Socio-economic determinants of HIV/AIDS in adolescents in rural western Uganda

MISSING IMAGE

Material Information

Title:
Socio-economic determinants of HIV/AIDS in adolescents in rural western Uganda
Physical Description:
xv, 320 leaves : ill. ; 29 cm.
Language:
English
Creator:
Stewart, Kearsley Alison
Publication Date:

Subjects

Subjects / Keywords:
Anthropology thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Anthropology -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 2000.
Bibliography:
Includes bibliographical references (leaves 303-318).
Statement of Responsibility:
by Kearsley Alison Stewart.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 024908606
oclc - 45805668
System ID:
AA00017691:00001


This item is only available as the following downloads:


Full Text










SOCIO-ECONOMIC DETERMINANTS OF HIV/AIDS
IN ADOLESCENTS IN RURAL WESTERN UGANDA



















By

KEARSLEY ALISON STEWART


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2000















Copyright 2000

by

Kearsley Alison Stewart





















In memory of my beloved grandfather, Robert J. Wood,
whose interest in travel encouraged me to journey without fear;

and in dedication to Margret Kajumbi,
whose dignity and determination was truly a gift,
and to Roggers Musinguzi,
whose passion for performance delighted and inspired everyone who met him.















ACKNOWLEDGMENTS

For my accidental introduction to Africa, I thank Jeannie Barnett for the warm

welcome I received in Aburi, Ghana; for generous and gentle guidance toward the M.A.

in African Studies, I thank Ned Alpers and Merrick Posnansky of U.C.L.A.; for sharing

his knowledge of and contacts in Hoima, many thanks to my fellow University of Florida

colleague Dan Ottenmoeller; for her mentorship and steady encouragement in Kampala, I

thank Elizabeth Marum of the CDC and the AIDS Information Centre; for their

extraordinary hospitality and fellowship in Hoima, I thank Bishop Wilson Turumanya of

the Hoima Church of Uganda Diocese and the entire Turumanya family; for her

mentorship and guidance into professional applied anthropology, I thank Dale Stratford

of CDC; for encouraging me to switch my focus from Archaeology to Medical

Anthropology, my thanks to George Armelagos of Emory University; and for unwavering

support, guidance, and friendship during the grant writing, fieldwork, and dissertation

writing phases, a BIG thank you to my dissertation Chair, Leslie Sue Lieberman of the

University of Florida.

My sincerest thanks to the following people who made selfless and generous

contributions to the development of the research proposal and the conduct of the

fieldwork: Rita Agarwal, Mary-Grace Alwano-Edyegu, Veronica Asaba, Rashid

Atugonza, Jane Babiiha, Grace Bantebya, Tom Barton, Nuuhu Bigirwa, Ben Biryahwaho,

Christine Businge, Levi Byaruhanga, Bob Downing, Douglas Feldman, Dr. Gamba-

Osiga, Rebecca Gearhart, Mary Goretti, Laura Guay, Holly Hansen, Katie Hogan,









Samuel Jalongo. Margaret Mary Kasaija, Joseph Kasaija, Newton Kateregga, I.B.

Katetegirwe, Joshua Kayondo, Lynn Khadiagala, Kenneth Alfred Kiiza, Joy Kwesiga,

Abwooli Tom Kyahurenda, Father Rudi Lehnertz, Anita Loughlin, MaryInez Lyons,

Norah Madaya, Patrick Madaya, Bomthon Mayaja, Nasur Musisi, Larry Marum, Janet

McGrath, Annet Mutagaywa, Simon Mugayo, Dara O'Connor, Helen O'Connor,

Omwony Ojwok, Ronald Otten, Dr. Oundo, Helen Pickering, Mark Rayfield, Pearl

Robinson, Charles Rwabukwali, John Rwomushana, Tim Scarnecchia, Janet Seeley,

David Serwadda. Nelson Sewankambo, Jan Shetler, Father William Ssozi, and Clyfus

Sunday.

My thanks to the staff of Creative Research and Evaluation Centre, Hoima

District Education Office, Hoima District Hospital and Laboratory, Hoima District

Medical Office, Hoima District Population Office, Hoima Integrated Community

Development Project, Hoima Resident District Commissioner, Meeting Point of Hoima,

Makerere Institute of Social Research, Makerere School of Medicine Library, Makerere

School of Public Health, Makerere Women's Studies Program, Nakasero Blood Bank,

Red Cross of Hoima, Uganda AIDS Research Sub-Committee, Uganda Ministry of

Health, Uganda Virus Research Institute, and White Fathers Catholic Diocese of Hoima

Special thanks to Jim Albury, computer consultant, University of Florida CIRCA,

for long hours of consultation formatting the dissertation; and to David J. Stewart,

manager, University of Georgia Sustainable Human Ecosystems Laboratory, for patience

and dedication digitizing many of the original maps presented here.

To my research team I say: Webale munno munno munno! Thank you so very

very very much: Robert Kaahwa, Margret Kajumbi, Rosette Kamanyi, Joel









Kibonwabake, Sarah Kyalisiima, Roggers Musinguzi, Faith Ochieng, Deborah

Rwahwire, Solomon Turumanya, and Enid Turumanya-Wamani.

Thank you to my family for love and support, my furry four-legged friend Kivu,

and for more than words can express, dis, merci mille fois.















TABLE OF CONTENTS

page

ACKNOW LEDGM ENTS .......................... .......... ....... ..................... ............ iv

L IST O F TA B LE S ................................................ .. ................................................ x

LIST O F FIGU RES ..................... ........ ... ... .. .... ... ... ........ ............. xiii

A BSTR A CT............ ................................. ........................................ ................ .... xiv

1 IN T R O D U C T IO N ...................... .. ........... .......... ...................... ..........................

O v erv iew ................................. .......................... ................................................1....... 1
G enesis of the Project .................. ........... ......... ....... .............. .... ................... 5
Choosing the R research Site ............................ ........ ...... .. .... ..................... 9
R research C learance............................................................................................... 11
Sources of Funding .................................... ........... ............... ........ .................... 12

2 LITERATURE REVIEW .............................. ................................14

Introduction ...................-........ ...... ........... ..... ............ ...... .... ................... 14
Epidemiology of HIV/AIDS in Uganda .................. .... .. ....... .............. 15
The Social Scientific Study of Sexual Behavior in Uganda ................................... 23
A anthropology and H IV /A ID S ......................................................-........ ................... 29

3 M E T H O D S ........................ ........ ...... ......... ....... .. .. .......................37

Sam pling D esign ........................................................................... ... 37
Sample Size................................................. 41
Randomization and Selection Process.................................. ................................ 42
Inform ed C onsent........................................................................... ....................... 47
E eligibility .... ................................................................ 47
Survey Instrument.................................. ... .................... 48
Research Assistants...................... ...... ............... ... 50
Translating the Survey Instrument......... ..................... .......................... 51
Data M management and Analysis ............................................ ... 53

4 SURVEY RESULTS: TESTING THE HYPOTHESIS .......................................... 56

Introduction............................................................... 56









General Socio-Demographic Variables ................................................ ............. 56
Sexual education, Behavior, and Health Variables.................................. .......... 60
Disaggregating Socio-demographic and Sexual Health Data by Gender/Sex ........... 63
Disaggregating Socio-demographic and Sexual Health Data by School Status......... 67
Comparing Socio-demographic Data by Gender/Sex and Sexual Activity .................. 70
Combining Variables to Create the Risk Profile Score ............................................ 73
Three-way Associations: Risk Profile Score, School Status, and Gender/Sex........... 80

5 VOLUNTARY COUNSELING AND TESTING.................................................82

Enrollment and Counseling Protocol for Hoima Adolescent VCT .......................... 82
Laboratory Protocol for Hoima VCT................................................................................. 83
A attendance at V C T ................................................................................................. 85
Socio-demographic VCT Results .................................................................... 87

6 THE REPRODUCTIVE LIFELINE TECHNIQUE................................................ 92

Introduction .................................... ......................................................................... 92
Assessing the Focus Group Discussion Method........................................................ 93
Challenging "Common Sense" Ideas about Adolescent Sexuality........................... 101
Developing Som thing N ew ....................................................................................... 105
The Reproductive Lifeline Technique ...................................................................... 108
N ew D ata, N ew Q uestions....................................................................................... 111

7 WORKING WITH UNEXPECTED DATA: THE DUHAGA SCHOOL VIDEO....114

When the End Was Really the Beginning................................................................... 114
Dilemm as of Seeing and Being Seen..................................................................... 117
Production V alues ...................................................................................................... 120
"The Youth Were Called upon to Take Care of Themselves" ..................................... 123

8 D ISC U SSIO N ........................................ ... ........ ... ...................................... 128

Introdu action ................................................................................................................. 12 8
Lessons Learned: Technical Issues of VCT..................................... ..................... 129
Lessons Learned: Counseling Issues and VCT........................................................ 131
Lessons Learned: Youth Issues and VCT................................................................. 133
Lessons Learned: Why the Youth Really Came for Testing .................................... 135
Lessons Learned: The Anthropologist and the Physician......................................... 140
Some Problems with Interpreting Survey Data on Sexual Behaviors ...................... 143
Assessing Discrepancies in Other Survey Data and "Honesty" of Hoima Respondents
.... ............................................................................................................................. 14 4
A Critical Variable for Risk Behaviors: Gender/Sex................................................ 147
A Critical Variable for Risk Behaviors: Education .................................................. 149
Challenging Local W wisdom ..................................................................................... 152
Ideas for Future Intervention-Linked HIV/AIDS Research in Africa...................... 155
Money is Necessary, but not Sufficient .................................................................... 157









APPENDICES

A R research C ontacts...................... ...................................... ........................................ 161

B Research Tim eline.. ..................... ...... ................. ........................................ 163

C Sampling Frame for Rural Interviews................................................................... 164

D Informed Consent for Youth Interviews (English) ..................................................165

E Informed Consent for Youth Interviews (Runyoro).................................................. 168

F Informed Consent for Voluntary HIV Testing and Counseling (English) ................171

G Informed Consent for Voluntary HIV Testing and Counseling (Runyoro) ..............174

H Survey Instrum ent (English) ..................................................................................1... 77

I Survey Instrum ent (Runyoro)..................................... ............................................ 231

J AIC Risk A ssessm ent Form .................... .. ...... .......................................... 290

K HIV-1/2 Serum Testing Algorithm............................. .. ................................292

L Transcript from Selected Portions of Duhaga Secondary School Video ..................293

M Letter from Rural Buhimba Girl about her Experience of VCT ..............................302

LIST OF REFERENCES.................. ......................................................................... 303

BIOGRAPH ICAL SKETCH ................................................................................... 319















LIST OF TABLES


Table Page

Table 2-1: Selected published papers describing epidemiological factors for STIs and
H IV /A ID S in U ganda........................................................ ....................... 17

Table 3-1: List of sampled schools for in-town survey.....................................................39

Table 3-2: Schedule of survey interviewing in rural Buhimba sub-county .........................40

Table 4-1: Selected socio-demographics (n=560 unless otherwise noted)..........................57

Table 4-2: Selected socio-demographics for Hoima District from the 1991 Population and
Housing Census for Hoima District, Uganda...............................................58

Table 4-3: Parent's education (n=560)............................................................................. 59

Table 4-4: Substance use by informant (n=560) ................................................................59

Table 4-5: Substance use by informant's three best friends ..........................................59

Table 4-6: Sexual experience and sexual health of Hoima youth, age 15-19 ......................61

Table 4-7: Ever use of any contraception by all women, age 15-19, Uganda......................61

Table 4-8: Ever use of condoms, by gender/sex, age 15-19, for Kampala District, Uganda, 62

Table 4-9: Percent distribution of respondents aged 15-19 who have ever had sex for
Kam pala, Jinja, and Lira Districts................................................................... 62

Table 4-10: Adolescent pregnancy and motherhood in Ugandan women aged 15-19 ..........63

Table 4-11: Percent distribution of respondents aged 15-19 by sexual status for Kampala,
Jinja, and Lira Districts............................................... .... .................................64

Table 4-12: Selected socio-demographic and sexual health data disaggregated by
gender/sex (n=560) ........................................... ..........................................65

Table 4-13: Selected socio-demographic and sexual health data disaggregated by school
status (n=560)................................................................................................. 68









Table 4-14: Selected socio-demographic and sexual health data disaggregated by
gender/sex for youth reporting no sexual activity (n=189) and for youth
reporting sexual activity (n=370)................................. .....................................71

Table 4-15: Criteria for risk profile score .............................................. ........................ 73

Table 4-16: Raw risk score for all sexually experienced youth (n=370).............................75

Table 4-17: Risk profile score for all sexually experienced youth (n=370).........................76

Table 4-18: Risk profile score by gender/sex (n=370).......................................................76

Table 4-19: Selected risk profile variables by gender/sex (n=370) .....................................78

Table 4-20: Risk profile score by school status (n=370) ...............................................79

Table 4-21: Selected risk profile variables by school status (n=370) ..................................79

Table 4-22: Risk profile score by school status for boys (n=183).......................................81

Table 4-23: Risk profile score by school status for girls (n=l 87)........................................81

Table 5-1: Frequency of youth from survey and VCT.........................................................88

Table 5-2: School status of survey youth and VCT ............................ .....................88

Table 5-3: Sexual experience of survey youth and VCT...................................................89

Table 5-4: Previous episode of a sexually transmitted infection for survey youth and VCT 89

Table 5-5: Gender of youth from survey and VCT..............................................................89

Table 5-6: Residence of youth from survey and VCT ........................................................90

Table 5-7: Primary reason young people requested VCT ..............................................90

Table 5-8: Residence and participation in VCT of girls from survey..................................91

Table 5-9: Residence and participation in VCT of boys from survey .................................91

Table 6-1: Dates, location, and other information about focus group interviews cited in
C h apter 6 ........................................... ............... ................... .....................9 5

Table 6-2: Mean age and location of sexual debut for youth aged 16-19............................102

Table 6-3: Sexual activity of female boarding students and village residents, aged 16-19 ...104

Table 6-4: Sources of sexual health information for youth aged 16-19.................................107









Table 6-5: Reproductive lifeline diagram of 40 year-old professional woman ...................109

Table 6-6: Personal data as reported by grandmotherss on own reproductive lifeline
diagram s ................................................. ...................... ... ..................... 112

Table 6-7: Frequency distribution by quartiles of age at first live birth for grandmotherss
reporting age at first live birth of first-born daughters.........................................113

Table 6-8: Mean age of first live birth for first-born daughters disaggregated by age of
(grand)m others at first live birth ........................................................................ 113















LIST OF FIGURES


Figure Page

Figure 1-1: Map of the African continent and the nation state of Uganda...........................2

Figure 1-2: Map of Uganda and Hoima District ....................................... ......................10

Figure 3-1: Map of locations of interviews in Buhimba sub-county...................................43

Figure 3-2: Map of Hoima town and Hoima schools........................................................45

Figure 5-1: Time Trend for HIV-1/2 VCT for both youth in survey and community adults.86

Figure 6-1: Map of reproductive lifeline interview sites. ................................................94















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

SOCIO-ECONOMIC DETERMINANTS OF HIV/AIDS
IN ADOLESCENTS IN RURAL WESTERN UGANDA

By

Kearsley Alison Stewart

August 2000
Chair: Leslie Sue Lieberman
Major Department: Anthropology

OBJECTIVE: This dissertation explored why young women, aged 15-19, have a rate of

HIV/AIDS six times higher than young men aged 15-19 in Uganda, eastern Africa. The

primary research goal assessed the predictive value of "number of years in school" on

"levels of sexual risk-taking" by youth in Uganda. SETTING: Research was conducted

in a western Ugandan town, population 4,000, and nearby rural area. Interviews were

conducted in community locations and voluntary, confidential, free HIV-1/2 antibody

rapid serum counseling and testing (VCT) were offered in the public District hospital.

DESIGN: From December, 1996 to December, 1997 survey, qualitative, biological, and

ethnographic data were collected from 550 adolescents, aged 15-19. A two-hour

interview covered issues broadly related to the transmission of HIV-1/2 including gender

roles, education, residence patterns, health status, friendships, leisure time, conjugal

partnerships, role models, household and personal resources, employment, and sexual

negotiation skills. In-depth key informant interviews and focus group discussions with








both youth and adults supplemented the survey data. Capillus, SeroCard, and Multi-Spot

products performed HIV-1/2 screening and confirmation testing for same-day counseling

and results. An experimental six-hour video on youth sexual health was recorded.

EPIDEMIOLOGIC OUTCOMES: 193 study youth and 176 community adults requested

VCT. Seroprevalence levels in these self-selected groups were approximately 1% for

youth and 13% for adults. RESULTS: HIV-1/2 rates for rural youth in western Uganda

are likely much lower than previously estimated by government sentinel surveillance

methods. We found no continued evidence of gender imbalance for HIV-1/2 in the 15-19

age group. Most behaviors related to risk of HIV-1/2 transmission were not significantly

different for gender/sex, but some were significantly related to school status. Significant

correlations appeared when data were disaggregated by both gender/sex and school

status. Levels of risk behaviors for boys were not related to school status. However,

more exposure to school was significantly related to lower levels of risk behaviors for

girls. Female rural farmers with no schooling and female school drop-outs had the

highest risk of HIV-1/2 transmission. CONCLUSION: VCT and behavioral change

programs for youth should tailor messages according to gender/sex, educational

attainment, and life-course experiences.













