The effect of misclassification error on reported cause-specific mortality fractions from verbal autopsy.

BACKGROUND Verbal autopsy (VA) studies are important for measuring cause-specific mortality in areas where medical certification of cause of death is uncommon. This paper explores the effects of misclassification errors on the results of verbal autopsy studies, and recommends ways to take misclassification errors into account in the interpretation of results. METHODS Mathematical formulae are derived for determining the size and direction of the error in cause-specific mortality estimates based on VA studies caused by misclassification. The levels of sensitivity and specificity found in currently available validation studies for childhood VA are examined. RESULTS There can be substantial errors in the estimates of the cause-specific mortality fraction derived from VA studies. The cause-specific mortality fraction itself has an important influence on the size of the error for given levels of sensitivity and specificity, and when the cause-specific mortality fraction is small, the size of the error depends more on specificity than on sensitivity. CONCLUSION Despite its drawbacks VA seems to be the most promising way of establishing cause of death when most deaths take place at home without medical attention. However, more validation studies on standardized instruments are required in order to collect information about sensitivity and specificity and subsequently improve the design of the instrument. At the same time, analysts need to take misclassification errors into consideration in ways outlined in this paper.