An ordinal measure of population health

We propose a population-level health index that addresses two of the main concerns of existing health measures: (1) the biases in self-assessed health measures, and (2) the difficulties associated with making comparisons across populations afflicted by a variety of conditions in a context in which multi-morbidity is high. Starting from a set of general axioms, we derive a partial order index that ranks health across populations based on the relative prevalence of various conditions within these populations, and that can be readily applied to many existing health surveys. We illustrate the use of our health measure by applying it to data from the National Health Interview Survey in order to examine health differences across racial groups in the U.S.