CHAPTER 1
INTRODUCTION


Overview

This dissertation explores the relationship between formal school education and

level of risk behaviors for adolescents aged 15-19 in a rural western district of Uganda

(see Figure 1-1). Three different methods of data collection-anthropological

(qualitative), demographic (survey), and epidemiologic (HIV-1/2 antibody test)-tested

the following null hypothesis: for the age cohort 15-19 in Hoima District there is no

significant difference in HIV-1/2 serostatus for three categories of adolescents. The

categories are as follows: (1) continuously enrolled in school; (2) dropped out of school

for at least 12 months; and (3) no school experience at all.

The immediate importance of answering this question stems from several

important facts. First, most HIV/AIDS prevention programs in Uganda develop only a

single approach targeting all adolescents. But is this an appropriate strategy for

HIV/AIDS prevention programs for Uganda's youth? The results presented here indicate

that the answer is no. For example, the data analyzed here reveal that the relationship

between school status and risk behaviors is a complicated one appearing to have little

effect on adolescent male sexual behavior but significant effects on adolescent female

sexual behaviors. Second, testing this hypothesis also promised new insights into why

young Ugandan women, aged 15-19, were six times more likely than young men of the

same age to test HIV-1/2 antibody positive (STD/AIDS Control Programme 1996:3).











UGANDA


Sudan


Democratic
Republic
of Congo


Kenya


Tanzania


Figure 1-1: Map of the African continent and the nation state of Uganda.


J
i









Biological vulnerability of adolescent women contributes to part of that lopsided HIV-1/2

ratio (Hankins 1996; Moss et al. 1991); but data presented here also bring to light the fact

that those sexual behaviors likely to lead to transmission of HIV for young women are

closely related to lack of access to educational opportunities. This is not true for boys.

Third, I also present data to challenge some widely held beliefs about adolescent

sexuality in Uganda. Specifically, I explored the notion that today's adolescents initiate

sexual activity at a younger age than did their parents and I found evidence that age at

sexual debut was not changed much over at least two generations. And finally, although

it was not an original research goal, I developed a protocol for introducing voluntary

HIV-1/2 antibody testing and counseling into rural areas in Uganda. As a result of my

research experience, I offer here specific ideas about the special counseling issues related

to HIV testing for adolescents. I discuss the value of committing scarce resources to

voluntary HIV testing and counseling for rural adolescents. For example, the data

presented here indicate a low, but proportionate seroprevalence rate for boys and girls.

This was an unexpected result that challenges researchers and health professionals to

explore fresh ideas about adolescent HIV-prevention strategies.

After this overview section, Chapter One describes in more depth the intellectual

genesis of this dissertation. Details about the research environment for the study of

HIV/AIDS in Uganda and procedures for obtaining research clearance describe some of

the logistical difficulties encountered during the fieldwork. Chapter Two contains a

review of literature. Rather than attempt to present a comprehensive discussion of

materials on HIV/AIDS in Africa, I elected to review the literature on several topics of

particular relevance to the development of this research project. Three areas stood out:









(1) the epidemiology of HIV/AIDS in Uganda; (2) the social scientific study of sexual

behaviors and AIDS in Uganda; and (3) theoretical and methodological literature on

anthropology and the study of HIV/AIDS in Africa. Chapter Three carefully presents the

methodologies used in this project. Details about the sampling design, calculation of

sample size, and the randomization process must appear in order for readers to assess my

argument that the data presented in this dissertation are generalizable to the national level

in Uganda, and thus are of interest to policy planners at the highest levels in the

government. A detailed discussion of criteria for eligibility to join the study and

procedures for informed consent is central to establishing the integrity of the research. I

discuss, as well, the design and procedures for a proper and meaningful translation of the

survey instruments.

Chapter Four presents the major results of the survey in a total of 23 tables. The

Chapter describes the general socio-demographic characteristics of the sample and

examines the critical variables in more detail. Most data are further disaggregated by

gender, school status, and sexual behaviors. A new variable, the risk profile score, is

developed from a composite assessment of sexual behaviors and breaks down into four

categories ranging from protective behaviors to dangerously risky behaviors. As briefly

introduced above, a major finding of Chapter Four is the contradictory effect of school

status on the sexual behaviors of adolescent boys and girls. Chapter Five presents the

voluntary HIV-1/2 antibody counseling and testing component of the study. I describe

technical details about the use of HIV-1/2 antibody rapid serum tests, laboratory

protocols, algorithm for determining and reporting individual diagnoses of HIV-1/2








status, pre- and post-test counseling procedures, plus protocols for maintaining client

confidentiality.

Chapter Six presents some of the results from focus group discussions conducted

with a variety of participants. We conducted over 20 focus group discussions with rural

and town youth, market women, elderly rural men, and professional women. A novel

procedure, the reproductive lifeline technique, is presented. This is a modified focus

group discussion exercise that facilitates a more meaningful conversation with older men

and women about the differences in sexual initiation and sexual behaviors between their

own lives and the lives of their children. Chapter Seven interprets some of the data

presented in the earlier chapters and offers reflexive and personal thoughts about my

experience of the fieldwork. Chapter Eight, a brief conclusion, is followed by an

Epilogue on the most unique event of the project-a day-long presentation (by local

secondary school students( of skits, poems, plays, debates, and games describing their

own interpretations of youth sexual health. In many ways, this is the most important

chapter of the entire dissertation. However, it is written in a decidedly nonacademic style

to underscore the creative contributions of the students, not my scholarly analysis. For

this reason, I have bracketed it off as an Epilogue. Appendices contain extensive detail

on research contacts, research timeline, sampling frames, all survey instruments in both

English and Runyoro, voluntary testing and counseling algorithms and questionnaires,

plus materials related to the secondary-school video presentation.


Genesis of the Project

I started my search for a research topic on HIV/AIDS in Uganda as a Ph.D.

student in medical Anthropology at the University of Florida. During the summers of









1992 and 1994, I traveled to Kampala, Uganda and sought out HIV/AIDS specialists at

the Makerere University School of Public Health located on the grounds of Mulago

Hospital, the largest government hospital in Uganda and one of the oldest hospitals in

eastern Africa. Many of the Ugandan professionals I met at Mulago were involved in

collaborative HIV/AIDS research with American universities, such as Case Western

Reserve University and Johns Hopkins University; American governmental agencies,

such as the Centers for Disease Control; or other governmental agencies, such as the

British Medical Research Council. These new contacts sent me to meet other Ugandan

and visiting researchers who were working for various Ugandan governmental agencies

and eventually I met people associated with local and international nongovernmental

organizations (NGOs), such as The AIDS Support Organization (TASO) and Save the

Children. See Appendix A for a list of research contacts.

I soon discovered that almost all HIV/AIDS-related research was conducted by

large collaborative teams of Ugandan and European or American biomedical physicians

who worked closely with biostatistians and clinical technicians and only occasionally

with social scientists or anthropologists. Invariably, I learned, when social scientists

were invited to participate in a large research project, they were brought onto the research

team long after the project design had been approved and the research hypotheses

formulated. Many sociologists and the few anthropologists I met complained that they

were excluded from high-level organizational meetings and when they were sent into the

field, it was often without the proper resources to fully accomplish their work. More

disturbing, however, was the complaint that when these social scientists finally were

given the opportunity to present their work to the biomedical team, their results were








undervalued and their recommendations ignored (for similar arguments see Dettwyler

1994; Packard and Epstein 1991). It was clear to me that to be most effective, HIV/AIDS

research needed the equal partnership-at all stages of research--of both the social and

biomedical sciences.

I took this as a goal for my own developing project. However, the more I pursued

the possibility of attaching myself and my Ph.D. research to a larger on-going project, the

more collaboration seemed an impossible goal. Finally, a senior sociologist on one of the

largest HIV/AIDS research programs in Uganda took me aside and frankly told me that

the possibility of an independent social scientist, such as a Ph.D. student, earning an

invitation to be attached loosely to a clinical field project, was probably one of the most

unlikely scenarios she could imagine. She conceded that many of the biomedical

researchers were not only skeptical of the potential value of cultural data for HIV/AIDS

research, but were actually fearful of results that might cast doubt on their own

conclusions. In desperation, I gave up the idea of attaching myself to a large ongoing

project and devoted myself instead to an independent project that would produce data of

interest to both epidemiologists and social scientists. Mindful of the irony of working

independently toward a goal of inter-disciplinary collaboration, I changed my networking

strategy away from established programs toward meeting the few academic independent

scholars working around Kampala.

Despite the scope of the epidemic and the resources devoted to prevention and

intervention, in 1992 I met surprisingly few scholars pursuing independent research on

HIV/AIDS in Uganda. At that time, most social scientific research on sexually

transmitted infections was in the context of policy-oriented research sponsored by various








NGOs. For example, in the late 1980s and early 1990s, masters theses on reproductive

health from the Departments of Statistics, Demography, or Social Work at Makerere

University were influenced heavily by the research goals and methods of the

Demographic and Health Survey, a United States Aid in Development (USAID)-funded

global research project (Blanc et al. 1996; Statistics and Inc. 1989; Statistics and Inc.

1996). Almost all of these theses were small-scale surveys of knowledge, attitudes, and

behaviors (KAB) using slightly modified Demographic Health Surveys (DHS) questions;

none broke fresh ground with new questions or innovative field methodologies (e.g.

Nabbosa-Nalugwa 1991; Najjumba-Kibira 1991). However, at that time, a few senior

women scholars were critiquing the prevailing methodologies and paradigms used in

most KAB research and their work laid an important foundation for the more broadly

conceived research that was to follow (Ankrah 1993; Lyons 1995; Obbo 1993).' By

1994, several social science graduate students from Europe, the U.S., and Uganda were

working on projects that historicized sexually transmitted diseases and the emergence of

HIV/AIDS in Uganda or that related sexual behavior to wider cultural and economic

issues of gender identity or religious beliefs (Fritz 1998; Morrow 1999; Rwabukwali

1997; Tuck 1997; Wimberley 1996). Yet none of this research on HIV/AIDS in Uganda

explicitly attempted to link biomedical and behavioral research. For that reason, it was

becoming increasingly clear to me that it was possible for an anthropologist to make an

original and significant contribution to the field of HIV/AIDS research in Uganda. The

question that remained was how to organize such a research plan.



'Often the only critique of NGO funded HIV/AIDS research in Uganda came from the scholars at the
Center for Basic Research (CBR) in Kampala, a nonprofit educational trust developed under the direction
of Mahmood Mamdani. For a published discussion, (see also Dicklitch 1998).









Choosing the Research Site

One researcher recommended that I choose a research site far from the

southwestern Districts of Masaka and Rakai where the bulk of HIV/AIDS investigators

were concentrated (see Figure 1-1).2 Instead of directly contributing to the growing

published data from these projects, my research strategy was to work in an area different

enough from the southwest region that I could test some of the emerging theories about

the spread of the epidemic. For example, I could test the theory that the spatial

distribution of HIV/AIDS was associated with geographical proximity to the "AIDS

highway" (Nunn 1996; Nzyuko 1997; Pickering 1997). I ruled out working in the eastern

regions of Uganda because there was already a high level of trade across the border with

Kenya. The areas north of Lake Kyoga were dangerous due to the decades-old

intermittent war with Sudan and years of civil war in Uganda's northern regions. The

southwestern borders with Rwanda and the Democratic Republic of Congo were also

unsettled because of the fall of Mobutu and the simmering aftermath of the 1993-1994

genocide.

This left only the western region. Thanks to a fellow graduate student from the

University of Florida I was introduced to the Anglican Bishop in Hoima District and was

promptly welcomed as a daughter, sharing a bedroom with one of his four daughters.

Literally with the Bishop's blessing, I set out in the summer of 1994 to persuade the local

District government and medical officials of the value of my project and to begin serious


2In Rakai District since 1989, Wawer (1999, 1998) of Columbia University and colleagues primarily from
Johns Hopkins University (Gray 1998), Makerere University (Sewankambo 1994), and the Uganda Virus
Research Institute (Konde-Lule 1997), are conducting a population-based cohort study of various clinical
and epidemiological aspects of the HIV/AIDS epidemic. The sample size in the first study was
approximately 2600 and the current enrollment is over 12,000. In Masaka District since 1989, with funding
from the UK Medical Research Council, Kengeya-Kayondo (1996), Mulder (1994), Nunn (1994), and
colleagues are conducting similar research with a sample size of approximately 10,000.









Sudan


Congo


Tanzania


Hoi


:


Figure 1-2: Map of Uganda and Hoima District.


10












Kenya












ma District





Hoima Town,


i,
--~--J--s
]~;;









preparations for the fieldwork. The actual field research, however, was not to begin for

another two-and-a-half years because of the complications of obtaining research

clearance for HIV research in Uganda and obtaining sufficient funding. A timeline of the

progress of the research is in Appendix B.


Research Clearance

The original version of the research proposal was submitted in late 1994 to the

Makerere School of Social Research (MISR), the first step for many social scientists in

obtaining research clearance in Uganda. Normally, most requests for research clearance

are forwarded from MISR to the Uganda National Council for Science and Technology

(UNCST) and then forwarded to the Office of the President for the final official

permission to conduct research. Because the proposal focused on HIV/AIDS, additional

review and special approval of the research protocols were needed from the Ugandan

medical community before the Office of the President would clear the research protocol.

The AIDS Research Sub-Committee, a standing committee of approximately 12 health

professionals and community advisors, serves as the local Institutional Review Board

(IRB) for all research related to HIV/AIDS in Uganda. I submitted my proposal to the

Sub-Committee.

Due to the large volume of proposals submitted to the Sub-Committee and the

difficulty of getting a quorum at meetings, it took over 18 months for the first review of

my proposal to move through the Sub-Committee. After submitting two more revisions

of the proposal, clearance was granted for the project in March 1997 and formal field

research and recruitment of informants began in April 1997. Local approval of the

research was also granted by the Medical Superintendent of Hoima Hospital, the District








Medical Officer, and the Resident District Commissioner, the highest-ranking political

official in Hoima District. In the United States, the study protocol was approved by the

University of Florida Health Center Institutional Review Board (IRB Project #389-96/97)

and by the Office for Protection from Research Risks at the National Institutes of Health

of the Department of Health and Human Services (Assurance #S-6233-14).

Between 1994 and 1996, this research proposal, including the informed consent

documents and the voluntary HIV-1/2 antibody rapid testing and counseling protocol,

was reviewed and approved by the following offices: (1) Department of Health and

Human Services, National Institutes of Health, Office for Protection Against Research

Risks, USA (2) University of Florida Institutional Review Board, USA (3) Government

of Uganda, AIDS Research Sub-Committee Institutional Review Board, Entebbe, Uganda

(4) Makerere Institute of Social Research, Kampala, Uganda (5) AIDS Information

Centre (AIC), Kampala, Uganda (6) Medical Superintendent, Hoima District Hospital,

Hoima, Uganda. The survey was conducted during May-August 1997, testing and

counseling during June-September 1997, and qualitative research from January-

November 1997.


Sources of Funding

I am grateful to many different sources for the funds to conduct the fieldwork

portion of this research project. In 1996 I received major dissertation awards from the

National Institutes of Mental Health Dissertation Research Grant in HIV/AIDS Research

(1-R03-MH56294-01 Al) and the National Science Foundation Dissertation

Improvement Grant (9632207). The NSF grant was co-reviewed and co-funded by the

Cultural and Physical Anthropology sections. The same year I received smaller awards






13


from the Wenner-Gren Foundation for Anthropological Research Predoctoral Grant, the

Woodrow Wilson-Johnson & Johnson Dissertation Grant in Women's Health, the Central

States Anthropology Society Leslie A. White Award, and the Society for the Scientific

Study of Sex Dissertation Award. In support of the dissertation-writing phase of the

research, in 1998 I received a Yardley Dissertation Fellowship from the College of

Liberal Arts and Sciences at the University of Florida.














CHAPTER 2
LITERATURE REVIEW


Introduction

The volume of published literature and unpublished government and

nongovernmental organization (NGO) reports on HIV/AIDS in Africa is so vast that a

meaningful literature review on the topic "AIDS in Africa" is nearly impossible. For the

purpose of this literature review, the field is narrowed to mostly published material about

the epidemic in Uganda. It is further narrowed to the two particular topics that framed

the research question of this dissertation: epidemiology of HIV/AIDS in Uganda and the

social scientific study of sexual behaviors in Uganda. A short discussion of the

anthropological study of HIV/AIDS follows.

The study of HIV/AIDS outside Africa has shifted away from behavioral and

demographic risk factors to molecular analyses of the viral diversity of HIV, adherence to

HIV medication regimens, and biological susceptibility to resistance to the new anti-

retroviral treatments (Padian 1998). However, the bulk of published work on HIV/AIDS

in Africa continues to focus on epidemiological and behavioral factors related to the

transmission of HIV-1/2. This is due to a number of factors: high cost of research on

molecular aspects of the virus, lack of proper research facilities in which to conduct such

research, the expense of the anti-retroviral therapies restricting access for most Africans,

and emerging evidence that behavioral interventions are effective in Africa. Despite

these obstacles, some scientists in Uganda are working at the molecular level to

understand genetic variability in HIV-1 (Janssens et al. 1997; Kaleebu et al. 2000;

14









Rayfield et al. 1998; Yirrell et al. 1998). As well, recent work on the impact of AIDS on

population structure and mortality rates (Low-Beer and Stoneburner 1997; Low-Beer et

al. 1997; Mulder et al. 1994; Nunn et al. 1997; Stoneburner et al. 1998; Stoneburner et al.

