Where the Burden Lies: A framework and evaluation of systematic error in measurement of the health effects of unsafe abortion

By

Caitlin E. Gerdts

Doctor of Philosophy in Epidemiology

University of California, Berkeley

Professor Jennifer Ahern, Chair

Background:

Measuring the incidence and sequellae of unsafe abortion is notoriously challenging. In the parts of the world where abortion is considered unsafe, it is often also illegal (or heavily restricted), and highly stigmatized. Women experiencing abortion related complications, for fear of severe legal, social, or religious repercussions, are, therefore, less likely than women experiencing other kinds of pregnancy-related complications to seek care in medical facilities, or disclose their experiences with abortion.

Biases are repeatedly discussed in accounts of post abortion care (PAC) 2 3,4 where researchers are interested in the proportion of cases resulting from induced vs. spontaneous abortion, often in similarly restrictive legal and social settings. Classification of PAC resulting from induced abortion as PAC resulting from spontaneous abortion is known to occur frequently, and the reverse is also thought to be common.

The field of global reproductive health needs a simple, straightforward, quantitative framework through which to assess the expected direction and magnitude of biases that exists in studies of unsafe abortion (and resulting sequellea). Such a framework would not only allow the researchers to better quantify bias in their own studies, but would aid readers ability to incorporate quantitative information about existing biases into their interpretation of results.

Methods:

Analyses investigating separate research aims related to systematic error in the measurement of unsafe abortion related mortality and morbidity will be conducted over three chapters, as follows: Chapter 1: Systematic Review of current estimates of unsafe abortion related mortality from 2000-2011, Chapter 2: Bias Framework and Multiple bias analysis of the proportion of maternal mortality resulting from unsafe abortion, and Chapter 3: Multiple bias analysis of unsafe abortion related post abortion care seekers in Zanzibar, Tanzania.

Discussion and Significance:

The results of the three preceding analyses suggest it is likely that unsafe abortion has been significantly underestimated as a cause of maternal death and post abortion care. These results have clear implications for increasing efforts aimed at the proven interventions which help to decrease abortion related mortality and morbidity: reducing unintended pregnancy, ensuring access to safe abortion services where it is legal, increasing access to safe abortion services where laws have the potential to be revised, and providing access to comprehensive post abortion care with contraceptive counseling in places where access to abortion remains highly restricted. These results also have implications for scientists committed to producing sound evidence in a field with endemic measurement challenges. Improving methods to quantify the direction and magnitude of systematic error in studies, and integrate such information into the interpretation of results concerning the burden of unsafe abortion-related mortality and morbidity is the necessary first step in understanding these grave public health concerns, and targeting interventions that appropriately address their underlying causes.