Confidence Intervals and Sample-size Calculations for the Sisterhood Method of Estimating Maternal Mortality

Abstract

The sisterhood method is an indirect method of estimating maternal mortality that has, in comparison with conventional direct methods, the dual advantages of ease of use in the field and smaller sample-size requirements. This report describes how to calculate a standard error to quantify the sampling variability for this method. This standard error can be used to construct confidence intervals and statistical tests and to plan the size of a sample survey that employs the sisterhood method. Statistical assumptions are discussed, particularly in relation to the effective sample size and to effects of extrabinomial variation. In a worked example of data from urban Pakistan, a maternal mortality ratio of 153 (95 percent confidence interval between 96 and 212) deaths per 100,000 live births is estimated.

High maternal mortality in rural south-west Ethiopia: estimate by using the sisterhood method

“…A more precise recommendation of the sample size estimation, which considers the margin of error, confidence level, power of the estimate, and the required number of maternal deaths of sisters, suggests a more detailed sample size determination [20]. The formula which calculates the number of maternal deaths required for reporting by respondents was determined as follows: r ≥ [Zα/2] 2 * [100÷% ME] 2 , where r is the number of sister deaths due to maternal causes that were required, Zα/2 is the standard normal deviate at a two-sided confidence level of 100[1-α], and the% ME is the percentage margin of error tolerated by the investigators.…”

“…From the formula we calculated [1.96] 2 * [100/10] 2 = 384 sister deaths due to pregnancy, childbirth, or 6 weeks after the pregnancy terminated. Hanely and colleagues [20] have suggested that with 80% statistical power for a community with a MMR > 750 per 100,000 live births, a report of ≥ 384 maternal deaths is expected from interviewing 8000 adult siblings. In 2000, the MMR estimate was 937 for Ethiopia [6].…”

“…We calculated the MMR as MMR =1-(P) 1/TFR , where P is the probability of surviving, which equals (1-Q), and TFR is the total fertility rate [20]. …”

Unmasking inequalities: Sub-national maternal and child mortality data from two urban slums in Lagos, Nigeria tells the story

“…It is nearly double the figure previously estimated for Lagos State of 545 per 100,000 live births [3] and is also considerably higher than the national figure of 814 [20]. The lifetime risk of maternal death (1 in 18) in these two communities is also more than double the estimate of 1 in 42 for Lagos State [3] and is also higher than the previous national estimate of 1 in 23 [21].…”

“…Since estimating the maternal mortality ratio was the primary objective of this study, by estimating a total fertility rate (TFR) of 5.4, this sample size was chosen to be large enough to detect a MMR of 500 with a margin of error of 20% and a confidence interval of 95% [20]. …”