Systematic Error in Measures of Induced Abortion Related Post-Abortion Care in Zanzibar, Tanzania: A Multiple Bias Analysis

Caitlin Gerdts, University of California, San Francisco
Alison Norris, Ohio State University

Our study sought to 1) establish a framework for identifying systematic error in hospital-based, post abortion care (PAC) data in settings where abortion is illegal; 2) employ this framework to examine reported cases of induced abortion among women seeking PAC at Mnazi Mmoja Hospital (MMH) in Zanzibar, Tanzania; and 3) employ multiple-bias analysis to “correct” for potential bias and generate a range of values for the true proportion of induced abortion related PAC. Between July 2010 and November 2010, 194 women, 15 or older seeking PAC services at MMH were enrolled. 4.4% of women reported seeking PAC as a result of induced abortion. We identified two sources of systematic error: Misclassification, and selection bias. Our of multiple-bias analysis of selection bias and misclassification in the proportion of PAC cases resulting from unsafe abortion showed substantial increases in the proportion of abortion related PAC when adjusted for biases in the data.