Religiosity and Abortion Perceptions in Three Zambian Provinces

RELIGIOSITY AND ABORTION PERCEPTIONS IN THREE ZAMBIAN PROVINCES

by Meghan C. Gallagher

A dissertation submitted to Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy

Baltimore, Maryland June 2018

DISSERTATION ABSTRACT

Background:

Access to safe abortion is critical to the health of women. In Zambia, where abortion laws are among the least restrictive in sub-Saharan Africa, an estimated 30% of the annual 398 maternal deaths per 100,000 live births are due to unsafe abortion, which is higher than many neighboring countries. Zambia is also a notably Christian country, with 98% of the population ascribing to

Protestant or Catholic denominations; further, Evangelicals comprise an estimated 25% of the overall population. These religious denominations tend to view abortion as a sin within their doctrine and Zambia is a country where negative attitudes toward abortion are common, often prompting women to access clandestine, and often unsafe, abortion services to avoid the perception of stigma that would potentially result from disclosure.

Objectives:

The objectives of this study were to first construct measures of personal abortion attitudes, support for Zambian abortion laws, and individual religiosity. Next, individual and community factors that influence individuals’ abortion perceptions (abortion attitudes and support for

Zambian abortion law) were investigated. Finally, the role that religion and stigma play in women’s understanding of abortion and how it informs their perceptions, decision-making, and social experiences with abortion were explored.

Methods:

This study was based on secondary analysis of a cross-sectional household sample of women of reproductive age (15-44) living within three Zambian provinces and in-depth interviews at two time points with 51 women who had terminated a pregnancy. Exploratory and confirmatory

ii factor analyses were conducted to validate the Duke University Religion Index (DUREL) as well as an abortion attitude scale. Multivariate logistic regression models were employed to measure the association between individual religiosity and perceptions of abortion as well as the influence of community religiosity and community abortion attitudes on individual abortion perceptions.

Finally, a directed content analysis approach was applied to semi-structured in-depth interviews with woman conducted immediately after they had terminated a pregnancy and again 3-4 months later.

Results:

The DUREL showed good internal consistency; however, the results of the EFA and CFA suggest that DUREL cannot be used as a scale to measure religiosity within this population, as the subscales of the DUREL are not explained by a common underlying construct, as evidenced by goodness of fit statistics. The findings support the validity of the three-item subscale measuring intrinsic religiosity. The validity assessment of the DUREL supports that the DUREL be analyzed as three independent subscales and not an aggregate measure of religiosity. Within this female Zambian population, the five survey questions pertaining to abortion attitudes do not function together as a valid scale.

The multivariate analyses do not find associations between religiosity and abortion perceptions at an individual level but find associations between religiosity at a community level and individual abortion perceptions.

iii

The analysis of the in-depth interviews found that many of the women interviewed viewed abortion is a sin in their religion. Though most believed that praying for forgiveness would lead to absolution, many women still noted feelings of guilt and shame. Several respondents felt that abortion is perceived as non-normative and unacceptable within their religious communities, which can lead to difficulty reintegrating.

Conclusion:

Although this is the first study of its kind, this research suggests that religiosity at a community level may play a role in perpetuating abortion stigma and influencing abortion perceptions for women. The qualitative research suggests that women perceive that they are being judged by their communities, notably their religious communities, even in situations where their abortion is unknown. This study provides initial evidence that community religiosity and attitudes may motivate abortion stigma. Further research is need to in sub-Saharan Africa to confirm these findings and better understand this important social dynamic.

COMMITTEE OF FINAL THESIS READERS

Committee Members:

Amy O. Tsui, PhD Professor, Johns Hopkins Bloomberg School of Public Health Department of Population, Family and Reproductive Health

Mary Elizabeth Hughes, PhD Associate Scientist, Johns Hopkins Bloomberg School of Public Health Department of Population, Family and Reproductive Health

Anne E. Burke, MD, MPH Associate Professor, Johns Hopkins University School of Medicine Department of Gynecology and Obstetrics Associate Professor, Johns Hopkins Bloomberg School of Public Health Department of Population, Family and Reproductive Health

Carol R. Underwood, PhD Assistant Professor, Johns Hopkins Bloomberg School of Public Health Department of Health, Behavior and Society

Alternate Committee Members:

Caitlin E. Kennedy, PhD, MPH Associate Professor, Johns Hopkins Bloomberg School of Public Health Department of International Health Joint Appointment in Department of Health, Behavior and Society

Caroline Moreau, MD, PhD Associate Professor, Johns Hopkins Bloomberg School of Public Health Department of Population, Family and Reproductive Health

ACKNOWLEDGEMENTS

I would like to express my gratitude for my wonderful advising duo. Drs. Amy Tsui and M.E. Hughes have been wildly supportive throughout my doctoral journey and accompanying demographic events. Their support, advice, encouragement, and humor were always appreciated and will be remembered fondly.

I am grateful to my committee members, Drs. Anne Burke and Carol Underwood, who took the time to read my work and provide constructive feedback during my preliminary and final examinations. I am also grateful to Dr. Caroline Moreau who provided insightful thoughts over the years and served on my departmental examination and as an alternate for my final defense. I appreciate the thoughtful feedback from Dr. Terri Powell who participated in my departmental and preliminary examinations. Thanks to Drs. Caitlin Kennedy and Stephane Helleringer who kindly committed to participate as alternate members of my committees.

I was fortunate to be a part of the Department of Population, Family, and Reproductive Health. I am very grateful for faculty that I had the opportunity to work with and learn from, especially Drs. Bill Mosher and Anne Riley. I am grateful for the ongoing support and assistance of Lauren Black. I am thankful for the funding I received toward my doctoral education, notably from the Laurie Schwab Zabin Award, the Carl Swan Shultz Fellowship, the Fellowship in Family Planning and Reproductive Health, the Edward J. Dehne Award, and the US Policy Communication Training Program from the Population Reference Bureau.

Without Ann Moore at the Guttmacher Institute and Jenny Cresswell and Veronique Filippi at the London School of Hygiene and Tropical Medicine, I would not have had such rich and informative data. Sara Casey and Therese McGinn laid the groundwork for my enthusiasm for research and I am grateful to have had the opportunity to work with them at Columbia University for many years. I am fortunate to currently have wonderful colleagues at Save the Children who tirelessly work toward expanding access to reproductive health services in humanitarian settings.

My community of fellow students has been a beacon of support, humor, and calm throughout the doctoral process. I am especially grateful to my GSD accountability group comprised of Amanda Gatewood, Lauren Okano, and Nomi Weiss-Laxer. Sara Riese was my first friend on day one of my public health journey and I am honored to continue on this path with her. Hannah Lantos, Michelle Hawks Cuellar, Suzanne Bell, Amanda Kalamar, Sahnah Lim, Sarah Peitzmeier and many others in the Pop Fam community were gracious with their time, humor, and grounding. Nicole Armstrong has been a source of inspiration and laughs since our two-week intensive course in life tables.

The road to my PhD has involved a wedding, a birth, and a full-time job. I am grateful to the village that has made this accomplishment possible. My community of longtime friends in Brooklyn and beyond have provided perspective, distraction, and ongoing love and support. Gillian Collymore has been a remarkable caregiver to my son. I am grateful for my large and

vi wonderful extended family who have no idea what I’m doing with my life, but love me all the same.

My sister is the funniest person I know and a phenomenal individual who has lent a hand or shared a laugh on more occasions than I can count. My parents instilled in me a desire for learning and a commitment to social justice that has carried me thorough life as well as the doctoral process. Roger and James keep me going every day and are my heart.

Finally, I am grateful to all the brave women in Zambia and beyond who give their time and tell their stories so that we can learn from them in hopes of better understanding pathways for improving the health and well-being of all women and families throughout the world. These are trying and complicated times and I am inspired by those who continue to push forward.

vii

TABLE OF CONTENTS DISSERTATION ABSTRACT ...... ii COMMITTEE OF FINAL THESIS READERS ...... v ACKNOWLEDGEMENTS ...... vi TABLE OF CONTENTS ...... viii LIST OF TABLES ...... x LIST OF FIGURES ...... xii CHAPTER ONE: INTRODUCTION ...... 1 Background and Significance...... 2 Zambian Context for Abortion ...... 3 Abortion Stigma ...... 5 Religiosity and Abortion ...... 9 Theory and Conceptual Framework ...... 11 Study Aims and Hypotheses ...... 14 Chapter One References ...... 15 CHAPTER TWO: METHODS ...... 20 Overview ...... 21 Quantitative Study Design ...... 22 Power Analyses ...... 25 Measures...... 26 Independent Variables ...... 26 Dependent Variables ...... 32 Treatment of Missing Data ...... 33 Analytic Methods (Aims 1 & 2) ...... 34 Analytic Methods: Aim 1a ...... 35 Analytic Methods: Aim 1b ...... 36 Analytic Methods: Aim 2 ...... 38 Qualitative Study Design ...... 40 Analytic Methods (Aim 3) ...... 42 Chapter Two Tables and Figures ...... 45 Abortion Attitudes ...... 56 Perception of Abortion Law ...... 56 Chapter Two References ...... 58

viii

CHAPTER THREE: AIM 1 ANALYTIC RESULTS ...... 62 Aim 1A Analytic Results: Validity and Reliability of DUREL ...... 63 Aim 1B Analytic Results...... 66 Chapter Three Tables and Figures ...... 71 Chapter Three References ...... 76 CHAPTER FOUR: AIM 2 ANALYTIC RESULTS ...... 77 Results ...... 78 Discussion ...... 85 Chapter Four Tables and Figures ...... 89 Chapter Four References ...... 97 CHAPTER FIVE: AIM 3 ANALYTIC RESULTS ...... 99 Results ...... 100 Discussion ...... 113 Chapter Five Tables and Figures ...... 114 Chapter Five References ...... 115 CHAPTER SIX: CONCLUSION ...... 116 Introduction ...... 117 Overview of Aim One ...... 117 Overview of Aim Two ...... 119 Overview of Aim Three ...... 121 Overall Conclusions of the Dissertation...... 122 Strengths and Limitations...... 126 Limitations ...... 126 Strengths ...... 128 Research Implications ...... 129 Public Health Implications ...... 130 Conclusion ...... 131 Chapter Six References ...... 133 APPENDIX ...... 136 APPENDIX A: In-Depth Interview Questionnaire ...... 137 APPENDIX B: Coding Scheme for Qualitative Data ...... 141 APPENDIX C: Institutional Review Board Determination ...... 145 CURRICULUM VITAE ...... 146

LIST OF TABLES

Table 2.1 Standard enumeration areas (SEAs) sampled per district, by province

Table 2.2 Response Rate by Province

Table 2.3 Power Calculation

Table 2.4 DUREL Subscales, item questions, and responses

Table 2.5 Religious Denomination Distributions Among Respondents

Table 2.6 Five abortion attitude item questions and responses

Table 2.7 Support for abortion laws, items and responses

Table 2.8 Distribution of respondents’ agreement with eight legal allowances for abortion

Table 2.9 Response patterns for legal allowances of abortion

Table 2.10 Flow diagram of missing data

Table 2.11 Patterns of missingness for outcome variables

Table 2.12 In-Depth Interviews by Type of Evacuation and Interview Completion

Table 3.1 DUREL Item Response Descriptive Statistics for Sample

Table 3.2 Means, SD, and coefficient alphas for Duke Religious Index scales

Table 3.3 Participant characteristics by lower vs. higher abortion attitudes

Table 3.4 Results of the Exploratory Factor Analysis

Table 4.1 Baseline characteristics of the sample

Table 4.2 Weighted* percent of women aged 15-44 who agreed with attitudinal statements on abortion

Table 4.3 Weighted* percent of women aged 15-44 support legalization of abortion for three or more indications

Table 4.4 Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “A woman has the right to decide whether or not to continue a pregnancy”

Table 4.5 Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “A woman has the right to decide whether or not to terminate a pregnancy”

Table 4.6 Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “Women should have access to safe abortion services”

Table 4.7 Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “Unmarried women should have access to safe abortion services”

Table 4.8 Estimated unadjusted and adjusted odds ratios from logistic regression analysis of support for legalization of abortion (≥3 indications)

Table 5.1 Descriptive characteristics of respondents at Time 1 and Time 2 Interviews

LIST OF FIGURES

Figure 1.1 The prevalence paradox: the social construction of deviance despite the high incidence of abortion

Figure 1.2 Conceptual framework of religiosity and abortion perceptions

Figure 2.1 Map of Zambia by provinces and districts

Figure 2.2 Single factor model of religiosity (DUREL)

Figure 2.3 Second-order single-factor model of the DUREL

Figure 2.4 Distribution of respondent’s abortion attitude responses, by item

Figure 2.5 Distribution of respondent’s number of favored legal allowances for abortion

Figure 2.6 Aim 2 Analytic Framework

Figure 3.1 Standardized parameter estimates of the single-factor model of the DUREL

Figure 3.2 Standardized parameter estimates of the second-order single-factor model of the DUREL

Figure 3.3 Distribution of Abortion Attitude Score

xii

CHAPTER ONE: INTRODUCTION

Background and Significance

Access to safe abortion is critical to the health of women. Unsafe abortion, a procedure for terminating a pregnancy performed by persons lacking necessary skills or in an environment that does not conform to minimal medical standards, is a leading and preventable cause of maternal mortality and morbidity worldwide.1, 2 Each year over 47,000 women die and an additional 5 million suffer from severe complications due to abortions that are unsafe.3 In sub-Saharan

Africa, more than 97 percent of abortions received by women are unsafe and unsafe abortion is responsible for at least 9 percent of all maternal deaths with over 1.6 million women hospitalized for complications due to these dangerous procedures each year.4-6

The reasons that women resort to unsafe abortion in sub-Saharan Africa are nuanced and complicated. Rates of unintended pregnancy in sub-Saharan Africa remain the highest in the world, although they have declined rapidly over the past two decades.7 Despite increasing contraceptive prevalence, this region still has the highest levels of unmet need for family planning.8-11 In most sub-Saharan African countries, abortions are illegal or highly regulated such that clandestine, unsafe abortion is the only option available. Even in settings where the procedure is legal, women often resort to unsafe abortion for reasons of discretion and confidentiality. Safe abortion is often prohibitively expensive for poor women, leading them to cheaper, unregulated alternatives.

Regulation and cost are important causes of unsafe abortion. However, in many instances it is not restrictive laws, but rather the stigma of abortion that encourages women to seek clandestine abortions.12 Stigma is an outward manifestation of a community’s cultural and religious norms

2 and involves negative judgement and shaming of behaviors considered non-normative 13 When living in an environment where abortion is highly stigmatized, women are willing to risk their health rather than suffer from social scorn and isolation.14, 15 Abortion is highly stigmatized by most major religions, especially Christianity. Several studies have shown that religion can drive shame-related stigma around various health issues; however, little research has been conducted to show the impact of religion on abortion attitudes and behavior in sub-Saharan Africa.16, 17

This research project sought to more thoroughly assess the relationship between religiosity, stigma, and women’s perceptions surrounding abortion to more clearly understand its role in perpetuating or minimizing unsafe abortion.

Zambian Context for Abortion

Zambia has one of the least restrictive abortion laws in sub-Saharan Africa, though it remains highly regulated. The 1972 Termination of Pregnancy Act permits the ending of a pregnancy for the following reasons: risk to the life of the pregnant woman; risk of injury to the physical or mental health of the pregnant woman; risk of injury to the physical or mental health of any existing children of the pregnant woman, or risk of physical or mental abnormalities to the unborn child.18 In 2005, Zambian lawmakers amended this code to include rape as a legal reason for seeking an abortion and to ensure that women who have been raped were no longer prosecuted for attempts to self-abort or abort illegally.19

Despite these comparatively liberal abortion laws, an estimated 30% of the annual 398 maternal deaths per 100,000 live births are due to unsafe abortion.19 Similarly, the incidence of abortion-

3 related near-miss complications is also high compared to legally restrictive settings, despite

Zambia’s more open abortion laws.20

Unsafe abortion remains prevalent in Zambia in part due to the stipulation that abortion be performed by a physician in a hospital with the consent of three registered medical providers, one of whom must be a specialist.18 Though only one physician is needed for consent in situations of emergency, the requirement remains a substantial barrier due to the sheer lack of physicians in the country- less than a third of the doctor-patient ratio recommended by the

WHO.21 A 2008 qualitative study using in-depth interviews with key stakeholders found that abortions by trained providers were available only at tertiary care facilities, making it difficult for women living far distances away to access these safe services.22

The institutional barriers to obtaining an abortion in Zambia are formidable. However, the most important barriers may be those that occur outside of a clinical setting and prevent women from even seeking an abortion. In order to seek out abortion services, women must first know that these services are available. A lack of knowledge about abortion laws presents another barrier, with many women in Zambia mistakenly believing that abortion is illegal and is not available in government hospitals or clinics.23 Once availability and legality of abortion is understood, women must be able to afford abortion services, yet many women cite the perceived burden of economic cost as primary obstacle to accessing safe abortion.24

Finally, and perhaps most importantly, many Zambians have negative attitudes toward abortion, believing that it is immoral, especially for adolescents and young people.22 Research in a variety

4 of geographical and cultural contexts has found that these negative attitudes, when pervasive at a population or community-level, lead to internalized stigma, which is the process of internalizing negative attitudes about abortion held by one’s community into negative beliefs about oneself.25

This stigma often leads to non-disclosure and secrecy around abortion behavior, interpersonal conflict with friends and relatives, and deterioration of physical and mental health.26-28 The desire to avoid stigma can lead women to prioritize clandestine abortion services, with less attention given to quality of care or safety.

Though abortion laws in Zambia should theoretically favor access to safe abortion services, a myriad of barriers persist including a lack of awareness regarding abortion laws, a fragile health system, and a highly stigmatized abortion environment. Though it is often difficult for women to access legal abortion services due to a dearth of trained health providers, the statistics suggest that women are choosing not to access formal abortion services and, to avoid disclosure of their unintended pregnancy and subsequent abortion, attempt to terminate pregnancy by measures taken on their own or through informal community services. Because a desire for secrecy may supersede the prioritization of health and safety, Zambian women are at risk for the potential negative health consequences of unsafe abortion.

Abortion Stigma

Stigmatization is a dynamic and contextual social process that leads to the disgrace of an individual due to a particular characteristic they hold that violates social expectations.27, 29-31 In

1963, Goffman defined stigma as “an attribute that is deeply discrediting,” reducing the bearer

“from a whole and usual person to a tainted, discounted one.”32 This definition has been

5 expanded over the past half century as it has been applied in various fields, including public health issues such as mental health, tuberculosis, cancer, HIV / AIDS, abortion, and health service provision for stigmatized groups.28, 33-44 However, two components of stigma have consistently appeared across disciplines- the perception of negative personal characteristics and the total devaluation of the bearer within society.29

Abortion stigma has been defined as “a negative attribute ascribed to women who seek to terminate a pregnancy that marks them, internally or externally, as inferior to ideals of womanhood”45 Abortion stigma, like all health-related stigma, develops across a wide variety of cultural and social contexts.30, 33 However, abortion stigma is not uniformly manifested; stigma is experienced in a “local world” consisting of one’s specific social relationships and cultural constructs.46 Therefore, stigma is a multilevel conceptualization and is created through interactions between individuals, within communities, in institutions, and in law and government structures..29, 45

Kumar et al. used Link and Phelan’s conceptualization of stigma to explain how abortion stigma is created and perpetuated.30, 45 Like all forms of stigma, abortion stigma arises due to the oversimplification of a complex situation. The termination of a pregnancy is highly contextual due to personal, societal, and cultural reasons. Often, however, the fact that abortion is a common procedure that occurs for 25% of all pregnancies worldwide is unknown or not part of common discourse, thus creating the façade that those who terminate pregnancies are a marked group.47 Widespread underreporting and misclassification of abortions by providers and clients has led to misperceptions about abortion prevalence, with only 35-60% of actual abortions being

6 reported in surveys.48-50 Because women feel stigmatized, they keep their abortion secret, which reinforces the misconception that abortion is a non-normative event. This leads to the idea that abortion is a deviant behavior and women who have had abortions face discrimination. Thus, we observe a “mutually reinforcing cycle of silence” that makes it challenging for people to know the true prevalence of abortion and thus prevents any degree normalization (see Figure 1.1).45

Figure 1.1. The prevalence paradox: the social construction of deviance despite the high incidence of abortion.

Women underreport and intentionally misclassify abortion behavior

Women fear Abortion is thought stigmatization for to be uncommon engaging in abortion (non-normative) behavior

Women who have A social norm is abortions face perpetuated that discrimination abortion is deviant

Kumar et al., 2009

An additional layer of complexity is added to the cycle of abortion stigma in that in many instances, women who have had abortions themselves often perpetuate abortion stigma within

7 their community. Existing literature suggest that women often feel that their abortion experience was justifiable due to their particular circumstances, but that others who have abortions have made immoral choices, thus discrediting their abortion decision.51, 52

Although abortion stigma persists in a variety of socio-cultural environments, it is exceptionally common in sub-Saharan Africa.47 Previous studies have found that behaviors related to abortion are also stigmatized, such as sexual behavior (premarital sex and adolescent sexuality) and postabortion care. Stigma is felt individually, at a medical level, and at a community level.

Individuals note feelings of fear, shame, and embarrassment for their behavior. At the medical level, women feel that providers can be judgmental, indifferent and/or disdainful. At a societal level, women are judged by family, peers, and their community for their morality and for violating real or perceived ethical norms.53-59

Though stigma results from a variety of societal beliefs, moral systems, and stereotypes, religion is one element that can influence the shame and discrimination that women face as a result of having an abortion. Though very little research currently examines associations between religiosity, stigma, and abortion, there is a strong body of literature showing associations between religion and other health-related stigma, such as HIV.16, 17, 55, 60 Religion plays an important role in daily life in most sub-Saharan African countries and research shows it prompts and sustains some of these health-related stigma.16, 17, 60

Religiosity and Abortion

Religious faiths vary on their stance toward abortion. However, among most major religions, viewpoints on abortion are closely tied to sexual morality and attitudes toward human life.61

Thus, these facets are commonly used when assessing individual attitudes toward abortion.31, 62

In Zambia, abortion is heavily influenced by Christian faith traditions, as 80% of the population is Protestant and 18% of the population is Catholic.63 Catholic doctrine specifically states that abortion is unacceptable in all cases, even in situations of fetal abnormalities, preservation of maternal mental health, and preservation of maternal life.64 Protestant faiths vary much more in their stance on abortion; some support abortion in cases of fetal abnormalities and to maintain the life and well-being of the mother, while others do not.65 However, even within religious dogma, the stance on reproductive health, and specifically abortion, has not been static over time.66-69 Very often, a variety of views coexist within the particular religious traditions. In understanding the position of any particular religious denomination on abortion, it is important not to look for a single “official” position in any religion, even in the most hierarchical religions.

Though religion is an important component in women’s lives, there are many other considerations and factors are involved in a woman’s fertility decisions, such as her educational level, the status of women within her society, and culturally embedded pathways.70

The roles of both religion and religiosity and their impact on abortion-seeking behavior as well as attitudes toward abortion have been studied in the United States. Research suggests that denominational affiliation is a statistically significant factor in explaining whether a young unmarried woman obtains an abortion, though the strength of the association pales in comparison to academic goals as a reason for or against abortion. Women who identify as conservative

Protestants are less likely to abort an out-of-wedlock pregnancy than women who identify as

Catholic or with a mainline Protestant denomination.67

A limited amount of research on the link between religion and abortion has been conducted outside of the United States. A study in Iran suggested that the current fertility decline is attributable to both contraception and abortion, despite the strict, theocratic image of Iranian society.71 Among Mexican Catholics, though the majority have stigmatizing attitudes toward abortion and negatively perceiving women who have had abortions, over four-fifths believed that abortion should be legal in at least some circumstances.72 Education had a liberalizing effect on abortion attitudes except among the most religious group for whom education had the opposite effect.72, 73 Among educated women in Cameroon, despite the shame associated with abortion, the humiliation was worse when birthing a child outside of marriage.74 Many studies have found that the religion of a health care provider can block access to safe abortion.57, 75-77

Though abortion stigma has been studied in a sub-Saharan African context, there has been no research to date that has explored the link between religion and abortion attitudes in sub-Saharan

Africa. This study will begin to fill this gap in understanding in two ways: by assessing the relationship between a woman’s religiosity and her attitudes and understanding around abortion, and by also exploring the influence of community-level religiosity and abortion perceptions on women’s own attitudes and understanding around abortion. It is important to examine the influence of religiosity at both individual and community levels, as religion generally exists simultaneously as a private relationship between the individual and the divine and the shared experience of communal worship and fellowship.

Theory and Conceptual Framework

Abortion knowledge, attitudes, and behaviors are influenced by a variety of factors at the individual level, such as age and educational attainment; the family level, such as wealth; and at the community level, including religiosity, culture, and stigma. Assessing the impact of religiosity on abortion requires taking these multilevel factors into account. The conceptual framework guiding this analysis is presented below in Figure 1.2. The model is influenced by the ecological framework that was originally developed by Urie Bronfenbrenner for use in child development.78 It has been adapted over time to address a host of areas within health research including health disparities,79 violence,80 and HIV.81 This model situates individuals within the multiple social contexts that they experience in their daily lives and considers the intricate interplay between individual, household, community, and societal factors.82, 83 In this case, it allows us to understand the range and layers of factors that contribute to an individual’s abortion perceptions. The overlapping rings in the model illustrate how factors at one level influence factors at another level. In addition to clarifying the factors that influence the outcome of interest, the model also suggests that in order to change individual-level abortion perceptions, it is necessary to work across multiple levels of the model at the same time, as opposed to focusing only on one level or one intervention. As abortion stigma is a social construction that is based on the interaction between an individual and society, the ecological model is necessary for understanding how individuals are influenced by their families and broader communities.

This particular ecological framework examines the relationship between individual, household, and community-level variables and the outcomes of interest, an individual’s attitudes toward

11 abortion and knowledge and support of abortion laws. The framework is based on existing evidence showing that these multilevel covariates influence one another in other contexts and so are likely to influence individual abortion perceptions in sub-Saharan Africa.

Figure 1.2. Conceptual framework of religiosity and abortion perceptions

Study Aims and Hypotheses

The aims of study are as follows:

Aim 1: Construct measures of personal abortion attitudes, support for Zambian abortion laws, and individual religiosity.

Sub-Aim Hypothesis 1a. Assess the validity and reliability of the No hypothesis. Duke University Religion Index (DUREL).

1b. Construct measure of personal abortion No hypothesis. attitudes and support for the legality of abortion.

Aim 2: Investigate the individual and community factors that influence individuals’ abortion perceptions (abortion attitudes and support for Zambian abortion law).

Sub-Aim Hypothesis 2a. Assess the relationships between 2a. Women with higher levels of religiosity will have individuals’ religiosity and their abortion more conservative abortion perceptions than women perceptions. with lower levels of religiosity.

