Population representativeness and interviewer-level heterogeneity in the Gaza Mortality Survey

In their Article published in The Lancet Global Health, Michael Spagat and colleagues describe the Gaza Mortality Survey as a population-representative, stratified two-stage cluster survey.1 The representativeness claim is central to the reliability of the resulting mortality estimates, which extrapolate 393 reported violent deaths in 2000 sampled households to 75 200 deaths at the population level. However, our independent re-analysis of the publicly released survey microdata and supplementary materials identifies multiple departures from the declared protocol and substantial interviewer-level outliers that call the realised sample’s representativeness into question and, consequently, undermine the validity of the population-level extrapolation.