1996; Stover and Way 1998) indicates that research on HIV/AIDS in Uganda has

advanced further than that of almost any other country in Africa.


Epidemiology of HIV/AIDS in Uganda

The first published epidemiologic description of HIV/AIDS in Uganda appeared

in the journal Lancet in 1985 (Serwadda et al. 1985). Serwadda and colleagues described

the symptoms and early patterns of transmission of HIV/AIDS, initially referred to as

HTLV-III infection. Their description set the stage for the parameters of the early social

scientific study of HIV/AIDS as well. Serwadda, a leading epidemiologist and professor

at the Makerere University School of Public Health in Kampala, described the primary

symptoms of HTLV-III infection as weight loss and chronic diarrhoea. This accounted

for the emergence of the local name, SLIM disease. The paper noted that HTLV-III

syndrome was rarely associated with Kaposi's sarcoma (KS), although KS was endemic

in the area of first outbreak-rural southwestern Uganda. Further, Serwadda and

colleagues noted that HTLV-III infection occurred predominantly in the "heterosexually

promiscuous population and there is no clear evidence to implicate other possible means

of transmission, such as by insect vectors or re-used injection needles" (:849). By noting

these factors-rare occurrence of KS, promiscuity, heterosexual transmission-Serwadda

et al, emphasized the apparently different epidemiologies of the disease in Africa, the

USA, and Europe. Unfortunately, this set the stage for a new, but misleading, concept,

"African AIDS," similar to the early typecasting of HIV/AIDS in the USA as a gay

disease that did not affect heterosexuals or women.









The research hypothesis and study design of this dissertation are rooted in the

more recent studies listed in Table 2-1. This table compares our methods and findings

with those of other similar studies in Uganda; it is intended neither as a meta-analysis

(Stroup et al. 2000) nor to estimate a common relative risk across different epidemiologic

study designs (Martin and Austin 2000). A meta-analysis of epidemiologic data on

HIV/AIDS in Uganda could provide a needed overview for the research community but

would need to take into consideration variability and gaps in the published data. For

example, variation in geographical location of the studies, sample size, and design

(longitudinal versus cross-sectional) is evident. Epidemiological data from self-reported

sexually transmitted infections (STIs) and laboratory confirmation of STIs are not

consistent in the data in Table 2-1, and self-reported prevalence may be higher or lower

than prevalence determined through laboratory analysis.

Other patterns in the epidemiological data also influenced my research design.

For example, the importance of further examining the relationship between urban and

rural residence on self-reported STI rates. Data cited in Table 2-1 clearly indicate a

significant difference in rates of STIs between urban and rural men and women (Ageyi

1992: urban male, 23%; rural male, 19%; urban female, 9%; rural female, 6%) and

between boys and girls (Statistics 1996: female age 15-19, 2.0%; male age 15-19, 3.9%).

Data implicating the number of sexual partners for young men (Musinguzi 1996:

Kampala male 15-19 with 5+ casual partners, 100%; Kampala male 15-19 with 0 casual

partners, 3%) also deserved further testing.














0

00 C
-E
2C
0





a w


0


0\o








0Cd
C- 4

oC


v0 co
-o t-L E t 7G .20 0 E o E

c oo
0 4L 5 o E 0


j ci WJ 9 3











>- -o
c 40 00
C.CC. c) o C




C- C 0




E2CO-6 ) )U co C


>Tr cO > z 2
cn



0 0 r- r
r C\ 00
II m ON



Q0 N


I I I I












- ~ r r


+


C 0 Ow 0
ex loN
crmgoi
~o sma


0N

o\o
- 3

s"o

mig
O5 2
o* 0- U,


-.' C)
* .C-


o\ <^
- ON U,
ON-o
i- 0'1





m8)o 5
-^
Cfl |
CC ON i-


-O
rs ^
4 N
BO 5


-o

* C


ca E



<--
o E
MOO ^


.. 0, M









C7,



cco
o o cm mu



















> ) > 5 Q >
-Ea t c c









c Z >-NS






_- (0 _( NS- _
(5 z -







=o 0, 005to5 o m O- V r
CC -



Cm o ot SoN:CO "ON












1 I1 1


C/
U3


0
o r~- oo N
osCT> (^
- os o S
-p I "
~o^ o o
1^ o^ o^ _


ON


0
U,
c





4 o
uuv
Ca


E i6

2 -5
S0 C
0tQ

E u.0f


4S U,5 I0 )
*" ~ 3 -0 0 I
-0 u -- om 3 m
-. ~ ~ ~ C coc 0 /i.
^ E + 0E t.C8) t
M O m cz V m mO"Oc
7; tz:P.Fo
14)
c~cc
ci, E aC Ec 2 ,e oC r r_
P-S ~ ~ "c^ v, ^cy S"- S ^ ;^-





x ^
Q.)
4) C
lCl
?3 u >ri X
E -5 c, cc t4 U m







0 Q04 C
fo ,m bOS ,2r ?


cl

C-o^ t c _c__



EC
'10 U


-:3
b- ^ 4)..I-
., 5 a 5 5 5m mSS *3S





oz u
co o n (



m =o V') E ook
ou 0n En m* m ^ ,C
1 C-S ^C2 *-4C4 E t ,
EC C






-)Q2- O>" ic) S.- < uM U



-I i O\ IO c- || _|
ONO ON '~
D03 03







oc U, 2'.oo








f) u
4)O
CC U.

om' oc o *S i 2 | ?

2 *i -c gcZ *r <

^S__ 1 ON1^ g


C C
o\ o0
c






"l Cl 0
A ._ rA .2 .
0 -4) -< o
r~El
Ao~~t
thN y


0

_,t,


I-
cy- t
^^ '
o\ E *g
S O C
tl -


























CC
u


-D


-C)


E
C


E
co
C a,


c c
oC o




- C


----
so




\ 00--
C7C
(N r r C
0 ~'~~t
Os-(N I
OsEsOS

,---- 3~-


I-
g
2e
D S

\S
r


00



m o


C)C
C 7a


S3 0
V c .S2 /C'























co c
E
0 cO







0









M ac c "c

E CO C 3 p o C O CO








O 00




z S ca C cc cw
- o E O! Ni


I / 1 1 I









Another concern with comparing the data in Table 2-1 is the variety of sampling

strategies. Comparison of sentinel surveillance epidemiological data to data from

community-based studies is difficult. While the overall quality of all of these cited

research projects is high-especially considering the logistical difficulties of conducting

public health research in Uganda-the sentinel studies generally are considered the

weakest because this research design leaves the largest chance for error (Mertens and

Low-Beer 1996). As described by the Ugandan Ministry of Health (STD/AIDS, 1996:1-

2), national HIV infection rates are based on the prevalence of HIV infection in pregnant

women attending antenatal clinics at 20 HIV sentinel surveillance sites distributed all

over Uganda. The Ministry of Health sends personnel to these 20 sites four times a year

to collect anonymous random blood samples from the routine blood draws standard for

all pregnant women attending government-funded antenatal clinic. No personal

identifiers are collected, only sex, age, and location.

My interest in exploring the relationship between gender and sexual risk is based

on the disproportionately high rates of HIV for females aged 15-19. In Table 2-1, these

data are cited from Kengeya-Kayondo (1996), Mulder (1995), Nunn (1994), Uganda

STD/AIDS (1996), Wawer (1991), and later in Konde-Lule (1997). Despite the striking

relationship between risk of HIV transmission and gender in this age group, very little

data were available at the time I was developing my original research proposal.

Moreover, few researchers were committing research resources to study the significant

difference in seroprevalence levels for young girls and men. In my conversations with

public health researchers in Uganda in the early 1990s, I discovered that many did not

believe adolescent sexual behaviors could play a significant role in the overall HIV

epidemic in Uganda. Adolescents, particularly young women, were constructed only as









receiving the virus from older sexual partners and there was little perceived need to

understand the epidemiological dynamics of adolescent sexuality on its own.

The study of adolescent sexuality in the context of the HIV epidemic in Uganda

attracted much more attention after a 1997 publication reported extensive data on the first

observed decline in HIV seroprevalence anywhere in Africa (Asiimwe-Okiror et al.

1997). Similar results had been briefly reported earlier (Konde-Lule 1995; Mulder et al.

1995). Asiimwe-Okiror and colleagues observed a sharp decline in HIV prevalence for

women attending antenatal clinics in two urban sentinel sites in the youngest age groups

(15-19). They attributed the decline to increased condom use and to a delay of two years

in the onset of sexual activity (Asiimwe-Okiror et al. 1997:1762). At the same time, the

authors cautioned that behind the good news, "stable or declining HIV prevalence in

antenatal clinic attendees may mask a high incidence often concentrated in young age

groups" (1997:1762).

In contrast to the conclusions of Asiimwe-Okiror and colleagues at the Ugandan

STD/AIDS programme, one of leading HIVAIDS research teams in Uganda reported that

declines in HIV prevalence were more likely the result of the natural dynamics of the

viral spread of HIV; that is, HIV-related mortality rates exceeded rates of new

seroconversions, an epidemiologic characteristic of "mature" epidemics (Wawer et al.

1997:1026, 1028). Wawer and colleagues at Makerere University Institute of Public

Health, Columbia University School of Public Health and Johns Hopkins University

School of Hygiene and Public Health caution against extrapolating trends in prevalence

to concurrent trends in incidence, particularly data from "mature" epidemics (1997:1029).

Stoneburner, Low-Beer, and colleagues offered a more positive position on the

use of surveillance data (such as that from the Asiimwe-Okiror study) to interpret the









decline in HIV seroprevalence (Low-Beer and Stoneburner 1997; Low-Beer et al. 1997;

Stoneburner et al. 1998; Stoneburner et al. 1996). They counter Wawer's position by

arguing that "the complex relationship between HIV incidence, prevalence, and mortality

may sometimes be obscured if studied only for short periods" (Stoneburner et al.

1998:226). Stoneburner and colleagues at the WHO, Ugandan Ministry of Health,

USAID, FHI, and elsewhere, developed an empirically-based model to forecast trends in

the epidemiology of infectious diseases, particularly HIV/AIDS in eastern Africa

(Stoneburner et al. 1996). Their conclusions argue against Wawer et al and support the

interpretation that some portion of the decline in HIV prevalence in Rakai District,

Uganda can be attributed to the reduction in high-risk sexual behaviors (Stoneburner et

al. 1998:227-228). This implies that behavioral intervention programs have been

successful and can continue to be successful in the future. For social scientists working

on HIV/AIDS in Uganda, this was a watershed moment that revitalized the research

community. As a result of these developments and of Asiimwe-Okiror's publication, a

new focus on youth became a permanent feature of most HIV research in Uganda

(Konde-Lule et al. 1997).


The Social Scientific Study of Sexual Behavior in Uganda

The current deluge of behavioral data on sexual practices in Uganda dates back

only to the mid-late 1980s when international public health agencies and Ugandan,

British, and American public health researchers (e.g.: Berkley et al. 1990; Serwadda et al.

1985; Wawer et al. 1991) were first invited by President Museveni to study the epidemic.

Previously marriage, reproduction, and sexual behavior in Uganda were studied mainly

by the colonial era missionary John Roscoe (1923), an occasional British anthropologist

(Beattie 1957, 1958 #861; 1965), and a handful of Ugandan researchers (Kisekka 1974;









Nyakatura 1970; Obbo 1980). Some relevant indicators, such as female fertility levels

and age at first birth, were probably gathered as early as 1948 in the first Ugandan

nationwide census, or by 1957 by the Family Planning Association of Uganda. However,

these types of data were not systematically collected until the 1988/89 Demographic and

Health Survey (Statistics and Inc. 1989: 2-4). Student theses from Makerere University

offer glimpses of these variables as well, but are not readily available and suffer from

uneven quality of data collection and analysis (e.g.: Baganizi 1991; Nabbosa-Nalugwa

1991; Najjumba-Kibira 1991; Nakamyuka 1982).

Without any significant baseline behavioral data from which to begin, researchers

relied on standardized sexual behavior surveys developed in Europe and the USA (e.g.

World Health Organization 1990). This recalls the earlier practice of colonial

missionaries, travelers, and administrators carrying James G. Frazer's standardized

questionnaire with them into the field as they systematically attempted to study all aspects

of native life based on western cultural categories (Frazer [1887] 1889 cited in Ray 1991:

23). Without question, the use of standardized sexual behavior surveys during the early

days of the HIV/AIDS epidemic in Uganda was vital for policy decisions and program

implementation, but it set a narrow course for the study of sexual behavior over the next

decade.

However, it is regrettable that little systematic social scientific knowledge is

available about the sexual practices of even a generation ago. Yet the continued

standardization of current survey research means that our awareness of any changes that

might have occurred during the HIV/AIDS epidemic will continue to be limited to

uncomplicated, and possibly incorrect, ideas about sexual behavior in Uganda. The

Demographic and Health Survey recognized this problem, and in 1995 they conducted an








additional survey in Uganda intended to supplement standard fertility and sexual

behavior data with more qualitative, open-ended questions broadly related to the

HIV/AIDS epidemic (Blanc et al. 1996). This important report is one of only three "in-

depth" research projects conducted by DHS in sub-Saharan Africa since 1988 (Macro and

Inc 1999). As well, some anthropologists and sociologists managed to promote context-

sensitive research on sexual behavior relatively early in the study of the epidemic (e.g.:

Ankrah 1993; McGrath et al. 1993; Ntozi and Kabera 1991; Obbo 1991; Olowo-Freers

and Barton 1992; Seeley et al. 1991), but the majority of these studies are without much

historical depth. Recently, some historians have begun to analyze sexual behavior in

Uganda (Lusembo 1990), but many of these historical studies focus on colonial policy

and social control programs implemented during epidemics of sexually transmitted

infections such as syphilis (Lyons 1994; Setel et al. 1999; Tuck 1997; Vaughan 1992).

At the beginning of the AIDS epidemic in eastern Africa in the middle 1980s,

there was no published evidence to suggest a relationship between education and

HIV/AIDS. Yet many Ugandans believed that higher education was positively correlated

with higher risk of HIV transmission. This belief is most likely due to the well-

publicized AIDS-related deaths-early in the epidemic--of civil servants, politicians,

and other prominent Ugandans. In neighboring Rwanda and further south in Malawi,

careful epidemiological work early in the epidemic suggested that this correlation was

significant, but only relevant early in the epidemic (Allen et al. 1991; Dallabetta et al.

1993). In both Uganda and Rwanda, people gossiped about wealthy men and their many

mistresses or spoke of the extra-marital affairs of politicians and civil servants and the

women of the "deuxieme bureaux" or "second offices." Perhaps it is not surprising that

much of the social scientific research in Africa, which studied nonbiologic or









nonepidemiologic variables related to the transmission of HIV, focused on the

relationship between wealth and HIV/AIDS, and the exchange of sexual intercourse for

money (Bailey et al. 1999; Berkley et al. 1989; Nzyuko et al. 1997; Serwadda et al. 1992;

Weiss 1993).

At the beginning of the epidemic in eastern Africa, Western biomedical scientists

identified prostitution (commercial sex work) and truck driving as risky occupations

(Bwayo et al. 1991; Carael et al. 1988; Kanki et al. 1992; Plummer et al. 1991),

proximity to highways as risky places to live (Nunn et al. 1996; Nzyuko et al. 1997;

Serwadda et al. 1992), vaginal application of herbal mixtures as a risky procedure

(Brown et al. 1993; Civic and Wilson 1996; Runganga and Kasule 1995; Sandala et al.

1995), and traditions such as widow inheritance as risky cultural practices (Hrdy 1987)

for the transmission of HIV/AIDS. This was the early picture of some of the social and

economic variables thought to be fueling the sudden explosion in the late 1980s and early

1990s of HIV/AIDS in the Great Lakes region of eastern Africa.

Biomedical and social scientific preoccupation with African "promiscuity" meant

that some groups (mostly women) were stigmatized and unfairly blamed for the epidemic

(Ankrah 1991; Schoepf 1992). At the same time, the study of the sexual transmission of

HIV focused on "high risk" groups and overlooked other critical aspects of transmission

dynamics, such as the risk for women living monogamously with a nonmonogamous

partner in a rural area. The definition of "group" and the methods for studying and

analyzing "groups" produced disagreement between epidemiologists and social scientists.

Similarly, social scientists challenged biomedical and epidemiological concepts of "risk"

by noting the culturally dependent aspects of the term (Douglas 1990; Douglas 1992;

Frankenberg 1993; Hayes 1992; Kendall 1992; Lupton 1993; Nelkin 1989). The work of









Treichler and Watney were early critiques of the misguided notion of a special kind of

"African AIDS" (Treichler 1989; Watney 1989). Watney notes that "African AIDS" was

presented as a harbinger of the epidemic in Europe and the USA (Watney: 1989:45).

Other scholars are even more direct in their interpretation of the disturbing early

preoccupation with risk groups and promiscuous behaviors; they argue "the racist

preconceptions of the researchers led them to conclusions that had no scientific

foundation" (Harrison-Chirimuuta and Chirimuuta 1997:166).