2b. Assess the relationships between the 2b1. Women living in communities with higher religiosity and abortion perceptions of an levels of religiosity will have more conservative individual’s residential community and abortion perceptions. her own abortion perceptions. 2b2. Women living in communities with more conservative abortion perceptions will have more conservative abortion perceptions themselves.

Aim 3: Explore the role that religion and stigma play in women’s understanding of abortion and how it informs their perceptions, decision-making, and social experiences with abortion.

Sub-Aim Hypothesis 3a. Explore the interplay of religion and stigma in women’s No hypothesis. experiences post-abortion. 3b. Explore how the roles of religion and stigma change between the No hypothesis. time immediately following an abortion and a point four months later.

Chapter One References

1. Ahman, E. and I.H. Shah, New estimates and trends regarding unsafe abortion mortality. Int J Gynaecol Obstet, 2011. 115(2): p. 121-6. 2. Grimes, D.A., et al., Unsafe abortion: the preventable pandemic. The Lancet, 2006. 368(9550): p. 1908-1919. 3. World Health Organization, Unsafe abortion: global and regional estimates of incidence of unsafe abortion and associated mortality in 2008. 2011. 4. Sedgh, G., et al., Induced abortion: incidence and trends worldwide from 1995 to 2008. The Lancet, 2012. 379(9816): p. 625-632. 5. Singh, S., J. Darroch, and L. Ashford, Adding it Up: The Costs and Benefits of Investing in Sexual and Reproductive Health 2014. Guttmacher Institute, 2014. 6. Singh, S. and I. Maddow-Zimet, Facility-based treatment for medical complications resulting from unsafe pregnancy termination in the developing world, 2012: a review of evidence from 26 countries. BJOG, 2015. 7. Sedgh, G., S. Singh, and R. Hussain, Intended and unintended pregnancies worldwide in 2012 and recent trends. Studies in Family Planning, 2014. 45(3): p. 301-314. 8. Bongaarts, J., The Impact of Family Planning Programs on Unmet Need and Demand for Contraception. Studies in Family Planning, 2014. 45(2): p. 247-262. 9. Wusu, O., Religious Influence on Non-Use of Modern Contraceptives among Women in Nigeria: Comparative Analysis of 1990 and 2008 Ndhs. J Biosoc Sci, 2015. 47(5): p. 593-612. 10. Asekun-Olarinmoye, E., et al., Barriers to use of modern contraceptives among women in an inner city area of Osogbo metropolis, Osun state, Nigeria. Int J Womens Health, 2013. 5: p. 647-55. 11. Tsui, A.O., et al., Managing unplanned pregnancies in five countries: Perspectives on contraception and abortion decisions. Global public health, 2011. 6(sup1): p. S1-S24. 12. Lithur, N.O., Destigmatising abortion: expanding community awareness of abortion as a reproductive health issue in Ghana. African journal of reproductive health, 2004. 8(1): p. 70-74. 13. Cockrill, K. and L. Hessini, Introduction: Bringing Abortion Stigma into Focus. Women & Health, 2014. 54(7): p. 593-598. 14. Cockrill, K. and A. Nack, “I'm Not That Type of Person”: Managing the Stigma of Having an Abortion. Deviant Behavior, 2013. 34(12): p. 973-990. 15. Cockrill, K., et al., Addressing Abortion Stigma Through Service Delivery. 2013.

16. Varas-Díaz, N., et al., Religion and HIV/AIDS stigma: Implications for health professionals in Puerto Rico. Global Public Health, 2010. 5(3): p. 295-312. 17. Zou, J., et al., Religion and HIV in Tanzania: influence of religious beliefs on HIV stigma, disclosure, and treatment attitudes. BMC public health, 2009. 9(1): p. 1. 18. The Termination of Pregnancy Act. 1972: Zambia. 19. Standards and Guidelines for Reducing Unsafe Abortion Morbidity and Mortality in Zambia, Government of the Republic of Zambia Ministry of Health, Editor. 2014: Lusaka, Zambia. 20. Owolabi, O.O., et al., Incidence of abortion-related near-miss complications in Zambia: cross-sectional study in Central, Copperbelt and Lusaka Provinces. Contraception, 2016. 21. Schatz, J.J., Zambia's health-worker crisis. The Lancet. 371(9613): p. 638-639. 22. Likwa, R.N., A. Biddlecom, and H. Ball, Unsafe abortion in Zambia. Issues in brief (Alan Guttmacher Institute), 2009(3): p. 1-4. 23. Geary, C.W., et al., Attitudes toward abortion in Zambia. International Journal of Gynecology & Obstetrics, 2012. 118: p. S148-S151. 24. Coast, E. and S.F. Murray, "These things are dangerous": Understanding induced abortion trajectories in urban Zambia. Soc Sci Med, 2016. 153: p. 201-9. 25. Major, B., et al., Report of the APA task force on mental health and abortion. American Psychological Association, 2008. 26. Major, B. and L.T. O'Brien, The social psychology of stigma. Annu. Rev. Psychol., 2005. 56: p. 393-421. 27. Link, B.G. and J.C. Phelan, Stigma and its public health implications. The Lancet, 2006. 367(9509): p. 528-529. 28. Shellenberg, K.M. and A.O. Tsui, Correlates of perceived and internalized stigma among abortion patients in the USA: An exploration by race and Hispanic ethnicity. International Journal of Gynecology & Obstetrics, 2012. 118, Supplement 2: p. S152- S159. 29. Norris, A., et al., Abortion stigma: a reconceptualization of constituents, causes, and consequences. Womens Health Issues, 2011. 21(3 Suppl): p. S49-54. 30. Link, B.G. and J.C. Phelan, Conceptualizing stigma, in Annual Review of Sociology. 2001. p. 363-385. 31. Shellenberg, K.M., L. Hessini, and B.A. Levandowski, Developing a Scale to Measure Stigmatizing Attitudes and Beliefs About Women Who Have Abortions: Results from Ghana and Zambia. Women & Health, 2014. 54(7): p. 599-616. 32. Goffman, E., Stigma: Notes on the management of spoiled identity. 2009: Simon and Schuster. 33. Link, B.G., et al., Measuring mental illness stigma. Schizophrenia Bulletin, 2004. 30(3): p. 511-541.

34. Major, B. and R.H. Gramzow, Abortion as stigma: cognitive and emotional implications of concealment. Journal of personality and social psychology, 1999. 77(4): p. 735. 35. Parker, R. and P. Aggleton, HIV and AIDS-related stigma and discrimination: a conceptual framework and implications for action. Social Science & Medicine, 2003. 57(1): p. 13-24. 36. Mokgethi, N.E., V.J. Ehlers, and M.M. van der Merwe, Professional nurses' attitudes towards providing termination of pregnancy services in a tertiary hospital in the north west province of South Africa. Curationis., 2006. 29(1): p. 32-39. 37. Moore, A.M., et al., Attitudes of women and men living with HIV and their healthcare providers towards pregnancy and abortion by HIV-positive women in Nigeria and Zambia. Afr J AIDS Res, 2015. 14(1): p. 29-42. 38. Chavkin, W., L. Leitman, and K. Polin, Conscientious objection and refusal to provide reproductive healthcare: A White Paper examining prevalence, health consequences, and policy responses. International Journal of Gynecology & Obstetrics, 2013. 123: p. S41- S56. 39. Joffe, C., Commentary: Abortion Provider Stigma and Mainstream Medicine. Women & Health, 2014. 54(7): p. 666-671. 40. Kleinman, A., et al., The social course of epilepsy: Chronic illness as social experience in interior China. Social Science & Medicine, 1995. 40(10): p. 1319-1330. 41. Jaramillo, E., Tuberculosis and Stigma: Predictors of Prejudice Against People with Tuberculosis. Journal of Health Psychology, 1999. 4(1): p. 71-79. 42. Kelly, P., Isolation and Stigma: The Experience of Patients With Active Tuberculosis. Journal of Community Health Nursing, 1999. 16(4): p. 233-241. 43. Long, N.H., et al., Fear and social isolation as consequences of tuberculosis in VietNam: a gender analysis. Health Policy, 2001. 58(1): p. 69-81. 44. Lebel, S. and G.M. Devins, Stigma in cancer patients whose behavior may have contributed to their disease. Future Oncol, 2008. 4(5): p. 717-33. 45. Kumar, A., L. Hessini, and E.M. Mitchell, Conceptualising abortion stigma. Cult Health Sex, 2009. 11(6): p. 625-39. 46. Yang, L.H., et al., Culture and stigma: Adding moral experience to stigma theory. Social Science and Medicine, 2007. 64(7): p. 1524-1535. 47. Sedgh, G., et al., Abortion incidence between 1990 and 2014: global, regional, and subregional levels and trends. The Lancet, 2016. 388(10041): p. 258-267. 48. Jones, R.K. and K. Kost, Underreporting of induced and spontaneous abortion in the United States: an analysis of the 2002 National Survey of Family Growth. Stud Fam Plann, 2007. 38(3): p. 187-97. 49. Jagannathan, R., Relying on Surveys to Understand Abortion Behavior: Some Cautionary Evidence. American Journal of Public Health, 2001. 91(11): p. 1825-1831. 50. Rossier, C., et al., Estimating clandestine abortion with the confidants method--results from Ouagadougou, Burkina Faso. Soc Sci Med, 2006. 62(1): p. 254-66. 17

51. Nickerson, A., R. Manski, and A. Dennis, A Qualitative Investigation of Low-Income Abortion Clients’ Attitudes Toward Public Funding for Abortion. Women & Health, 2014. 54(7): p. 672-686. 52. Sorhaindo, A.M., et al., Qualitative Evidence on Abortion Stigma from Mexico City and Five States in Mexico. Women & Health, 2014. 54(7): p. 622-640. 53. Tagoe-Darko, E., “Fear, Shame and Embarrassment”: The Stigma Factor in Post Abortion Care at Komfo Anokye Teaching Hospital, Kumasi, Ghana. Asian Social Science, 2013. 9(10). 54. Levandowski, B.A., et al., Investigating social consequences of unwanted pregnancy and unsafe abortion in Malawi: The role of stigma. International Journal of Gynecology & Obstetrics, 2012. 118: p. S167-S171. 55. Neuman, M., C.M. Obermeyer, and M.S. Group, Experiences of stigma, discrimination, care and support among people living with HIV: a four country study. AIDS Behav, 2013. 17(5): p. 1796-808. 56. Orner, P., et al., A qualitative exploration of HIV-positive pregnant women's decision- making regarding abortion in Cape Town, South Africa. SAHARA-J: Journal of Social Aspects of HIV/AIDS, 2010. 7(2): p. 44-51. 57. Holcombe, S.J., A. Berhe, and A. Cherie, Personal Beliefs and Professional Responsibilities: Ethiopian Midwives' Attitudes toward Providing Abortion Services after Legal Reform. Stud Fam Plann, 2015. 46(1): p. 73-95. 58. Larsson, S., et al., The discourses on induced abortion in Ugandan daily newspapers: a discourse analysis. Reprod Health, 2015. 12: p. 58. 59. Izugbara, C.O., C. Egesa, and R. Okelo, 'High profile health facilities can add to your trouble': Women, stigma and un/safe abortion in Kenya. Soc Sci Med, 2015. 141: p. 9-18. 60. Sambisa, W., AIDS stigma and uptake of HIV testing in Zimbabwe. 2008. 61. Jelen, T.G., The Subjective Bases of Abortion Attitudes: A Cross National Comparison of Religious Traditions. Politics and Religion, 2014. 7(03): p. 550-567. 62. Cresswell, J.A., et al., Women's knowledge and attitudes surrounding abortion in Zambia: a cross-sectional survey across three provinces. BMJ Open, 2016. 6(3): p. e010076. 63. Central Statistical Office (CSO) [Zambia], M.o.H.M.Z., and ICF International. , Zambia Demographic and Health Survey 2013-14. 2014, Central Statistical Office, Ministry of Health, and ICF International: Rockville, Maryland, USA. 64. See, T.H. Catechism of the Catholic Church. Article 5 The Fifth Commandment May 2, 2016]; Available from: http://www.vatican.va/archive/ccc_css/archive/catechism/p3s2c2a5.htm. 65. Tomkins, A., et al., Controversies in faith and health care. The Lancet, 2015. 386(10005): p. 1776-1785. 66. Kissling, F., Religion and abortion: Roman catholicism lost in the pelvic zone. Women's Health Issues, 1993. 3(3): p. 132-137.

67. Adamczyk, A., The effects of religious contextual norms, structural constraints, and personal religiosity on abortion decisions. Social Science Research, 2008. 37(2): p. 657- 672. 68. Hirsch, J.S., Catholics using contraceptives: Religion, family planning, and interpretive agency in rural Mexico. Studies in Family Planning, 2008. 39(2): p. 93-104. 69. Stephens, M., et al., Religious perspectives on abortion and a secular response. Journal of religion and health, 2010. 49(4): p. 513-535. 70. Maguire, D.C., Abortion and Religion, in The Wiley Blackwell Encyclopedia of Gender and Sexuality Studies. 2016, John Wiley & Sons, Ltd. 71. Erfani, A. and K. McQuillan, Rates of induced abortion in Iran: The roles of contraceptive use and religiosity. Studies in Family Planning, 2008. 39(2): p. 111-122. 72. McMurtrie, S.M., et al., Public opinion about abortion-related stigma among Mexican Catholics and implications for unsafe abortion. International Journal of Gynecology & Obstetrics, 2012. 118: p. S160-S166. 73. Rosenhouse-Persson, S. and G. Sabagh, Attitudes toward abortion among catholic Mexican-American women: The effects of religiosity and education. Demography, 1983. 20(1): p. 87-98. 74. Johnson-Hanks, J., The lesser shame: abortion among educated women in southern Cameroon. Social Science & Medicine, 2002. 55(8): p. 1337-1349. 75. Aniteye, P. and S.H. Mayhew, Shaping legal abortion provision in Ghana: Using policy theory to understand provider-related obstacles to policy implementation. Health Research Policy and Systems, 2013. 11(1). 76. Lema, V.M., Conscientious objection and reproductive health service delivery in sub- Saharan Africa. Afr J Reprod Health, 2012. 16(1): p. 15-21. 77. Rehnstrom Loi, U., et al., Health care providers' perceptions of and attitudes towards induced abortions in sub-Saharan Africa and Southeast Asia: a systematic literature review of qualitative and quantitative data. BMC Public Health, 2015. 15: p. 139. 78. Bronfenbrenner, U., Contexts of child rearing: Problems and prospects. American psychologist, 1979. 34(10): p. 844. 79. Reifsnider, E., M. Gallagher, and B. Forgione, Using ecological models in research on health disparities. Journal of Professional Nursing, 2005. 21(4): p. 216-222. 80. Krug, E.G., et al., The world report on violence and health. The lancet, 2002. 360(9339): p. 1083-1088. 81. Baral, S., et al., Modified social ecological model: a tool to guide the assessment of the risks and risk contexts of HIV epidemics. BMC public health, 2013. 13(1): p. 1. 82. Bronfenbrenner, U. and P.A. Morris, The ecology of developmental processes. 1998. 83. Bronfenbrenner, U. and G.W. Evans, Developmental science in the 21st century: Emerging questions, theoretical models, research designs and empirical findings. Social development, 2000. 9(1): p. 115-125.

CHAPTER TWO: METHODS

Overview

This study utilizes a mixed-methods design; both quantitative and qualitative approaches were employed to explore the impact of religiosity on abortion perceptions in three Zambian provinces. All data come from evaluation components of Preventing Unwanted Maternal Deaths from Unsafe Pregnancies (EVA-PUMDUP), a program started in 2011 with the goal of preventing maternal death and unsafe abortion through the augmentation and amelioration of abortion service outlets and trained providers in fourteen low-income countries. The program and its evaluation were funded by the United Kingdom’s Department for International

Development (DFID). Marie Stopes International and Ipas implemented the program. In Zambia, support and clinical training were provided to over 85 health facilities that provide safe and comprehensive abortion and family planning services.

The quantitative component of the study consists of secondary data analysis of a 2016 population-based survey of 1,497 women of reproductive age in three provinces of Zambia. The qualitative component will involve a secondary data analysis of in-depth interviews (IDIs) conducted with 51 abortion clients at two points in time: immediately after their procedure and

3-4 months after their procedure.

Mixed methods are particularly important given the topic of abortion. Quantitative methodologies are limited in their ability to fully describe the complexities of abortion attitudes and behaviors, particularly given the silence around abortion due to social stigma. Population research has shown that abortion rates are difficult to measure accurately due to inaccurate registration systems, the reluctance of women to report abortion events in a survey, or a

21 combination of the aforementioned.1-4 Since it is impossible to accurately assess abortion history and thus its relationship to abortion attitudes via a cross-sectional survey, the qualitative element of this study provides needed information from women who have had a recent abortion. In addition, these in-depth interviews enrich the understanding of the relationship between the scales developed under quantitative Aim 1 and provide context to the relationship between religiosity and abortion perceptions as developed in Aim 2, thus greatly improving the insights gained from the quantitative survey by exploring topics that could not be expanded upon using a quantitative methodology. The triangulation of data collection methods (in-depth interviews and the quantitative survey) strengthen the reliability and validity of the study by providing the opportunity to observe data convergence or divergence in hypothesis testing.5

Quantitative Study Design

The community-based household survey was designed by the London School of Hygiene and

Tropical Medicine as an element of an evaluation of the Preventing Unwanted Maternal Deaths from Unsafe Pregnancies (PUMDUP) program. The evaluation component assessed improvements in the quality of abortion services available as well as women’s ability to access and afford these services. A cross-sectional baseline survey was conducted in 2014 followed by a cross-sectional endline survey in 2016. This analysis used data from the 2016 endline survey.

The study is a secondary data analysis of existing, de-identified data. This study did not qualify as human subjects research as defined by the Department of Health and Human Services regulations 45 CFR 46.102 by the Johns Hopkins Bloomberg School of Public Health

Institutional Review Board (IRB) Office on January 24, 2018 and did not require IRB oversight.6

(See Appendix C for IRB determination.)

The 2016 survey was granted ethical approval by three Institutional Review Boards (IRBs): the

London School of Hygiene and Tropical Medicine, the University of Zambia Biomedical

Research Ethics Committee (UNZABREC) and the Population Council. Prior to their involvement in the survey, the risks and benefits were explained to participants and written informed consent was obtained. Any information learned over the course of the survey administration was strictly confidential. Prior to analysis, all identifiers were stripped from the data. Interviewers underwent extensive training on administering a survey, handling sensitive questions, and confidentiality. All interviews were conducted in private. If a respondent indicated a need for particular services during the interview, they were referred to appropriate support.

The endline survey was administered between March and May 2016 to women of reproductive age (15-44) throughout the Central, Copperbelt and Lusaka provinces of Zambia, where the program facilities were concentrated. These provinces contain 41.7% of the overall population of

Zambia’s ten provinces (13,092,666) and are each examples of low, medium, and high population density. Central Province has a population of 1,307,111 and 13.4 people/km2;

Copperbelt Province has a population of 1,972,317 and 62.5 people/km2; and Lusaka Province has a population of 2,191,225 and 100.4 people/km2.7 Each province is made up of districts that are further subdivided into wards (Figure 2.1).

In addition to these administrative units, during the 2000 population census, each ward was further subdivided into areas called census supervisory areas (CSAs), and in turn each CSA into standard enumeration areas (SEAs). A SEA is defined by the Central Statistical Office (CSO)-

Zambia as a convenient geographical area with an average size of 130 households or 600 people.8

A multistage sampling procedure with probability proportional to the estimated number of households in each cluster was adopted for this survey. The sample frame was the list of SEAs developed from the 2010 population census, with the number and size of households within each

SEA enumerated. As the study arose from a programmatic evaluation, the level of program intensity was included for each SEA allowing for assessment of the intervention. Intensity was determined by proportion of the population served by trained health providers in that SEA.

Probability proportional to size was used to select 51 SEAs, or clusters, from which 30 households were selected in each via systematic random sampling. Within each household, one woman of reproductive age (15-44 years) was selected from among all eligible women. In the event that 2 or more eligible women were found in the selected household, a Kish grid was used to randomly select one of them for the interview. A Kish grid is a technique used to maintain equal-probability sampling by selecting a respondent at random when more than one eligible respondent is found in a sampled address or household.9 The intended number of women to be interviewed in this study was 1500. Table 2.1 shows the province, district, level of programmatic intensity and total number of SEAs sampled per district.

The structured survey questionnaire contained questions regarding sociodemographic characteristics, reproductive and contraceptive history, and fertility preferences. The wording used in the Demographic and Health Surveys (DHS) was maintained where possible in order to facilitate comparisons. A set of questions adapted from a previous study on knowledge of and attitudes towards abortion conducted by Geary et al.10 was included, as was the 5-item Duke

University Religion Index (DUREL).11

Women were interviewed in person, with the fieldworker entering her responses into a tablet computer. The enumerators who administered the survey were females of reproductive age, most had previous survey experience and all had the ability to communicate fluently in English and at least one local language. All participants provided written informed consent prior to participating in the survey.

In total, 1531 household visits were made during the endline survey, of which 1497 resulted in successful interview completion, giving an overall response rate of 97.78%. The 34 nonresponse households were all refusals of participation (Table 2.2).

Power Analyses

In statistics, the power of a test is the probability of correctly rejecting the null hypothesis when the null hypothesis is false or, simply put, the ability to detect statistical significance.12 This depends on the strength of the association between the variables of interest and the variance of each of those variables. For relationships that are less precise or that have a weaker association,

25 larger sample sizes are needed to detect real differences. Generally, a power level of 80% is an acceptable standard for statistical studies.

There are not previous studies in sub-Saharan Africa that examine the relationship between religiosity and abortion perceptions, thus no prior data were available to use as a comparison.

Sample sizes required to detect a significant difference in variable estimates at a power of 0.80 and 0.90 were calculated using the sum of the five personal abortion attitude questions. The minimum score is 5 and the maximum score is 25, with higher numbers indicating greater acceptance of abortion. Sample size calculations were conducted in Stata 14 using the sampsi command. The sample size was multiplied by the DEFF (design effects), which is a measure of how different the clustered sample is from a simple random sample. The most conservative sample size needed to detect a 20% change in the mean score of abortion attitudes is 932 (α =

0.01 and power = 0.90). Based on the results from the power calculations in Table 2.3 the analytic sample size of 1,432 is more than sufficient to ensure an adequate degree of power.

Measures

Independent Variables

Duke University Religion Index (DUREL)

The Duke University Religion Index (DUREL) is a five-item questionnaire measuring three core dimensions of religious involvement including organizational religious activity, measured by frequency of communal religious attendance, i.e. churchgoing (ORA / one item); non- organizational religious activity, measured by frequency of time spent alone in religious

26 activities such as prayer, meditation, or Bible study (NOR / one item); and intrinsic religiosity, measured by three questions pertaining to how religion guides one’s life (IR / three items) .13

The inclusiveness of DUREL, along with its brevity, makes it a valuable measure to be used in large-scale surveys, measuring three core dimensions of religiosity with five questions, as compared to others that measure a greater number of dimensions, but with many more questions.11, 14-16

The process of scoring the DUREL is relevant for analysis and for interpretation. Initially, the developers of the DUREL advised creating a continuous total score that ranged from 5 (lowest possible level of religiosity) to 27 (highest possible level of religiosity), in addition to considering the three subscale scores.13 However, due to recent work, the developers of the

DUREL advise against creating a total summary score by adding up the three subscales. They recommend examination of each subscale score independently in separate regression models to examine relationships to health outcomes.11 They argue that, if subscales are taken as a single model, multiple collinearity between them may cause a misinterpretation of the effects accompanying each subscale. Moreover, the subscales could cancel the effects of one another if combined into a single score or analysis.11 In 2000, Sherman and colleagues similarly recommended that the differing relationships of the three subscales of the DUREL with health outcomes justify examining the subscale scores separately rather than examining solely total scores, though they did not ultimately recommend against creating a summary score.17

The DUREL was developed on a model of three factors, organizational religiosity (OR), non- organizational religiosity (NOR), and intrinsic religiosity (IR), that theoretically comprise one

27 overarching latent variable of “religiosity.” OR is derived from responses to Item One of the

DUREL, NOR from responses to Item Two, and the IR from Items Three, Four and Five. A one- factor model has been demonstrated to fit data reliably for several demographic groups. We continued the investigation of the psychometric properties of the DUREL by testing and comparing a theoretical model postulating a simple single factor, “religiosity” (Figure 2.2), to one postulating three correlated factors (OR, NOR, IR) underlying one larger factor (religiosity)

(Figure 2.3).

An objective of Aim 1 is to assess the factorial validity of the DUREL through psychometric testing. Though the DUREL has been used previously in Zambia as a measure of religiosity, it has not been validated as a scale for use among women of reproductive age.18, 19

Individual-Level Covariates

Education

Level of education is measured as a categorical variable based on the highest level of school attended. Level of education ranges from no education to university degree and beyond. The four categories are: no education, primary school, secondary school, and post-secondary schooling

(including university, professional training programs, etc.).

Denomination

Denomination is assessed via each woman’s response to the open-ended question, “What is your religion?” Response options for this question were Catholic, Protestant, Muslim, and Other.

Distributions are seen Table 2.5.

Those respondents who stated their religion as Muslim (5) and Other (3) are excluded from the analytic sample, as they represent 0.49% (n=7) of total respondents, which is an insufficient sample size for sub-analyses.

Age

Age was measured as a continuous variable, based on respondents stated age. The survey was limited to women of reproductive age, as defined, in this instance, as ranging from 15-44. In cases of non-response, age was captured based on the individual’s self-reported year of birth. In

Aim 2 analyses, age functions as a categorical variable with the following designations: adolescents (15-19), youth (20-24), younger women of reproductive age (25-34), and older women of reproductive age (35-44).

Gravidity

Gravidity is defined as the number of pregnancies a woman has in her lifetime.20 When used in conjunction with parity, it can depict a woman’s reproductive history. Gravidity is measured as a continuous variable via the totaling from a series of questions regarding the number of living children, deceased children, stillbirths, spontaneous miscarriages, and induced abortions a woman has experienced in her lifetime. For this analysis, gravidity is trichotomized as women who have never been pregnancy, women who have had 1 – 3 pregnancies and women who have had 4 or more pregnancies. Gravidity can be a more specific measure than parity in study populations that include adolescents, youth, and unmarried women.21, 22

Marital Status

Women were asked about their marital history and dichotomized into two groups- ever married / cohabited and never married. The ever married / cohabited group included women who were widowed or divorced. The never married group consisted of women who had never been married or cohabited with a male partner.

