The measurement and estimation of tuberculosis mortality.

Tuberculosis (TB) ranks among the 10 principal causes of death and disability worldwide, largely on the basis of mortality estimates. These estimates have been derived by a variety of methods, from a limited database. Here we review the data and methods used to measure and estimate TB mortality in adults, assess the strengths and weaknesses of each and suggest ways to improve current mortality statistics. In principle, deaths attributable to TB can be obtained directly from national vital registration (VR) systems. However, only 59 of 213 countries in 2005 (including three in the World Health Organization Africa Region and one in the South-East Asia Region) had VR systems that reported TB deaths, corresponding to just 10% of all estimated deaths attributable to TB. Until comprehensive, national VR systems are established, an interim solution is to carry out verbal autopsies within sample VR schemes. The number of TB deaths from VR should ultimately converge with deaths recorded in national TB control programmes. At present, deaths in treatment cohorts cover a small subset of all estimated TB deaths (<13% in 2006), as deaths are missed among patients who are never diagnosed, who default or fail treatment, and among patients with untreated recurrent TB or TB sequelae. In contrast, some deaths recorded during treatment are not due to TB. To ensure convergence between cohort monitoring and VR, definitions of causes of death--including TB as an associate cause in deaths from human immunodeficiency virus/acquired immune-deficiency syndrome--should be standardised, so that both systems adhere to the International Classification of Diseases.