By the late 1990s, however, most social scientists and many biomedical scientists

had arrived at the consensus that the major co-factors for the emergence of the HIV/AIDS

epidemic in the Great Lakes area of eastern Africa were endemic levels of ulcerative

STIs, immune-compromising aspects of chronic malaria, and social disruption from

migration and years of civil war (Green 1994:9-12, 212-213, 233; Hunt 1989; Hunt 1996;

Kalipeni 2000). But the specific mixture of co-factors necessary to sustain the epidemic

remained complicated and no single category of factors-sexual behaviors, geographical

(dis)locations, other health problems-could reliably predict the direction of HIV

prevalence or incidence.

For example, after Kampala, Masaka, and Mpigi Districts, Rakai District (see

Figure 1-2) has the largest number of documented cases of both adult and pediatric AIDS

(STD/AIDS and Uganda 1999:15). In contrast to Kampala, Masaka, and Mpigi, Rakai is

not directly on the "AIDS highway" but is near enough that some of the social and

economic effects of the highway are evident in this rural area (Pickering et al. 1997).

However, the social disruption and military devastation that plagued Rakai District

before, during, and after the 1978-1979 civil war, are central to any explanation of the

early appearance of HIV/AIDS in this area. It was overrun by both the Ugandan and









Tanzanian military during the Liberation War that toppled Idi Amin in 1979 (Smallman-

Raynor and Cliff 1991). Of all culturally or economically defined population groups,

soldiers typically sustain the highest rates of HIV infection and their mobility contributes

to the rapid transmission of HIV over large and often remote areas (Dodge 1990).

Although the evidence is sparse, Hunt would argue that male labor migration also

contributed to the early epidemic in Rakai (Hunt 1996:1293-1295). Add the already high

level of endemic STIs in Rakai District to this complicated scenario and this compelling

scenario suggests why HIV/AIDS first appeared in epidemic proportions in Rakai. By

contrast, Hoima District was not located near any major roads and nor did it normally

serve as an important trans-shipment point for the movement of consumer goods or

military personnel. Therefore choosing Hoima District as the research site would

produce data which could challenge or confirm the social disruption/AIDS highway

theory of HIV/AIDS.

Up to this point, the epidemiological and social scientific literature shaped my

choice of a dependent or response variable (HIV serostatus), the age cohort (15-19), and

the research site (Hoima). What remained was choosing the independent or explanatory

variables. To ensure the generalizability of my data, it was clear I needed to duplicate the

standard sexual behavioral questions found in most HIV/AIDS survey research in

Uganda. I correlated my questionnaire with those of DHS and the Statistics Department

at Makerere University (Blanc et al. 1996; Statistics and Inc. 1989; Statistics and Inc.

1996) and the STD/AIDS Control Programme in the Ugandan Ministry of Health

(Musinguzi et al. 1996). But I was not confident that my questionnaires would produce

any significant new insights into sexual behaviors in Uganda.









Until the AIDS epidemic, the primary focus of the demographic study of

education as an independent variable was to demonstrate the influence of education on

attitudes toward and use of contraceptive technology (Agyei and Migadde 1995). The

positive relationship between level of education and use of modern contraceptives has

been demonstrated all over the world, and this relationship holds true in Uganda as well

(Statistics and Inc. 1996:45, 52). Education as an independent variable related to the HIV

epidemic in eastern Africa began as a demonstration of the positive relationship between

level of education and risk of HIV infection (Allen et al. 1991). (This is discussed in

more detail in Chapter 6.) Early in the epidemic, this relationship was strongest for well

educated white-collar urban men, although in these cases, education was most likely a

proxy for one of the early underlying causes of HIV/AIDS in adult males in eastern

Africa: high socio-economic status which allowed multiple commercial sexual partners.

Later, education as an independent variable was measured only as number of years in

school (Smith et al. 1999); or, if a study age cohort included adolescents, the research

tended to look only at youth currently enrolled in school (Kinsman et al. 1999; Shuey et

al. 1999; Twa-Twa 1997; Venier et al. 1998). These studies failed to measure the actual

effect of education on adolescent HIV-related risk behaviors and to include all categories

of youth in the same study-not just the easy-to-reach school goers. Even today,

HIV/AIDS behavioral intervention and prevention programs for adolescents in Africa

tend to develop materials first for school youth, and then when resources permit, to

modify the materials for school drop-outs or illiterates (Muller et al. 1999).


Anthropology and HIV/AIDS

Any meaningful contribution to the reduction of HIV transmission must attempt

an understanding of sexual cultures. Therefore it seems logical that anthropologists









should lead the way in the study of sexual behaviors and HIV/AIDS in Africa. Yet this is

not the case, why? Sobo suggests that anthropological AIDS research is not well-

integrated into the broader field of medical anthropology because many anthropologists

who work on HIV/AIDS ignore current theoretical debates in medical anthropology.

According to Sobo, it sometimes appears that anthropologists who study HIV/AIDS have

created their own isolated sub-field (1999:11). Okami and Pendleton (citing Tuzin and

Frayser) argue that the lack of theory in the anthropological study of sexual behaviors is

because "sexuality is typically ascribed the status of illegitimate child in the study of

marriage, reproduction, and kinship" (1994:85).

Perhaps this paradox flows less from a lack of desire for developing theory for an

audience of medical anthropologists than from fundamental differences between the

social scientific study of behaviors (Auerbach 1998; Catania et al. 1995; Pisani et al.

1998) and the ethnographic study of cultures (Barth 1989). At a 1990 Wenner-Gren

Symposium on "AIDS Research: Issues for Anthropological Theory, Method, and

Practice" held in Colorado, the meeting opened up with thoughts on how AIDS research

could contribute to theory and practice in anthropology and ended with conversations

about what anthropology could contribute to the fight against AIDS (Treichler 1999:218).

As an example, the study of sexual behaviors in Uganda is designed to collect data (and

develop theory) to create change at a variety of levels-individual condom use (Kamya et

al. 1997); safer sex practices between married partners (Serwadda et al. 1995); extended

family support of HIV positive family members (Seeley et al. 1993); integrating HIV

education into secondary school curriculum (Shuey et al. 1999) and in the mosques

(Kagimu et al. 1998). On the other hand, the ethnographic study of sexual behaviors in

eastern Africa is designed to promote a profoundly complex and symbolically rich









description (and theory) of human sexuality (Weiss 1993:24-27). In fact, these

descriptions are so symbolically rich that if the researcher presented their interpretation of

the data to their informants, the informant might resist or even contradict the researcher's

interpretive conclusions.

Another possible explanation for the relative disarray of the literature published

by anthropologists on AIDS in Africa might be the haphazard way in which much of the

anthropological study of AIDS was accomplished in the early days of the epidemic. It

was often an appendage to a health-related project already in progress by an

anthropologist in the field. For example, Taylor was an anthropologist working in

Rwanda in the mid-late 1980s surprised to find himself suddenly balancing two

perspectives: one, the study of behavior with the aim to change the informant's actions

and two, the study of culture with the aim to change the perception, by outsiders, of the

informant's behaviors (1990).

In his anthropological fieldwork in the mid-late 1980s on personhood in Rwanda,

Taylor had the opportunity to witness the impact of the first attempts by the Rwandan

government and various NGOs to introduce condoms as AIDS prevention. He observed

strong resistance by many of the Rwandans with whom he spoke about the use of

condoms. At the same time, a survey conducted in 1988 funded by the WHO reported

that 97% of the respondents were aware that HIV/AIDS could be transmitted sexually,

but an astounding 0% reported condom use, although 69% of men and 47% of women

were familiar with condoms and their use against HIV/AIDS (Carael, 1988 cited by

Taylor 1990:1023). Taylor then explored the meaning of risk in Rwanda (based on

community, not individual, health) and the conceptual and literal importance of "flow"

and "blockage" during sexual intercourse. He predicted that current health education









campaigns designed to increase condom usage in Rwanda would be unsuccessful because

the notion of personal, individual risk was weak in Rwanda. But more importantly, the

use of condoms interrupted a critical aspect of sexual intercourse for Rwandans the

exchange of sexual fluids. Taylor offered some speculative suggestions for applying

these insights to health education programs, but his main intent was to explain to western

biomedical and public health practitioners the logic behind the apparently "illogical"

behavior of failing to use condoms in the midst of the AIDS epidemic in Rwanda. Here

was Anthropologist as sympathetic Translator. Soon after the publication of this piece,

however, Taylor developed his work into a book integrating these perspectives (1992).

Armelagos and colleagues suggest that the continued maintenance of the two

distinct models medical anthropologists traditionally employ to understand, explain, and

improve human-disease interactions-cultural and biological-prevents the emergence of

an integrated medical anthropology (Armelagos et al. 1990:354). This chasm is

especially acute among anthropologists who study HIV/AIDS. The biocultural

perspective addresses this divide. It combines the cultural and biological perspectives of

medical anthropology by modifying the epidemiological model of host-pathogen-

environment to include an analysis of all these elements at the scales of both the

individual actor and the population or economic system (1990:355). Further, the

biocultural model uses an "actor-based" model of coping that emphasizes the choices and

constraints of an individual actor enmeshed in a larger interacting system of health and

economy (1990:358).

More recently, Armelagos and colleagues have updated and expanded the

biocultural approach. Smith and Thomas write of the "biocultural metamorphosis" which

attempts to move beyond the "complementarity of behavioral and biological responses"









to embrace "the complexity of this interaction" (Smith and Thomas 1998:463). The

study of HIV/AIDS seems particularly well-suited to embrace this complexity and

challenge the call to biocultural anthropologists to set this metamorphosis in motion.

While HIV is an infectious disease, it is also a chronic disease with an extremely long

latency period. To further complicate the situation, the trajectory of a patient's viral load

(amount of HIV virus replicating in the body) from initial infection to death begins with

an accelerated period of extreme infectiousness, peaks, troughs, and peaks again many

times depending on a variety of biological, psychological, economic, and environmental

factors. A recent paper based on data from Uganda concludes that viral load is the chief

predictor of the risk of HIV-1 heterosexual transmission. Furthermore, transmission is

rare when the viral load is < 1500 copies of HIV-1 RNA per milliliter of blood (Quinn et

al. 2000:921). The interaction of these modulating biological factors with the

complexities of human sexuality and a fluctuating backdrop of individual and macro-

level economic forces suggests that no single explanatory model could fully account for

the infinite combinations of factors which lead one person to seroconversion while

another remains seronegative.

Rather than be overwhelmed by the seeming impossibility of accounting for all

factors related to HIV/AIDS in Uganda, Wallman and colleagues enthusiastically

embrace the perspective that everything is related, in some way or another, to AIDS

(Wallman et al. 1996). For Wallman, the task is to sort out the relevant proximate causes

of vulnerability to HIV/AIDS as defined by both the citizens of Uganda and by her

training as a Social Anthropologist specializing in health-related behaviors. In a

remarkable collaboration of Ugandan residents of a Kampala shanty-town, European-

trained professional Ugandans, and European social scientists and physicians, Wallman et









al aim to contribute to the reduction of the transmission of HIV in a densely-populated

market area of Kampala called Kamwokya by understanding behaviors which prevent

women from seeking early treatment of STIs (Wallman et al. 1996:2-4). Over a period of

two years (1994-1996), Wallman and colleagues used a wide variety of social scientific

research techniques, ranging from rich ethnographic observation of public life in the

study area to a community created video, to assess an individual's risk of HIV.

They discovered that "risk" had a variety of meanings in a variety of contexts, and

women shuffled these risks according to their personal assessment of the situation at

hand. For example, protecting oneself from the biological risk of HIV by insisting on a

condom might mean risking an public reputation as a fertile woman and the advantages

that reputation confers (Wallman et al. 1996:229); different women at different life stages

will handle this choice differently. The amount of control one woman felt she had over

her "disease-environment" was often related to the amount of responsibility that the city

government of Kamwokya took for the health standards in the streets outside her house.

For women for whom the city did nothing to control mosquitoes, rats, garbage, and

human waste, the attention devoted to minimizing her personal risk was often less than

the woman who was better able to control her compound and the surrounding area

(Wallman et al. 1996:229-230). These observations provided important insights for

health educators planning yet another AIDS campaign about taking personal

responsibility to protect one's self against HIV/AIDS; safe sex messages were actually

counter-productive at that point in the epidemic (Wallman et al. 1996:229).

In contrast to the achievements of a large research team, the work of Michelle

Renaud demonstrates the contributions that a single field researcher can make to the

anthropological literature as well as the virtues of attempting a sustained treatment of one









issue in a monograph (1997). Her book, Women at the Crossroads: A Prostitute

Community's Response to AIDS in Urban Senegal, draws on her 1991-1992

anthropological doctoral research and offers a keenly observed description of the daily

lives of a community of prostitutes in urban Senegal. It is a richly detailed account of a

successful HIV/AIDS behavioral intervention program that promoted condom use and

offered aggressive treatment of STIs to commercial sex workers. Presented as a series of

character sketches of the women who share a compound of small rooms at the edge of the

dusty town, Renaud eschews the standard didactic strategy of alternating snipets of

interview dialogue with anthropological analysis. Instead, the answers to the questions

she posed are woven together with an almost novelistic flair. We learn which lessons

these women brought home from AIDS seminars to reduce their risk of HIV

transmission. We learn which lessons were less useful and how the women compensated

by acting together as a group to convince reluctant clients to use condoms. Remarkably

enough, prostitution is legal in Senegal, and those women who register with the local

police and visit the STI clinic every two weeks are not harassed by police seeking bribes.

This allows the women to pursue their trade in relative security, but it doesn't necessarily

translate into acceptance by their families or the community.

Secondly, the book is a rare look into the moral ambiguities prostitutes face in a

predominantly Muslim country where the struggle to survive in a tough economy is

balanced against the struggle to be a good Muslim. Renaud goes beyond merely

recording the number of transactions per prostitute per day to engaging client and

prostitute as well as "respectable" men and women in dialogue about the relationship

between religion and commercial sex. How does one ensure passage into heaven? Is it

enough to perform the haj? Does God understand that for many the only alternative to









prostitution is begging? If a woman saves her children from poverty, doesn't that absolve

her in God's eyes of her earthly sins? Renaud demonstrates how much the economy of

local merchants and neighbors depends on facilitating the work of the prostitutes. Even

the STI laboratory technician managed to profit from the twice-monthly clinic visits by

selling lunch to the prostitutes from his mother's tea hut behind the clinic. Although

Renaud's work is not an intervention, it is a profoundly rich analysis of what happens to

the intervention "message" when people come home and live the AIDS lesson they

learned that day at the clinic.














CHAPTER 3
METHODS


Sampling Design

The main sampling objective of this research project was to produce a body of

data about the current sexual behaviors and risk of HIV-1/2 transmission of Ugandan

adolescents in the western district of Hoima that could be generalizable to the national

level. I also needed data to test my null hypothesis: for youths aged 15-19 in Hoima

District, there is no statistically significant correlation between level of risk behavior and

number of continuous years in school. In other words, I was interested in understanding

the relationship between formal education and sexual risk behaviors. To make the

sampling frame rigorous enough to produce comparable results with other studies

conducted at the national level, both the survey and epidemiological data had to be

collected in such a way that other social scientists, public health officials, and policy

makers would have confidence in the results. That meant I needed to follow a sampling

design similarly rigorous to recent Ministry of Health studies (Musinguzi 1996;

STD/AIDS Control Programme 1998) or Demographic Health Surveys (Statistics 1996).

To achieve this end, I conducted a probability sample where the selection of participants

was random and the reliability of the sample strategy reduced sampling error as much as

possible (Cochran 1977; Henry 1990; Som 1996). Fortunately, a complete enumeration

of the total population of Uganda was conducted in a house-to-house population census

in 1991 by the Ministry of Finance and Economic Planning and the Department of








Statistics at Makerere University (Statistics Department 1992). According to the 1991

Population and Housing Census, there were 22,428 youths aged 15-19 in Hoima District

(Statistics and Uganda 1992:p:8). The results are generally considered to be of high

quality and are among the best population-based demographic statistics available on any

country in sub-Saharan Africa. Sample size is discussed in the next section below.

In addition to a random sample, I needed a stratified sample. I divided the total

universe of youths in Hoima District by the three strata of the independent variable:

school-goer, school-leaver, and no-school. However, it was not sufficient to merely

sample each strata in direct proportion to their actual distribution in the general

population. Of all youth in Hoima District aged 15-19 years, approximately 15% are

enrolled in or have completed high school (Statistics Department 1992:14). If I

conducted a truly simple random sample, there was a strong chance that I would not end

up with a sufficient number of school-going youth in the sample.

More importantly, however, I suspected that school going adolescents, who

generally had greater access to consumer goods, exerted a far greater influence on the

contours of local adolescent culture than their small numbers might otherwise predict. I

suspected that school-goers, quite visible in town and village in their neatly pressed

school uniforms, carrying book bags, and wearing the latest fashions in shoes and

hairstyles, would set the local standard for adolescent behaviors. In addition, our

research team observed them at discos and video clubs setting the local brand choice for

alcohol. I hypothesized that the importance of their consumer behavior and access to

cash might be an indicator of their influence on local youth sexual behaviors and sexual

networks as well. I needed to devise a sampling strategy which would capture a larger









proportion of secondary school students in the sample than was actually represented in

the general population. Therefore I oversampled the school-going population by a factor

of two and attempted to distribute the three strata evenly in the total sample; that is, one-

third school-goers, one-third school-leavers, and one-third no-school. In the end,

however, the actual proportion of adolescents who participated in the survey was school-

goer 36%, school-leaver 42%, and no-school 22%. Chapter Three discusses the

demographic results in greater detail.

