Accuracy of indirect estimates of maternal mortality: a simulation model.

A simulation model was developed to test the accuracy of indirect estimates of maternal mortality (the sisterhood method). The model generated a first generation of grandmothers, a second generation of mothers (with brothers and sisters), and a third generation of children (births). In the second generation, maternal mortality was introduced. Empirical values for the parameters of fertility and mortality were taken from a prospective survey in Senegal (Niakhar). Results based on 100 simulations of the same situation revealed several limitations of the sisterhood method: The indirect estimates could fall as far as 33 percent from the true values on individual cases; the indirect estimates tended to be systematically higher than the direct estimates; their range was wider, as were their confidence intervals; and biases were particularly strong for the younger age groups of respondents. Reasons for these biases are explored.