Household-Level Covariates

Wealth

Each woman’s household economic status was measured using a the five-quintile DHS Wealth

Index, which has been used extensively and effectively in a myriad of contexts and settings, including Zambia.23-25 The index is a composite measure of a household's cumulative living standard. It is calculated from a principal components factor analysis of household assets, such as televisions and bicycles; materials used for housing construction; and types of water access and sanitation facilities.26 Assets that are more unequally distributed across households are given more weight than assets more widely available in a population.27

Respondents to the survey used in this study were not asked the complete panel of wealth-related questions used in the DHS. Therefore, all the variables were recoded as binary responses (have the asset or do not) and a PCA was conducted on all household wealth questions. We assumed the first principal component to be a measure of economic status.28 After removing variables not contributing to the overall wealth score, a second PCA was conducted and the resulting factor score was then divided into quintiles to represent the poorest to the richest groups.26, 29

Community-Level Covariates

Urban / Rural

Whether the enumeration area in which the respondent resides is urban or rural will be determined by the SEAs, which were stratified separately by urban and rural areas within each province.8 An urban area in Zambia is defined as a locality of 5,000 inhabitants or more, in which a majority of the labor force not in agricultural activities.30

Community Religiosity and Abortion Perceptions

Community-level religiosity and abortion will be measured by aggregating individual females’ responses within each SEA. Specifically, community-level religiosity will be measured via the enumeration-area average of the religiosity scores assessed by the DUREL, knowledge and support of abortion law will be measured via the enumeration-area average of the abortion knowledge and support scales, and abortion attitudes will be assessed by the enumeration-area average of the abortion attitude scale.

The aggregate measures will be calculated separately for each woman excluding her own responses; hence, each community-level measure is constructed from the responses of up to 29 women within the same enumeration area. Previous reproductive health studies in sub-Saharan

Africa have used SEAs, which are statistical constructs, as a proxy for an individual respondent’s true community.23, 31, 32

Each community-level variable was trichotomized to depict low, medium, and high levels of each dimension of religiosity and abortion attitudes within the community.

Dependent Variables

Through scale development, measures were constructed to assess individual abortion perceptions as an objective of Aim 1a, using questions regarding personal abortion attitudes and support for abortion law in Zambia. Abortion attitudes refer to the manner in which individuals perceive the morality or correctness of abortion. These attitudes can be expressed both publicly and privately.

People experience a wide variety of abortion attitudes and where they might be in favor of abortions in certain situations, they may be opposed to it in others. Previous studies have examined the underlying factor structure of abortion stigma, but a validated measure for abortion perceptions has not been specifically developed.33 The 5 self-reported items were analyzed individually in two previous studies, but not as a scale or index.10, 34

Personal Abortion Attitudes

The five questionnaire items about abortion attitudes focused on the morality of abortion, women’s rights to abortion, and access to abortion for both married and unmarried women

(Table 2.6). Each of these items had a 5-point response set ranging from “strongly disagree” to

“strongly agree.” Though these five elements do not exhaust the dimensions of abortion attitudes, they are adapted from previous studies used to examine abortion attitudes in the

Zambian context.10, 34

The agreement with the statement “abortion is immoral” was recoded so that the directionality matched the other four questions (lower scores equating with more conservative attitudes).

Response distributions can be seen in Figure 2.4. “Abortion is immoral” was dropped from analysis due to its notable skew; only 3% of respondents disagreed with the statement.

Support for Abortion Law

Support for abortion laws (abortion legality) will be measured using yes/no responses to eight possible legal allowances for abortion that the respondent believes should be legal in Zambia

(Table 2.7). These allowances have been used in previous studies examining knowledge of abortion laws.10, 34-37

Table 2.8 shows respondent agreement with each specific legal allowance for abortion. Table 2.9 shows the various response patterns to personal support for specific legal allowances for abortion. There were 69 unique patterns of favored legal allowances. The most common pattern was the desire for the illegality of abortion in all instances (n=721, 50.4%) followed by those who believed that abortion should be legal in all instances (n=295, 20.6%). The next most common pattern was the desire for abortion to be legal in the case of three allowances: when pregnancy threatens the life of the woman, when there is a health risk to the woman, and when there is a health risk to the child (n=150, 10.5%). Based on response patterns and the overall distribution of responses (Figure 2.5), this outcome is dichotomized in analysis with 3 or more supported allowances for the legality of abortion as the cutoff.

Treatment of Missing Data

Of the 1531 households sampled, there were 34 refusals (2.2%). There is no information known about women who refused to participate other than their location. Chi-square tests found no

33 significant difference in refusal rates between urban and rural dwellers (p=0.56) nor province of residence (p=0.51).

A conservative approach was taken to missing data. Respondents were excluded if they did not respond to any of the five DUREL questions (n=3), any of the five abortion attitude questions

(n=18), or any element of the opinion toward abortion law questions (n=36) (Table 2.10).

Including the non-respondents, a total of 5.9% of those surveyed were missing key information to the study objectives and were excluded for analysis. Aim 2 analysis excludes an additional two women who were the only respondents sampled in their SEA, rendering the concept of community meaningless.

Patterns of missingness for the series of outcome variables can be seen in Table 2.11. Missing data are well-distributed across questions; thus, data were determined to be completely missing at random. Chi-square tests were used to compare overall missingness across key covariates and found no significant differences between those excluded from the analysis due to missing data and those included.

Analytic Methods (Aims 1 & 2)

Aim 1A sought to assess the reliability and validity of the 5-item Duke University Religion

Index. Aim 1B constructed measures of personal attitudes toward abortion and support for

Zambian abortion laws. Aim 2A assessed the relationship between individuals’ religiosity and their overall abortion perceptions. Aim 2B evaluated the relationships between the religiosity and abortion perceptions of an individual’s residential community and her own abortion perceptions.

Analytic Methods: Aim 1a

Aim 1a assessed the validity and reliability of the DUREL. For initial descriptive purposes, a summary religiosity score was created and dichotomized at the median value (<24 considered low religiosity and ≥24 considered high religiosity). Differences in religiosity by key demographic variables were examined using two-tailed chi-squared tests for binary variables and one-way ANOVA with pairwise comparison of means for categorical data exceeding two groups; significance was set at α<0.05. Descriptive analyses were conducted in STATA Version

14.238

After conducting descriptive analyses, Cronbach’s α reliability coefficient was calculated to assess the internal consistency of the DUREL for the total score (five-item measure of religiosity) and the IR score.39 The OR and NOR DUREL items have an item response range of one to six, whereas the three IR items have a range of one to five. Because of the mixed ranges of responses, the items are standardized to have a mean of 0 and a variance of 1. Values of

Cronbach’s α greater than 0.7 were considered acceptable. Due to the ordinal nature of the

DUREL responses, principal components analysis (PCA) was conducted on a polychoric correlation matrix to assess construct validity of the DUREL. PCA was performed using Stata

14.2.38 The criteria for factor assessment included: Eigen values greater than 1.0, percent variance explained, and parallel analysis.40

To test the validity of the DUREL as a one-factor model, a confirmatory factor analysis (CFA) was conducted. CFA is used to confirm a pattern of relationships predicted on the basis of theory

35 or previous analytic results.40 In this model, five observed variables, which are the responses to the five items of the DUREL, contribute directly to the latent variable of religiosity (Figure 2.2).

CFA was then conduced on items Three, Four, and Five to test the validity of the IR subscale.

Next, the psychometric properties of the DUREL were tested via CFA as a second-order single factor model, with three correlated factors (OR, NOR, IR) underlying the single factor of religiosity (Figure 2.3).

To evaluate the one-factor model and the second-order single factor model, a weighted least squares mean and variance (WLSMV) estimator was used, as recommended for categorical indicators whose distribution is not multivariate normal.41 Adequacy of model fit was assessed using the root-mean-square error of approximation (RMSEA), with values <0.10 as indicating acceptable fit.42, 43 In addition, the Tucker-Lewis Index (TLI) and Comparative Fit Index (CFI) values >0.95 were used to denote good fit and >0.90 indicating that the model is plausible.44-46

The pattern of factor loadings were considered in addition to goodness of fit statistics. If individual items or subscales had high levels of uniqueness (>0.50) or if they did not load highly on one factor (<0.40) they may be better suited as independent scales (OR, NOR, and IR). All

EFA and CFA tests were conducted using MPlus Version 8.41

Analytic Methods: Aim 1b

Aim 1b sought to construct a measure of abortion perceptions using questions regarding personal attitudes toward abortion and support for Zambian abortion laws. For initial descriptive purposes, a summary score was created for the five abortion attitude questions and the eight questions regarding the desired permissibility of Zambian abortion law. The summary scores were

36 dichotomized at the median value and differences by key demographic variables were examined using two-tailed chi-squared tests for binary variables and one-way ANOVA with pairwise comparison of means for categorical data exceeding two groups; significance was set at α<0.05.

An inductive approach was taken to determine how to conceptualize abortion perceptions based on the data. Thus, a principle components analysis (PCA) on a polychoric correlation matrix was followed by an exploratory factor analysis (EFA). The criteria for factor assessment included:

Eigen values greater than 1.0, percent variance explained, and parallel analysis.40 EFA was then conducted using a promax rotation and a maximum likelihood estimator. Items were considered for elimination if they had high levels of uniqueness (>0.50) or if they did not load highly on one factor (<0.40).44

Once items were eliminated, PCA and EFA were re-conducted to ensure satisfactory uniqueness values and factor loadings. Next, a confirmatory factor analysis (CFA) was conducted. A weighted least squares mean and variance (WLSMV) estimator was used, as recommended for categorical indicators whose distribution is not multivariate normal.41 Adequacy of model fit was assessed using the root-mean-square error of approximation (RMSEA), with values <0.10 as indicating acceptable fit.42, 43 In addition, the Tucker-Lewis Index (TLI) and Comparative Fit

Index (CFI) values >0.95 were used to denote good fit and >0.90 indicating that the model is plausible.44-46 The pattern of factor loadings were considered in addition to goodness of fit statistics. All EFA and CFA tests were conducted using MPlus Version 8.41 Cronbach’s alpha reliability coefficient was used to assess the scale’s internal consistency.

Analytic Methods: Aim 2

Aim 2 investigated the individual and community factors that influence individuals’ abortion perceptions. Aim 2A examines the relationship between individuals’ personal religiosity and their abortion perceptions, while aim 2B assesses the relationships between the religiosity of an individual’s residential community and her own abortion perceptions.

To understand the relationship between religiosity and abortion perceptions, the analysis uses four standalone questions pertaining to abortion attitudes and one question pertaining to perceptions of abortion legality. The analytic framework is shown in Figure 2.6. All analyses were conducted using Stata Version 14.2 statistical software to adjust for complex survey design effects and sampling weights.38

Initial bivariate analyses were conducted to examine the unadjusted association between the independent variables specified previously and the dependent variables (abortion perceptions).

Tests of means across subgroups were conducted to ensure there are sufficient numbers of observations in each category.

To assess the relationship between religiosity and abortion perceptions, block modeling was used. Block modeling is an approach whereby groups of variables are added in stages based on patterns occurring in the relationships between them. In this case, multivariate regression models were estimated using this step-by-step process according to the clustering indicated in the analytic framework (Figure 2.6): individual, household, and community variables.

To test Hypothesis 2A, a logistic regression model was estimated (Equation 1), in which the unadjusted association (훽1) between individual religiosity (푋1) and each of the abortion perception scales developed in aim 1 (푌푖) will be estimated. Equation 2 tests the same association while controlling for individual and household variables, 푋푖, with i=2, …, M, where M is the individual and household control variables and 휀푖 is the individual error term.

푌푖 = 훽0 + 훽1푋1 + 휀푖 (Equation 1)

푌푖 = 훽0 + 훽1푋1 + ⋯ + 휷풎푿풎 + 휀푖 (Equation 2)

Sub-aim 2B examined the effect of community religiosity and abortion perceptions on individual abortion perceptions. To test hypotheses 2B, the effect of community religiosity and community abortion perceptions on individual abortion perceptions, a random intercepts multilevel model was employed (Equation 3). A random intercepts model is a model in which slopes are fixed, but intercepts are allowed to vary. Therefore, the scores on the dependent variable (abortion perceptions) for each individual observation are predicted by the intercept that varies across groups (SEAs). In the model, there are J clusters with a different number of women, nj, in each cluster. Community religiosity at a cluster level (푊1) is the key independent variable. The outcome is individual abortion perceptions (푌푖푗).

푌푖푗 = 훽0푗 + 훽1푗푋1푖푗 + 휷풎풋푿풎풊풋 + 훾1푊1푗 + 휸풌푾풌풋 + 휇푘 + 휀푖푗 (Equation 3)

Qualitative Study Design

Secondary data analysis was performed on semi-structured in-depth interviews that were collected in late 2014 by the Guttmacher Institute, as a component of the evaluation of the

Preventing Unwanted Maternal Deaths from Unsafe Pregnancies (PUMDUP) program.

PUMDUP launched in 2011 with the goal of preventing maternal death and unsafe abortion through the augmentation and amelioration of abortion service outlets and trained providers in

Zambia, as well as thirteen other low-income countries. The in-depth interviews were conducted with women accessing termination of pregnancy (TOP) and post-abortion care (PAC) for induced abortions in public hospitals and private clinics. A convenience sample was used; women who were at the facility for abortion services on the days of the interviews were invited to participate. A longitudinal design was employed such that each respondent was enrolled initially at the point of care (T1) and was followed up 3-4 months later for a second interview

(T2). There was a 70% follow-up rate for the T2 interview overall, with 63% of PAC clients and

78% of TOP clients responding to a follow-up interview (Table 2.12). All interviews were completely confidential and respondents’ names were never linked to their audio recordings or transcripts. Data collection was conducted with ethical approval from the Guttmacher Institute’s

IRB. This study did not qualify as human subjects research as defined by the Department of

Health and Human Services regulations 45 CFR 46.102 by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board (IRB) Office, thus did not require IRB oversight.6

(See Appendix C for IRB determination).

The interviewers were women of reproductive age themselves with at least a basic background in health. All interviewers had previous experience conducting in-depth interviews and received a

40 practical three-day training that involved a detailed overview of the interview guide and skill- building in confidentiality, probing for further information or elaboration, and remaining non- judgmental regardless of personal opinions.

The in-depth interviews explored the respondent’s abortion experience, reason(s) for having the abortion, financial and social costs of the abortion, the effect of the abortion on personal relationships, role of religion in abortion choice, mental health / outlook on life, and the effect of abortion laws in the decision to terminate the pregnancy. The interview guide remained the same for T1 and T2 interviews in order to observe changes in responses and themes over time (see

Appendix A).

Interviews were conducted in the preferred language of the respondent (English, Bemba, Nyanja, and Tonga). Interviewers probed for further information when respondents stated something that was unclear or did not elaborate on a response. The average interview length was 60 minutes.

The taped interviews were transcribed by hand, first into the language in which they took place and then translated into English. The English translations were reviewed for accuracy and completeness by the study coordinator. It is to be noted that this process is susceptible to the introduction of ambiguity and sometimes error either through incorrect transcription or through inaccurate translation (because the word or concept does not exist in English, the intended meaning in the local language was not clear or because the translator chose the wrong word when a more suitable word exists). This weakness is intrinsic to any data collection done in a language different than the language of analysis.47 A Guttmacher Institute staff member input the

41 text into NVivo Version 11, software that supports qualitative and mixed methods research, such as interviews, focus group discussions, open-ended survey responses, and articles.48

Though the data were originally collected to better understand the social, health, and economic consequences of safe abortion as compared to unsafe abortion, this study analyses these data with a broader scope which includes religion and stigma. Secondary data analysis is less common in qualitative research as compared to quantitative research. However, secondary analysis of qualitative research has been shown to be useful when undertaking a sensitive area of research or accessing a hard to reach population.49 This study involves speaking with women about their recent abortion, thus is both a sensitive topic area and a sometimes elusive population.

Analytic Methods (Aim 3)

To analyze the transcripts, a directed content analysis approach was applied. Unlike the more traditional and inductive grounded theory approach in which the theories are constructed from the data itself, in directed content analysis initial coding allows themes to emerge from the data.50, 51 The goal of coding was to organize it into categories that allow the comparison of data within and between these categories and thus help to develop theoretical concepts.52 A preliminary codebook was developed using an iterative coding process. Key themes were identified using line-by-line open coding within NVivo Version 11. Theoretical memo writing was employed throughout, connecting concepts and their potential relationships to additional concepts.

The transcripts were first read by two researchers to identify overarching themes for the creation of draft codebooks. Following the steps put forth by Zhang & Wildemuth,53 the two researchers worked in tandem to develop the final codebook. The codebook was organized by general themes and sub-themes. Eight transcripts were coded separately by the two researchers using the draft codebooks and the results were discussed to revise the codebooks, adding, deleting or collapsing codes as necessary.

Once codebooks were finalized, coding was performed in NVivo independently by two researchers; selected transcripts were coded by a third researcher to ensure reliability and validity of the coding. Researcher triangulation is a process by which a second researcher with experience in inductive qualitative coding will conduct a separate analysis of the data. In addition to improving precision, researcher triangulation creates a broader and more complex understanding of the phenomenon being studied.54 The consistency of coding was assessed by inter-coder reliability, calculated as the number of agreements divided by the total number of agreements and disagreements. Disagreements were discussed and resolved until the inter-rater agreement was 90% or greater. Finally, the data were analyzed and presented using the words of the respondents.

For sub-aim 3A, Time 1 and Time 2 interviews were aggregated in order to examine the interconnectedness of religion and abortion stigma. This qualitative aim allows for a better understanding of the interplay between religiosity and stigma as felt by Zambian women who have recently experienced an abortion.

The analysis of sub-aim 3B took advantage of the longitudinal nature of the study design and examined changes in the role of religion and stigma in a woman’s abortion experience trajectory.

Women were asked about both religion and community reactions to their abortion experience at the first interview, immediately after the abortion, and again at the follow-up interview several months later; thus, the evolution of their feelings around these notions can be observed. Using transition codes from the codebook, Time 1 interviews were compared to Time 2 interviews and the differences in themes at the two time points examined.

The qualitative analysis of Aim 3 elaborates upon and provides greater context to the findings from Aim 2. The in-depth nature of the interviews combined with the longitudinal structure provided context on the impact of stigma and religiosity beyond one quantitative point in time.

The themes analyzed in Aims 2 and 3 are consistent; however, the in-depth interviews come exclusively from women who have had abortions, which provided specific insights into how abortion stigma is felt. Further, unlike the survey data, changes in the dynamic between religiosity and stigma and their overall impact can be observed over time.

Chapter Two Tables and Figures

Figure 2.1. Map of Zambia by provinces and districts

Table 2.1. Standard enumeration areas (SEAs) sampled per district, by province Level of Total SEAs Province District Intensity sampled Chibombo Low 2 Kapiri Mposhi Low 2 Mkushi Low 2 Central Mumbwa Low 2 Serenje Low 2 Kabwe Medium 3 Lufwanyama None 2 Masaiti None 2 Mpongwe None 2 Chililabombwe Low 3 Chingola Low 3 Copperbelt Kalulushi Low 3 Luanshya Low 3 Mufulira Low 3 Kitwe Medium 3 Ndola Medium 3 Luangwa None 2 Chongwe Low 2 Lusaka Kafue Medium 3 Lusaka High 4

Table 2.2. Response Rate by Province Result of Visit Province Successful Response rate Refusal interview (%) Central 383 7 98.3% Copperbelt 793 17 98.0% Lusaka 321 10 97.0% Total 1497 34 97.7%

Table 2.3. Power Calculation Sum of Abortion Attitude Questions (1-5 scale; min = 0, max = 25) 20% Sample Alpha Power Mean Standard Total sample Increase required DEFF (α) (1-β) Score Deviation size required (n1 + n2) 0.01 106 737 0.05 0.8 12.70 3.79 15.24 70 6.95 487 0.10 56 390 0.01 134 932 0.05 0.9 12.70 3.79 15.24 94 6.95 654 0.10 78 702

Table 2.4. DUREL Subscales, item questions, and responses Subscale Item Response Never Once a year or less Item 1: How often do you attend church or other A few times a year OR religious meetings or spiritual gatherings? A few times a month Once a week More than once/week Rarely or never A few times a month Item 2: How often do you spend time alone in Once a week NOR religious activities, such as prayer, meditation or Two or more times / week Bible/Koranic study? Daily More than once a day Item 3: In my life, I experience the presence of the Divine (i.e., God, Allah) Definitely not true Item 4: My religious beliefs are what really lie Tends not to be true behind my whole approach to life. (i.e. my religion IR Unsure is the foundation for how I live my life) Tends to be true Item 5: I try hard to carry my religion over to all Definitely true of me other dealings in life (i.e. my work, socializing with friends and neighbours) OR = Organizational Religiosity NOR = Non-Organizational Religiosity IR = Intrinsic Religiosity

Table 2.5. Religious Denomination Distributions Among Respondents Religion n (%) Catholic 248 (16.6) Protestant 1241 (82.9) Muslim 5 (0.3) Other 3 (0.2)

Table 2.6. Five abortion attitude item questions and responses QUESTIONS RESPONSE “Abortion is immoral” “A woman has the right to decide whether or not to continue a pregnancy.” Strongly Disagree 1 “A woman has the right to decide whether or not to Disagree 2 terminate a pregnancy.” Neutral 3 “Women should have access to safe abortion Agree 4 services.” Strongly Agree 5 “Unmarried women should have access to safe abortion services.”

Figure 2.2. Single factor model of religiosity (DUREL)

Figure 2.3. Second-order single-factor model of the DUREL

Figure 2.4 Distribution of respondent’s abortion attitude responses, by item

Table 2.7. Support for abortion laws, items and responses QUESTIONS RESPONSE Do you think abortion currently should be legal in Zambia for each of the following reasons:

For any reason? (Regardless of reason) NO 0 YES (skip to end) 1 If a woman does not want another child? NO 0 YES 1 If a woman cannot afford another child? NO 0 YES 1 If the pregnancy was a consequence of incest? NO 0 YES 1 If the pregnancy was a consequence of rape? NO 0 YES 1 If the pregnancy was a consequence of child abuse? NO 0 YES 1 If there is a health risk to the child? NO 0 YES 1 If there is a health risk to the woman? NO 0 YES 1 If the pregnancy threatens the life of the woman? NO 0 YES 1

Table 2.8. Distribution of respondents’ agreement with eight legal allowances for abortion Response Yes No n (%) n (%)

Woman does not want another child 329 (22.8%) 1103 (77.0%) Woman cannot afford another child 344 (24.0%) 1088 (76.0%) Pregnancy was a consequence of incest 375 (26.2%) 1057 (73.8%) Pregnancy was a consequence of rape 385 (26.9%) 1047 (73.1%) Pregnancy was a consequence of child abuse 392 (27.4%) 1040 (72.6%) There is a health risk to the child 536 (37.4%) 896 (62.6%) There is a health risk to the woman 606 (42.3%) 826 (57.7%) Pregnancy threatens the life of the woman 648 (45.2%) 784 (54.8%)

*Respondents indicating that abortion should be legal for any reason are included in the “Yes” category for each case

Table 2.9. Response patterns* for legal allowances of abortion Pregnancy Woman Woman Pregnancy There is a There is a was a Pregnancy Pregnancy cannot does not threatens health risk health risk consequence was a was a afford want the life of to the to the of child consequence consequence another another the woman woman child abuse of rape of incest child child Frequency Percent no no no no no no no no 721 50.42 yes yes yes yes yes yes yes yes 295 20.63 yes yes yes no no no no no 150 10.49 yes yes no no no no no no 49 3.43 yes no no no no no no no 43 3.01 yes yes yes yes yes yes no no 19 1.33 yes yes yes yes no no no no 10 0.7 no no no no no yes no no 9 0.63 no no no no yes no no no 8 0.56 yes yes yes yes yes yes yes no 7 0.49 no no no yes yes yes no no 5 0.35 yes yes yes no no yes no no 5 0.35 yes yes yes no yes no no no 5 0.35 yes yes yes yes yes no no no 5 0.35 no no no yes no no no no 4 0.28 no no no yes yes no no no 4 0.28 no yes no no no no no no 4 0.28 yes yes no no no yes no no 4 0.28 yes yes no no yes no no no 4 0.28 yes yes yes yes no yes no no 4 0.28 no no no no no no yes yes 3 0.21 no no no no yes yes no no 3 0.21 yes yes yes no no no yes no 3 0.21 yes yes yes yes yes no yes yes 3 0.21 no no no no no no no yes 2 0.14 no no no no no no yes no 2 0.14 no no no yes no yes no no 2 0.14 no no no yes yes yes yes yes 2 0.14 no no yes no no no no no 2 0.14 no no yes yes yes no no no 2 0.14 yes no no no no no no yes 2 0.14 yes yes no no no no yes no 2 0.14 yes yes no no no no yes yes 2 0.14 yes yes no yes yes no yes yes 2 0.14 yes yes no yes yes yes no no 2 0.14 yes yes no yes yes yes yes no 2 0.14 yes yes yes no no no yes yes 2 0.14 yes yes yes no yes no yes yes 2 0.14 yes yes yes yes no no no yes 2 0.14 yes yes yes yes no no yes yes 2 0.14 yes yes yes yes yes no yes no 2 0.14 Yes yes yes yes yes yes no yes 2 0.14 *There are 27 additional response patterns that were each unique to a single respondent (n=1, 0.07%)

Figure 2.5 Distribution of respondent’s number of favored legal allowances for abortion

Table 2.10. Flow diagram of missing data n N Total households visited · 1531 Refusal to participate in survey 34 1497 Identified religion as Muslim (n=5) or Other (n=3) 8 1489 Missing response to at least one DUREL question 3 1486 No response to at least one abortion attitude question 18 1468 No response to at least one abortion legality question 36 1432 Total analytic sample 1432

Table 2.11. Patterns of missingness for outcome variables

Abortion Attitudes Perception of Abortion Law

Continue pregnancy Terminate pregnancy toAccess abortion Unmarried access Doesnot want child Cannotafford child N Missing

Abortion immoral Respondents

Incest Rape Child abuse Health risk to child Health risk to woman Threatens life woman + + + + + + + + + + + + + 0 1432 + + + + + + + o + + + + + 1 6 + + + + + + + + + + o + + 1 5 + + + + o + + + + + + + + 1 4 + + + o + + + + + + + + + 1 4 o + + + + + + + + + + + + 1 3 + + + o o + + + + + + + + 2 3 + + + + + + + + o + + + + 1 2 + + + + + + + + + + + o o 2 2 + + + + + o o + + + + + + 2 2 + + + + + + + + + + o o o 3 2 + + + + + + + + + o + + + 1 1 + + + + + o + + + + + + + 1 1 + + o + + + + + + + + + + 1 1 + + + + + + + + + + o + o 2 1 + + + + + + + + + + o o + 2 1 + + + + + + + + + o o + + 2 1 + + + + + + + o + + + + o 2 1 + + + + + + + o + + o + + 2 1 + + + + + + + o + o + + + 2 1 + + + + + + + o o + + + + 2 1 + + + + + + o o + + + + + 2 1 + + + + + o + + + + o + + 2 1 + + + + + o + + + o + + + 2 1 + + + + o + + o + + + + + 2 1 + + + + + + + o o o + + + 3 1 + + + + + + o o o + + + + 3 1 + + + + + + + + o + o o o 4 1 + + + + + + + o o + o o o 5 1 + + + + + o o o + + o o o 6 1 + + + + o o o o o o o o o 9 1 o o o o o o o o o o o o o 13 1

Figure 2.6. Aim 2 Analytic Framework

Table 2.12. In-Depth Interviews by Type of Evacuation and Interview Completion Time 1 Time 2 N (%) N (%) Termination of Pregnancy (TOP) Manual Vacuum Aspiration (MVA) 9 (17.6%) 5 (13.2%) Medical Abortion (MA) 17 (33.3%) 16 (42.1%) Post-Abortion Care (PAC) Low severity 13 (25.5%) 9 (23.7%) Moderate/severe 12 (23.5%) 8 (21.1%) TOTAL 51 (100.0%) 38 (100.0%)

Chapter Two References

1. Jones, R.K. and K. Kost, Underreporting of induced and spontaneous abortion in the United States: an analysis of the 2002 National Survey of Family Growth. Stud Fam Plann, 2007. 38(3): p. 187-97. 2. Jagannathan, R., Relying on Surveys to Understand Abortion Behavior: Some Cautionary Evidence. American Journal of Public Health, 2001. 91(11): p. 1825-1831. 3. Lara, D., et al., Measuring induced abortion in Mexico: a comparison of four methodologies. Sociological Methods & Research, 2004. 32(4): p. 529-558. 4. Philipov, D., et al., Induced abortion in Russia: recent trends and underreporting in surveys. European Journal of Population/Revue européenne de Démographie, 2004. 20(2): p. 95-117. 5. Abowitz, D. and T.M. Toole, Mixed Method Research: Fundamental Issues of Design, Validity, and Reliability in Construction Research. 2010. 6. US Department of Health and Human Services, Title 45 Code of Federal Regulations Part 46, Protection of Human Subjects. Subpart A: Federal Policy for the Protection of Human Subjects DHHS Policy for Protection of Human Research Subjects. 56 FR 28003. 1991, June. 7. Central Statistical Office [Zambia], National Descriptive Tables, in 2010 Census of Population and Housing. 2012: Lusaka. 8. Central Statistical Office (CSO) [Zambia], M.o.H.M.Z., and ICF International. , Zambia Demographic and Health Survey 2013-14. 2014, Central Statistical Office, Ministry of Health, and ICF International: Rockville, Maryland, USA. 9. Kish, L., A procedure for objective respondent selection within the household. Journal of the American statistical Association, 1949. 44(247): p. 380-387. 10. Geary, C.W., et al., Attitudes toward abortion in Zambia. International Journal of Gynecology & Obstetrics, 2012. 118: p. S148-S151. 11. Koenig, H.G. and A. Büssing, The Duke University Religion Index (DUREL): A Five- Item Measure for Use in Epidemological Studies. Religions, 2010. 1(1): p. 78-85. 12. Ellis, P.D., The essential guide to effect sizes: Statistical power, meta-analysis, and the interpretation of research results. 2010: Cambridge University Press. 13. Koenig, H., G.R. Parkerson Jr, and K.G. Meador, Religion index for psychiatric research. 1997. 14. Pearce, L.D., G.M. Hayward, and J.A. Pearlman, Measuring Five Dimensions of Religiosity across Adolescence. Review of religious research, 2017. 59(3): p. 367-393. 15. Hoge, R., A validated intrinsic religious motivation scale. Journal for the scientific study of religion, 1972: p. 369-376. 16. Allport, G.W. and J.M. Ross, Personal religious orientation and prejudice. Journal of personality and social psychology, 1967. 5(4): p. 432.