In addition, the study employed a multi-stage design to ensure adequate coverage

of both semi-urban (town) and rural (village) adolescents living in Hoima District. In

Hoima town, all seven day and boarding schools were sampled (Table 3-1) and all the

town wards were canvassed for interviews with school-leavers and no-schoolers. For the




Table 3-1: List of sampled schools for in-town survey

Name of school in Hoima town Total # males in Total # females in Total # males Total # females
school school interviewed interviewed
for survey for survey
Bwikya Muslim Secondary 225 225 14 14
Duhaga Secondary 279 225 28 31
Hoima High 279 261 14 14
Hoima Modem Secondary 100 150 7 7
Hoima Progressive Secondary 190 110 13 14
Hoima Secondary 53 52 7 7
Kitara Secondary 265 200 5 3
St. Andrea Kaahwa College 60 140 14 14
TOTAL 102 104






40


Table 3-2: Schedule of survey interviewing in rural Buhimba sub-county

Parish (LC2) Village (LC1) Building Total boys Total girls
interviewed interviewed
Kyabatalyia Buhimba West Gomborola 7 5
Headquarters
Buhimba Secondary 6 6
School
Kibararu Primary 5 6
School
Ruhunga Kitoole Kitoole Primary 6 6
School
Kijugunya Catholic Church 6 6
Rwoga Catholic and Anglican 10 5
Churches
Musaija Mukuru East Musaija Mukuru Solomon Iguru 4 10
Primary School
Bujalya Bujalya Primary 11 16
School
Musaija Mukuru West Kigaya Kigaya Primary 6 7
School
Kisiiha Kisiiha Trading Center 14 14
Karama Catholic Church 5 7
Kinongozi Kabale Catholic Church 5 7
Kisenyi Kisenyi Primary 6 8
School
Ngogoma Ngogoma Church 13 14
TOTAL 104 117


rural interviews, we interviewed in all six parish divisions of one sub-county of Hoima

District (Table 3-2 above; also see Figure 3-1 Map of Buhimba Sub-County below, page

43). We conducted the selection of all informants without replacement so that none

could be interviewed twice (see also Appendix C). The strata were two of the main

independent variables, school status (in-school, drop-out, no-school) and gender.

Sampling was carried out in each stratum to ensure proportional representation of boys

and girls and adequate coverage of the three school strata.








Sample Size

Published epidemiological data guided my decisions for estimating sample size.

As discussed below in Chapter 3 in the section on Sample Size, there were no published

seroprevalence data from Hoima. I had to guess at the sample size I would need in order

to achieve a minimum number of observations in each cell in a chi-square table. For

example, a 2x2 table with gender and HIV serostatus would need a minimum of 5

observations in the HIV positive cells for both boys and girls. I assumed that the

HIV/AIDS seroprevalence data on rural adolescents reported in Nunn (1994) could be

used as a proxy for determining my own sample size. Seroprevalence data from Fort

Portal could have been more relevant (Killian 1999, Kipp 1995) because this German-

funded project looked specifically at the 15-19 age cohort, but they declined to release

any of their preliminary data before publication. The intent of my research design was to

make an original contribution to the literature by producing seroprevalence data for this

understudied age cohort. I also wanted to distinguish my data from those data that were

collected through sentinel surveillance at an antenatal clinic or hospital (Asiimwe-Okikor

1997; Berkley 1989; Hudson 1988; Uganda STD/AIDS 1996, 1999) or through partial

(Kengeya-Kayondo 1996; Nunn 1994) or 100% coverage of a community (Wawer 1991,

1997, 1998, 1999, 2000). My sampling strategy was designed to collect a random sample

of all youth aged 15-19 in Hoima schools and in rural Buhimba sub-county, a design no

other researcher had used previously to study adolescent HIV/AIDS in Uganda.

Sample size was based on a number of factors (Lemeshow et al. 1990). Rates of

HIV-1/2 seroprevalence in the 15-19 cohort in the Masaka study served to approximate a

rate for Hoima because they were the only available published seroprevalence data based

on laboratory diagnostic testing for this age cohort in Uganda (Nunn 1994). Other factors








determining sample size were the minimum number of case observations required for chi

square analysis at .05 significance and project resources. The null hypothesis stated that

there was no significant difference in HIV-1/2 antibody serostatus for youth aged 15-19

according to school status (in-school, drop-out, no-school). To successfully use chi

square analysis to test the dependent variable (HIV-1/2 antibody test result) against the

independent variable (school status), a minimum of five cases (HIV-1/2 antibody

positive) is required in each cell or stratum of the independent variable for a p value of

.05. This calculates to a total of 15 cases. In Masaka, the 1990 seroprevalence data for

males aged 15-19 was 0.4%, while for females aged 15-19 it was 6.7% (Nunn 1994:83).

In this study females were the gender of interest, therefore a minimum sample size of 224

was needed to obtain 15 female cases. Resources allowed an increase to 300 males and

300 females. Forty pilot surveys were excluded from analysis and the final sample size

was 560, of which 288 were females and 272 were males.


Randomization and Selection Process

The geographical areas for the interviews and the individual study participants

were chosen randomly by a variety of techniques. Hoima District (LC5) is

administratively divided into two counties (LC4), four sub-counties (LC3), five parishes

(LC2), and many villages (LC1) of about 200 persons each or less. Based on these

divisions, Hoima town was chosen as one research area and rural Buhimba sub-county as

the second research area. The rural interview area was determined randomly by a coin

toss (result: Bugahya county) and the lottery method (result: Buhimba sub-county).

Interviews were conducted in two to three villages in each of the five parishes in

Buhimba sub-country (see Figure 3-1 Map of Buhimba Sub-County). Finally, villages







Hoima Town


Survey Interview Sites
/\/Main road


Figure 3-1: Map of locations of interviews in Buhimba sub-county.








(LC 1) were chosen if they had vacant primary school or church facilities in which

interviews could be conducted. For the actual selection of informants, I used several

different strategies to create the random sampling frame for the final sample. For the

currently enrolled students, I reviewed the school enrollment list for each school for

levels Secondary 1 through Secondary 4 (S1-S4). These grades correlate roughly with

grades 9-12 in the American educational system. I selected a name on the first

registration list and then chose every 10th name until I had a list of 30 randomly selected

names. If the final gender balance was not 1:1, I adjusted the list by choosing the name

or names of every next 10th boy or girl as needed. We then asked each student on our list

if they wanted to participate in the interview. Almost all students agreed to the interview.

If someone was absent, or declined to participate in the interview, I returned to the

Headmaster's office and chose another name at random from the registration list.

Creating a sampling frame for the rural areas and the town interviews was more

difficult (see Figure 3-2). There are no lists of adolescent residents living in the rural

areas or in town, so we could not generate a simple random sample of participants.

Instead we chose the participants based on a cluster sample of their village or town ward.

For the rural area, we interviewed equal numbers of adolescents in each of the six

parishes (LC2) of Buhimba sub-county (LC3). In this area, the homesteads are dispersed,

often separated by one kilometer or more from each other. We sent a letter to the local

government representative (LC1) explaining the research project and requested that they

inform the young people of the village of our arrival the next week. In some villages we

walked house to house and requested to speak to any adolescent aged 15-19 in the home.

Each of the seven research assistants set off in different directions from the central







45



crE
)
E 2
-E -
E -o to 8
0 0

-6 -CU mco cm o
CJ I 0
S 1







o0 CO
CO

a -.



a.
.5 E
m 0 00


1 / =











0/
U* II












Oo
I /


4-


< *


o-
(0
So0
(D C.
' 0) C
(0 U cC


Figure 3-2: Map of Hoima town and Hoima schools.








interview location usually a trading center, church, or primary school to enlist two or

three informants for their day's fieldwork. Often the research assistants exchanged

informants to avoid any bias which might be the result of a particular research assistant

choosing a certain type of informant over another.

Other days we would arrive to find a small group of young people waiting for us

in the trading center and we would randomly choose the informants for that day. Often,

young people would trickle in as the day progressed. We had to balance our strategy of

randomly interviewing every second or third youth with the possibility that not enough

people would show up, often making us reluctant to turn away anyone eligible for an

interview. Most often the sampling strategy was a mix of all of these procedures. Boys

responded more often than girls to the call by the LC for participation and we often

resorted to door to door requests for young women to keep the balance of interviews

between boys and girls equal. Canvassing door to door was everyone's least favorite

strategy because it was time consuming and exhausting. In addition, most of the local

community leaders preferred not to have us walking all over the village disturbing people

in their homesteads. For this reason, we were quite fortunate to have such dedicated LC

representatives in all the villages who worked very hard at assembling seven boys and

seven girls for a full day of interviews for locating a certain type of youth, for example,

housegirls, hair salon girls, boda boda boys (bicycle taxi drivers), construction workers,

etc. This strategy was adopted because we needed several attempts to secure an

interview with these busy and often elusive town youth. The other research assistants

systematically looked for unemployed youth who were loitering in town or around their

neighborhoods.








Informed Consent

Before an adolescent participated in an interview, the informed consent document

was read aloud by the interviewer to each individual informant in a private setting

(Appendices D and E). All research instruments and consent forms were offered in

English and Runyoro, the local language. Each informant verbally indicated to the

interviewer that they wanted to participate and that they understood the procedures to

maintain confidentiality as well as the intimate nature of some of the survey questions.

The informants then signed or placed a fingerprint on two identical consent documents

and kept one copy for themselves. Similar procedures were followed for the voluntary

HIV-1/2 antibody rapid diagnostic serum testing and counseling (VCT) (Appendices F

and G). An additional signature was required for youth under age 18 from a parent or

guardian indicating that the parent or guardian agreed to the test for the youth and that the

results were confidential and released only to the youth.


Eligibility

One of the primary goals of the research was to establish the relationship between

school experience and HIV-1/2 related risk behavior for the 15-19 age group. Therefore

we had to be certain that all youth enrolled in the study were aged 15-19 and that we were

clearly measuring the effects of schooling on their sexual behavior. For youth who could

state their age, we also asked the year of birth to be sure the reported age was correct.

Most youth with some school experience could confidently name the year of their birth

and sometimes the month and day. But this was a more difficult task for illiterate rural

and town youth who rarely need to know their precise age. If mothers were near by, they

were consulted; but more often we resorted to mnemonic techniques to establish the year









of birth. For example, the current president of Uganda, Yoweri Museveni, waged a

successful battle in this area in 1986, the memory of which serves as a time marker for

most rural residents of Buhimba. If a youth could confidently say what grade they were

enrolled in when Museveni's troops liberated the area, then we could calculate their age.

Or, if a boy had already paid more than one year of graduated tax, he was older than 19

years and thus ineligible for the study. If we were unable to determine year of birth, the

youth was ineligible for the survey.

A second eligibility algorithm distinguished three categories of school status. A

school-goer is a youth continuously enrolled in school with no breaks since entering his

or her first year of primary school. A school-leaver has attended at least five years of

basic primary school but has not attended school for a minimum of 18 months prior to the

survey. A youth designated as no-school has less than four years of primary school

experience.


Survey Instrument

The questionnaire was developed based on existing questionnaires I gathered

from various on-going surveys in Uganda (Blanc et al. 1996; Statistics and Inc. 1989;

Statistics and Inc. 1996), plus my own original questions (Appendices H and I). The

most common way to proceed in surveys is to conduct focus group discussion or key

informant interviews first and then draw up the survey based on the qualitative findings.

Because I felt that the more significant information would come out of the qualitative

techniques, and since there was an abundance of extant surveys, I decided to start with

the survey portion. I reasoned that this would allow me the opportunity to become

familiar with the community, but more importantly, the community could become









familiar with me and hopefully come to trust me as they saw me interact with local

officials. I could also use the survey to test some common sense ideas parents held about

adolescent sexuality. Adult ideas about adolescent sexuality are often based on the

adults' own informal "survey" of watching youth and gauging behaviors but not really

discussing these issues in-depth with them. Using data reported by the youth, I could get

an overall sense of the contours of adolescent sexual behavior. I would then use that

information as a springboard to go beyond the usual questions in the focus groups and

ask compelling questions about the contradictory behaviors reported in the survey. I felt

that the more formulaic information from the survey was easier to gather and would

allow me time to become a trusted figure in the community and ultimately lead me into

important questions for the qualitative portion of the research.

The pre-coded survey measured standard demographic characteristics plus

work/employment, access to cash, social life, tobacco, alcohol and substance abuse,

sexual education, sexual debut, sexual relationships, contraceptive use, STIs, and

HIV/AIDS knowledge. The questionnaire was carefully translated and back translated

between English and Runyoro (described in more detail below). Special attention was

given to precise and socially proper translation of words and concepts likely to offend or

have multiple meanings, such as boy/girlfriend, husband/wife, lover, sexual intercourse,

marriage, consent, and rape.

Due to the sensitive nature of many of the survey questions, it was essential to

find a quiet and secure place for the interview. This also ensured confidentiality and

promoted an atmosphere for truthful answers. In Hoima town, interviews were

conducted in private rooms at the research office, in empty meeting rooms at the Hoima









Rural Training Center, or scattered outside on school grounds under shade trees. In

Buhimba sub-county, interviews were conducted in empty primary school classrooms or

unused church facilities. To date, no complaint of breech of confidentiality for either the

survey or the voluntary HIV/AIDS testing and counseling component (discussed in

Chapter Four) has been reported to the local interview team. Refusal rate for the survey

was less than one percent.


Research Assistants

With only rudimentary ability to speak the local language Runyoro, I knew I

needed research assistants to help me administer the survey and conduct focus group and

key informant interviews. The general consensus in Kampala among research

professionals was that the best research assistants were those who were formally trained

at the School of Social Work and Administration at Makerere University. While I agreed

that graduates from this program were well trained, I suspected that that training might

not necessarily have prepared the students for the rigorous field conditions we would

certainly confront, especially during the month of rural interviewing. Most students who

are fortunate enough to attend the university come from an elite class and most are

educated at the best boarding schools far from their home districts. Even if these students

had been back recently in their home villages, they probably did not feel at ease walking

around the rural homesteads far from their own homes in the central trading centers.

In contrast to these privileged university students, I suspected that there were

talented, ambitious young people living in Hoima District who simply did not have the

means to make it to the university. But with a little training, I was willing to gamble that

they would probably fulfill the requirements of field research assistant as well as those









recommended university graduates. I advertised the position at local Red Cross

meetings, spoke in churches and schools, canvassed the hospital and local clinics, and

hung out at the sports club. Within three weeks, I interviewed over 30 candidates and

chose 12 for a one-week intensive training course in research methods and interviewing

techniques. The course covered survey interviewing, ethnographic interviewing,

translation and transcription, determining informant eligibility, plainly explaining the

goals of the research, proper codes of conduct such as respect for informants, maintaining

strict confidentiality, and professionalism. At the end of the training, I observed each of

the 12 candidates in a one-hour mock interview in English and chose seven assistants-

four men and three women. At the beginning of the training session, I made sure that all

the candidates were aware that there were more applicants than positions and that not

everyone could be offered a job with the project.

Working with local residents was important for two reasons. First, it was an

effective way of gaining the trust of the community because the locally recruited research

assistants also served as public relations personnel who were often called upon to answer

questions about the project by their neighbors. Second, it demonstrated my commitment

to the community by equipping local young people with skills that they might be able to

use again in the future.


Translating the Survey Instrument

Translating the lengthy survey from my American English into "polite" but easily

understandable Runyoro was an unexpectedly intense experience. For three weeks the

research team and I plodded question by question through the English survey, engaging

in long and often heated debates over the correct choice of vocabulary and even









orthography. Ultimately it proved to be a critical process. I could have simply hired one

person to translate the survey from English into Runyoro, but what emerged from the

experience was not only a technically correct and conversational translation, but a good

sense of teamwork and an intimate familiarity with the survey instrument. We developed

confidence in the survey and knew we were a team ready to take on the formidable task

of 600 interviews on a sensitive topic.

The group translation process served yet another purpose. I learned that not

everyone in the group of seven assistants was equally fluent in Runyoro. I had assumed

that anyone I interviewed in Hoima would be fluent in Runyoro if they were born here

and had been educated locally through secondary school. (Of course I perfunctorily asked

them in the interview if they were fluent and they all assured me they were.) I wrote one

of the survey questions in English on a big sheet of blank newsprint taped to the wall of

our rented office. Then the group would debate among themselves the best translation

and one of the assistants, or myself, would write the Runyoro translation on the poster

paper for further discussion about spelling and punctuation. By then, I could "hear"

Runyoro fairly well and knew the general rules of grammar, so I managed to write the

translation if the group members spoke clearly and slowly. So I was shocked to discover

that I was better at writing Runyoro than one of my chosen assistants. It turned out that

she could not write Runyoro at all. She was conversant in Runyoro, but had never

written it, and did not really speak it at home. However, because she was the only

interviewer with a B.A degree from Makerere University, I was reluctant to let her go..

I suffered a few sleepless nights wondering what to do about it. I was worried

that her weak command of written Runyoro would compromise her ability to carry out









interviews in the rural areas. Female interviewers were much more difficult to find than

male interviewers, and she did so well in the English interview, I did not want to replace

her with a less impressive female candidate from among those I had rejected initially.

