Comparability in Population Based Surveys: Concepts, Methods and Analysis

Emre Ozaltin, Harvard Initiative for Global Health
Christopher J.L. Murray, Harvard University
Ajay Tandon, Harvard University

This paper examines the problem of cross-population comparability in national health surveys, introduces a framework for conceptualizing the issue, and outlines a strategy for correction. Reported findings from the literature are used to illustrate that self-reported health measures may give misleading results in the absence of adjustments. The problem is conceptualized as shifts in response category boundaries across populations when responding to categorical self-report questions. For instruments to have a common metric in different populations where the same response corresponds to the same level of health, we suggest the use of new survey instruments that incorporate anchoring vignettes and a new statistical model for data analysis, the categorical hierarchical ordered probit (CHOPIT) model. We use data from 71 countries in the 2002 World Health Survey (WHS) to examine the success of implementation of vignettes in these countries and to compare CHOPIT adjusted results with unadjusted results.

  See extended abstract

Presented in Session 111: Innovative Techniques in Data Collection and Analysis II