17. Sherman, A., et al., A multidimensional measure of religious involvement for cancer patients: The Duke Religious Index. Supportive Care in Cancer, 2000. 8(2): p. 102-109. 18. Pandya, S.P., Effect of a spiritual education programme in developing altruism and prosocial behaviour among children. International Journal of Children's Spirituality, 2017. 22(3-4): p. 220-238. 19. Peltzer, K. and S. Pengpid, Correlates of illicit drug use among university students in Africa and the Caribbean. Journal of Psychology in Africa, 2016. 26(4): p. 390-393. 20. Creinin, M.D. and H.N. Simhan, Can we communicate gravidity and parity better? Obstetrics & Gynecology, 2009. 113(3): p. 709-711. 21. Johnson-Hanks, J., The lesser shame: abortion among educated women in southern Cameroon. Social Science & Medicine, 2002. 55(8): p. 1337-1349. 22. Dahlback, E., et al., Pregnancy loss: spontaneous and induced abortions among young women in Lusaka, Zambia. Cult Health Sex, 2010. 12(3): p. 247-62. 23. Elfstrom, K.M. and R. Stephenson, The Role of Place in Shaping Contraceptive Use among Women in Africa. PLOS ONE, 2012. 7(7): p. e40670. 24. Murray, N., et al., Factors related to induced abortion among young women in Edo State, Nigeria. Studies in Family Planning, 2006. 37(4): p. 251-268. 25. Mutombo, N. and P. Bakibinga, The effect of joint contraceptive decisions on the use of Injectables, Long-Acting and Permanent Methods (ILAPMs) among married female (15– 49) contraceptive users in Zambia: a cross-sectional study. Reproductive health, 2014. 11(1): p. 51. 26. DHS Program. Wealth Index. Available from: http://dhsprogram.com/topics/wealth- index/. 27. McKenzie, D., Measure inequality with asset indicators. Bureau for Research and Economic Analysis of Development, Centre for International Development, BREAD, 2003. 28. Houweling, T.A., A.E. Kunst, and J.P. Mackenbach, Measuring health inequality among children in developing countries: does the choice of the indicator of economic status matter? International journal for equity in health, 2003. 2(1): p. 8. 29. Vyas, S. and L. Kumaranayake, Constructing socio-economic status indices: how to use principal components analysis. Health Policy Plan, 2006. 21(6): p. 459-68. 30. World Urbanization Prospects, the 2014 revision. 2014; Available from: https://esa.un.org/unpd/wup/DataSources/. 31. Entwisle, B., et al., Community and Contraceptive Choice in Rural Thailand: A Case Study of Nang Rong. Demography, 1996. 33(1): p. 1-11. 32. Stephenson, R., et al., Contextual influences on modern contraceptive use in sub-Saharan Africa. Am J Public Health, 2007. 97(7): p. 1233-40. 33. Shellenberg, K.M., L. Hessini, and B.A. Levandowski, Developing a Scale to Measure Stigmatizing Attitudes and Beliefs About Women Who Have Abortions: Results from Ghana and Zambia. Women & Health, 2014. 54(7): p. 599-616. 59

34. Cresswell, J.A., et al., Women's knowledge and attitudes surrounding abortion in Zambia: a cross-sectional survey across three provinces. BMJ Open, 2016. 6(3): p. e010076. 35. McMurtrie, S.M., et al., Public opinion about abortion-related stigma among Mexican Catholics and implications for unsafe abortion. International Journal of Gynecology & Obstetrics, 2012. 118: p. S160-S166. 36. Owolabi, O.O., et al., Incidence of abortion-related near-miss complications in Zambia: cross-sectional study in Central, Copperbelt and Lusaka Provinces. Contraception, 2016. 37. Izugbara, C.O., C. Egesa, and R. Okelo, 'High profile health facilities can add to your trouble': Women, stigma and un/safe abortion in Kenya. Soc Sci Med, 2015. 141: p. 9-18. 38. StataCorp, L., Stata Statistical Software: Release 14. College Station, TX: Stata-Corp LP 2015. 39. Santos, J.R.A., Cronbach’s alpha: A tool for assessing the reliability of scales. Journal of extension, 1999. 37(2): p. 1-5. 40. DeVellis, R.F., Scale development: Theory and applications. Vol. 26. 2012: Sage publications. 41. Muthén, L., Mplus User’s Guide, (Muthén & Muthén, Los Angeles). Mplus User's Guide,(Muthén & Muthén, Los Angeles), 2010. 42. Browen, M. and R. Cudeck, Alternative ways of assessing model fit in testing structural equation model. 1993, Newbury Park, CA: Sage. 43. Steiger, J.H., Structural model evaluation and modification: An interval estimation approach. Multivariate behavioral research, 1990. 25(2): p. 173-180. 44. Kim, J.-O. and C.W. Mueller, Factor analysis: Statistical methods and practical issues. Vol. 14. 1978: Sage. 45. Netemeyer, R.G., W.O. Bearden, and S. Sharma, Scaling procedures: Issues and applications. 2003: Sage Publications. 46. Bentler, P.M., Comparative fit indexes in structural models. Psychological bulletin, 1990. 107(2): p. 238. 47. Twinn, S., An exploratory study examining the influence of translation on the validity and reliability of qualitative data in nursing research. Journal of Advanced Nursing, 1997. 26(2): p. 418-423. 48. QSR International Pty Ltd, NVivo qualitative data analysis Software. 2015. 49. Long-Sutehall, T., M. Sque, and J. Addington-Hall, Secondary analysis of qualitative data: a valuable method for exploring sensitive issues with an elusive population? Journal of Research in Nursing, 2011. 16(4): p. 335-344. 50. Charmaz, K., Constructing grounded theory. 2014: Sage. 51. Hsieh, H.-F. and S.E. Shannon, Three approaches to qualitative content analysis. Qualitative health research, 2005. 15(9): p. 1277-1288.

52. Maxwell, J.A., Qualitative research design: An interpretative approach. 1996, Thousand Oaks, CA: Sage. 53. Zhang, Y. and B.M. Wildemuth, Qualitative Analysis of Content. Applications of Social Research Methods to Questions in Information and Library Science, 2016: p. 318. 54. Tong, A., P. Sainsbury, and J. Craig, Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups. International journal for quality in health care, 2007. 19(6): p. 349-357.

CHAPTER THREE: AIM 1 ANALYTIC RESULTS

This chapter presents the results for the Aim 1 analysis, which uses population-based survey data from women of reproductive age in three Zambian provinces. Recall that the goals of Aim 1 were to assess the validity and reliability of the Duke University Religion Index and construct a measure of personal abortion attitudes and support for the legality of abortion.

To better understand the relationship between religiosity and health indicators in Zambia and demographically similar sub-Saharan African countries, it is necessary for investigators to use well-defined models and validated measures. This study examined the psychometric properties of the DUREL among women of reproductive age in three Zambian provinces. The adapted questions were easily completed by the survey respondents. Based on current knowledge, this is the first time that the DUREL has been validated in Zambia and not used as a summary score.

The attitudinal questions have been used in previous studies as standalone questions but have not been examined in the literature as a scale.1, 2

The results of the Aim 1 analysis establish the outcome measures that will be subsequently used in for analyses in Aim 2.

Aim 1A Analytic Results: Validity and Reliability of DUREL

Participants were women of reproductive age representing three Zambian provinces. Their ages ranged from 15-44 years (mean age, 27.5 ±7.7) and 56.9% lived in urban areas. Responses to the

DUREL are presented in Table 3.1

The sample showed the full range of possible scores for each item, i.e. scores from 1-6 for Items

One (OR) and Two (NOR), and scores of 1-5 for Items Three, Four, and Five (IR). Respondents’ aggregated DUREL scores ranged from 5 to 27 (M = 23.22, SD = 2.80).

Cronbach’s alpha for the five-item measure of religiosity was calculated for the entire sample (N

=1432). Results indicated that sample had good internal consistency overall (α = 0.78). The

Cronbach’s alpha for the three-item IR subscale was also indicated strong internal consistency (α

= 0.89). The means, standard deviations, and alpha for each subscale of the DUREL are reported in Table 3.2.

The value of the Kaiser–Meyer–Olkin measure of sampling adequacy was 0.778. The Bartlett test of sphericity was statistically significant (χ2 = 2858.93; P <0.001). Together, these indicate a lack of correlation between variables such that the exploratory factor analysis could be performed. The results of the polychoric PCA, scree plot, and parallel analysis indicate that a

1‑factor solution fits the data. One factor was identified, explaining 61.89% of the variance with an eigenvalue of 3.09. The loadings of items 1 to 5 on this factor were 0.2511, 0.3444, 0.5273,

0.5269, and 0.5125, respectively.

The model fit indices suggested that the single factor model was a not a good fit to the data: degrees of freedom (df) = 5, chi-square (X2) = 100.92, RMSEA = 0.116, CFI = 0.995, and TLI =

0.989. All standardized factor loadings were significant at the level of p<0.001. However, standardized factor loading values for Item 1 (church attendance) was exceptionally low at 0.326 while factor values for Item 2 (NOR; individual prayer) was 0.461. The remaining standardized

64 factor loadings were within an acceptable range: 0.935 (Item Three; presence of God), 0.950

(Item Four; religion as approach to life), and 0.880 (Item Five; religion influencing dealings in life). The goodness of fit statistics and the factor loadings offer evidence that a one-factor model is inappropriate due to a high RMSEA and low factor loadings on Items 1 and 2 (the OR and

NOR subscales (Figure 3.1).

CFA was then conduced on items Three, Four, and Five to test the validity of the IR subscale.

The model fit indices reflect a saturated model and suggest that the three items load well onto one underlying factor, IR: degrees of freedom (df) = 3, chi-square (X2) = 25781.54, RMSEA

<0.001, CFI = 1.000, and TLI = 1.000. All standardized factor loadings were significant at the level of p<0.001 and greater than 0.5: Item 3 = 0.934, Item 4 = 0.951, and Item 5 = 0.881.

After ensuring the validity of the IR subscale, the second-order single factor model was specified. The model fit indices suggest that the model does not appropriately fit the data: degrees of freedom (df) = 5, chi-square (X2) = 425.073, RMSEA = 0.242, CFI = 0.977, and TLI

= 0.954.

The DUREL, as well as the IR subscale, showed good internal consistency similar to previous studies using the DUREL in various populations.3-5 However, the results of EFA and CFA suggest that neither a one factor model nor a second-order single factor model are appropriate to measure religiosity The subscales of the DUREL are not explained by a common underlying construct, as evidenced by goodness of fit statistics and the poor factor loadings of OR and NOR.

However, the findings do support the validity of IR as a latent construct for Items 3-5. This

65 supports the current recommendation of Koenig and Büssing that the subscales should be analyzed independently and not as an aggregate measure of religiosity.6

Though these findings align with the recommendations of the DUREL’s creators, the results differ from several previous validation studies of the DUREL. Several previous studies validated the DUREL as a single factor model.3-5, 7 Another study found validated the DUREL as a second- order single factor model for both English and Spanish-speaking study participants.8 For this reason, it is important to continue to assess the dimensionality of religiosity constructs within different populations to ensure applicability and accuracy.

Aim 1B Analytic Results

The objective of Aim 1B was to construct a measure of abortion perceptions using questions regarding personal attitudes toward abortion and legal permissibility of abortion. These questions were included in the survey in an ad hoc manner based on previous use in the baseline survey and a previous study of abortion in Zambia.

A preliminary factor analysis was conducted which included each of the 5 attitudinal questions and each of the 8 legal exceptions for abortion. Despite showing good internal consistency (α =

0.80), all attitudinal questions loaded onto a unique factor and all legal exceptions for abortion loaded onto a separate factor, with all loadings exceeding 0.40 for the corresponding factor yet below 0.40 for the other factor. This indicated that legal permissibility should be analyzed separately from abortion attitudes; thus, the subsequent analysis focuses specifically on abortion attitude questions.

For initial descriptive purposes, an additive summary score was created from the five abortion attitude questions. The five abortion attitude questions were on a 5-point Likert Scale, with 1 indicating the least permissive and 5 indicating the most permissive attitudes toward abortion.

Thus, total possible scores ranged from 5-25. Within the analytic sample, summary scores ranged from 5-23, with higher scores noting a higher level of support for abortion. The overall distribution is skewed slightly (Figure 3.3).

The summary scores were dichotomized at the median value of 12 and grouped into lower and higher levels of abortion support (lower support indicated by ≤12 and higher support indicated by >12). The mean was 12.59, thus the results are the same for both mean and median.

Differences by key demographic characteristics were examined using two-tailed chi-squared tests for binary variables and one-way ANOVA with pairwise comparison of means for categorical data exceeding two groups; significance was set at α<0.05.

The median summary score for combined abortion attitudes and law was 12 (IQR: 10-16). There were significant differences in abortion attitude summary scores by all key demographic characteristics, with the exception of religious denomination (p<0.05) (Table 3.3).

For the five statements of abortion attitudes, Cronbach’s alpha was calculated for the entire sample (N =1432). Results indicated that sample had reasonable internal consistency overall (α =

0.70).

The value of the Kaiser–Meyer–Olkin measure of sampling adequacy, which measures the sampling adequacy for each variable in the model and the complete model, was 0.654, indicating that sampling is mediocre (where 1.0 is ideal). The Bartlett test of sphericity, used to test for equal variances, was statistically significant (χ2 = 2027.86); P <0.001). Together, this indicates a lack of correlation between variables such that the exploratory factor analysis could be performed. The results of the polychoric PCA, scree plot, and parallel analysis indicate that a

1‑factor solution fits the data.

The results of the EFA conducted on the 5 abortion attitude items indicated a single-factor structure. Item 1, agreement with the statement “abortion is immoral,” exhibited a high uniqueness value (>0.50) and did not load highly (<0.40) onto the single factor. This was the sole question where a higher response `on the Likert scale correlated with reduced support of abortion. Further, existing literature suggests that within sub-Saharan African countries where the populations are predominantly Christian, even when there is legal and medical support for abortion, the morality of abortion is questioned.1, 9-12 Additionally, Items 2 and 3 did not meet the criteria for uniqueness below 0.50; however, they both had loadings exceeding 0.40, so remained in the analysis due to the limited number of overall items.

Item 1 was dropped and prior analyses were reconducted. Chronbach’s alpha increased to 0.78, still indicating good overall consistency. The Kaiser–Meyer–Olkin measure of sampling adequacy remained virtually the same at 0.656, continuing to suggest that sampling is mediocre.

The Bartlett test of sphericity was statistically significant (χ2 = 2015.062); P <0.001).

The results of the PCA and EFA again suggested a single-factor model. The percent variance explained increased from 48% in the 5-item scale to 60% in the 4-item scale and the alpha coefficient for the 4-item scale was 0.78. However, items 2 and 3 continued to have uniqueness values exceeding 0.50. EFA results are presented in Table 3.4. The model fit indices suggested that the model was a not a good fit to the data: degrees of freedom (df) = 2, chi-square (χ2) =

490.570, RMSEA = 0.421, CFI = 0.950, and TLI = 0.851.

Thus, the five questions regarding abortion attitudes do not share a latent construct and must be analyzed as individual variables.

Aim 1B Discussion

The results of this analysis suggest that the five survey questions regarding abortion attitudes do not function together as a valid scale within the study population. To our knowledge, these questions have not been previously validated as a scale in any context. They were selected for the survey based on previous use in several Zambian studies and were initially chosen to represent particular domains of beliefs and attitudes regarding abortion: morality, women’s rights, access to care, and support for legality.1, 2 Additionally, they were used in the baseline survey and needed to remain the same for the endline survey as a requisite for the evaluation that was the original objective for this data collection.

To develop a reliable and valid scale that measures the underlying construct of attitudes toward abortion would be an important contribution to the understanding of abortion stigma, particularly in countries like Zambia where rates of unsafe abortion are high. It would be ideal to incorporate

69 questions extending beyond these chosen five, ideally incorporating broader questions around community stigma from scales that have been validated in sub-Saharan Africa.13 Beyond expanding the questions asked regarding abortion attitudes, a broader population of respondents would allow for comparisons across gender, country, or region. Ultimately, however, the five identified questions were not designed to function as a scale and further investigations are needed to develop more accurate population-based measures of abortion attitudes.

Chapter Three Tables and Figures

Table 3.1. DUREL Item Response Descriptive Statistics for Sample N (Total = 1432) Percent (%) Item 1 (ORA) Never 9 0.6 Once a year or less 12 0.8 A few times a year 16 1.1 A few times a month 59 4.1 Once a week 596 41.6 More than once/week 740 51.7 Item 2 (NORA) Rarely or never 106 7.4 A few times a month 32 2.2 Once a week 147 10.3 Two or more times/week 348 24.3 Daily 737 51.5 More than once a day 62 4.3 Item 3 (IR1) Definitely not true 6 0.4 Tends not to be true 13 0.9 Unsure 22 1.5 Tends to be true 522 36.5 Definitely true of me 869 60.7 Item 4 (IR2) Definitely not true 2 0.1 Tends not to be true 17 1.2 Unsure 26 1.8 Tends to be true 572 39.9 Definitely true of me 815 57.0 Item 5 (IR3) Definitely not true 8 0.6 Tends not to be true 22 1.5 Unsure 28 2.0 Tends to be true 557 38.9 Definitely true of me 817 57.0

Table 3.2. Means, SD, and coefficient alphas for Duke Religious Index scales Scale Mean SD Alpha

OR 5.40 0.79 - NOR 4.23 1.22 - IR 13.59 1.72 0.89 Total 23.22 2.80 0.78

Figure 3.1 Standardized parameter estimates of the single-factor model of the DUREL

Figure 3.2. Standardized parameter estimates of the second-order single-factor model of the DUREL

Figure 3.3. Distribution of Abortion Attitude Score

Table 3.3. Participant characteristics by lower vs. higher abortion attitudes*

Summary Abortion Attitudes Total Lower Higher Characteristic n=1,432 n=721 n=711 p-value

Age, mean (SD) 27.5 (7.7) 28.4 (7.9) 26.6 (7.4) 0.071

Religion, % Protestant 83.2 84.3 82.0 0.239 Catholic 16.8 15.7 18.0

Gravidity, % 0 18.7 16.0 21.4 <0.001 1-3 50.4 48.1 52.7 4+ 30.9 35.9 25.9

Ever Married, % No 24.0 20.0 28.1 <0.001 Yes 76.0 80.0 71.9

Education, % Never attended 4.2 5.1 3.2 0.022 Primary school 36.7 39.4 34.0 Secondary school 52.4 49.7 55.1 Post-secondary school 6.7 5.8 7.6

Wealth Quintile, % 1 20.2 20.5 19.8 0.010 2 20.0 21.9 18.1 3 20.3 22.1 18.4 4 20.8 20.0 21.7 5 18.7 15.5 21.9

Residence, % Urban 56.8 52.3 61.5 <0.001 Rural 43.2 47.7 38.5 *Dichotomized at Median (12) / Mean (12.59)

Table 3.4. Results of the Exploratory Factor Analysis

Exploratory Factor Analysis

Loading Uniqueness 5 Items Factor 1: Abortion attitudes Item 1: “Abortion is immoral.” 0.025 0.999 Item 2: “A woman has the right to decide whether or not to continue a pregnancy.” 0.549 0.699 Item 3: “A woman has the right to decide whether or not to terminate a pregnancy.” 0.663 0.561 Item 4: “Women should have access to safe abortion services.” 0.779 0.393 Item 5: “Unmarried women should have access to safe abortion services.” 0.751 0.436 Item (overall scale α=0.71)

4 Items Factor 1: Abortion attitudes Item 2: “A woman has the right to decide whether or not to continue a pregnancy.” 0.548 0.699 Item 3: “A woman has the right to decide whether or not to terminate a pregnancy.” 0.661 0.564 Item 4: “Women should have access to safe abortion services.” 0.780 0.392 Item 5: “Unmarried women should have access to safe abortion services.” 0.752 0.435 Item (overall scale α=0.78)

Chapter Three References

1. Geary, C.W., et al., Attitudes toward abortion in Zambia. International Journal of Gynecology & Obstetrics, 2012. 118: p. S148-S151. 2. Cresswell, J.A., et al., Women's knowledge and attitudes surrounding abortion in Zambia: a cross-sectional survey across three provinces. BMJ Open, 2016. 6(3): p. e010076. 3. Lucchetti, G., et al., Validation of the Duke Religion Index: DUREL (Portuguese version). J Relig Health, 2012. 51(2): p. 579-86. 4. Dobrowolska, B., et al., Validation of the Polish version of the Duke University Religion Index (PolDUREL). Pol Arch Med Wewn, 2016. 126(12): p. 1005-1008. 5. Hafizi, S., et al., The Duke University Religion Index (DUREL): validation and reliability of the Farsi version. Psychol Rep, 2013. 112(1): p. 151-9. 6. Koenig, H.G. and A. Büssing, The Duke University Religion Index (DUREL): A Five- Item Measure for Use in Epidemological Studies. Religions, 2010. 1(1): p. 78-85. 7. Bentley, J., Z. Ahmad, and J. Thoburn, Religiosity and posttraumatic stress in a sample of East African refugees. Mental Health, Religion & Culture, 2013. 17(2): p. 185-195. 8. Taylor, P.W., Psychometric properties of the Duke University Religion Index, English and Spanish versions, for Hispanic‑American Women, in Psychology. 2013, San Diego State University. 9. Harries, J., et al., Conscientious objection and its impact on abortion service provision in South Africa: A qualitative study. Reproductive Health, 2014. 11(1). 10. Rehnstrom Loi, U., et al., Health care providers' perceptions of and attitudes towards induced abortions in sub-Saharan Africa and Southeast Asia: a systematic literature review of qualitative and quantitative data. BMC Public Health, 2015. 15: p. 139. 11. Tagoe-Darko, E., “Fear, Shame and Embarrassment”: The Stigma Factor in Post Abortion Care at Komfo Anokye Teaching Hospital, Kumasi, Ghana. Asian Social Science, 2013. 9(10). 12. Aniteye, P. and S.H. Mayhew, Shaping legal abortion provision in Ghana: Using policy theory to understand provider-related obstacles to policy implementation. Health Research Policy and Systems, 2013. 11(1). 13. Shellenberg, K.M., L. Hessini, and B.A. Levandowski, Developing a Scale to Measure Stigmatizing Attitudes and Beliefs About Women Who Have Abortions: Results from Ghana and Zambia. Women & Health, 2014. 54(7): p. 599-616.

CHAPTER FOUR: AIM 2 ANALYTIC RESULTS

This chapter presents the results for the Aim 2 analysis, which uses population-based survey data from women of reproductive age in three Zambian provinces. Recall that the goals of Aim 2 were to assess the relationships between individual’s religiosity and their abortion perceptions as well as to assess the relationships between the religiosity and abortion perceptions of an individual’s residential community and her own abortion perceptions. It is important to examine the influence of religiosity at both individual and community levels, as religion generally exists simultaneously as an inward expression of faith or relationship with the divine as well as an outward, shared experience of communal worship with others who have aligned beliefs.