Eventually I decided to let her conduct the English interviews with the school girls who

chose English over Runyoro as the language of the interview and give her more time to

improve her Runyoro skills. I reasoned when we were ready to move out into the

villages where very few people would choose English over Runyoro, I would have a

colleague observe her progress in Runyoro and then decide whether she was able to write

down Runyoro well enough for the open-ended questions. Eventually, her skills in

Runyoro returned and she passed her "exam" with the local government officer for

culture. In the end, she proved to be one of the most conscientious field workers in the

group.

The final step was to "back" translate the Runyoro version of the questionnaire

into English. By comparing the second back-translated "local" English version with the

original American English version I was able to determine if my original meaning was

lost or distorted by the Runyoro translation. The District Information Officer, a Runyoro

specialist, offered to do the back translation. We pretested the final survey amongst

ourselves, and then in pilot interviews with the younger brothers and sisters of the

research assistants.


Data Management and Analysis

I entered all pre-coded survey and epidemiological data in the field using EPI-

INFO 6.0 (Centers for Disease Control, Atlanta, GA, USA). SPSS 9.0 (Statistical

Products and Service Solutions, Chicago, IL, USA) and SAS 6.03 (Statistical Analysis









System, Cary, NC, USA) performed more complete data analysis in the office.' One-way

ANOVA tests calculated significant associations at the 0.05, 0.01, and 0.001 levels

between the main dependent variable (risk profile as a continuous numeric variable) and

a variety of independent variables. However, Bartlett's test for homogeneity of variance

indicated a high level of between and within group variance of the sum of squares values.

This suggested the need to transform the data. In addition, Tukey's post-hoc test statistic

indicated the possibility of Type I errors. Standard transformations (square root, natural

log) of the data set failed to pass Bartlett's test. For these reasons, we did not run a

regression analysis or MANOVA to determine the predictive strength of any of the

independent variables for the dependent variable "risk profile." The most appropriate

analysis may be log-linear analysis (regression analysis for categorical data); we will

explore this test at a later date.

The risk profile variable was transformed from a continuous numeric variable to a

categorical ranked ordinal variable and analyzed using standard nonparametric

techniques and test statistics. The Kruskal-Wallis Analysis of Variance test

approximating the one-way ANOVA, Pearson chi-square, cross tabulations and the

Spearman Rank Correlation Coefficient were chosen. If the Pearson chi-square test

statistic indicated a significant P value of < 0.05 for an association between two variables,

then raw unstandardized residual differences between observed and expected counts in

individual cell counts were analyzed for a more precise understanding of the significance

of the overall association. Multivariate data are reported as significant at the 0.05 level if


'For SAS and general ANOVA statistical procedures I consulted Madrigal (1998) and Weinberg and
Goldberg (1990). For categorical data analysis and log linear procedures I followed Afifi and Clark (1996),
Agresti (1996), Fienberg (1980), and Le (1998). For a specific discussion of SPSS software I consulted
Bryman and Cramer (1997) and Morgan and Griego (1998).






55


the adjusted residual value is 2 1.9 or < (-1.9). Frequency data by gender are reported as

significant if P < 0.05 by Pearson X2 or Fisher's exact test. These data are discussed in

the next Chapter.














CHAPTER 4
SURVEY RESULTS: TESTING THE HYPOTHESIS


Introduction

This chapter presents an overview of selected socio-demographic and behavioral

characteristics that describe the 560 youth in our sample (see Appendices H and I for the

survey instrument). For variables such as religion, ethnicity, marital status, and

adolescent pregnancy, our sample conforms closely with many of the results obtained

during the 1991 Population and Housing Census, Hoima District, Uganda (October 1992)

for the 15-19 age cohort (Statistics and Uganda 1992). Other examples of the vigor of

our sample are discussed below. We feel confident, therefore, that our sample is

representative of the entire district and is generalizable to the national level.


General Socio-Demographic Variables

Table 4-1 reports a few general characteristics of the youth sample. Gender/sex

and age are proportionately well represented in this sample. Table 4-2 offers selected

statistics for comparison from the 1991 Population and Housing Census, Hoima District,

Uganda (October 1992) (Statistics and Uganda 1992). As discussed in Chapter 3,

residence and education were sub-strata of the sampling design, and their well-balanced

proportions in Table 4-1 are due to our particular sampling strategy. The representation

of religion and ethnicity in our youth sample is generally standard for the western region

of Uganda, except for two characteristics. First, our sample has a higher rate of Moslem

youth than government results in Table 4-2 because we specifically over-sampled a










Table 4-1: Selected socio-demographics (n=560 unless otherwise noted)


Frequency (%)
Gender/sex
Male 272 (48.6)
Female 288 (51.4)
Age
15 61 (10.9)
16 101 (18.0)
17 151 (27.0)
18 159(28.4)
19 88(15.7)
Residence
Rural Buhimba 217(38.8)
Hoima School 203 (36.2)
Hoima Town 140 (25.0)
Education
None (P4 or lower) 122 (21.8)
Drop out (P5 or higher and out of school at least 18 months) 234 (41.8)
Continuously enrolled (in-school) 204 (36.4)
Religion
Catholic 200 (35.7)
Anglican/Protestant 260 (46.4)
Muslim 45 (08.0)
Other 55 (09.9)
Ethnicity
Munyoro 454(81.1)
Muganda 27 (04.8)
Other 79(14.1)
Marital Status
Never married 468 (83.6)
Married/Living with partner 78 (13.9)
Widowed 1 (00.2)
Divorced/Separated 13 (02.3)
Adolescent pregnancy
You or partner now or ever pregnant (yes) (n=560) 119 (21.3)
You or partner now or ever pregnant (yes) (n=370 all sexually experienced youth) 119 (32.2)
Average cash per month from any source
0 $US 241 (43.0)
1 5 $US 61(10.9)
6 10 $US 84(15.0)
11 20 $US 90(16.1)
> 20 $US 84(15.0)
If work for cash is steady or permanent (n= 17)
Housework 17(14.5)
Farmwork 21 (17.9)
Alcohol production/distribution 5 (04.3)
Construction, labor 44 (37.6)
Market seller or own business 30 (25.7)









Table 4-2: Selected socio-demographics for Hoima District from the 1991 Population and
Housing Census for Hoima District, Uganda

(%)
Religion (age 15-29)
Catholic 46.4
Church of Uganda (Protestant/Anglican) 47.2
Moslem 04.8
Other 01.6
Ethnicity (all ages)
Munyoro 75.9
Muganda 01.2
Other 22.9
Marital Status (age 15-19, n=22,428)
Never married 73.8
Married/Living with partner 24.4
Widowed 00.2
Divorced/Separated 01.4
Not stated 00.2
Adolescent pregnancy (age 15-19, n= 11,071)
Ever pregnant (yes) 31.4

Source: (Statistics and Uganda 1992)




Moslem school in Hoima town for informants. We also allowed for more responses in

the "other" category than the government census and this might account for our lower

response to "Catholic." Second, our sample has a higher rate of youth who claim their

ethnicity as Baganda than the government reports for all ages; perhaps our higher

percentage is an artifact of our younger 15-19 age cohort-government statistics for this

variable were not available disaggregated by age group. Our data on female fertility

disaggregated to sexually experienced youth (n=370) compare well with the 1991 census

data. Lastly, the variables access to cash for the entire sample and types of employment

for youth working in a permanent or steady job, are included in Table 4-1.

Table 4-3 reports the level of parent's education; our results are not surprising

with fathers reported to have a higher level of education than mothers. However, it is










interesting to note that adolescents are more likely to be unsure of their father's education

than their mother's. Table 4-4 describes self-reported use of cigarettes, alcohol (both


Table 4-3: Parent's education (n=560)


Frequency (%)


Father's education
Post secondary school 56 (10.0)
Secondary school (S 1-S6) 147 (26.3)
Primary 5-7 133 (23.8)
Up to primary 4 or never attended 92 (16.4)
Do not know 132 (23.5)
Mother's education
Post secondary school 28 (05.0)
Secondary school (SI-S6) 108 (19.3)
Primary 5-7 165 (29.5)
Up to primary 4 or never attended 164 (29.3)
Do not know 95 (16.9)







Table 4-4: Substance use by informant (n=560)
Frequency (%)
Smoke cigarettes (yes) 30 (05.4)
Ever drink alcohol (yes) 234 (41.8)
Ever tried illegal drugs (yes) 7 (01.3)







Table 4-5: Substance use by informant's three best friends
Frequency (%)
Smoke cigarettes (yes) n=548 103 (18.8)
Ever drink alcohol (yes) n=551 163 (29.6)
Ever tried illegal drugs (yes) n=552 19 (03.4)


commercial and locally made beers and liquors), and illegal drugs (mainly marijuana).

Table 4-5 describes substance use for three of the informant's best friends. Best friends

are reported to use more cigarettes and illegal drugs, but alcohol consumption of best









friends is reported to be lower than that of the informants themselves. It is generally

anticipated that adolescents will report higher substance use for friends than for

themselves. Consequently, it is difficult to interpret the relationship between these two

tables, except to note that alcohol use is fairly common among youth aged 15-19.


Sexual education, Behavior, and Health Variables

To date, few population-based surveys have been conducted in Hoima District. It

is even more difficult to find comparison data on sexual behaviors disaggregated by age

in either the published literature or privately circulated governmental and

nongovernmental reports. This paucity of comparison data is compounded by our focus

in this study on the variable "age at first sexual intercourse" because most population-

based surveys ask women about age at first live birth, not sexual debut. Furthermore,

very little survey data is available on the first sexual experiences of young men anywhere

in Africa (Setel 1996). For these reasons, much of the data presented here offer an

entirely new and detailed perspective on the sexual behavior of 15-19 year olds in Hoima

District.

Table 4-6 describes some basic data on sexual health and sexual experience.

Two-thirds of Hoima youth aged 15-19 report experience with sexual intercourse,

approximately four out of those five sexually experienced youth intend to continue to

engage in sexual intercourse over the next 12 months. Almost one-quarter report three or

more different sexual partners in their lifetime, one-quarter report two different lifetime

partners, and half of the 370 who reported sexual activity have had only one partner in

their lifetime. Of those 260 youth who lost a family or household member to HIV/AIDS,










Table 4-6: Sexual experience and sexual health of Hoima youth, age 15-19

Frequency (%)
Ever had penetrative sexual intercourse (yes) (n=560) 370 (66.1)
Male yes 183 (67.3%); Female yes 187 (65.1%)_
Intend to have penetrative sexual intercourse within next 12 months (yes) (n=560) 300 (53.6)
Number of different sexual partners over lifetime (n=370)
1 187 (50.5)
2 82 (22.2)
3 42(11.4)
4-6 39(10.5)
7 or more 20 (05.4)
You or partner ever pregnant (yes) (n=370 of sexually active male and female
respondents) 119(32.2)
Contraceptives ever used (n=370), multiple answers allowed
Condom use reported by either male or female 196 (53.1)
Male yes 96 (25.9%); Female yes 99 (26.8%)
Birth control pill 28 (07.6)
Injectibles 9 (02.4)
Diaphram 1 (00.3)
Rhythm or withdrawal or other 11 (03.0)
None 154(41.7)
Ever had STI infection (yes) (n=560) 20 (03.6)
Male yes 9 (3.3%); Female yes 11 (3.8%)
Lost a family or household member to HIV/AIDS (yes) (n=560) 260 (46.4)
Relationship of family or household AIDS patient to you (n=260)
Biological or step-parent 18 (06.9)
Older relative 190 (73.1)
Same-age relative 45 (17.3)
Other 7 (02.7)
Know someone your age who is sick with/or died from HIV/AIDS (yes) (n=559) 275 (47.3)


Table 4-7: Ever use of any contraception by all women, age 15-19, Uganda

Contraceptives ever used (n=1,606), multiple answers allowed %
Condoms (04.9)
Birth control pill (02.5)
Injectibles (00.5)
Diaphram (00.0)
Rhythm or withdrawal or other (10.0)
None (83.5)

Source: (Statistics and Macro International 1995, p. 47), frequency counts not available









the great majority lost an older relative, while almost 7% lost a parent. None of our

Hoima data from Table 3-6 (5th row) on contraceptive use compare well with 1995 DHS

data in Table 4-7, but when the variable "condom use ever" is disaggregated by

gender/sex, our data do compare favorable with the 1996 STD/AIDS Control Programme

data in Table 4-8. Both our data and that of the Ministry of Health report at least one-







Table 4-8: Ever use of condoms, by gender/sex, age 15-19, for Kampala District, Uganda,

Male Female
Condom ever used (yes) 25.0% 24.0%

Source: (STD/AIDS Control Programme, 1996, p. 24)




quarter of the youth population aged 15-19 have ever used a condom. Again, comparing

Hoima data from Table 4-6 with data from other regions in Uganda in Table 4-9, we see


Table 4-9: Percent distribution of respondents aged 15-19 who have ever had sex for
Kampala, Jinja, and Lira Districts

Kampala Jinja Lira
(central) (eastern) (northern)
Ever sex 45.5% 55.0% 60.0%
N 117 131 102

Source: amended from 1996 Uganda Ministry of Health data, (Musinguzi et al. 1996:19,
35, 50)









that reported sexual activity for the 15-19 age cohort is highest in Hoima District. Table

4-10 presents other comparative data as a measure of where adolescent sexual behavior in

Hoima District fits into the wider demographic data on Uganda. Possible reasons for

these differences are considered in the discussion below in Chapter 7.







Table 4-10: Adolescent pregnancy and motherhood in Ugandan women aged 15-19

Eastern Central Northern Western
Already a mother or
currently pregnant 51.1% 42.7% 41.4% 37.2%
N 350 502 344 411

Source: (Statistics and Inc. 1996:40)




Disaggregating Socio-demographic and Sexual Health Data by Gender/Sex

The often obvious influences of gender/sex makes it one of the most commonly

disaggregated variables in survey research. This section describes the relationship

between gender/sex and selected adolescent behaviors and characteristics from the Hoima

survey data. However, studying gender/sex as a variable has an additional purpose here:

the evaluation of gender/sex is also one of the most important post facto assessments of

the quality of sexual behavior survey data. If responses to survey questions, especially

sensitive ones regarding sexual behaviors, are mutually consistent for both men and

women, then the survey data are most likely of the highest quality (Ferry et al. 1995:28).

For example, in the Hoima data from Table 4-12 below, males and females

equally report: ever had sexual intercourse (male 67.3%, female 65.1%); condoms used









with first sexual partner (male 33.9%, female 38.5%); ever had a sexually transmitted

infection (male 3.3%, female 3.8%). Males and females report similar mean ages at first

sexual intercourse and at first parenthood (means not significantly different by

Independent Samples T test). Furthermore, when data from the 1996 Ministry of Health

survey in Table 4-9 are disaggregated by gender/sex and presented in Table 4-11, we see

by comparison the Hoima data are even more remarkable for their balanced gendered

perspective on adolescent sexual debut.


Table 4-11: Percent distribution of respondents aged 15-19 by sexual status for Kampala,
Jinja, and Lira Districts

Kampala Jinja Lira
Gender M F M F M F
Never sex 57% 52% 56% 34% 50% 30%
Ever sex 43% 48% 44% 66% 50% 70%
N 117 244 131 211 102 150

Source: amended from 1996 Uganda Ministry of Health data, (Musinguzi et al. 1996:19,
35, 50)



The influence of gender/sex is sometimes expected but also sometimes surprising.

