and using mortality data in humanitarian emergencies. A primer for non epidemiologists
  • Precision, time, and cost: a comparison of three sampling designs in an emergency setting
  • The complex emergency database: a global repository of small-scale surveys on nutrition, health and mortality. Welte A, editor
  • Monitoring mortality in forced migrants—can Bayesian methods help us to do better with the (little) data we have?
  • Bayes-LQAS: classifying the prevalence of global acute malnutrition
  • A weakly informative default prior distribution for logistic and other regression models
  • Bayesian hypothesis testing for psychologists: a tutorial on the Savage-Dickey method
  • Incorporating Bayesian Ideas into health-care evaluation
  • Space-time smoothing of complex survey data: small area estimation for child mortality
  • Excess mortality in refugees, internally displaced persons and resident populations in complex humanitarian emergencies (1998-2012)—insights from operational data