Results

Distributions of the religiosity variables and relevant covariates are shown in Table 4.1. The survey population had high levels of religiosity in all three dimensions. Organizational religiosity

(OR) is measured by frequency of church attendance and more than half of women in the sample

(51.7%) attend church more than one time per week. Similarly, more than half of the sample

(55.8%) spends time alone in religious activities, i.e. prayer, at least once daily; an indicator of non-organizational religiosity (NOR). More than half of the population (52.7%) had high

Intrinsic Religiosity (IR), an indicator of how religion guides one’s life and measured as a composite score from a three-item validated scale.

The relationship between individual religiosity, as measured by church attendance (OR), time spend in solitary prayer (NOR), and intrinsic religiosity (IR), and four abortion attitude outcomes and an index measuring individual perceptions of legal permissibility for abortion will be examined, controlling for key individual, household, and community covariates (Table 4.1).

Adjusted bivariate associations of individual and community religiosity variables and covariates with four abortion attitude variables are presented in Table 4.2; the associations with legal permissibility are presented in Table 4.3 There is variation in associations for the religiosity variables and covariates with all five outcome variables. Responding affirmatively to “A woman has the right to decide whether or not to continue a pregnancy” is higher for women with higher levels of religiosity (OR, NOR, and IR) than those with the lowest levels. However, this pattern is reversed for the other three attitudinal statements. Across outcomes, older age, higher gravidity, marriage, lower wealth, lower education, and living in rural areas show conservative attitude patterns. However, these relationships are non-linear; for instance, 46.1% of women who’ve attended post-secondary school believe that a woman should have the right to continue her pregnancy as compared to 60.5% of women who have attended secondary school.

Given the hierarchical nature of the data, as the respondents women are clustered within standard enumeration areas (SEAs) or sampling clusters, initial analyses of the community-level effects on individual abortion perceptions (attitudes and legal permissibility) used multilevel modeling techniques. Intercept-only models via a generalized linear latent and mixed model (GLLAMM) were employed. This model allowed for the assessment of the level of variation between SEAs in the average odds of my outcomes of interest, from which the intraclass correlation coefficients

(ICCs), or the proportion of total variance that is accounted for by differences between SEAs.

As the results of the intercept-only models produced very small ICC values for all outcomes, the random intercept model was not required for the data and the community variables were included

79 in standard logistic models. The clustering of observations within villages was accounted for using robust estimation of variance.

Multivariate logistic regression models are shown for each of the four attitudinal outcomes of interest (Tables 4.4 – 4.7) and the single legal permissibility outcome in Table 4.8. In each table,

Model 1 presents unadjusted odds ratios for each of the independent variables and covariates.

Model 2 presents adjusted odds ratios of key individual-level religiosity measures as well as individual- and household-level covariates. Model 3 adds community-level covariates to Model

2- urban vs. rural and community-level religiosity covariates. In Model 4 there is the addition of a community-level attitudinal question; in the case of the four attitudinal outcomes (Tables 4.4 –

4.7), community-level legal permissibility is included in Model 4 and for the single legal permissibility outcome, community-level attitudes regarding access to safe abortion is used

(Table 4.8).

The multivariate logistic regression model for the individual response to the question “A woman has the right to decide whether or not to continue a pregnancy” is presented in Table 4.4. The individual religiosity dimensions were not associated with attitudes toward a woman’s right to decide to continue a pregnancy in any of the models. When examining individual- and household-level associations (Model 2), secondary education (OR=1.91) and status within the highest wealth quintile (OR=1.75) were significantly associated with higher odds of supporting a woman’s right to decide to continue her pregnancy as compared to those without schooling and the poorest quintile. Women aged 25-34 had reduced odds of supporting a woman’s right to decide to continue her pregnancy as compared to adolescents aged 15-19 (OR=0.50). In Models

3 and 4, with the incorporation of community-level variables, wealth was no longer statistically significant; however, age expanded such that women aged 20-24 and 25-34 both had lower odds of supporting a woman’s right to choose to continue a pregnancy (Model 4: age 20-24, OR=0.58; age 25-34, OR=0.50). Education remains significant in Models 3 and 4, with primary school and secondary school attendance increasing one’s odds of supporting a woman’s right to choose to continue her pregnancy (Model 4: primary school, OR=1.93; secondary school, OR=2.15).

Community-level church attendance is associated with reduced odds of supporting a woman’s right to decide to continue a pregnancy (midrange church attendance, OR=0.70; high church attendance, OR=0.60). This indicates that those women living in sampling clusters where churchgoing is frequent are less likely to personally support the right of a woman to choose to continue her pregnancy. Unexpectedly, high community-level support for more permissive abortion laws is also associated with reduced odds of individual support for a woman’s right to choose to continue a pregnancy (OR=0.73).

Next, the multivariate logistic regression model for the individual response to the question “A woman has the right to decide whether or not to terminate a pregnancy” is presented in Table

4.5. Again, the individual religiosity dimensions were not associated with attitudes toward a woman’s right to decide to continue a pregnancy in any of the models. Though there were several significant univariate associations in the unadjusted model (Model 1), when examining the individual- and household-level adjusted model (Model 2), it is only one’s status within the highest (OR=2.16) and next highest (OR=1.79) wealth quintile that is significantly associated with higher odds of supporting a woman’s right to decide to terminate her pregnancy as

81 compared to those in the poorest wealth quintile. However, in with the incorporation of community-level variables in Model 3, the odds ratios were attenuated, and the next highest wealth quintile was no longer statistically significant. In Model 4, with the incorporation of community-level legal permissibility, wealth is no longer significantly associated with a woman’s right to decide to terminate her pregnancy.

In both Models 3 and 4, community-level church attendance (Model 4: midrange, OR=0.67; high, OR=0.76) is associated with lower odds of supporting a woman’s right to decide to terminate a pregnancy, meaning that women living in sampling clusters where churchgoing is frequent are less likely to personally support the right of a woman to choose to terminate her pregnancy. In the full model, Model 4, living in a sampling cluster with high community-level support for more permissive abortion laws is associated with increased odds of individual support for a woman’s right to choose to terminate a pregnancy (OR=1.33).

The multivariate logistic regression model for the third abortion attitude outcome, agreement with the statement, “Women should have access to safe abortion services” is presented in Table

4.6. In examining the individual- and household-level adjusted model (Model 2), we observe that women ages 25 and older have decreased odds of supporting access to safe abortion services than women aged 15-19 (age 25-34, OR=0.51; age 35+, OR=0.35). Women who have attended post-secondary schooling have 2.6 times greater odds of supporting access to safe abortion services than women who have not attended school. When examining the full model that includes all community-level variables (Model 4), education is no longer statistically significant.

However, women aged 25 and older continue to have lower odds of supporting access to safe

82 abortion (age 25-34, OR=0.47; age 35+, OR=0.30). Women who have children have higher odds of supporting access to safe abortion than women who do not (1-3 children, OR=1.59; 4+ children, OR=2.18).

In Model 3, high intrinsic religiosity within the community is associated with lower odds of individual support for safe abortion access (OR=0.55); however, when community legal permissibility is incorporated in Model 4, this association is no longer significant. In both

Models 3 and 4, community-level non-organizational religiosity (engagement in solitary prayer, meditation, etc.) is significantly associated with higher odds of support for safe abortion access

(Model 4: midrange prayer, OR=1.60; high prayer, OR=1.67). Likewise, in Model 4, high levels of desired legal permissibility for abortion within a community are significantly associated with higher odds of support for safe abortion access (OR=2.37).

The multivariate logistic regression model for the final abortion attitude outcome, agreement with the statement, “Unmarried women should have access to safe abortion services” is presented in Table 4.7. In the individual- and household-level adjusted model (Model 2), women aged 25-34 (OR=0.50) and 35 and older (OR=0.31) have reduced odds of supporting access to safe abortion services for unmarried women; this association remains significant in Models 3 and

4 (age 25-34, OR=0.46; age 35+, OR=0.26). Women in the highest wealth quintile have two times greater odds of supporting access to safe abortion services for unmarried women than women in the poorest wealth quintile (Model 2); however, once controlling for community-level variables this association is weakened and is no longer significant. In the full model (Model 4),

83 those in the middle wealth quintiles have lower odds of supporting access to safe abortion for unmarried women (next lowest, OR=0.66; middle, OR=0.45; next highest, OR=0.57).

Several community-level variables are significantly associated with attitudes surrounding access to safe abortion for unmarried women. In Models 3 and 4, women living in urban areas have higher odds of supporting access to safe abortion for the unmarried (Model 3, OR=2.21; Model

4, OR=1.9). In Model 3, middle (OR=1.40) and high levels (OR=2.38) of community non- organizational religiosity (solitary prayer) are significantly associated with higher odds of support for safe abortion access for the unmarried; in Model 4, only high levels of non- organizational religiosity are significantly associated (OR=1.67). In Model 3, high community intrinsic religiosity is significantly associated with support for access to safe abortion for unmarried women as compared to low levels (OR=0.53); in Model 4, both middle (OR=0.74) and high levels (OR=0.71) of intrinsic religiosity are significantly associated. Women living in communities with high levels of support for legal permissibility of abortion have 2.48 times greater odds of supporting access to safe abortion services for unmarried women.

The final outcome variable, legal permissibility of abortion was dichotomized into two groups- low- those who thought that abortion should be illegal or legal for two or fewer indications and high- those who thought that abortion should be legal for three or more indications (Table 4.8).

Community attitudes around access to safe abortion was used to measure community-level attitudes as predictor variable. There were several significant univariate associations in the unadjusted model (Model 1). In examining the individual- and household-level adjusted model

(Model 2), we observe that those who have ever been married have lower odds of legal support

84 for abortion than those who have never been married (OR=0.54). Those who have attended post- secondary school (OR=12.20) and those in the highest wealth quintile (OR=3.61) have higher odds of greater legal permissibility for abortion those who have never attended school or are in the lowest wealth quintile, respectively. These significant associations in Model 2 are carried forward in Models 3 and 4 when community-level covariates are added to the model.

Middle and high levels of community churchgoing are associated with lower odds of legal permissibility of abortion as compared to low levels of community churchgoing in Model 3

(midrange churchgoing, OR=0.65; high churchgoing, OR=0.64); in Model 4, only middle levels of community churchgoing are significantly associated with reduced odds of legal permissibility of abortion (OR=0.74). Community non-organizational religiosity is significantly associated with higher odds of legal permissibility in both Models 3 and 4 (Model 4: middle, OR=1.46; high,

OR=2.42), while high levels of community intrinsic religiosity reduces the odds of support for legal permissibility in Models 3 and 4 (Model 4: high, OR=0.53). Finally, living in a community where acceptance for access to safe abortion is middle or high increases the odds of support for legalization for abortion than those living in communities where acceptance for access to safe abortion is low (middle, OR=1.72; high, OR=1.53).

Discussion

No previous quantitative studies have examined the role of religiosity on attitudes and beliefs regarding abortion in a sub-Saharan African context. However, religion plays an important role in daily life in most sub-Saharan African countries, and it has been shown to prompt and sustain health-related stigma within sub-Saharan African societies.1-3 Researchers have previously linked

85 individual religiosity, measured primarily through church attendance, to patterns of sexual health behaviors and attitudes. In most instances, though notably in Western settings, these studies have shown that higher levels of personal religiosity are associated with lower contraceptive use, more traditional attitudes regarding sexual and reproductive health, and increased fertility.4-8 In sub-

Saharan Africa, still in the earlier stages of fertility transition, these limited studies have had more varied results. In several instances, those attending church more regularly have had higher contraceptive use and more progressive attitudes toward sexual and reproductive health than those who did not.9-11 In this case, living in a community with higher church attendance led to generally more conservative attitudes toward abortion; however, living in communities where people reported high levels of solitary prayer was associated with more liberal attitudes toward abortion.

The findings from this study are particularly notable, as they do not show an association between religiosity and abortion perceptions at an individual level but find associations between religiosity at a community level, implying that one’s own religious beliefs may be less relevant in influencing their attitudes and beliefs surrounding abortion than the religious beliefs of the communities where they live. This suggests that it may not be religion itself driving abortion attitudes, but rather stigma generated from perceptions of community religious norms. Women living in communities with high churchgoing behavior are less likely to support a woman’s right to decide whether to continue or terminate a pregnancy and are also less likely to support the legal permissibility of abortion. Similarly, women who live in communities where intrinsic religiosity is high among the population are less likely to themselves support access to abortion for all women, married and unmarried. Non-organizational religiosity at a community level, i.e.

86 solitary prayer, has the opposite effect on individual abortion perceptions. Women living in communities where other women report higher levels of non-organizational religiosity are more likely to have more progressive attitudes toward abortion access, including for unmarried women. This finding is not surprising, seeing as non-organizational religiosity is less outwardly observable than churchgoing and intrinsic religiosity. Where women might note the observable piety of others, it is impossible to observe the extent to which fellow community members engage in solitary prayer; thus, it is less likely to drive stigma. Further, previous studies have suggested that non-organizational prayer is not associated with abortion stigma and is linked to easing of guilt associated with abortion among women who have had abortions, thus suggesting that prayerful women might also be more empathetic generally to the struggles of others.12-14

In Zambia, much like other countries in Eastern and Southern Africa, the growth of the

Pentecostal and charismatic Christian movements has been notable in the past decades and have a great deal of influence on religious and societal norms, including sexual and reproductive health beliefs and outcomes.15, 16 Previous studies have shown differences in attitudes and behaviors toward contraception between Catholics and mainline Protestants as compared to

Evangelicals, with the later having more conservative beliefs.17-19 As this survey was structured to match the Demographic and Health Survey (DHS), religious denomination was not disaggregated beyond generalized Protestantism, thus not allowing further exploration into the nuance of denomination as pertaining to abortion perceptions.20

In addition to community religiosity driving individual abortion perceptions, perceptions toward abortion within a community are associated with individual attitudes toward abortion. Living in a

87 community where there is strong support for a more permissive legal framework for safe abortion is associated with greater individual support for all four measures of abortion attitudes- right to continue pregnancy, right to terminate pregnancy, access to safe abortion, and access to safe abortion for unmarried women. Likewise, community support for access to safe abortion is associated with increased support for the legal permissibility of safe abortion.21 Living in a community where people are more accepting of abortion leads to greater person support for abortion. This is potentially explained by a disruption of the cycle of stigma whereby abortion is normalized in a community and thus no longer perceived as deviant.22, 23

Though covariates of significance varied across the analyses of multiple outcomes, those of significance are consistent with previous literature on facilitators and barriers to reproductive health attitudes and behaviors in sub-Saharan Africa, notably in regards to contraceptive use.24-29

The covariates associated with more liberal perceptions of abortion variables were younger age, greater education, higher wealth, not being married, having children, and urban residence.

However, covariate associations were not consistent across outcomes and in some cases showed changes in directionality.

Chapter Four Tables and Figures

Table 4.1. Baseline characteristics of the sample

Total Percent Characteristic N=1,430 (%)

Church Attendance Less than once per week 96 6.7

Once per week 595 41.6

More than once per week 739 51.7

Time Spent Alone Less than once per week 138 9.7

in Religious At least once weekly 494 34.6

Activities (Prayer) At least once daily 798 55.8

Intrinsic Low IR 74 5.2

Religiosity (IR) Midrange IR 602 42.1

High IR 754 52.7

Age 15-19 231 16.2

20-24 381 26.6

25-34 500 35.0

35+ 318 22.2

Gravidity 0 266 18.6

1-3 722 50.5

4+ 442 30.9

Ever Married No 343 24.0

Yes 1087 76.0

Religion Protestant 1190 83.2

Catholic 240 16.8

Education, Never attended 60 4.2

Highest level Primary school 526 36.8

attended Secondary school 748 52.3

Post-secondary school 96 6.7

Wealth Quintile, Lowest 289 20.2

Lowest to Next lowest 287 20.1

highest Middle 290 20.3

Next highest 298 20.8

Highest 266 18.6

Residence Urban 812 56.8

Rural 618 43.2

Table 4.2. Weighted* percent of women aged 15-44 who agreed with attitudinal statements on abortion A woman has the A woman has the Unmarried women right to decide right to decide Women should should have access Characteristic whether or not to whether or not to have access to safe to safe abortion continue a terminate a abortion services services pregnancy pregnancy % p-value % p-value % p-value % p-value Church Attendance Less than once per week 48.7 0.339 38.8 0.564 40.3 0.540 32.8 0.037 Once per week 58.2 37.6 36.2 32.7 More than once per week 55.5 35.6 33.8 25.4

Time Spent Alone Less than once per week 53.3 0.974 40.1 0.623 32.2 0.116 26.2 0.130 in Religious At least once weekly 57.1 36.3 32.5 26.8 Activities (Prayer) At least once daily 55.7 36.4 37.7 30.7

Intrinsic Low IR 53.0 0.351 42.0 0.325 29.1 0.582 26.9 0.093 Religiosity (IR) Midrange IR 58.6 40.1 36.4 27.0 High IR 54.3 33.1 35.1 30.6

Age 15-19 63.7 0.309 39.1 0.115 39.0 0.001 35.5 0.001 20-24 55.2 42.6 46.1 36.6 25-34 55.0 37.0 32.1 26.4 35+ 53.7 27.2 25.2 19.3

Gravidity 0 56.4 0.766 43.1 0.015 40.9 0.007 36.4 0.017 1-3 56.9 39.4 36.9 29.6 4+ 54.4 28.1 28.9 23.0

Ever Married No 57.8 0.591 43.7 0.024 42.7 0.006 35.9 0.009 Yes 55.6 34.7 33.2 26.9

Religion Catholic 59.6 0.611 34.1 0.893 34.9 0.281 28.8 0.253 Protestant 55.3 37.2 36.8 29.3

Education, Never attended 38.8 0.043 29.3 0.048 24.8 0.001 21.4 0.107 Highest level Primary school 53.3 29.4 33.6 28.7 attended Secondary school 60.5 40.7 35.0 28.4 Post-secondary school 46.1 47.1 54.4 39.9

Wealth Quintile, Lowest 52.4 0.333 28.1 0.002 37.8 0.007 29.4 0.031 Lowest to Next lowest 47.7 27.2 32.7 24.3 highest Middle 52.9 32.8 26.7 21.6 Next highest 61.6 41.8 34.6 29.8 Highest 61.7 47.5 45.5 38.5

Residence Rural 51.6 0.177 29.8 0.043 32.7 0.001 23.1 0.001 Urban 57.8 39.4 36.3 31.1

Community OR Low 59.8 0.057 41.9 0.139 34.6 0.827 29.1 0.536 Middle 55.2 31.0 31.1 23.8 High 51.7 34.6 40.0 33.0

Community NOR Low 60.0 0.148 40.6 0.080 29.3 0.053 25.7 0.008 Middle 50.2 32.3 36.3 26.3 High 55.8 34.5 45.3 38.0

Community IR Low 57.6 0.022 41.1 <0.001 36.2 0.010 31.7 0.017 Middle 54.5 34.5 39.4 29.6 High 55.6 32.8 29.6 24.0 *Adjusted for sample design p-values significant at p<0.05 (in boldface)

Table 4.3 Weighted* percent of women aged 15-44 support legalization of abortion for three or more indications

Characteristic Abortion should be legal for 3 or more indications p-value Church Attendance Less than once per week 34.1 0.735 Once per week 41.6 More than once per week 40.6

Time Spent Alone Less than once per week 31.4 0.016 in Religious At least once weekly 35.7 Activities (Prayer) At least once daily 45.5

Intrinsic Low IR 39.4 <0.001 Religiosity (IR) Midrange IR 32.8 High IR 47.1

Age 15-19 41.7 0.600 20-24 46.5 25-34 36.4 35+ 39.9

Gravidity 0 55.4 <0.001 1-3 40.4 4+ 31.9

Ever Married No 57.1 <0.001 Yes 35.9

Religion Catholic 50.0 0.036 Protestant 38.6

Education, Never attended 15.1 <0.001 Highest level Primary school 33.6 attended Secondary school 42.5 Post-secondary school 80.8

Wealth Quintile, Lowest 29.2 <0.001 Lowest to Next lowest 29.1 highest Middle 30.1 Next highest 41.7 Highest 65.9

Residence Rural 31.0 <0.001 Urban 44.3

Community OR Low 43.0 0.021 Middle 32.4 High 44.5

Community NOR Low 36.4 <0.001 Middle 36.9 High 53.0

Community IR Low 45.1 0.068 Middle 42.3 High 32.4 *Adjusted for sample design p-values significant at p<0.05 (in boldface)

Table 4.4. Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “A woman has the right to decide whether or not to continue a pregnancy” Model 4: Adjusted Model 3: Adjusted Model 2: Adjusted Odds Ratio – Odds Ratio – Model 1: Unadjusted Odds Ratio – Individual, Household, Covariate Individual, Household Odds Ratio Individual & Community Religiosity & Community Household Covariates & Legal Permissibility Religiosity Covariates Covariates OR CI OR CI OR CI OR CI Individual-Level Religiosity Variables Church (OR) REF: < weekly Once weekly 1.47 0.8-2.7 1.31 0.7-2.5 1.33 0.7-2.4 1.26 0.7-2.2 2+ times weekly 1.31 0.7-2.5 1.16 0.6-2.1 1.21 0.7-2.2 1.14 0.6-2.1

Pray Alone REF: < weekly (NOR) At least weekly 1.16 0.7-2.0 1.12 0.7-1.7 1.13 0.7-1.8 1.11 0.7-1.7 At least daily 1.10 0.7-1.7 1.12 0.7-1.7 1.16 0.7-1.9 1.18 0.7-1.9

Intrinsic Religiosity REF: low IR (IR) Midrange IR 1.26 0.5-2.9 1.31 0.6-3.0 1.40 0.6-3.0 1.48 0.7-3.1 High IR 1.06 0.4-2.7 1.09 0.4-2.7 1.14 0.4-2.9 1.16 0.5-2.9

Other Individual-Level Variables Age REF: 15-19 20-24 0.70 0.4-1.1 0.60 0.3-1.1 0.57† 0.3-1.1 0.58† 0.3-1.1 25-34 0.69 0.4-1.1 0.50† 0.3-1.0 0.49* 0.2-1.0 0.50† 0.2-1.0 35+ 0.66 0.3-1.2 0.51 0.2-1.2 0.51 0.2-1.2 0.53 0.2-1.2

Gravidity REF: 0 1-3 1.02 0.7-1.5 1.46 0.9-2.5 1.54 0.9-2.6 1.54 0.9-2.6 4+ 0.92 0.5-1.6 1.70 0.9-3.3 1.74 0.9-3.5 1.71 0.9-3.4

Ever Married REF: No Yes 0.91 0.7-1.3 1.14 0.8-1.6 1.09 0.8-1.5 1.04 0.7-1.5

Religion REF: Catholic Protestant 0.84 0.6-1.1 0.88 0.6-1.3 0.88 0.6-1.3 0.89 0.6-1.3

Education, REF: No school Highest level Primary school 1.78 0.8-4.0 1.66 0.9-3.2 1.84† 0.9-3.7 1.93† 1.0-3.9 attended Secondary school 2.41* 1.1-5.4 1.91† 0.9-3.9 2.05† 0.9-4.4 2.15† 1.0-4.8 Post-secondary school 1.35 0.6-3.1 1.06 0.5-2.3 1.12 0.5-2.7 1.16 0.5-2.9

Household-Level Variables Wealth Quintile, REF: Lowest Lowest to Next lowest 0.83 0.5-1.4 0.87 0.5-1.5 0.85 0.5-1.5 0.85 0.5-1.5 highest Middle 1.02 0.6-1.7 1.08 0.6-1.8 1.08 0.6-2.1 1.11 0.6-2.2 Next highest 1.46 0.8-2.6 1.60 0.9-2.9 1.58 0.7-3.5 1.64 0.7-3.8 Highest 1.46† 1.0-2.2 1.75* 1.0-2.9 1.64 0.8-3.4 1.79 0.8-4.0

Community-Level Variables Residence REF: Rural Urban 1.28† 1.0-1.7 - - 0.96 0.6-1.5 1.04 0.7-1.6

Community REF: Low Churchgoing Middle 0.83 0.6-1.2 - - 0.77 0.6-1.1 0.70* 0.5-0.9 High 0.72† 0.5-1.0 - - 0.68* 0.5-1.0 0.60* 0.4-0.9

Community REF: Low Pray Alone Middle 0.67* 0.5-0.9 - - 0.72† 0.5-1.0 0.90 0.7-1.2 High 0.84 0.6-1.2 - - 0.91 0.6-1.4 1.16 0.8-1.7

Community REF: Low Intrinsic Religiosity Middle 0.88 0.6-1.3 - - 1.05 0.6-1.8 1.00 0.6-1.7 High 0.92 0.6-1.5 - - 1.00 0.6-1.7 0.91 0.6-1.4

Community REF: Low Legal Permissibility Middle 0.75 0.5-1.1 - - - - 0.68* 0.5-0.9 High 1.03 0.7-1.4 - - - - 0.73† 0.5-1.1 Significant at †p<0.10, *p<0.05

Table 4.5. Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “A woman has the right to decide whether or not to terminate a pregnancy” Model 4: Adjusted Model 3: Adjusted Model 2: Adjusted Odds Ratio – Odds Ratio – Model 1: Unadjusted Odds Ratio – Individual, Household, Covariate Individual, Household Odds Ratio Individual & Community Religiosity & Community Household Covariates & Legal Permissibility Religiosity Covariates Covariates OR CI OR CI OR CI OR CI Individual-Level Religiosity Variables Church (OR) REF: < weekly Once weekly 0.95 0.6-1.5 0.99 0.6-1.6 1.01 0.6-1.6 1.04 0.6-1.7 2+ times weekly 0.87 0.5-1.5 0.97 0.6-1.5 1.03 0.6-1.7 1.06 0.6-1.8

Pray Alone REF: < weekly (NOR) At least weekly 0.85 0.5-1.3 0.92 0.6-1.4 0.90 0.6-1.3 0.93 0.6-1.4 At least daily 0.85 0.5-1.4 1.05 0.7-1.7 1.07 0.7-1.7 1.07 0.7-1.7

Intrinsic Religiosity REF: low IR (IR) Midrange IR 0.92 0.5-1.7 1.03 0.6-1.7 1.10 0.6-1.9 1.12 0.6-2.0 High IR 0.68 0.4-1.2 0.75 0.4-1.3 0.79 0.5-1.4 0.81 0.5-1.5

Other Individual-Level Variables Age REF: 15-19 20-24 1.16 0.6-2.1 1.15 0.6-2.3 1.09 0.5-2.2 1.09 0.5-2.2 25-34 0.92 0.7-1.2 0.89 0.6-1.4 0.85 0.5-1.3 0.85 0.5-1.3 35+ 0.58* 0.3-1.0 0.73 0.3-1.6 0.71 0.3-1.7 0.71 0.3-1.7