It is expectedly statistically insignificant for these variables in Table 4-12: parent's

education, ethnicity (except the response Muganda), and religion. Gender/sex is

statistically significant for these variables at P < 0.001, and this is also expected: 55.9%

of girls versus 29.4% of boys have access to no cash at all; only 5.9% of girls versus

24.6% of boys have access to $20 or more per month; girls more often receive gifts for









Table 4-12: Selected socio-demographic and sexual health data disaggregated by
gender/sex (n=560)

Frequency (% or mean)
Males Females
(n=272) (n=288)
Mean age of sample (17.4) (17.1)
Ever had penetrative sexual intercourse (yes) 183 (67.3) 187 (65.1)
Mean age at first sex n=183 (15.0) n=187 (15.3)
***Have you or your partner ever been pregnant (yes) 32 (17.5) 87 (46.5)
Mean age at first parenthood (16.6) (16.2)
Residence
Rural Buhimba 100 (36.8) 117 (40.6)
Hoima School 99 (36.4) 104 (36.1)
Hoima Town 73 (26.8) 67 (23.3)
Education
None (P4 or lower) 56 (20.6) 66 (22.9)
Drop out (P5 or higher and out of school at least 18 months) 116 (42.6) 118 (41.0)
Continuously enrolled (in-school) 100(36.8) 104 (36.1)
Father's education
Post secondary school 23 (08.5) 33(11.5)
Secondary school (S -S6) 71 (26.1) 76 (26.4)
*Primary 5-7 75 (27.6) 58 (20.1)
Up to primary 4 or never attended 54 (19.9) 51 (17.7)
Do not know 49(18.0) 70(24.3)
Mother's education
Post secondary school 10(03.7) 18(06.3)
Secondary school (SI-S6) 47(17.3) 61(21.2)
Primary 5-7 82 (30.1) 83 (28.8)
Up to primary 4 or never attended 84 (30.9) 84 (29.2)
Do not know 49 (18.0) 42 (14.6)
Ethnicity
Munyoro 226 (83.1) 228 (79.2)
***Muganda 7 (02.6) 20 (06.9)
Other 39(14.3) 40 (13.9)
Religion
Catholic 97 (35.7) 103 (35.8)
Anglican/Protestant 136 (50.0) 124 (43.1)
Muslim 20 (07.4) 25 (08.7)
*Other 19(06.9) 36 (12.4)
***Average cash per month from any source
0 $US 80 (29.4) 161 (55.9)
1 5 US 30(11.0) 31 (10.8)
6 10 $US 42(15.4) 42(14.6)
11 20 $US 53 (19.5) 37(12.8)
220 $US 67 (24.6) 17(05.9)
If work for cash is steady or permanent
**Housework 3(01.1) 14(04.8)
*Farmwork 15(05.5) 6(02.1)
Alcohol production/distribution 3 (01.1) 2 (00.7)
***Construction, labor 43 (15.8) 1 (00.3)
Market seller or own business 16(05.9) 14(04.8)









Table 4-12-continued


Frequency (%)
Males Females
(n=272) (n=288)
Substance Use
***Ever smoke cigarettes (yes) 27 (09.9) 3 (01.0)
*Ever drink alcohol (yes) 126 (46.3) 108 (37.5)
Ever taken illegal drugs (yes) 6 (02.2) 1 (00.3)
Male circumcision, female labial elongation (yes) 25 (09.2) 22 (07.6)
*Age at first sexual intercourse
11-14 years 59 (32.3) 39 (20.9)
15-17 years 115(62.8) 139 (74.3)
18-19 years 9 (04.9) 9 (04.8)
Current relationship status
None 145 (53.3) 140 (48.6)
One casual or steady partner 104 (38.2) 100 (34.7)
*Several casual or steady partners 4(01.5) 0
***Living with partner 9 (03.3) 36 (12.5)
Formally married (traditional, church, civil) 9 (03.3) 11 (03.8)
Separated or widowed 1 (00.4) 1 (00.4)
Ever had STI (yes) 10 (03.7) 10 (03.5)
***Received money or gifts for first sex (yes) 13 (04.8) 141 (49.0)
***Gave money or gifts for first sex (yes) 116(42.6) 17 (06.0)
Used condoms with first sexual partner (yes) 62 (33.9) 72 (38.5)

*P < 0.05, **P < 0.01, ***P < 0.001. For categorical response data such as yes/no
responses, Pearson x2 test or Fisher's exact test for cell counts < 5; for continuous
response data such as age, Independent Samples T test (Levene's test for equality of
variances and t-test for equality of means).


sexual intercourse than boys (49.0% of girls versus 4.8% of boys); and boys report being

a parent fewer times than girls (17.5% of sexually active boys versus 46.5% of sexually

active girls). The significant differences in "if work for cash is steady" is also expected.

Some of the unexpected statistically insignificant variables related to gender/sex

include: ever had penetrative sexual intercourse; mean age at first sex; mean age at first

parenthood; and prevalence of STIs. The following variables are surprising for the

strength or weakness of their statistical significance for gender/sex: consumption of

alcohol is just barely statistically significant for gender/sex (37.5% of girls have ever had

alcohol versus 46.3% of boys; P < 0.05); likewise, the overall difference in age at first









intercourse is only slightly significant, most of that significance is due to more boys

reporting an earlier sexual initiation than girls (32.3% of boys at age 11-14 versus 20.9%

of the girls; P < 0.05); the disparity between males and females in reporting a live-in

partner is unexpectedly wide (3.3% of boys report a live-in partner versus 12.5% of girls,

P < 0.001). Describing the effect of gender/sex on access to cash in more detail also

gives surprising results. Overall, 72% of males but only 44% of females typically have

access to cash, either as pocket money from parents and family members or as earned

salary. However, gender/sex is not a significant factor for access to moderate amounts of

cash ($1 $20). For example, overall 38.2% of girls and 45.9% of boys report similar

access to cash in this moderate monthly income range. For adolescents in this area,

gender disparity in access to cash is most pronounced for youth with no cash at all or,

conversely, with large amounts of cash.


Disaggregating Socio-demographic and Sexual Health Data by School Status

As discussed in previous sections on sampling design and sample size in Chapter

3, the null hypothesis predicted that for youths aged 15-19 in Hoima District, there would

be no statistically significant correlation between HIV serostatus (later changed to level

of risk behavior) and number of continuous years in school. Similar to the influence of

gender/sex above, the relationship of school status to selected socio-demographic

variables is sometimes expected but also sometimes surprising. Table 4-13 below

presents the same variables listed in Table 4-12 above, only disaggregated by school

status. Criteria for the sub-strata of "school goer," "school leaver," "no-school" are

discussed in several sections of Chapter 3 above (Sampling design, Randomization and

selection process, and Eligibility). For school status, results that are statistically









significantly different, but expected, are: parents' education, all categories of ethnicity

and religion, types of employment for cash, and current relationship status. The

difference in mean age according to school status is significant, but this was not expected.

School leavers in this sample are on average 4 months older than school goers, and youth

without any school experience at all are on average 4 months younger than school goers.

In all likelihood, this is probably not an important difference. The variables "ever had

sexual intercourse" and "mean age at first sex" were not significant for gender/sex in








Table 4-13: Selected socio-demographic and sexual health data disaggregated by school
status (n=560)

Frequency (%)
School goer School Leaver No School
(n= 204) (n=234) (n=122)
***Mean age of sample (17.1) (17.5) (16.8)
***Ever had penetrative sexual intercourse (yes) 117(57.4) 176 (75.2) 77 (63.1)
*Mean age at first sex (15.3) (15.3) (14.7)
***Have you or your partner ever been pregnant (yes) 2 (01.7) 78 (44.3) 39 (50.6)
Mean age at first parenthood (15.5) (16.5) (16.3)
Residence
Rural Buhimba 139(59.4) 77(63.1)
Hoima School 204 (100.0)
Hoima Town 95 (40.6) 45 (36.9)
***Father's education
Post secondary school 39(19.1) 15(06.4) 2(01.7)
Secondary school (S1-S6) 80(39.2) 56 (23.9) 11(09.0)
Primary 5-7 36(17.6) 68 (29.1) 29(23.8)
Up to primary 4 or never attended 15 (07.4) 47(20.1) 43 (35.2)
Do not know 34 (16.7) 48 (20.5) 37 (30.3)
***Mother's education
Post secondary school 20 (09.8) 5(02.1) 3(02.5)
Secondary school (SI-S6) 71(34.8) 32(13.7) 5 (04.1)
Primary 5-7 62(30.4) 75 (32.1) 28 (23.0)
Up to primary 4 or never attended 24 (11.8) 85 (36.3) 59 (48.3)
Do not know 27(13.2) 37(15.8) 27 (22.1)









Table 4-13-continued


Frequency (%)
School goer School Leaver No School
(n= 204) (n=234) (n=122)
***Ethnicity
Munyoro 142 (69.6) 213 (91.0) 99(81.1)
Muganda 17(08.3) 6 (02.6) 4 (03.3)
Other (Munyankole, Mutooro, Rwandese, 45 (22.1) 15(06.4) 19(15.6)
Muchiga, Alur, Mugungu, other)
***Religion
Catholic 56 (27.4) 88 (37.6) 56 (45.9)
Anglican/Protestant 105 (51.5) 109 (46.6) 46(37.7)
Muslim 30 (14.7) 9 (03.8) 6 (04.9)
Other (Born Again, 7t Day Adventist, Baptist, 13 (06.4) 28 (12.0) 14 (11.5)
Pentecostal, other)
***Average cash per month from any source
0 $US 150(73.5) 61(26.1) 30(24.6)
1 5 $US 13(06.4) 32 (13.7) 16(13.1)
6 <10 $US 12(05.9) 44(18.8) 28 (23.0)
11 20 $US 13 (06.4) 48 (20.5) 29 (23.8)
> 20 $US 16(07.8) 49(20.9) 19(15.5)
*If work for cash is steady or permanent
Housework 0 7(03.0) 10(08.2)
Farmwork 7 (03.4) 6 (02.6) 8 (06.6)
Alcohol production/distribution 1 (00.5) 3 (12.8) 1 (00.8)
Construction, labor 9(04.4) 21 (09.0) 14(11.5)
Market seller or own business 3 (01.5) 18(07.7) 9 (07.4)
Substance Use
**Ever smoke cigarettes (yes) 1 (00.5) 17 (07.3) 12 (09.8)
Ever drink alcohol (yes) 81 (39.7) 100 (42.7) 53 (43.4)
Ever taken illegal drugs (yes) 3 (01.5) 2 (00.9) 2(01.6)
Male circumcision, female labial elongation (yes) 32 (15.7) 9 (3.8) 6 (4.9)
*Age at first sexual intercourse
11-14 years 24 (20.5) 45 (25.6) 29 (37.7)
15-17 years 88 (75.2) 119(67.6) 47 (61.0)
18-19 years 5 (04.3) 12(06.8) 1 (01.3)
***Current relationship status
None 107 (52.4) 113(48.3) 65 (53.3)
One casual or steady partner 95 (46.6) 73 (31.2) 36 (29.5)
Several casual or steady partners 2 (01.0) 2 (00.9) 0
Living with partner 0 32(13.6) 13(10.7)
Formally married (traditional, church, civil) 0 13 (05.6) 7 (05.7)
Separated or widowed 0 1 (00.4) 1 (00.8)
Ever had STI (yes) 6 (02.9) 9(03.8) 5(04.1)
Received money or gifts for first sex (yes) 52(25.5) 74(31.6) 28(22.9)
Gave money or gifts for first sex (yes) 46 (22.5) 57 (24.4) 30 (25.0)
***Used condoms with first sexual partner (yes) 59 (50.4) 56(31.8) 19(24.7)

*P < 0.05, **P < 0.01, ***P < 0.001. For categorical response data, Pearson x2 test or
Fisher's exact test for cell counts < 5; and for continuous response data such as age, one-
way ANOVA (Post hoc test: Bonferroni).









Table 4-12; however, they are significant for school status. Within the category school

status, there are important differences in sexual behaviors. School goers report less

sexual experience (117 reported versus 135 expected, P < 0.001) while school leavers

report more sexual experience (176 reported versus 154 expected, P < 0.001). For age at

first sexual intercourse, fewer school goers report initiating sexual debut at age 11-14 (24

reported versus 30 expected, P < 0.05) while those with no school experience more

frequently report initiating sexual debut at age 11-14 (29 reported versus 20 expected, P

< 0.01). For ages 15-17, more school goers than expected report initiating sexual debut

(88 reported versus 79 expected, P < 0.05), fewer youth of no school experience report

initiating sexual debut at age 15-17 (47 reported versus 53 expected, P < 0.05), while

almost all youth without school experience have initiated sexual debut by age 18-19 (1

reported versus 4 expected, P < 0.05). And finally, the exchange and receipt of gifts was

highly significant for gender/sex, but not significant for school status.


Comparing Socio-demographic Data by Gender/Sex and Sexual Activity

Table 4-14 compares selected socio-demographic data by gender/sex according to

self-reported experience with sexual activity. Surprisingly few variables are statistically

significantly different between the group of 189 youth who reported no sexual activity

ever, and those 370 youth who reported at least one experience with sexual intercourse.

When significance is found, the variable in the left column is usually more significant for

one gender than for the other. For example, differences in school status were more

significant when comparing the level of sexual activity for girls than boys. In other

words, the incidence of sexual activity for boys was not related to their level of

education, although male drop-outs are more likely to report sexual activity and in-school









boys are more likely to report no sexual activity. For girls, the differences are more

pronounced but in the same direction as the boys; that is, in-school girls are less likely to

report sexual activity whereas female drop-outs have the highest level of reported sexual

debut. Other significant relationships in Table 4-14 are access to cash from any source

per month, substance use, and current relationship status. As with education, the variable




Table 4-14: Selected socio-demographic and sexual health data disaggregated by
gender/sex for youth reporting no sexual activity (n=1 89) and for youth reporting sexual
activity (n=370)

Frequency (% or mean)
Males Females Males Females
No sex No sex Yes sex Yes sex
(n=89) (n=100) (n=183) (n=187)
Mean age (16.8) (16.3) (17.6) (17.4)
**Education
None (P4 or lower) 18(20.2) 27 (27.0) 38 (20.8) 39 (20.9)
Drop out (P5 or higher and out of school at 30(33.7) 27(27.0) 86(47.0) 90(48.1)
least 18 months)
Continuously enrolled (in-school) 41 (46.1) 46 (46.0) 59 (32.2) 58 (31.0)
Father's education
Post secondary school 6(06.7) 13 (13.0) 17(09.3) 20(10.6)
Secondary school (S1-S6) 29(32.6) 28 (28.0) 42 (23.0) 47 (25.1)
Primary 5-7 20 (22.5) 15(15.0) 55 (30.1) 43 (23.0)
Up to primary 4 or never attended 18 (20.2) 23 (23.0) 36(19.6) 28(15.0)
Do not know 16(18.0) 21(21.0) 33(18.0) 49(28.1)
Mother's education
Post secondary school 2 (02.2) 8 (08.0) 8 (04.3) 10(05.3)
Secondary school (S1-S6) 20 (22.5) 25 (25.0) 27 (14.7) 36 (19.3)
Primary 5-7 27(30.3) 24(24.0) 55 (30.1) 59(31.6)
Up to primary 4 or never attended 25 (28.1) 29(29.0) 59(32.3) 54(28.8)
Do not know 15(16.9) 14(14.0) 34(18.6) 28(15.0)
Ethnicity
Munyoro 73 (82.0) 77(77.0) 153 (83.6) 150 (80.2)
Muganda 4(04.5) 7(07.0) 3 (01.6) 13(07.0)
Other 12(13.5) 16(16.0) 27(14.8) 24(12.8)
Religion
Catholic 34 (38.2) 36 (36.0) 63 (34.4) 66 (35.3)
Anglican/Protestant 42 (47.2) 38 (38.0) 94(51.4) 86(46.0)
Muslim 7 (07.9) 7(07.0) 13(07.1) 18(09.6)
Other 6 (06.7) 19(19.0) 13 (07.1) 17(09.1)









Table 4-14--continued


Frequency (% or mean)
Males Females Males Females
No sex No sex Yes sex Yes sex
(n=89) (n=100) (n=183) (n=187)
**Average cash per month from any source
0 $US 34(38.2) 57(57.0) 46(25.1) 103 (55.1)
1 5 $US 25 (28.1) 15(15.0) 29 (15.8) 30 (16.0)
6 10 $US 13(14.6) 19(19.0) 35(19.1) 37(19.8)
11 20 $US 10 (11.2) 5(05.0) 38 (20.8) 7(03.7)
> 20 $US 7 (07.9) 4 (04.0) 35 (19.2) 10(05.4)
If work for cash is steady or permanent
Housework 0 5(05.0) 3(01.6) 9(04.8)
Farmwork 4(04.5) 2(02.0) 11(06.0) 4(02.1)
Alcohol production/distribution 1 (01.1) 0 2(01.1) 2(01.1)
Construction, labor 14(15.7) 1 (01.0) 29(15.8) 0
Market seller or own business 5 (05.6) 5 (04.0) 11 (06.0) 9(04.8)
Substance Use
* Ever smoke cigarettes (yes) 3(03.4) 2 (02.0) 24 (13.1) 1 (00.5)
*** Ever drink alcohol (yes) 23 (25.8) 25 (25.0) 103 (56.3) 82 (43.9)
Ever taken illegal drugs (yes) 0 0 6 (03.3) 1 (00.5)
Male circumcision, female labial elongation (yes) 8 (9.0) 4 (4.0) 18(9.8) 17(9.1)
***Current relationship status
None 80 (89.9) 90 (90.0) 65 (35.5) 49 (26.2)
One casual or steady partner 9(10.1) 10(10.0) 95(51.9) 90 (48.2)
Several casual or steady partners 0 0 4 (2.2) 0
Living with partner 0 0 9(4.9) 36 (19.3)
Formally married (traditional, church, civil) 0 0 9 (4.9) 11 (05.8)
Separated or widowed 0 0 i (0.6) 1 (00.5)

*P < 0.05, **P < 0.01, ***P < 0.001. For categorical response data such as yes/no
responses, 3-way crosstabs analysis by Pearson x2 test or Fisher's exact test for cell
counts < 5; for continuous response data such as age, Independent Samples T test
(Levene's test for equality of variances and t-test for equality of means).



"average cash per month from any source" is significant for one gender but not the other.

For girls, sexual behavior is remarkably similar in the two groups according to access to

cash. However, for the boys, access to cash is significantly related to sexual activity;

boys with little access to cash are more likely to have no sexual experience, while those

young men with significant amounts of cash per month report more sexual activity.

Finally, it is not surprising that those youth, both boys and girls, who report no current









relationship also report less sexual activity than those youth who report a current partner

or partners.


Combining Variables to Create the Risk Profile Score

Because it was not possible to follow the original research design (to directly

compare HIV serostatus with school status for the entire youth sample population), I

needed another way to study the hypothesis. As a compromise method for assessing the

relationship between education and HIV serostatus, I developed a more general measure

of risk-related behaviors-the risk profile score. Risk profile is a numeric scoring of a

cluster of HIV-related risk behaviors associated with first sexual intercourse and up to

three recent sexual partners (Table 4-15). This provides a rough method for intra-sample

comparison of cumulative levels of HIV/AIDS risk related to adolescent sexual behavior.


