PLoS Medicine
(Accessed 18 April 2010)
http://medicine.plosjournals.org/perlserv/?request=browse&issn=1549-1676&method=pubdate&search_fulltext=1&order=online_date&row_start=1&limit=10&document_count=1533&ct=1&SESSID=aac96924d41874935d8e1c2a2501181c#results

Mortality Measurement Matters: Improving Data Collection and Estimation Methods for Child and Adult Mortality
Colin Mathers, Ties Boerma

Abstract
The accurate measurement and estimation of mortality levels, trends, causes, and differentials are a cornerstone of public health. Child and adult mortality rates, often summarized in a life expectancy measure, are key indicators of levels of health and development. The preferred source of mortality data is prospective measurement through continuous registration of deaths, as is done in civil registration systems. But in many countries, especially those with poorly developed statistical systems and higher levels of mortality, retrospective measurement in households and surveys is the principal vehicle for data collection. All methods of data collection suffer from two generic problems: omission of events and dating errors. During the past few decades, demographers have developed and used a range of methods to improve data collection, assess levels of bias, and correct for such biases [1]–[3]. In three papers published in this issue of PLoS Medicine [4]–[6], Murray, Rajaratnam and colleagues revisit these analytical methods and techniques and present improved methods for the analysis of mortality data collected through death registration, censuses, or household surveys.