Gravidity REF: 0 1-3 0.86 0.5-1.4 1.14 0.6-2.2 1.18 0.6-2.2 1.22 0.6-2.3 4+ 0.52* 0.3-0.8 0.94 0.5-1.7 0.99 0.6-1.7 1.02 0.6-1.8

Ever Married REF: No Yes 0.69* 0.5-0.9 0.95 0.6-1.5 0.90 0.6-1.4 0.88 0.5-1.4

Religion REF: Catholic Protestant 1.14 0.7-1.8 1.4 0.8-2.2 1.43 0.8-2.5 1.47 0.8-2.5

Education, REF: No school Highest level Primary school 1.00 0.4-2.4 0.97 0.5-1.8 1.04 0.3-3.2 1.04 0.3-3.2 attended Secondary school 1.65 0.9-3.2 1.08 0.5-2.4 1.14 0.5-2.5 1.15 0.5-2.5 Post-secondary school 2.15 0.8-5.9 1.18 0.4-3.6 1.22 0.4-3.8 1.17 0.4-3.7

Household-Level Variables Wealth Quintile, REF: Lowest Lowest to Next lowest 0.96 0.5-1.7 0.97 0.5-1.8 0.93 0.5-1.9 0.94 0.5-1.9 highest Middle 1.23 0.8-2.0 1.24 0.8-1.9 1.17 0.6-2.2 1.18 0.6-2.2 Next highest 1.83* 1.1-3.1 1.79* 1.1-3.0 1.71 0.9-3.3 1.63 0.8-3.2 Highest 2.31* 1.4-3.8 2.16* 1.2-3.8 1.99† 1.0-4.1 1.79 0.8-3.8

Community-Level Variables Residence REF: Rural Urban 1.53* 1.1-2.2 - - 0.95 0.6-1.4 0.90 0.6-1.3

Community REF: Low Churchgoing Middle 0.62* 0.4-1.0 - - 0.63* 0.4-0.9 0.67* 0.5-0.9 High 0.73 0.5-1.2 - - 0.70* 0.5-1.0 0.76† 0.6-1.0

Community REF: Low Pray Alone Middle 0.70† 0.5-1.1 - - 0.92 0.7-1.3 0.89 0.6-1.3 High 0.77 0.4-1.3 - - 0.85 0.5-1.3 0.75 0.5-1.2

Community REF: Low Intrinsic Religiosity Middle 0.75 0.4-1.4 - - 0.86 0.5-1.4 0.86 0.5-1.4 High 0.70 0.4-1.3 - - 0.77 0.5-1.2 0.85 0.6-1.3

Community REF: Low Legal Permissibility Middle 0.99 0.7-1.4 - - 0.99 0.7-1.4 High 1.78* 1.1-2.8 - - 1.33† 1.0-1.8 Significant at †p<0.10, *p<0.05

Table 4.6. Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “Women should have access to safe abortion services” Model 4: Adjusted Model 3: Adjusted Model 2: Adjusted Odds Ratio – Odds Ratio – Model 1: Unadjusted Odds Ratio – Individual, Household, Covariate Individual, Household Odds Ratio Individual & Community Religiosity & Community Household Covariates & Legal Permissibility Religiosity Covariates Covariates OR CI OR CI OR CI OR CI Individual-Level Religiosity Variables Church (OR) REF: < weekly Once weekly 0.84 0.5-1.5 0.66 0.3-1.3 0.64 0.3-1.2 0.69 0.4-1.3 2+ times weekly 0.75 0.4-1.3 0.67 0.3-1.3 0.64 0.3-1.2 0.70 0.4-1.3

Pray Alone REF: < weekly (NOR) At least weekly 1.02 0.7-1.6 1.08 0.7-1.7 0.94 0.6-1.5 1.03 0.6-1.7 At least daily 1.27 0.8-2.0 1.40 0.9-2.2 1.13 0.8-1.7 1.14 0.8-1.7

Intrinsic Religiosity REF: low IR (IR) Midrange IR 1.39 0.9-2.3 1.60 0.9-2.9 1.51 0.8-3.0 1.58 0.8-3.3 High IR 1.32 0.7-2.4 1.43 0.7-2.9 1.43 0.7-3.0 1.57 0.7-3.4

Other Individual-Level Variables Age REF: 15-19 20-24 1.33 0.8-2.2 1.11 0.7-1.7 1.12 0.7-1.7 1.12 0.7-1.7 25-34 0.74 0.5-1.1 0.51† 0.3-1.0 0.48* 0.2-1.0 0.47* 0.2-1.0 35+ 0.53* 0.4-0.8 0.35* 0.2-0.8 0.30* 0.1-0.7 0.30* 0.1-0.7

Gravidity REF: 0 1-3 0.85 0.6-1.2 1.38 0.8-2.3 1.45 0.9-2.4 1.59† 0.9-2.7 4+ 0.59* 0.4-0.9 1.64 0.8-3.4 1.93† 1.0-3.8 2.18* 1.1-4.3

Ever Married REF: No Yes 0.67† 0.4-1.0 0.87 0.5-1.5 0.93 0.5-1.7 0.87 0.5-1.6

Religion REF: Catholic Protestant 0.92 0.7-1.2 1.00 0.7-1.4 1.04 0.8-1.4 1.08 0.8-1.5

Education, REF: No school Highest level Primary school 1.54 0.6-3.7 1.60 0.8-3.3 1.54 0.8-2.9 1.53 0.8-3.0 attended Secondary school 1.64 0.6-4.3 1.17 0.5-3.0 1.08 0.4-2.7 1.08 0.4-2.8 Post-secondary school 3.63* 1.2-11.0 2.60† 0.8-8.1 2.35 0.8-6.6 2.12 0.8-5.8

Household-Level Variables Wealth Quintile, REF: Lowest Lowest to Next lowest 0.80 0.5-1.3 0.81 0.5-1.3 0.74 0.5-1.2 0.78 0.5-1.2 highest Middle 0.60† 0.3-1.1 0.66 0.4-1.2 0.54* 0.3-0.9 0.54* 0.3-1.0 Next highest 0.87 0.4-1.7 0.97 0.4-2.1 0.72 0.4-1.4 0.62 0.3-1.2 Highest 1.38 0.8-2.4 1.56 0.8-3.0 1.20 0.6-2.5 0.89 0.4-1.8

Community-Level Variables Residence REF: Rural Urban 1.17 0.7-2.0 - - 1.63† 1.0-2.7 1.36 0.8-2.2

Community REF: Low Churchgoing Middle 0.85 0.4-1.7 - - 0.86 0.5-1.5 1.03 0.7-1.5 High 1.25 0.6-2.6 - - 0.83 0.4-1.5 1.07 0.6-1.8

Community REF: Low Pray Alone Middle 1.38 0.7-2.6 - - 1.89* 1.2-3.0 1.60* 1.0-2.5 High 2.00† 1.0-4.0 - - 2.49* 1.2-5.0 1.67† 0.9-3.0

Community REF: Low Intrinsic Religiosity Middle 1.14 0.6-2.1 - - 0.88 0.6-1.3 0.90 0.7-1.2 High 0.74 0.3-1.6 - - 0.55* 0.3-1.0 0.72 0.5-1.2

Community REF: Low Legal Permissibility Middle 1.47 0.8-2.8 - - - - 1.16 0.7-2.1 High 2.96# 1.8-5.0 - - - - 2.37* 1.5-3.8 Significant at †p<0.10, *p<0.05, #p<0.001

Table 4.7. Estimated unadjusted and adjusted odds ratios from logistic regression analysis of agreement with statement “Unmarried women should have access to safe abortion services” Model 4: Adjusted Model 3: Adjusted Model 2: Adjusted Odds Ratio – Odds Ratio – Model 1: Unadjusted Odds Ratio – Individual, Household, Covariate Individual, Household Odds Ratio Individual & Community Religiosity & Community Household Covariates & Legal Permissibility Religiosity Covariates Covariates OR CI OR CI OR CI OR CI Individual-Level Religiosity Variables Church (OR) REF: < weekly Once weekly 0.99 0.5-2.0 0.80 0.3-2.1 0.80 0.3-2.0 0.86 0.4-2.1 2+ times weekly 0.70 0.4-1.4 0.61 0.3-1.5 0.57 0.2-1.3 0.61 0.3-1.4

Pray Alone REF: < weekly (NOR) At least weekly 1.02 0.6-1.7 1.05 0.6-1.9 0.92 0.5-1.7 1.03 0.6-1.9 At least daily 1.25 0.7-2.2 1.34 0.7-2.4 1.10 0.6-2.0 1.11 0.6-1.8

Intrinsic Religiosity REF: low IR (IR) Midrange IR 1.01 0.5-2.0 1.17 0.5-2.6 1.18 0.5-2.8 1.31 0.6-3.0 High IR 1.20 0.6-2.3 1.40 0.6-3.2 1.51 0.6-3.6 1.77 0.7-4.3

Other Individual-Level Variables Age REF: 15-19 20-24 1.05 0.6-1.9 0.92 0.6-1.5 0.89 0.5-1.5 0.90 0.5-1.6 25-34 0.65† 0.4-1.1 0.50* 0.3-1.0 0.46* 0.2-0.9 0.46* 0.2-0.9 35+ 0.43* 0.3-0.7 0.31* 0.1-0.8 0.26* 0.1-0.7 0.26* 0.1-0.7

Gravidity REF: 0 1-3 0.74 0.5-1.1 1.06 0.6-1.9 1.13 0.6-2.1 1.28 0.7-2.5 4+ 0.52* 0.3-0.8 1.21 0.5-2.7 1.49 0.7-3.3 1.76 0.8-3.9

Ever Married REF: No Yes 0.66* 0.4-1.0 1.06 0.6-2.0 1.08 0.6-2.1 0.93 0.5-1.9

Religion REF: Catholic Protestant 0.97 0.6-1.5 1.05 0.6-1.8 1.15 0.7-2.0 1.22 0.7-2.2

Education, REF: No school Highest level Primary school 1.48 0.5-4.1 1.46 0.6-3.5 1.48 0.6-3.4 1.52 0.7-3.6 attended Secondary school 1.46 0.5-4.0 0.86 0.3-2.4 0.82 0.3-2.3 0.86 0.3-2.4 Post-secondary school 2.44† 0.9-6.8 1.34 0.4-4.1 1.25 0.4-3.6 1.15 0.4-3.2

Household-Level Variables Wealth Quintile, REF: Lowest Lowest to Next lowest 0.77 0.5-1.2 0.79 0.5-1.3 0.61* 0.4-1.0 0.66† 0.4-1.0 highest Middle 0.66 0.4-1.2 0.77 0.4-1.3 0.44* 0.3-0.8 0.45* 0.3-0.8 Next highest 1.02 0.6-1.8 1.37 0.7-2.6 0.69 0.4-1.2 0.57* 0.3-1.0 Highest 1.50 0.9-2.6 2.02† 1.0-4.2 1.01 0.5-2.0 0.70 0.4-1.3

Community-Level Variables Residence REF: Rural Urban 1.50 0.9-2.5 - - 2.21* 1.3-3.8 1.9* 1.1-3.1

Community REF: Low Churchgoing Middle 0.76 0.4-1.3 - - 0.76 0.4-1.3 0.89 0.6-1.3 High 1.20 0.6-2.3 - - 0.84 0.5-1.5 1.07 0.7-1.6

Community REF: Low Pray Alone Middle 1.03 0.6-1.8 - - 1.40† 1.0-2.0 1.35 0.6-1.0 High 1.77 0.9-3.6 - - 2.38* 1.3-4.4 1.67* 1.0-2.7

Community REF: Low Intrinsic Religiosity Middle 0.90 0.5-1.7 - - 0.76 0.5-1.2 0.74* 0.6-1.0 High 0.68 0.4-1.3 - - 0.53* 0.3-0.9 0.71† 0.5-1.0

Community REF: Low Legal Permissibility Middle 0.99 0.6-1.8 - - - - 0.80 0.4-1.4 High 2.93# 2.0-4.2 - - - - 2.48# 1.7-3.6 Significant at †p<0.10, *p<0.05, #p<0.001

Table 4.8. Estimated unadjusted and adjusted odds ratios from logistic regression analysis of support for legalization of abortion (≥3 indications) Model 4: Adjusted Model 3: Adjusted Model 2: Adjusted Odds Ratio – Odds Ratio – Model 1: Unadjusted Odds Ratio – Individual, Household, Covariate Individual, Household Odds Ratio Individual & Community Religiosity & Community Household Covariates & Legal Permissibility Religiosity Covariates Covariates OR CI OR CI OR CI OR CI Individual-Level Religiosity Variables Church (OR) REF: < weekly Once weekly 1.38 0.8-2.4 1.29 0.6-2.8 1.23 0.6-2.7 1.34 0.6-2.9 2+ times weekly 1.32 0.8-2.1 1.22 0.5-2.7 1.21 0.5-2.8 1.27 0.6-2.9

Pray Alone REF: < weekly (NOR) At least weekly 1.21* 0.7-2.2 1.06 0.5-2.2 0.94 0.5-1.9 1.02 0.5-2.1 At least daily 1.82 1.2-2.9 1.37 0.6-2.9 1.11 0.5-2.4 1.14 0.5-2.5

Intrinsic Religiosity REF: low IR (IR) Midrange IR 0.75 0.4-1.5 0.75 0.4-1.6 0.71 0.3-1.4 0.68 0.3-1.4 High IR 1.37 0.9-2.2 1.43 0.7-2.9 1.46 0.7-3.1 1.45 0.7-3.2

Other Individual-Level Variables Age REF: 15-19 20-24 1.21 0.9-1.6 1.32 0.8-2.3 1.30 0.7-2.2 1.34 0.8-2.3 25-34 0.80 0.5-1.3 0.92 0.5-1.6 0.84 0.5-1.5 0.84 0.5-1.5 35+ 0.93 0.6-1.5 1.56 0.7-3.2 1.33 0.6-2.8 1.38 0.7-2.8

Gravidity REF: 0 1-3 0.55* 0.4-0.8 1.01 0.5-1.9 1.05 0.5-2.0 1.09 0.6-2.1 4+ 0.38* 0.2-0.7 0.80 0.4-1.8 0.92 0.4-2.1 0.97 0.4-2.2

Ever Married REF: No Yes 0.42* 0.3-0.7 0.54† 0.3-1.0 0.58* 0.3-01.0 0.58* 0.3-0.8

Religion REF: Catholic Protestant 0.63† 0.4-1.1 0.70 0.4-1.2 0.73 0.4-1.4 0.75 0.4-1.4

Education, REF: No school Highest level Primary school 2.84† 0.9-8.8 2.64 0.7-10.4 2.54 0.7-9.3 2.46 0.7-9.1 attended Secondary school 4.14* 1.3-13.3 2.23 0.5-9.0 2.03 0.5-7.6 1.96 0.5-7.4 Post-secondary school 23.64# 7.9-70.5 12.20* 3.0-49.5 11.3# 3.3-38.0 10.20* 3.0-34.4

Household-Level Variables Wealth Quintile, REF: Lowest Lowest to Next lowest 1.00 0.6-1.7 0.94 0.5-1.6 0.89 0.5-1.5 0.90 0.5-1.5 highest Middle 1.04 0.6-1.8 1.06 0.6-1.8 0.99 0.5-1.8 1.15 0.6-2.0 Next highest 1.73† 0.9-3.2 1.56 0.7-3.4 1.34 0.6-3.0 1.48 0.7-3.3 Highest 4.69# 2.8-8.0 3.61# 1.9-6.8 3.22* 1.6-6.3 3.13# 1.7-5.7

Community-Level Variables Residence REF: Rural Urban 1.78* 1.1-3.0 - - 1.27 0.9-1.9 1.21 0.8-1.8

Community REF: Low Churchgoing Middle 0.63 0.3-1.2 - - 0.65* 0.5-0.9 0.74* 0.6-1.0 High 1.06 0.5-2.3 - - 0.64† 0.4-1.0 0.74 0.5-1.1

Community REF: Low Pray Alone Middle 1.02 0.5-2.0 - - 1.73* 1.1-2.8 1.46† 1.0-2.2 High 1.97† 0.9-4.3 - - 2.75# 1.7-4.3 2.42* 1.5-3.9

Community REF: Low Intrinsic Religiosity Middle 0.89 0.4-1.8 - - 0.75 0.5-1.2 0.77 0.5-1.2 High 0.58 0.3-1.2 - - 0.49* 0.3-0.8 0.53* 0.4-0.8

Community REF: Low Acceptance for Access Middle 2.23* 1.3-3.8 - - - - 1.72* 1.2-2.4 to Abortion High 3.14# 1.8-5.4 - - - - 1.53† 1.0-2.4 Significant at †p<0.10, *p<0.05, #p<0.001

Chapter Four References

1. Zou, J., et al., Religion and HIV in Tanzania: influence of religious beliefs on HIV stigma, disclosure, and treatment attitudes. BMC public health, 2009. 9(1): p. 1. 2. Varas-Díaz, N., et al., Religion and HIV/AIDS stigma: Implications for health professionals in Puerto Rico. Global Public Health, 2010. 5(3): p. 295-312. 3. Sambisa, W., AIDS stigma and uptake of HIV testing in Zimbabwe. 2008. 4. de Visser, R.O., et al., Associations between religiosity and sexuality in a representative sample of Australian adults. Arch Sex Behav, 2007. 36(1): p. 33-46. 5. Goldscheider, C. and W.D. Mosher, Patterns of contraceptive use in the United States: the importance of religious factors. Studies in Family Planning, 1991. 22(2): p. 102-115. 6. Mosher, W.D., L.B. Williams, and D.P. Johnson, Religion and fertility in the United States: New patterns. Demography, 1992. 29(2): p. 199-214. 7. Zhang, L., Religious affiliation, religiosity, and male and female fertility. Demographic Research, 2008. 18: p. 233-262. 8. Hayford, S.R. and S.P. Morgan, Religiosity and fertility in the United States: The role of fertility intentions. Social Forces, 2008. 86(3): p. 1163-1188. 9. Agadjanian, V., Religious denomination, religious involvement, and modern contraceptive use in southern Mozambique. Studies in Family Planning, 2013. 44(3): p. 259-274. 10. Yeatman, S.E. and J. Trinitapoli, Beyond denomination: The relationship between religion and family planning in rural Malawi. Demographic Research, 2008. 19: p. 1851- 1881. 11. Avong, H., Relationship Between Religion and Use of Modern Contraceptives among the Atyap in Kaduna State. Nigeria. Research on Humanities and Social Sciences, 2012. 2(8): p. 82-89. 12. Harries, J., et al., Delays in seeking an abortion until the second trimester: a qualitative study in South Africa. Reprod Health, 2007. 4: p. 7. 13. McMurtrie, S.M., et al., Public opinion about abortion-related stigma among Mexican Catholics and implications for unsafe abortion. International Journal of Gynecology & Obstetrics, 2012. 118: p. S160-S166. 14. Adamczyk, A., The effects of religious contextual norms, structural constraints, and personal religiosity on abortion decisions. Social Science Research, 2008. 37(2): p. 657- 672. 15. Meyer, B., Christianity in Africa: From African Independent to Pentecostal-Charismatic Churches. Annual Review of Anthropology, 2004. 33(1): p. 447-474. 16. McQuillan, K., When Does Religion Influence Fertility? Population and Development Review, 2004. 30(1): p. 25-56.

17. Agadjanian, V. and S.T. Yabiku, Religious Belonging, Religious Agency, and Women's Autonomy in Mozambique. Journal for the Scientific Study of Religion, 2015. 54(3): p. 461-476. 18. Doctor, H.V., J.F. Phillips, and E. Sakeah, The influence of changes in women's religious affiliation on contraceptive use and fertility among the Kassena-Nankana of Northern Ghana. Studies in Family Planning, 2009. 40(2): p. 113-122. 19. Wusu, O., Religious Influence on Non-Use of Modern Contraceptives among Women in Nigeria: Comparative Analysis of 1990 and 2008 Ndhs. J Biosoc Sci, 2015. 47(5): p. 593-612. 20. Central Statistical Office (CSO) [Zambia], M.o.H.M.Z., and ICF International. , Zambia Demographic and Health Survey 2013-14. 2014, Central Statistical Office, Ministry of Health, and ICF International: Rockville, Maryland, USA. 21. Belton, S., et al., Attitudes towards the legal context of unsafe abortion in Timor-Leste. Reproductive Health Matters, 2009. 17(34): p. 55-64. 22. Kumar, A., L. Hessini, and E.M. Mitchell, Conceptualising abortion stigma. Cult Health Sex, 2009. 11(6): p. 625-39. 23. Izugbara, C.O., C. Egesa, and R. Okelo, 'High profile health facilities can add to your trouble': Women, stigma and un/safe abortion in Kenya. Soc Sci Med, 2015. 141: p. 9-18. 24. Geary, C.W., et al., Attitudes toward abortion in Zambia. International Journal of Gynecology & Obstetrics, 2012. 118: p. S148-S151. 25. Elfstrom, K.M. and R. Stephenson, The Role of Place in Shaping Contraceptive Use among Women in Africa. PLOS ONE, 2012. 7(7): p. e40670. 26. Singh, S., J. Darroch, and L. Ashford, Adding it Up: The Costs and Benefits of Investing in Sexual and Reproductive Health 2014. Guttmacher Institute, 2014. 27. Sumer, Z.H., Gender, Religiosity, Sexual Activity, Sexual Knowledge, and Attitudes Toward Controversial Aspects of Sexuality. J Relig Health, 2015. 54(6): p. 2033-44. 28. Cresswell, J.A., et al., Women's knowledge and attitudes surrounding abortion in Zambia: a cross-sectional survey across three provinces. BMJ Open, 2016. 6(3): p. e010076. 29. Bankole, A., et al., Differences in unintended pregnancy, contraceptive use and abortion by HIV status among women in Nigeria and Zambia. International Perspectives on Sexual and Reproductive Health, 2014. 40(1): p. 28-38.

CHAPTER FIVE: AIM 3 ANALYTIC RESULTS

This chapter presents the results for the Aim 3 analysis, which uses in-depth interviews with women who have terminated a pregnancy to explore the role that religion and stigma play in women’s understanding of abortion and how it informs their perceptions, decision-making, and social experiences with abortion.

After a woman has an abortion, there are a variety of issues with which she must grapple.

Economic, social, and health factors often change because of an abortion and women cope and move forward in a variety of ways. This thematic analysis uncovered both immediate and longer- term attitudinal and behavioral responses to abortion in the context of religion. It explores ideas of sin, redemption, and religious reintegration at individual and community levels.

Results

Description of Participants

The descriptive characteristics of the sample are detailed in Table 5.1. A total of 51 women were interviewed immediately following their abortion procedure and prior to discharge from the point of service. Just over half of the women presented for postabortion care (PAC) services and the remainder presented for termination of pregnancy (TOP). To distinguish, PAC services treat complications from unsafe induced abortions that require urgent medical attention, complications from spontaneous abortion (miscarriage), and complications from safe abortion that require additional medical treatment.1 TOP indicates that a woman sought a safe abortion and went to the health facility for the service.

100

Basic demographic characteristics were not systematically collected by all interviewers nor were they explicitly included in the interview guide, thus certain descriptive information is unknown for many of the women interviewed. The 51 participants ranged in age from 16 to 32 years and just under half of all women interviewed were students. More than two-thirds of the participants were unmarried, with a small subset widowed or divorced. Over half of the women lived with family members, either their biological parents or extended family. The women’s relationships with their male partner implicated in the pregnancy and subsequent abortion were varied. For over one-third of the women interviewed, the implicated male partner was unmarried. Seven respondents conceived with partners in relationships with other women, while four were in

“sugar daddy” relationships with the male partner. There were three women interviewed who conceived as a result of rape.

There was a 75% follow-up rate, with 38 women participating in a second interview 3 to 4 months after their initial interview. A slightly higher proportion of TOP clients completed a second interview, as compared to PAC clients.

By the second interview many partner relationships that existed at the first interview had dissolved and new relationships had formed in some cases; apparently, changes in marital status and partner’s relationship status are linked to attrition. All adolescent respondents, aged 19 or younger, completed their second interview and only one young person under the age of 25 was lost to follow-up.

101

The interviews revealed a complex interplay between long-held, conservative religious beliefs around abortion and rationalizations that allowed most women to come to terms with their abortions and their spiritual state. while a minority continued to feel irredeemable and experience immense, ongoing guilt.

In particular, four key themes pertaining to religiosity emanated from the thematic analysis and are discussed here: the notion of abortion as a sin, God’s role in understanding future fertility, prayer and forgiveness, and changes in religious participation over time post-abortion. Several other important themes surfaced from the interviews, including relationships with male partner and family members, delays in accessing abortion services, perceptions of quality of care at the health facility, and economic ramifications of the abortion. These additional themes are only discussed in the context of this study as they pertain to religiosity and community.

Abortion as a sin

Nearly all women characterized their abortion as a sin in the eyes of God and their religion.

Though women were not asked specifically about their religious denomination, Zambian society is overwhelmingly Christian (98%), and many Christian sects view abortion as a sin or moral transgression.2-4 Several respondents noted that premarital sexual intercourse and conception outside of marriage are also sinful acts. Many women reference religious texts, namely the Bible, as evidence for the sinful nature of abortion, but more commonly, respondents have internalized the notion of abortion as a sinful act and this is supported through messages heard at church and from family, friends, and the broader community.

102

Our family church is Jehovah’s Witness and when you terminate a pregnancy it is a sin. My

mum was shouting at me saying if my dad knows he will really shout at me. My sisters said

what I did was wrong and I had committed a very big sin.

(27 years old, Divorced, Employed)

It’s not allowed. They [Christian leaders] don’t allow it, they say its sin. They say that it is

sin for a woman to terminate [a pregnancy].

(21 years old, Married, Not Employed)

However, the perceptions of the sinful nature of abortion vary widely among respondents. For some women, the sinfulness of abortion is minimized by the compounding effects of their day- to-day challenges of life, including poverty, the raising of other children, and partnership dissolution. They recognize that abortion is thought to be immoral in their religion, but their reasons for aborting were justified and thus, the magnitude of the sin is minimized.

Somehow, I think I have…I think I have killed somehow. Sometimes I encourage myself and

sometimes I feel I have really made a big sin. Maybe if that child was there today or

tomorrow that child maybe is the one who can keep me in future or maybe become president

of Zambia. But it happened this way. I feel if I was… if I had money… I was working a good

job…if I had an advantage of maybe being educated more I would have managed, that’s

what I think because if I was comfortable I could have managed being pregnant. Maybe I

would be able to provide for the baby and if my husband was there he would also provide.

103

We would both help each other and put this child on formula, good food and everything then

I would carry on with this pregnancy…now because of this life [I aborted].

(32 years old, Married, Employed)

Even in the three instances of rape, women noted that terminating a pregnancy was a sin, but seemed to also recognize that their decision to terminate was justifiable.

When it comes to religion, terminating is a sin. It is a sin but for me I know I had to sin, but it

wasn’t my fault. Even the day they gave me the pills I had to say “sorry God I have to do

this” …because I didn’t even know the father, [which] in Zambian religion is a sin.

(19 years old, Unmarried, Student; conception resulted from rape)

For a much smaller subset of women, the sin of abortion was much more severe. Several equated the termination of pregnancy with murder and felt that they deserved punishment. Others spoke of the need to atone for the sin by promising something to God in return.

You just feel that you have sin in your life. That you killed a small baby, and God is going to

punish you.

(21 years old, Married, Not Employed)

They say if you sin willfully God cannot forgive you and I have done this knowing that it’s

sin…. I will be reading the bible every day.

(Age Unknown, Unmarried, Student)

104

Future Fertility

For over half of the women interviewed a recurring theme was their future fertility as something decided upon by God. For these women, God determined the number of children a woman would have in her lifetime, though this number was not known to the woman herself. Women’s interpretation of this concept manifested very differently in how it pertained to their abortion.

Some women feared that they would not have any more children and that the fetus which was aborted had been their sole opportunity for childbearing. Women did not frame concerns about future fertility as a punishment by God, or even a source of atonement; however, they did express sadness and guilt for potentially relinquishing their only opportunity for child bearing.

I am worried. There are people who have only one child so I’m also thinking that the child I

terminated is maybe my first and last child.

(Unmarried, Age and Employment Unknown)

Interviewer: When you think of having children again how does that make you feel?

Respondent: Bad.

Interviewer: What do you mean bad?

Respondent: What if that was the only child that I was supposed to have then I mess up like

this?

(Unmarried, Employed)

Among the women interviewed who spoke about their future fertility, most felt generally optimistic about having children in the future and didn’t feel that having had an abortion would

105 change the plan intended for them. They also believed that God determined the number of children they would have in a lifetime but did not feel that anything that they did throughout their life, including having had an abortion, would change God’s plans for the number of children they would have.

I: Because of the way that you terminated your pregnancy, do you think you will have

children in future or what?

R: I will have.

I: How do you know that you will have?

R: Because God is the one that gives children.

(18 years old, Employed)

I don’t know if I will have a child or not. I have just told myself that it’s only God who knows,

having a child or not it’s only God who knows.

(Unmarried Student)

Prayer, Forgiveness, and Acceptance

Prayer arises in the transcripts as essential in framing the respondents’ abortion experiences.

Respondents report praying throughout the abortion process. Many noted praying during the procedure itself both for their own health and safety and also for the pregnancy to be successfully terminated. Praying around the initial decision to terminate the pregnancy or not did not factor into any of the interviews.

106

I had this fear I was just praying hard that please my God help me I should not die or

something bad happen to me… I wasn’t even thinking of anything I was just praying to my

God that it should be terminated.

(Unmarried, Student)

The majority of respondents presented with a common pattern in which they acknowledged abortion was a sin, prayed for forgiveness, and felt forgiven. Some of these women believed that they were forgiven as a result of God understanding and sympathizing with their uniquely challenging situations in life; however, most implied that simply praying to God for forgiveness would absolve them. This theme was noted in both first and second interviews, indicating that many respondents felt absolved of their perceived sin immediately after their abortion procedure, while the process for others required ongoing prayer and time to pass.

Ok, at first, I used to feel guilt like when at church. ok, imagine you have sinned, you have

done something very bad, and the again you pretend like you are very much committed to

church and the like, so used to feel guilt. But I learnt one thing that, no matter how big the

sin you have committed, when you ask God to forgive you, he does. So I asked for forgiveness

from God and I am sure that my God answered me. He has forgiven me, so I stopped feeling

guilt now. I just go freely.

(21 years old, Unmarried, Student)

I prayed when terminating saying, “God should forgive me” and he has forgiven me.

(28 years old, Married, Not Employed)

107

I have told my mother, “please forgive me I will never do what I have done again. I beg.”

She said, “I have forgiven you already, what you need to do is pray hard to the almighty,

because the termination you have done is a sin, God should forgive you, then that will be it, if

God forgives you then everything else will be fine, so for you what you need to do now, is

pray hard that God forgives you because this sin is huge. It’s murder.”

(Unmarried Student)

Two-thirds of the women interviewed noted feelings of shame and guilt regarding the termination of their pregnancy; however, two-thirds also implied that they felt secure in their decision to have an abortion and to no longer be pregnant. Thus, a portion of the women interviewed expressed confidence in their decision to terminate, but also expressed simultaneous feelings of shame. Women who were more reluctant to terminate their pregnancy or who were coerced by their partner or relatives seemed to experience more severe and lasting feelings of guilt. The tension between guilt and security with the decision is notably observed in the women’s’ solitary prayer behaviors. Though several respondents felt that they would ultimately be forgiven for their termination, they also articulated feelings of guilt and shame that lingered, even while praying. For some women, this guilt appeared to dissipate or was not mentioned at the second interview. Others mentioned that the guilt persisted despite ongoing prayers and simultaneous feelings of absolution.

Respondents’ prayers extended beyond their health and repentance. Many women prayed that their communities would continue to accept them after terminating their pregnancy. In cases

108 where the abortion was a secret known to few other people, women prayed that their experience would not be known by or disclosed to others as time passed.

It’s just about prayer like I earlier said, it’s just prayer… so that people who liked me don’t

run away from me.

(20 years old, Not Employed)

If you know that you have wronged and you pray to God not everyone will be talking about

it.

(20 years old, Not Employed)

Solitary prayer is an integral component of the respondents’ life. It is depicted as primarily conversational and is used to ask for aid, guidance, and forgiveness. Though community prayer at church is mentioned, it is solitary prayer which seems to be the more direct channel of communication with God, which subsequently offers great individual comfort during a complicated and stressful period of life.

Transitions in Religious Participation Over Time

Taking advantage of two rounds of interviews, changes in the participants attitudes toward religion and their religious behaviors, including the concepts in the aforementioned themes, were observable. Nearly half of the women who participated in a follow-up interview noted changes in their religious participation, notably in prayer in churchgoing behaviors.

109

For most women, their individual prayer with God does not cease as a result of their abortion. In fact, several women acknowledge that their faith and participation have become stronger over time as a result of their abortion. In the process of asking for forgiveness and feeling forgiven, their engagement in solitary and communal prayer (churchgoing) increased, even in scenarios where they initially felt much guilt and isolation.

“I would say now…I’m even more closer to God than I was before…like after the

termination, I would feel so guilty about it and would ask God to forgive me. I have been

going to church, reading the bible like that…I would say I have gotten… I’m more close to

God now.”

(21 years old, Unmarried, Student)

Other women eventually re-engage with their faith communities but note that time alone to pray and sort through notions of guilt and forgiveness is necessary prior to fully reengaging with their religious communities and previous levels of religious participation.

"Before I terminated, I was very much committed at church, so after I terminated the

pregnancy, I [had] very much guilt. So, my husband used to remind me like, ‘aren’t you

going for prayers’ I would always say ‘aa- I am not going.’ Or maybe when eating he would

ask me to pray but I used to be guilty saying ‘I will pray some other time.’ So after sometime,

I just told myself that what I did is wrong so let me just confess to my God for forgiveness

that’s when I started going for prayers. Then I started praying own my own, just like that.

Now I have continued going for prayers.”

110

(28 years old, Married, Not Employed)

I feel that maybe at the moment I have to be…I have to be home first…I want to be home first

to sort out myself… then I will go at church to apologize.

(Age and Marital Status Unknown, Employed)

Though many women successfully reengage with their religious communities, there are many women who report no longer feeling that they are a part of their church communities. In most cases, their abortions are unknown to their fellow churchgoers, and yet they still feel a sense of shame and that they are being judged by their religious communities. Many note feeling fearful that their religious community will inherently know that they have terminated a pregnancy and shame them or exclude them from participation. They report feeling the risk of stigmatization, even though none of them reported any actual backlash from their church communities nor the disclosure of their abortion to fellow church members.

Respondent: I never felt comfortable anymore at church, I would feel guilty to go to church.

It took time for me to start going at church. I didn’t expect that I would be feeling guilty. I

just thought it was going to be normal.

(19 years old, Unmarried, Employment Unknown)

All that I can say is that I feel guilty even when going to church, am not free [at ease],

thinking everyone knows whatever took place, so even when you know the way we women

111

talk someone brings up a topic you think they are setting [making fun of] you and all that

so…

(Age unknown, unmarried, employed)

I feel guilty going to church. Sometimes…like last week I went but even before the service

finished I came out. I don’t know… I didn’t feel I belonged there.

(19 years old, Unmarried)

Interviewer: How did your relationship with [church community] change or maybe on your

own, maybe the issue of going to worship?

Respondent: They are happy with me but I don’t know what they think

Interviewer: What about you, terminating the pregnancy and praying?

Respondent: To the… to the group study I don’t go, I only go on Sunday

Interviewer: How do you feel after terminating a pregnancy and then you go for prayers?

Respondent: I feel embarrassed

Interviewer: Why do you feel embarrassed?

Respondent: That they look down on me. That my prayers are fake. My prayers are fake.

(18 years old, employed)

Many of these women who feel stigmatized have not fully reengaged with their religious communities in the 3-4 months after their abortion, though in most cases, they have maintained solitary prayer as they work through their feels of guilt and isolation with the goal of ultimately reintegrating more securely into their faith communities.

112

Discussion

Most women reported feeling that abortion was a sin in their religion. Though most of the women interviewed believed that praying for forgiveness would lead to absolution, many women still noted feelings of guilt and shame. They reported feeling that their abortion behavior was non-normative and unacceptable within their religious communities, which could lead to difficulty reintegrating into their church communities.

113

Chapter Five Tables and Figures

Table 5.1 Descriptive characteristics of respondents at Time 1 and Time 2 Interviews T1 (N=51) T2 (N=38) n % n %

Type of Abortion PAC 29 56.9 20 52.6 TOP 22 43.1 18 47.4

Age 19 or younger 5 9.8 5 13.2 20-25 16 31.4 15 39.5 25+ 12 23.5 11 28.9 Age unknown 18 35.3 7 18.4

Employment Student 22 43.1 10 26.3 Employed 12 23.5 11 28.9 Not Employed 12 23.5 12 31.6 Unknown 5 9.8 5 13.2

Marital Status Married 8 15.7 8 21.1 Unmarried 30 58.8 20 52.6 Divorced 6 11.8 2 5.3 Widowed 1 2.0 1 2.6 Unknown 6 11.8 7 18.4

Living Arrangement Alone 5 9.8 3 7.9 Biological parents 17 33.3 9 23.7 Non-parent family 10 19.6 6 15.8 Partner 5 9.8 7 18.4 Unknown 14 27.5 13 34.2

Relationship Status of Partner Partner - married 8 15.7 6 15.8 Partner - unmarried 19 37.3 13 34.2 Partner – other relationship 7 13.7 6 15.8 Sugar daddy 4 7.8 1 2.6 Rape 3 5.9 1 2.6 Unknown 10 19.6 11 28.9

114

Chapter Five References

1. Ipas. Postabortion Care. 1998-2018; Available from: http://www.ipas.org/en/What-We- Do/Comprehensive-Abortion-Care/Postabortion-Care.aspx. 2. Central Statistical Office (CSO) [Zambia], Ministry of Health (MOH) [Zambia], and ICF International, Zambia Demographic and Health Survey 2013-14. 2014, Central Statistical Office, Ministry of Health, and ICF International: Rockville, Maryland, USA. 3. Stephens, M., et al., Religious perspectives on abortion and a secular response. Journal of religion and health, 2010. 49(4): p. 513-535. 4. Tomkins, A., et al., Controversies in faith and health care. The Lancet, 2015. 386(10005): p. 1776-1785.

115

CHAPTER SIX : CONCLUSION

116

Introduction

Unsafe abortion is a critical public health issue that disproportionately affects the morbidity and mortality of women living in sub-Saharan Africa.1-3 Even in Zambia, where abortion laws are among the least restrictive on the continent, an estimated 30% of maternal deaths are a result of unsafe abortion.4 Thus, it is likely that stigma reinforces the idea that abortion is deviant, thus a desire to keep the act of abortion secret, even if that means resorting to unsafe, clandestine practices. As religion and religious institutions have been previously associated with perpetuating stigma, examining the role that religion plays in this cycle is necessary.

This dissertation examined the association between religiosity and abortion perceptions among women in three Zambian provinces. The mixed methods approach allowed the opportunity to link the results of cross-sectional survey data from women of reproductive age with in-depth interviews of women from the same populations who had recently had abortions. In this chapter, the findings for each aim are summarized followed by overall conclusions, strengths, limitations, and implications.

Overview of Aim One

Chapter Three aimed to assess the validity and psychometric properties of the Duke University

Religion Index (DUREL) and sought to construct scales based on five key questions assessing individual attitudes regarding abortion.

The DUREL had not been previously validated within the Zambian context. The DUREL showed good internal consistency, as it has in a variety of other populations.5-7 However, the

117 results of the EFA and CFA suggest that DUREL cannot be used as a scale to measure religiosity within this population, as the subscales of the DUREL are not explained by a common underlying construct, as evidenced by goodness of fit statistics and the poor factor loadings of

OR and NOR. The findings do support the validity of the three-item subscale measuring intrinsic religiosity. The validity assessment of the DUREL supports the current literature and the recommendation that the DUREL be analyzed as three independent subscales and not an aggregate measure of religiosity.8

Within this female Zambian population, the five survey questions pertaining to abortion attitudes do not function together as a valid scale. To our knowledge, these questions have not been previously validated as a scale in any context and were not selected for this purpose. They represent particular domains of beliefs and attitudes regarding abortion: morality, women’s rights, access to care, and support for legality and were a straightforward way to assess change in specific attitudes over time using the endline evaluation from which these data origniated.9, 10

As a result of findings in Chapter Three, the key independent variable of religiosity was measured via the three separate constructs of the DUREL: organizational religiosity (frequency of churchgoing), non-organizational religiosity (time spent alone in prayer), and intrinsic religiosity (degree of personal religious commitment or motivation).8 Abortion attitudes were analyzed as individual outcome variables and not as a unified scale.

118

Overview of Aim Two

This study is among the first to quantitatively explore associations between religiosity and abortion attitudes in a sub-Saharan African context. The findings from this study are particularly notable, as they do not show an association between religiosity and abortion perceptions at an individual level but find associations between religiosity at a community level and individual abortion perceptions, implying that a woman’s own religious beliefs and practices are less relevant in influencing her attitudes and beliefs regarding abortion than the broader religious beliefs and practices of the community where she lives. This suggests that it is not religion itself driving abortion attitudes, but rather stigma generated from individual interpretations of community religious norms.

Women living in communities where the frequency of church attendance is high are less likely to personally support a woman’s right to decide whether to continue or terminate a pregnancy and are also less likely to support the legal permissibility of abortion. Similarly, women who live in communities where intrinsic religiosity is high among the population are less likely to themselves support access to abortion for all women, married and unmarried. Non-organizational religiosity at a community level, i.e. solitary prayer, has the opposite effect on individual abortion perceptions. Women living in communities where other women report higher levels of solitary prayer are more likely to have more progressive attitudes toward abortion access, including for unmarried women. Where women might note the observable piety of others, it is impossible to observe if those around you engage solitary prayer; thus, it is less likely to drive stigma. In supporting this premise further, previous studies have suggested that prayer is not associated with abortion stigma and is linked to easing of guilt associated with abortion among

119 women who have had abortions, thus suggesting that prayerful women might also be generally more sympathetic to the struggles of others.11-13

Beyond community religiosity, living in a community where there is strong support for a more permissive legal framework for safe abortion is associated with heightened individual support for all four measures of abortion attitudes- right to continue pregnancy, right to terminate pregnancy, access to safe abortion, and access to safe abortion for unmarried women. Living in a community where people are more accepting of abortion and want it to be legal and accessible is associated with more liberal personal attitudes toward abortion. This is potentially explained by a disruption of the cycle of stigma whereby abortion is normalized in a community and thus no longer perceived as deviant.14, 15

The significance of covariates varied in significance and association across the analyses of multiple outcomes. In cases of statistical significance, younger, unmarried, more educated women living in cities had more liberal attitudes toward abortion. Wealth was negatively associated with more liberal abortion attitudes pertaining to an unmarried woman’s right to safe abortion but was positively associated with greater legal permissibility of abortion. The associated covariates are generally consistent with previous literature on facilitators and barriers to reproductive health attitudes and behaviors in sub-Saharan Africa, notably in regards to contraceptive use.2, 9, 10, 16-18

120

Overview of Aim Three

This thematic analysis of in-depth interviews at two time points with Zambian women who had an abortion uncovered both immediate and longer-term attitudinal and behavioral responses to abortion in the context of their religious beliefs. A total of 51 women were interviewed immediately following their abortion procedure and prior to discharge from the health facility.

Just over half of the women presented for postabortion care (PAC) services and the reminder presented for termination of pregnancy (TOP). The 51 participants ranged in age from 16 to 32 years and just under half of all women interviewed were students.

Four key themes surfaced in the analysis of the interviews: the notion of abortion as a sin, God’s role in understanding future fertility, prayer and forgiveness, and changes in religious participation over time post-abortion. Nearly all women categorized abortion as a sin within their religion and believed themselves to have sinned. Though a few women cited the Bible as evidence for the sinful nature of abortion, most respondents had internalized the notion of abortion as a sinful act and this notion was supported through messages heard at church and from family, friends, and the broader community.

Many women believed their future fertility to be determined by God and some worried that the fetus they had aborted was their only opportunity for child-bearing. This concept was not seen as punishment by God, but rather something that was predetermined, unknown to them, and outside of their control.

121

Prayer was a fundamental element of most women’s abortion experiences and respondents noted praying throughout the process- initially for their own health and safety and later, postabortion, for forgiveness. The majority of respondents, acknowledging abortion as a sin, reported praying for forgiveness and feeling absolved. Women also prayed for acceptance and smooth reintegration into their communities. Though women generally felt forgiven for their perceived sin, several noted they still struggled with guilt, despite most feeling secure in their decision to have had an abortion.

Some women struggled to reintegrate into their religious communities and often used solitary prayer as a means to resume participation in religious activities. Though their abortions are nearly all secret and unknown to their religious communities, respondents still felt a sense of shame and judgement. Many feared that their religious community would innately know that they had terminated a pregnancy and publicly shame them or exclude them from participation.

They reported feeling the risk of stigmatization, even though there is no mention of backlash or disclosure of abortion to religious communities.

Overall, most women reported feeling that abortion is a sin in their religion. Though most believed that praying for forgiveness would lead to absolution, many women still noted feelings of guilt and shame. Respondents feel that their abortion was non-normative and unacceptable within their religious communities, which could lead to difficulty reintegrating.

Overall Conclusions of the Dissertation

When performed in accordance with WHO procedures, abortion is a safe, low-risk medical procedure. However, when abortions are provided by insufficiently trained individuals using

122 dangerous methods in unsanitary conditions, a woman’s health and life are put at great risk.19 In sub-Saharan Africa, more than 97 percent of abortions received by women are unsafe and unsafe abortion is responsible for at least 9 percent of all maternal deaths.1-3

Even in Zambia, where abortion laws are less restrictive than in neighboring countries still approximately 30% of the annual 398 maternal deaths per 100,000 live births are due to unsafe abortion, suggesting that this public health issue extends beyond mere access to safe abortion services.4 The literature suggests that the stigma surrounding abortion drives women to seek clandestine abortions outside of the formal health sector to ensure that their procedure remain secret.20, 21

Stigma is an outward manifestation of a community’s social norms and involves negative judgement and shaming of behaviors considered non-normative 22 When living in an environment where abortion is highly stigmatized, women are willing to risk their health rather than suffer social consequences.23, 24 Further, abortion is highly stigmatized by most major religions, especially Christianity. Several studies have shown that religion can drive shame- related stigma around various health issues; however, the impact of religion on abortion attitudes and behavior in sub-Saharan Africa has not been widely studied.25, 26

This study found that, indeed, higher levels of community religiosity in the form of churchgoing and intrinsic religiosity are associated with more conservative attitudes on abortion. Further, qualitative findings revealed that women were often reluctant, in the short-term, to reengage with religious communities after having an abortion for fear that they would be judged or their secret

123 would be revealed. This supports the notion that outward manifestations of religiosity aggregated at a community level propel the prevalence paradox as described in Chapter One. The findings suggest that religion may be a contributing factor in perpetuating the social norm that abortion is deviant and thus women fear stigmatization and underreport their abortions. Thus, much like the cycle of stigma proposed by Kumar et al. and elaborated upon in Chapter One, abortion continues to be thought of as uncommon (non-normative), and the cycle of stigma continues and is reinforced14

The study did not find associations between individual religiosity and abortion perceptions, indicating that it is not religious doctrine or one’s own relationship with God that leads to more conservative viewpoints, but rather one’s residence in communities where outward manifestations of religiosity are high and people express more conservative personal attitudes toward abortion. The qualitative interviews supported this notion, with the female participants speaking of their close and familiar relationship with God, via prayer and forgiveness, but often feeling a great deal of guilt or shame in terms of how they would be perceived socially.

Though living in communities with higher reported churchgoing and intrinsic religiosity was associated with more conservative abortion perceptions, living in communities where people reported greater frequency of solitary prayer had the reverse pattern. Where time spent alone in prayer was greater, women were more likely to have liberal perceptions around abortion. Women interviewed after their abortions noted that solitary prayer was essential to their feelings of forgiveness and redemption, as well as their reintegration into their communities. Praying alone is a private activity, unlike churchgoing and intrinisic religiosity, which can be observed in other

124 people. This aligns with previous studies which have noted that prayer facilitates empathy and reduces emotional reactivity and bias towards others, which could potentially lead to greater feelings of acceptance within the community and liberalized abortion attitudes for individals.27, 28

In both quantiative and qualitative studies, women asserted that abortion was immoral. Among women surveyed, only 3% disagreed with the notion that abortion is immoral and nearly all those interveiwed mentioned the immorality of abortion, as it pertained to both their religion and their communities. However, women’s perceptions of abortion beyond immorality were very nuanced and liberal. This suggests that despite conservative interpretations of abortion by religious groups and individuals, there remains open thinking around abortion, particualry as it pertains to access to safe abortion.

Although this is the first study of its kind, this research suggests that religiousity at a community level may play a role in perpetuating abortion stigma and influencing abortion perceptions for women. The qualitative research suggests that women perceive that they are being judged by their communities, notably their religious communities, even in situations where their aboritions were unknown to anyone but themselves. Interventions to reduce stigma are currently minimal.21

This study provides initial evidence that community religiosity and attitudes may motivate abortion stigma. Further research is need to in sub-Saharan Africa to confirm these findings and better understand this important social dynamic. This pathway cannot be ignored when addressing the public health risks assocaited with unsafe abortion.

125

Strengths and Limitations

This study is the first to examine the role of religiosity in attitudes and understanding around abortion in sub-Saharan Africa. Though it will contribute important evidence to the study of abortion in religious contexts, the study also has inherent limits.

Limitations

Secondary Data Analysis

Because the data used for this study were collected for another purpose, they do not contain all the information relevant to stigma religion and abortion. Mainly in the qualitative interviews, certain questions were not asked regarding religiosity and probing was not conducted, as the goals of the original research, a program evaluation, were more general than the goals of this study which examines the impact of religiosity and social stigma on abortion attitudes and knowledge. The survey questionnaire did not include a validated scale to measure abortion stigma,29 as it would have substantially lengthened the time of the survey. Thus, the findings of this study are constrained by the data collected and measures used for the original investigation.

Cross-Sectional Data

The quantitative data in this study are cross-sectional and thus reflect the respondents’ inputs at one single time point. Establishing causal linkages is not possible. Longitudinal data would provide a better sense of respondents changing attitudes over time and would also allow for a better assessment of reliability and validity of the scales.

126

Lack of Direct Measure of Abortion Stigma

Though there are validated scales to measure abortion stigma, they were not able to be included in the population-based survey.29 To that end, community stigma is gauged by an aggregation of individual abortion perceptions at the cluster level of sampling. Though this is a proxy for measuring community attitudes it is less precise that directly measuring individual and community levels of abortion stigma.

Social Desirability Bias

Social desirability bias is a concern within the survey and in the qualitative interviews. Because abortion is highly stigmatized in Zambia, it is possible that reporting of abortion attitudes in the survey might skew women toward reporting more negative attitudes and desire for stricter laws in the survey. Similarly, the women who participated in the qualitative interviews might emphasize certain elements of their abortion experience, knowing it is a socially stigmatized behavior, which is a reality for all studies on abortion.

Lack of Objective Measures of Abortion Behavior

Regardless of whether abortion is legal or illegal in a country, the under-reporting of induced abortions in surveys is largely prevalent. Previous studies show that the extent of under-reporting on health-related outcomes is substantial, particularly in regards to sexual and reproductive health.30-32 This phenomenon is particularly relevant for abortion, where studies that compare survey responses to medical records show high levels of abortion under-reporting.33-35 In countries where abortion is legal, directly asking women about their abortion experience resulted

127 in under-reporting of up to 70%.36 In countries where abortion is not fully legal, the underreporting is likely higher.1

Due to a small sub-sample of women reporting any abortion behavior combined with the problematic nature of asking questions about abortion in administered surveys, I eliminated the variable from my analysis. Thus, I am unable to examine the relationship between religiosity and abortion behavior, only knowledge and attitudes.

Strengths

Comprehensive measure of religiosity

The DUREL is a measure of religiosity that incorporates the major dimensions of religiosity, has high validity and reliability, and has been confirmed in a variety of samples by independent investigative teams. There have not been studies to date that examine associations between religiosity and abortion in sub-Saharan Africa and the DUREL allowed for a concise set of questions to be added to a survey to provide a measure of religiosity.

Population-based sample

This study uses a reasonably large, population-based sample in a typically hard-to-reach population. This is particularly useful in a population skewed toward higher levels of religiosity, as without a large sample, power would be diminished.

Longitudinal qualitative data

It is rare to have qualitative interviews with the same respondents at two points in time with minimal attrition, particularly for groups at risk of experiencing high levels of social stigma.

128

Mixed methods

This study includes both quantitative and qualitative methods, which allow for the findings to be representative of the overall population, while ensuring that the experiences of individual women to are heard. Though the qualitative and quantitative data are not precisely parallel populations, as the women interviewed post-abortion were not a part of the survey, the qualitative interviews complement the quantitative data and provide greater context to the overall findings. Since abortion is highly stigmatized, the qualitative interviews are important for eliciting information that is much more nuanced and detailed than what is learned from a survey.

Research Implications

Associations between religiosity and abortion attitudes and behaviors have not been previously studied in a sub-Saharan African context. This study shares evidence to suggest that community level stigma generated by observable religiosity drives more conservative attitudes regarding abortion than one’s individual religiosity. Further studies are needed both inside and outside of

Zambia to confirm, better measure and understand this phenomenon and how to best address it through more targeted programmatic and policy interventions.

Measures of religiosity, including the use of the DUREL, should be further developed and regularly included in population-based surveys addressing reproductive health outcomes.

Because of religion-driven stigma, both real and perceived, that exists in the context of sensitive issues, including abortion, contraception, and childbearing, it is important that the role of religion in health is measured when possible and that improved measures are developed in the process.

This particular study focused on a population identifying as Christian, both Protestant and

129

Catholic, in a country with a notably conservative religious nature. In countries where religious identity and participation is more heterogenous, it is important to ensure that measures for religiosity are also sensitive enough to capture nuance. It is easy to rely on age-old tropes to explain the role of religion in health; however, a broader evidence base can and should be developed.

Similarly, there is a need to further develop valid measures of abortion attitudes and to use these measures within population-based surveys. The questions used in this study were used in order to be consistent with two previous studies but were not validated at any point as a scale.

Investigating more valid and comprehensive measures of abortion perceptions would contribute greatly to the knowledge base and understanding what drives abortion attitudes and stigma at a personal level. A 18-item validated scale to measure abortion stigma already exists and should be used in the context of population-based community surveys to better understand stigma. 29

Public Health Implications

The effect of religiosity on abortion attitudes and legal knowledge has been profoundly understudied, particularly in the sub-Saharan African context. Because both religion and abortion are both deeply personal experiences highly influenced by socio-cultural dynamics, they are commonly neglected in mental and public health interventions as to avoid the possibility of raising sensitivities or causing discomfort among key stakeholders, donors, or governments.

Though religion and abortion are so often perceived as sensitive and personal topics, both are important elements in the daily lives of most people throughout the world.

130

However, with the improved understanding of the role that religiosity plays in shaping attitudes and understanding around abortion, programmatic interventions related to unsafe abortion can more accurately identify both the barriers and opportunities that faith communities create for health improvements at a population level. By working with religious leaders and religious communities and incorporating messages into public health campaigns that acknowledge the important role that religion plays in the lives of many, the health risks of unsafe abortion can be better understood and improved upon within communities.

This study sought to separate the effects of religiosity and community stigma in order to better understand the interplay between religion and stigma at both the individual and community levels and finds that communal aspects of religiosity have a greater influence on attitudes regarding abortion than one’s own religiosity

This study aims to inform other public health researchers and health practitioners working in the realm of reproductive health to acknowledge and address religion in our communities without hesitation. Further, it aims to support national-level reproductive health policy strategies that can reduced the prevalence of unsafe abortion.

Conclusion

Unsafe abortion is an important public health issue globally and accounts for a large proportion of maternal deaths, especially in environments where safe abortion services are not legally permissible or, in the case of Zambia, highly stigmatized. Zambia is predominantly Christian and is a highly religious country. This dissertation provides evidence that religiosity at a community

131 level is associated with women’s own attitudes toward abortion and the extent to which they believe it should be legal in their country. Religiosity at an individual level is not associated with these outcomes of interest, suggesting that stigma is driven by outward expressions of religion as opposed to individual religious beliefs and practices. Our results can be used to develop improved culturally-relevant measures of abortion attitudes and behaviors and to inform further research and program and policy interventions that address this important health issue.

132

Chapter Six References

1. Sedgh, G., et al., Induced abortion: incidence and trends worldwide from 1995 to 2008. The Lancet, 2012. 379(9816): p. 625-632. 2. Singh, S., J. Darroch, and L. Ashford, Adding it Up: The Costs and Benefits of Investing in Sexual and Reproductive Health 2014. Guttmacher Institute, 2014. 3. Singh, S. and I. Maddow-Zimet, Facility-based treatment for medical complications resulting from unsafe pregnancy termination in the developing world, 2012: a review of evidence from 26 countries. BJOG, 2015. 4. Standards and Guidelines for Reducing Unsafe Abortion Morbidity and Mortality in Zambia, Government of the Republic of Zambia Ministry of Health, Editor. 2014: Lusaka, Zambia. 5. Lucchetti, G., et al., Validation of the Duke Religion Index: DUREL (Portuguese version). J Relig Health, 2012. 51(2): p. 579-86. 6. Dobrowolska, B., et al., Validation of the Polish version of the Duke University Religion Index (PolDUREL). Pol Arch Med Wewn, 2016. 126(12): p. 1005-1008. 7. Hafizi, S., et al., The Duke University Religion Index (DUREL): validation and reliability of the Farsi version. Psychol Rep, 2013. 112(1): p. 151-9. 8. Koenig, H.G. and A. Büssing, The Duke University Religion Index (DUREL): A Five- Item Measure for Use in Epidemological Studies. Religions, 2010. 1(1): p. 78-85. 9. Geary, C.W., et al., Attitudes toward abortion in Zambia. International Journal of Gynecology & Obstetrics, 2012. 118: p. S148-S151. 10. Cresswell, J.A., et al., Women's knowledge and attitudes surrounding abortion in Zambia: a cross-sectional survey across three provinces. BMJ Open, 2016. 6(3): p. e010076. 11. Harries, J., et al., Delays in seeking an abortion until the second trimester: a qualitative study in South Africa. Reprod Health, 2007. 4: p. 7. 12. McMurtrie, S.M., et al., Public opinion about abortion-related stigma among Mexican Catholics and implications for unsafe abortion. International Journal of Gynecology & Obstetrics, 2012. 118: p. S160-S166. 13. Adamczyk, A., The effects of religious contextual norms, structural constraints, and personal religiosity on abortion decisions. Social Science Research, 2008. 37(2): p. 657- 672. 14. Kumar, A., L. Hessini, and E.M. Mitchell, Conceptualising abortion stigma. Cult Health Sex, 2009. 11(6): p. 625-39. 15. Izugbara, C.O., C. Egesa, and R. Okelo, 'High profile health facilities can add to your trouble': Women, stigma and un/safe abortion in Kenya. Soc Sci Med, 2015. 141: p. 9-18. 16. Elfstrom, K.M. and R. Stephenson, The Role of Place in Shaping Contraceptive Use among Women in Africa. PLOS ONE, 2012. 7(7): p. e40670.

133

17. Sumer, Z.H., Gender, Religiosity, Sexual Activity, Sexual Knowledge, and Attitudes Toward Controversial Aspects of Sexuality. J Relig Health, 2015. 54(6): p. 2033-44. 18. Bankole, A., et al., Differences in unintended pregnancy, contraceptive use and abortion by HIV status among women in Nigeria and Zambia. International Perspectives on Sexual and Reproductive Health, 2014. 40(1): p. 28-38. 19. Ganatra, B., et al., Global, regional, and subregional classification of abortions by safety, 2010–14: estimates from a Bayesian hierarchical model. The Lancet, 2017. 390(10110): p. 2372-2381. 20. Lithur, N.O., Destigmatising abortion: expanding community awareness of abortion as a reproductive health issue in Ghana. African journal of reproductive health, 2004. 8(1): p. 70-74. 21. Hanschmidt, F., et al., Abortion Stigma: A Systematic Review. Perspect Sex Reprod Health, 2016. 48(4): p. 169-177. 22. Cockrill, K. and L. Hessini, Introduction: Bringing Abortion Stigma into Focus. Women & Health, 2014. 54(7): p. 593-598. 23. Cockrill, K. and A. Nack, “I'm Not That Type of Person”: Managing the Stigma of Having an Abortion. Deviant Behavior, 2013. 34(12): p. 973-990. 24. Cockrill, K., et al., Addressing Abortion Stigma Through Service Delivery. 2013. 25. Varas-Díaz, N., et al., Religion and HIV/AIDS stigma: Implications for health professionals in Puerto Rico. Global Public Health, 2010. 5(3): p. 295-312. 26. Zou, J., et al., Religion and HIV in Tanzania: influence of religious beliefs on HIV stigma, disclosure, and treatment attitudes. BMC public health, 2009. 9(1): p. 1. 27. Butler, M.H., B.C. Gardner, and M.H. Bird, Not just a time-out: change dynamics of prayer for religious couples in conflict situations. Fam Process, 1998. 37(4): p. 451-78. 28. Vasiliauskas, S.L. and M.R. McMinn, The effects of a prayer intervention on the process of forgiveness. Psychology of Religion and Spirituality, 2013. 5(1): p. 23-32. 29. Shellenberg, K.M., L. Hessini, and B.A. Levandowski, Developing a Scale to Measure Stigmatizing Attitudes and Beliefs About Women Who Have Abortions: Results from Ghana and Zambia. Women & Health, 2014. 54(7): p. 599-616. 30. Glynn, J.R., et al., Assessing the validity of sexual behaviour reports in a whole population survey in rural Malawi. PLoS One, 2011. 6(7): p. e22840. 31. Schroder, K.E., M.P. Carey, and P.A. Vanable, Methodological challenges in research on sexual risk behavior: II. Accuracy of self-reports. Ann Behav Med, 2003. 26(2): p. 104- 23. 32. Moseson, H., et al., Reducing under-reporting of stigmatized health events using the List Experiment: results from a randomized, population-based study of abortion in Liberia. Int J Epidemiol, 2015. 44(6): p. 1951-8. 33. Anderson, B.A., et al., The validity of survey responses on abortion: evidence from Estonia. Demography, 1994. 31(1): p. 115-32.

134

34. Jones, E.F. and J.D. Forrest, Underreporting of abortion in surveys of U.S. women: 1976 to 1988. Demography, 1992. 29(1): p. 113-26. 35. Jones, R.K. and K. Kost, Underreporting of induced and spontaneous abortion in the United States: an analysis of the 2002 National Survey of Family Growth. Stud Fam Plann, 2007. 38(3): p. 187-97. 36. Jagannathan, R., Relying on surveys to understand abortion behavior: some cautionary evidence. Am J Public Health, 2001. 91(11): p. 1825-31.

135

APPENDIX

136

APPENDIX A: In-Depth Interview Questionnaire

1. Can you tell me—was this a planned pregnancy? That is, did you want another child at the time the pregnancy occurred? 2. Were you using any method of contraception at the time you became pregnant? IF YES: a. What method were you using? How long had you been using it? b. Tell me more about what your experiences were like using that method. c. What do you think went wrong? IF NO: d. What were your reasons for not using any family planning method? (Probe on lay beliefs) 3. Were there any changes in your life in the last 12 months that made it especially hard for you to take care of yourself? (Probe on if she lost her job, if she had a baby in the last 12 months, if she moved more than twice, if she got married/divorced, if she was the victim of a crime, if she fell behind on her bills, if she fell behind on her rent/house payments, if she was using drugs/alcohol) 4. When was your last period? a. Please tell me reasons that led to you having the termination at that point in your pregnancy. (Probe more about reasons for delay) 5. When did you realize that you didn’t want to have the baby? What were the circumstances that made you realize that? Is there any health reason that has led you to decide you didn’t want the baby? a. Who did you talk to about your desire to end the pregnancy that you recently ended? Who was consulted? b. How was the decision made to consult him/her/them? c. What advice did they give you? (Probe on each person respondent consulted) d. Please tell me about how the decision was made to seek the termination. e. What were the greatest influences for you in deciding to go where you did for your termination? 6. The following questions ask about how you felt yesterday on a scale from 0 to 10. Zero means you did not experience the feeling “at all” yesterday while 10 means you experienced the feeling “all of the time” yesterday. I will now read out a list of ways you might have felt yesterday. A. How about happy? [0-10] Please tell me more. B. How about worried? [0-10] Please tell me more. C. How about depressed? [0-10] Please tell me more.

7. What did you anticipate your experience would be like? a. Did you have any fears related to the termination?

137

b. What consequences did you anticipate experiencing from having a termination? 8. Once the decision was made to terminate the pregnancy, please tell me how you went about what to do. How much time passed before you did something? What did you do first? IF SELF-INDUCED: a. Tell me about what made you decide to try to do something to yourself to end the pregnancy. b. What was used to try and terminate the pregnancy? c. Did you try to induce the termination by taking any pills? Please tell me more. (Probe on what kind of pills she took, where she got the pills, how she took them, who instructed her on how to take them.) d. Was medication purchased from a chemist? i. Who purchased the medication? ii. Did that person purchase what you wanted him/her to? e. Please tell me about your experience with obtaining and using the drugs. What kind of complications did you experience? IF WOMAN WENT TO ANY PROVIDER: a. Who made the decision to go to this provider? What were the reasons for choosing this particular provider? b. Please tell me about your journey getting to the provider. How long did it take to travel there? Did anyone accompany you? c. Were you given a choice of abortion method when you came here? d. How do you feel you were treated by the person who provided the termination? What makes you say that? Were you given pain killers for your termination? What were you given? 9. Were additional steps required to complete the termination? Please tell me about those. 10. How long have you been at this facility? How long did you have to wait to obtain care once you arrived at the facility? Why do you think you were not attended to sooner? 11. Who accompanied you here? Please tell me how it came to be that [PERSON] accompanied you to the facility. How difficult has it been for [PERSON] accompanying you? (Probe on time off of work, responsibilities deferred, costs incurred) 12. Please tell me about the costs you or anyone supporting you has incurred to date to obtain this termination, including monies you or anyone supporting you spent on any step(s) you took before getting to this facility. (Probe on out-of-pocket costs of medical care, medical supplies woman had to bring with her, food and accommodation that possibly had to be purchased for her or anyone who may have accompanied her, as well as transport/transfers.) a. What costs have you incurred to obtain care at this facility? How were the costs of services provided by the health facility communicated to you? Did the provider explain what these fees were? Did the provider explain why s/he was asking for this fee?

138

13. Where did the money come from to pay for those costs? a. Did you have to borrow any money? b. Did you or your guardian have to forego/delay paying for other household costs because of these costs? (Probe on food security, children’s schooling and children’s care) c. Did you or your guardian have to sell any household assets/property, farm/business assets or an entire farm/business itself to come up with the money? d. How long did it take for you or your guardian to assemble these funds? e. Have these costs left you without any savings? What do you think the effects of that will be? 14. How much do you think is the appropriate amount that women and their families should have to pay for an abortion? 15. Prior to coming to this facility, please tell me about any challenges you’ve had carrying out your regular activities due to the pregnancy. (Probe on childcare, household responsibilities, and school/work.) How has this affected your bedroom issues? 16. Did you or any of your household members lose any income during this time as a result of the decision to seek a pregnancy termination? Altogether, how much would you say that your household lost in income as a result? 17. Did anyone in your household have to forgo going to school during this time as a result of the decision to seek a pregnancy termination? Please tell me more. (Probe on how many children, for how long, etc.) 18. Now I’d like to speak to you about consequences of your termination of the pregnancy that don’t involve your health or finances. When I say consequences, I don’t mean only negative consequences, I mean positive consequences, too. Who knows you’ve had this termination? a. Who has your relationship changed with as a result of your termination? In what ways? (Probe on partner, in-laws, relatives, friends, school teachers, work mates) Do you think it will change with anyone else? In what way? i. Since your family knows, what will happen when you go back home? How does that make you feel? ii. What do you think your family or anyone in your community will do as a result of your termination? How does that make you feel? b. What has your family done/will they do as a result of your abortion? What are the cultural traditions will bring to bear on this event? c. How has your termination affected how you relate to others? (Probe on partner, children, other family members, community, religion) How do you think it will affect how you relate to others? d. Do you think anyone has made you feel bad for having decided to have a termination? Have you experienced any physical or verbal abuse that you think is a result of the fact that you had a termination? e. Have you felt especially supported by anyone in relation to your abortion? Can you tell me more about this?

139

f. How has the abortion made you feel about having more children in the future? (Probe on concerns about getting pregnant again in the future.) g. How has the abortion changed how you think of yourself? (Probe on positive and negative effects.) How does your religion affect how you think about this choice and what it means about you? h. Did anyone have to lie to cover up the termination? Please tell me more. 19. Do you intend to start a contraceptive method at this time? a. Which method? (Probe on woman’s reasons for choosing that method) b. How will you go about getting it? c. What barriers do you anticipate experiencing in relation to getting the method? d. What barriers do you anticipate experiencing in relation to using the method? e. Did you feel any pressure/coercion to select a method? f. Did you feel any pressure/coercion to begin using a method? 20. Please tell me: what would your life have been like if you had not had this abortion? 21. Have you heard about any laws regarding abortion in Zambia? a. What is your understanding of Zambia’s abortion law? b. How did this influence your decision to terminate the pregnancy? c. How did this influence your decision on where to terminate the pregnancy?

140

APPENDIX B: Coding Scheme for Qualitative Data

1. Response to pregnancy a. Abuse - verbal b. Abuse - physical c. Pressure to continue 2. Health complications a. HIV+ b. Late term abortion 3. Not first abortion 4. Reason for abortion a. School b. Job c. Financial d. Family e. Other child f. Infidelity g. Partner does not want pregnancy or baby 5. Partner reason for abortion a. School b. Job c. Financial d. Family e. Other child f. Infidelity g. Doesn't want child 6. Non-consensual abortion 7. Partner's role a. Helped procure materials-medicine b. Paid for abortion c. Advice d. Supported mother's decision e. Partner not informed 8. Lack of reproductive health knowledge 9. Abortion story 10. Assistance-support-advice a. Parent(s) b. Partner's Role c. Other Family-Friend d. Other 11. Fear a. Abandonment

141

b. Fear of dying 12. Community information 13. 1st Delay-deciding to seek medical care a. Fear of questioning by providers b. Fear of arrest c. Getting money d. Recognizing severity 14. 2nd Delay-reaching appropriate obstetric facility a. Lack of trained providers b. Distance c. Transportation 15. 3rd Delay-at facility a. Long wait time b. Insufficient staff 16. Quality of Care a. Kind b. Informative c. Trustworthy d. Good reputation-known for services e. Unkind f. Uninformative g. Satisfaction with services h. Dissatisfaction with services i. Confidentiality violation 17. Debt accrued 18. Perceived benefits of having abortion a. Education opportunities b. Employment opportunities c. Investment in health 19. Physical a. Continued medical complications from abortion b. Fear of future fertility (medical) c. Abuse - verbal d. Abuse - physical e. Missed work 20. Emotional a. Shame b. Loneliness c. Regret d. Depression-anxiety e. Relief f. Secure

142

g. Productive 21. Social a. Isolation b. Lies to family and friends c. Secrets d. Blame e. Exclusion f. Mocking g. Reprimanding h. Conflict with partner i. Pregnancy is for married people 22. Barriers to contraceptive use a. Partner b. Access c. Plans for celibacy d. Stigma (i.e. perceptions of promiscuity) e. Perceived side effects f. Want child 23. Using contraception during conception 24. Start using contraception a. Implant b. IUD c. OCP d. Injectable e. Condoms 25. Plans to use contraception in future a. Implant b. IUD c. OCP d. Injectable e. Condoms 26. Non-consensual contraception 27. Discontinue contraception 28. Perception of law a. Completely illegal b. Legal in some cases c. Doesn't know the law 29. Fear of being arrested 30. Denomination a. Protestant b. Catholic c. Evangelical

143

31. Abortion is a sin 32. God can forgive a. Feel forgiven b. Do not feel forgiven c. Penance 33. God gives a set number of children 34. Feeling of hypocrisy 35. Church attendance 36. Prayer 37. Faith 38. Perceptions of judgment from religious community 39. Follow-up visit 40. Resumption of sexual activity 41. Relationship status with partner a. Still together b. Dissolution c. Poor quality d. New partner 42. Expansion of people who know about abortion 43. Change in religious participation 44. Change in employment 45. Notable change in narrative from T1 46. Key quotes

144

APPENDIX C: Institutional Review Board Determination

145

CURRICULUM VITAE

146

MEGHAN C. GALLAGHER 38 Wyckoff Street, Apartment 3L Brooklyn, NY 11201 (302) 668–7352 [email protected]

EDUCATION

2013 - 2018 PhD, Population and Health Department of Population, Family and Reproductive Health Johns Hopkins University Bloomberg School of Public Health Methods Concentration: Demographic Methods, Study Design Certificates in Demographic Methods and Population & Health

May 2007 Master of Public Health (MPH) Forced Migration and Health Program, Heilbrunn Department of Population and Family Health Mailman School of Public Health, Columbia University

May 2002 Bachelor of Science (BS), Foreign Service. Minor: Justice and Peace Studies Edmund A. Walsh School of Foreign Service, Georgetown University

PROFESSIONAL EXPERIENCE

July 2016 Senior Specialist, Monitoring and Evaluation – Present Emergency Health and Nutrition, Save the Children USA • Lead the development of the monitoring and evaluation framework for the Reproductive Health in Emergencies initiative • Provide monitoring and evaluation support to field programs and capacity building of country office staff. • Establish Health Management Information Systems (HMIS) for data collection in fragile states. • Build an institutional culture of program documentation and research.

Sept 2013 Research Assistant - Aug 2013 Johns Hopkins University Bloomberg School of Public Health • UN Population Division – WHO Department of Reproductive Health and Research Global Abortion Policies Project: Research abortion policies in francophone African countries. • Bill and Melinda Gates Institute for Population and Reproductive Health: Conduct research on family planning policies and demographic trends in the Democratic Republic of Congo, Burkina Faso, and Senegal (2013 – 2015). • National Children’s Study Life Course Health Science Working Group: Assist in developing longitudinal measures for child health within a life course framework (2014 – 2015).

Sept 2008 Senior Research, Monitoring, and Evaluation Officer - Aug 2013 RAISE (Reproductive Health Access, Information and Services in Emergencies) Initiative at Columbia University

147

• Provide expert technical guidance to NGO partners in DR Congo, Pakistan, Chad, and Sudan on the collection and use of quality data for improved family planning and reproductive health services in settings affected by conflict and natural disasters. • Develop tools and training curriculum to improve project partners’ capacity for family planning data collection and use. • Strengthen the capacity of international NGO project partners and field-based project staff to develop and use routine monitoring systems for the analysis and use of clinical data. • Conduct baseline and endline program assessments using population-based surveys, health facility assessments, and community-based participatory research. • Document and publish project data in collaboration with field partners.

June 2007 Research Associate, Evidence-Based Strategies, Comparative Effectiveness/Population Studies – Sept 2008 Pfizer Inc. • Conduct population-based health outcomes research using national data sets and large insurance claims databases. • Interpret, write and disseminate research findings to marketing department and consumer populations. • Create educational materials for physicians and consumers based on significant research findings.

Sept. 2004 English and Economics Professor, Centre Universitaire de Labé, Guinea – July 2005 Teachers’ For Africa Fellowship Program International Foundation for Education and Self Help • Develop new English language curriculum. • Provide pedagogical consultation for the recently established Business English program.

June 2002 Lead Teacher, 7th Grade E.S.L. Algebra, Biology & Health Education – June 2004 James Lick Middle School, San Francisco, CA Teach for America • Develop strategies to incorporate adolescent health into multi-disciplinary core curriculum. • Represent teachers on council that governs policies and positions relating to the school.

INTERNATIONAL CONSULTING EXPERIENCE

May 2014 Public Health Consultant Seattle, WA – Present Bill and Melinda Gates Foundation

October 2006 Consultant – Learning Plus Initiative Kigali, Rwanda & New York –May 2007 UNICEF • Design institutional analysis tools and needs assessment mechanisms to assess the capacity of East African schools. • Pilot tools in five schools in Rwanda and develop framework for performance assessment

May 2006 American Jewish World Service Public Health Fellow Thailand – August 2006 MAP Foundation for the Health and Knowledge of Ethnic Labour

148

• Collaborate with the International Organization of Migration to conduct focus group discussions on knowledge, attitudes, and behaviors relating to HIV and gender among migrants living along the Thai-Burma border. • Develop a program plan and curriculum for a newly-established HIV prevention program in the Phang-Nga province of Thailand.

SKILLS

Computer: PC/MAC, SPSS, Stata, R, Epi Info, Word, Excel, Power Point. Languages: English (native), French professional fluency.

HONORS AND AWARDS

2018-2019 Edward J. Dehne Award in Population Dynamics 2016-2017 Recipient, US Policy Communication Training Program, Population Reference Bureau 2016-2018 Recipient, Fellowship in Family Planning and Reproductive Health 2015-2016 Recipient, Carl Swan Shultz Fellowship 2014-2015 Recipient, Department of Population, Family and Reproductive Health Research Day Award 2014-2015 Recipient, Laurie Schwab Zabin Award for Population and Family Planning

SELECTED PUBLICATIONS

Chukwumalu, K., Gallagher, M. C., Baunach, S., & Cannon, A. (2017). Uptake of postabortion care services and acceptance of postabortion contraception in Puntland, Somalia. Reproductive health matters, 1-10.

Casey, S. E., Gallagher, M. C., Makanda, B. R., Meyers, J. L., Vinas, M. C., & Austin, J. (2011). Care- seeking behavior by survivors of sexual assault in the Democratic Republic of the Congo. Am J Public Health, 101(6), 1054-5.

Casey, S. E., Chynoweth, S. K., Cornier, N., Gallagher, M. C., & Wheeler, E. E. (2015). Progress and gaps in reproductive health services in three humanitarian settings: mixed-methods case studies. Conflict and Health,9(Suppl 1), S3.

Siegler A; Roberts L, Balch E, Bargues E, Bhalla A, Bills C, Dzeng E, Epelboym Y, Foster T, Fulton L, Gallagher M, Gastolomendo JD, Giorgi G, Habtehans S, Kim J, McGee B, McMahan A, Riese S, Santamaria-Schwartz R, Walsh F, Wahlstrom J, Wecleles J. (2008). Media coverage of violent deaths in Iraq: an opportunistic capture-recapture assessment. Prehospital and disaster medicine, 23(04), 369-371.

PRESENTATIONS

"The Influence of Religiosity on Abortion Perceptions in Three Zambian Provinces" Population Association of America Annual Meeting, Denver, CO. April 2018 [Poster Presentation]

"Post-Abortion Contraception Among Women in Zambia: Missed Opportunities" Population Association of America Annual Meeting, Chicago, IL. April 2017 [Poster Presentation]

149

"Restrictive Abortion Laws in Sub-Saharan Africa: A Legacy of Colonization" Population Association of America Annual Meeting, Washington, DC. April 2016 [Poster Presentation]

“Restrictive Abortion Laws: A Legacy of Colonization?” International Conference on Family Planning, Bali, Indonesia. January 2016

"Marriage, Cohabitation, and Family Planning Patterns in Ethiopia" Population Association of America Annual Meeting, San Diego, CA. May 2015

SERVICE

Teaching Assistant, Johns Hopkins University, Demographic Methods, Issues in Survey Research Design, and Social and Economic Aspects of Human Fertility Manuscript reviewer for the American Journal of Public Health Manuscript reviewer for Global Public Health Manuscript reviewer for Disaster Medicine and Public Health Preparedness Guest lecturer, Mailman School of Public Health, Columbia University

150