Table 4-15: Criteria for risk profile score

Risk variable Risk value
Ever had penetrative sexual intercourse (yes) +1
Age at first intercourse
11-14 years +1
15-17 years 0
18-19 years -1
Used condoms with first sexual partner (yes) -1
#Of up to three most recent sexual partners)
Nonconsensual intercourse +1
Condom used
All the time -2
Some of the time -1
Never +1
You or your partner drink alcohol before intercourse
All the time +2
Some of the time +1
You or your partner use illegal drugs before intercourse
All the time +2
Some of the time +1









Table 4-15--continued

Risk variable Risk value
Ever had STI symptoms) (yes) +1
Tell your partner about your STI symptoms)
Yes -1
No +1
Total number lifetime sexual partners
1 +1
2 +2
3 +3
> 4 +4
Blood transfusion (yes) +1

#Scored separately for each individual partner up to most recent three partners










For each specific behavior known or suspected to increase risk of HIV

transmission, one point is added to the risk profile score: initiation of penetrative sexual

intercourse, early first sexual intercourse (age 11-14 years), nonconsensual intercourse,

intercourse without a condom, alcohol or drug use by either partner before sexual

intercourse, presence of STI symptoms, number of partners, and blood transfusion.

Scarification and tattooing have not been demonstrated to pose risk of HIV transmission

and we did not ask about these procedures in the youth survey. For each specific

behavior known or suspected to decrease risk of HIV transmission, one or more points

were deducted: delayed first sexual intercourse, condom use, and informing partner about

STI symptoms. The points are then summed to calculate the risk profile score. The raw

risk scores for all sexually experienced youth (n=370) are presented in Table 4-16.

Because the risk score is a composite of categorical variables, it is also a categorical









Table 4-16: Raw risk score for all sexually experienced youth (n=370)


variable and therefore not suitable for linear regression or ANOVA analyses. Loglinear

regression analysis would be more appropriate for these data (Agresti 1996). To make

analysis by Pearson's chi square more meaningful, we divided the range of risk scores

from (-4) to (15) into four categories of risk behavior: very protective, moderately

protective, risky, and dangerous. Frequency data for the risk profile for the sub-sample

"all sexually experienced youth" (n=370) are presented in Table 4-17. About half of the

sample scored very protective or moderately protective behavior and about half scored

risky or dangerous behavior. Although not intended to be predictive of actual HIV

serostatus, all adolescents in the study with positive HIV test results were school leavers

who scored in the category "risky behavior" (see Appendix K).









Table 4-17: Risk profile score for all sexually experienced youth (n=370)

Risk profile category Frequency (%)
Very protective behavior ((-4)-0) 119 (32.2)
Moderately protective behavior (1-3) 63 (17.0)
Risky behavior (4-9) 171 (46.2)
Dangerous behavior (10-15) 17(04.6)


The consistency of our sample is demonstrated again in Table 4-18: risk profile

score is not significant when disaggregated by gender/sex. Table 4-18 also suggests that

the variables chosen in Table 4-15 are, as a composite, a valid representation of which of

lifetime sexual partners." As expected, boys report more sexual partners in their lifetime

and girls report fewer sexual partners. Almost twice as many girls as boys report only




Table 4-18: Risk profile score by gender/sex (n=370)

Frequency (%)
Risk profile results for all sexually experienced youth Males Females
(n=183) (n=187)
Very protective behavior 55 (30.1) 64 (34.2)
Moderately protective behavior 32 (17.5) 31 (16.6)
Risky behavior 87 (47.5) 84 (44.9)
Dangerous behavior 9 (04.9) 8 (04.3)

*P < 0.05, **P < 0.01, ***P < 0.001. Pearson x2 test.



one lifetime partner (64.7% of females versus 36.8% of males; P < 0.001), many more

boys than girls report four or more lifetime partners (25.3% of males versus 5.9% of

females; P < 0.001). However, one surprising association to note is more boys report

initiation of sexual intercourse at a younger age than girls (32.3% of males in the 11-14

age group versus 20.9% of females; P < 0.01). Further, although the absolute values are

behaviors to analyze when describing risk of adolescent HIV transmission. For example,









in Table 4-19, the only variable that is highly significant for gender/sex is "total number

small, males are less likely to report a STI symptom to their partners) than females (5

males versus 8 females). In addition, all females who reported STI symptoms to their

partner were married (traditional, civil, or church ceremonies), whereas only one married

man reported symptoms to his partner (data not shown). As with many variables reported

above, the risk profile score is significant when disaggregated by school status (Table 4-

20). As with Table 4-19 above, Table 4-21 disaggregates the risk profile data for school

status. In Table 4-21, few risk profile variables are statistically significant for school

status. While the variable "total number of lifetime sexual partners" was not significant

for school status overall, a cell by cell analysis of the observed versus the expected counts

reveals some significant differences. School goers were most likely to report a single

sexual partner (68 reported versus 58 expected, P < 0.05), while those without school

experience reported fewer single sexual partners (32 reported versus 39 expected, P <

0.05) and were most likely to report four or more sexual partners (18 reported versus 12

expected, P < 0.01). The variables that account for the significant difference in risk

behaviors by school status in Table 4-20 mainly describe differences in condom usage

and age at first sexual intercourse. In contrast to Tables 4-18 and 4-19 where significant

differences in risk profile between girls and boys are due to a single variable, in Table 4-

20 and 4-21, differences in risk profile according to school status are more consistent

significant in several different variables, thus suggesting that there are more differences

in sexual behaviors (risk profile) according to school status than according to gender.






78


Table 4-19: Selected risk profile variables by gender/sex (n=370)

Frequency (%)
Males Females
(n=183) (n=187)
Age at first sexual intercourse
11-14 years 59 (32.3) 39 (20.9)
15-17 years 115(62.8) 139 (74.3)
18-19 years 9 (04.9) 9 (04.8)
Used condoms with first sexual partner (yes) 62 (33.9) 72 (38.5)
Ever had STI symptoms) (yes) 9 (04.9) 11 (05.9)
Tell your partner about your STI symptoms) (yes) 5 (55.6) 8 (72.7)
***Total number lifetime sexual partners
1 68 (36.8) 121 (64.7)
2 41 (22.5) 41(21.9)
3 28(15.4) 14(07.5)
2 4 46(25.3) 11(05.9)
**Of current partner:
Was first sexual intercourse consensual (yes) 173 (97.2) 162 (89.5)
Of current partner:
Condom used:
All the time 37 (20.2) 40 (21.4)
Some of the time 45 (24.6) 48 (25.7)
Never 98 (53.6) 95 (50.8)
Missing data 3 (01.6) 4(02.1)
*Of current partner:
You take alcohol before sexual intercourse
All the time 1 (00.5) 1 (00.5)
Some of the time 22 (12.0) 6 (03.2)
Never 157(85.9) 175 (93.6)
Missing data 3 (01.6) 5 (02.7)
Of current partner:
You use illegal drugs before sexual intercourse
Some of the time 1 (00.5) 2(01.1)
Never 180 (98.4) 180 (96.3)
Missing data 2 (01.1) 5(02.6)
Blood transfusion (yes) 5 (02.7) 10(05.3)

*P < 0.05, **P < 0.01, ***P < 0.001. Pearson x2 test or Fisher's exact test for cell counts












Table 4-20: Risk profile score by school status (n=370)


Frequency (%)
***Risk profile results for all sexually experienced youth School goer School Leaver No School
(n=117) (n=176) (n=77)
Very protective behavior 56 (47.9) 47 (26.7) 16 (20.8)
Moderately protective behavior 24(20.5) 28 (15.9) 11 (14.3)
Risky behavior 36 (30.8) 93 (52.8) 42 (54.5)
Dangerous behavior 1 (00.8) 8 (04.6) 8(10.4)


*P < 0.05, **P < 0.01, ***P < 0.001. Pearson x2 test.







Table 4-21: Selected risk profile variables by school status (n=370)


Frequency (%)
School goer School Leaver No School
(n=204) (n=234) (n=122)
*Age at first sexual intercourse
11-14 years 24 (20.5) 45 (25.6) 29 (37.7)
15-17 years 88 (75.2) 119(67.6) 47 (61.0)
18-19 years 5(04.3) 12(06.8) 1 (01.3)
***Used condoms with first sexual partner (yes) 58 (49.6) 56 (31.8) 19 (24.7)
Ever had STI symptoms) (yes) 6(05.1) 9(05.1) 5(06.5)
*Tell your partner about your STI symptoms) (yes) 2 (33.4) 9 (100.0) 2 (40.0)
Total number lifetime sexual partners
1 68(58.1) 88(50.0) 32(41.6)
2 24 (20.5) 43 (24.4) 16(20.8)
3 11 (09.4) 20(11.4) 11 (14.2)
S4 14(12.0) 25(14.2) 18(23.4)
Of current partner:
Was first sexual intercourse consensual (yes) 100 (85.5) 160(90.9) 75 (97.4)
***Of current partner:
Condom used:
All the time 34(29.1) 35(19.9) 9(11.7)
Some of the time 38 (32.5) 36 (20.5) 20 (26.0)
Never 40 (34.2) 105 (59.6) 48 (62.3)
Missing data 5 -
Of current partner:
You take alcohol before intercourse
All the time 1 (00.9) 1 (00.6) 0
Some of the time 4(03.4) 15(08.4) 11 (14.3)
Never 106 (90.6) 160 (91.0) 66 (85.7)
Missing data 6 -









Table 4-21--continued


Frequency (%)
School goer School Leaver No School
(n=204) (n=234) (n=122)
Of current partner:
You use illegal drugs before intercourse
Some of the time 2(01.7) 1(00.6) 0
Never 109 (93.2) 175 (99.4) 77 (100.0)
Missing data 6
Blood transfusion (yes) 6 (05.1) 6(03.4) 3 (03.9)

*P < 0.05, **P < 0.01, ***P < 0.001. Pearson x2 test or Fisher's exact test for cell counts
<5.




Three-way Associations: Risk Profile Score, School Status, and Gender/Sex

To this point, the analysis of the relationship between risk profile and school

status or gender/sex has demonstrated that the main difference in sexual behaviors

between girls and boys is in the number of lifetime sexual partners while the main

difference in sexual behaviors across categories of school status is frequency of use of

condoms. If Table 4-17 is further disaggregated by both gender/sex and school status, the

analysis is even more precise. The data in Tables 4-22 and 4-23 clearly indicate that the

variability that contributed to the significant p score in Table 4-20 came only from

differences in the sexual behaviors of girls, not boys. Remarkably, the level of risk

behaviors reported by all sexually active boys in Table 4-22 is consistently similar,

irrespective of the boy's relationship to school. In sharp contrast, Table 4-23 indicates

that the reported sexual behaviors of school girls accounts for two-thirds of all young

women in the protective behavior category, while the riskiest and most dangerous levels

of sexual behavior are reported by female school leavers and girls without any access to

education at all.






81







Table 4-22: Risk profile score by school status for boys (n=183)


*P < 0.05, **P < 0.01, ***P < 0.001. Pearson x2 test.






Table 4-23: Risk profile score by school status for girls (n=187)


*P < 0.05, **P < 0.01, ***P < 0.001. Pearson x2 test.


Frequency (%)
Risk profile results for all sexually experienced boys School goer School Leaver No School
(n=59) (n=86) (n=38)
Very protective behavior 20 (33.9) 26 (30.2) 9 (23.7)
Moderately protective behavior 13 (22.0) 13 (15.1) 6 (15.8)
Risky behavior 25 (42.4) 42 (48.9) 20 (52.6)
Dangerous behavior 1 (01.7) 5 (05.8) 3 (07.9)


Frequency (%)_
***Risk profile results for all sexually experienced girls School goer School Leaver No School
(n=58) (n=90) (n=39)
Very protective behavior 36 (62.0) 21 (23.3) 7 (17.9)
Moderately protective behavior 11 (19.0) 15 (16.7) 5 (12.8)
Risky behavior 11(19.0) 51 (56.7) 22 (56.4)
Dangerous behavior 0 3 (03.3) 5 (12.9)














CHAPTER 5
VOLUNTARY COUNSELING AND TESTING


Enrollment and Counseling Protocol for Hoima Adolescent VCT

When a survey was completed, the interviewer briefly explained the availability

of and procedures for the free voluntary confidential HIV/AIDS counseling and testing

(VCT) component at the local hospital. The interviewer stressed that no one was

required to take the test, the results were confidential (although not anonymous), results

were released only to the youth, and testing and counseling was also available at the

AIDS Information Centre (AIC) in Kampala. Each survey participant was given an

information sheet explaining VCT with dates and times for testing and counseling, as

well as a mandatory release form for the parent or guardian to sign for youth under the

age of 18. Rural survey participants were reimbursed for the cost of traveling 24 miles

round-trip to the hospital (about $2.50 US) and offered lunch. School survey participants

were offered a smaller transport reimbursement (about $0.50 US) and lunch. Town

survey participants were offered only lunch. Due to popular demand, free anonymous

counseling and testing was made available to anyone in Hoima District who requested the

service during June-September 1997. A local nurse who informally counseled AIDS

patients at the hospital was given special training in HIV/AIDS counseling procedures

from AIC in Kampala. A laboratory technician employed by Hoima Hospital received

technical training in rapid testing protocols and laboratory procedures from AIC.









Procedures for registration, informed consent, data collection, pre- and post-test

counseling, general reproductive health counseling, and medical referrals followed AIC

protocol for rapid testing (Kassler 1997). Upon arrival, clients were greeted at the same

registration desk as other hospital patients and directed to our private reception,

counseling, and phlebotomy areas. This reduced the possibility that clients would be

recognized in the general waiting room as attending the testing clinic. Two group

counseling sessions of ten clients each were scheduled twice a week for four months.

Pre-test counseling was given in a group session with special emphasis on personal risk

assessment and clarifying the meaning of reactive and nonreactive test results. In the pre-

test session, clients were encouraged to formulate personal strategies for both reactive

and nonreactive test results. Next, a risk assessment form based on a similar document

used at AIC in Kampala (Appendix J) and informed consent (Appendices F and G) were

collected from each individual client in private. After the blood draw, the group gathered

together again for prevention counseling and general reproductive health information

while they waited for the test results. Results were given privately to individuals,

condoms were distributed, and special counseling and medical referrals were offered to

clients with reactive test results or other special health issues. The average time from

arrival to departure was three hours. Of 377 persons tested, less than 3% failed to remain

in the hospital for their test results and post-test counseling.


Laboratory Protocol for Hoima VCT

Five-ten ml venous blood was collected by venipuncture using Vacutainer serum

separation tubes (Beckon Dickson, Franklin Lakes, NJ, USA). Specimens were

centrifuged on-site in groups of five. The waiting period from blood draw to centrifuged









separation was never more than 30 minutes. Our testing algorithm followed 1996

recommendations and best practices established by the AIC, Kampala, Uganda (Mary-

Grace Alwano-Edyegu, Director AIC, pers. comm.). This algorithm and the numerical

laboratory results are presented in Appendix K. This algorithm follows WHO/UNAIDS

recommendations for HIV Testing Strategy III designed for the diagnosis of

asymptomatic persons in areas with less than 10% prevalence. Therefore the algorithm

used in this research follows the most stringent of all WHO/UNAIDS recommended

algorithms for HIV-1/2 testing (1997) and is designed to minimize false positives and

false negatives and to maximize the accuracy of diagnoses reported to clients. The

current AIC algorithm has been modified slightly (Downing 1998; Kassler 1998).

All specimens were first screened with Capillus HIV-1/2 (Cambridge Diagnostics,

Galway, Ireland), a direct agglutination antibody assay performed on a capillary slide

(Beelaert 1994; Kassler et al. 1996; WHO/UNAIDS 1998). Preliminary field testing of

the Capillus electronic reader was unsatisfactory. Therefore, test results were determined

by comparing the density of the assay reaction against the manufacturer's printed visual

scale. Specimens unambiguously negative on Capillus were reported immediately to the

counselor as nonreactive. Specimens unambiguously positive on Capillus were

confirmed immediately on Serocard, an ELISA test (Trinity Biotech USA, Jamestown,

NY). If unambiguously positive on Capillus and positive on Serocard, results were

reported to the counselor as reactive. If unambiguously positive on Capillus but negative

on Serocard, a third test was immediately performed using Multispot HIV-1/2 (Sanofi

Pasteur, Paris, France) as a tie-breaker. If indeterminant or slightly positive on Capillus,

specimens were immediately retested on Serocard. If the slightly positive or









indeterminate Capillus specimen was negative on Serocard, it was reported to the

counselor as nonreactive with a recommendation for a retest in three months. If the

slightly positive or indeterminate Capillus specimen was positive on Serocard, it was

immediately retested on Multispot as the tie-breaker.


Attendance at VCT

Figure 5-1 below shows the trend over time of participation in HIV-1/2 VCT at


VCT Clients
Scommunt
ruryouth
I schyouth
Itowyouth


'. CO D M 0 t'- 1 CO W W- CO 'M0 a
Co Co Co Co r. r- o Co C Co Co Co Co CoOQ
00 0 0 00 0 0 0Q Q
Date


Figure 5-1: Time Trend for HIV-1/2 VCT for both youth in survey and community adults


40,





30.

in

20
S20.
:-
0



10.




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ELBDFMEIP_DOLLT3 INGEST_TIME 2013-10-10T01:12:52Z PACKAGE AA00017691_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES