Sexual Behavior, HIV, and Fertility Trends: A Comparative Analysis of Six Countries

Transcription

1 Sexal Behavior, HIV, and Fertility Trends: A Comparative Analysis of Six Contries Phase I of the ABC Stdy Prepared by: Rth Bessinger Priscilla Akwara Daniel Halperin This report was prepared by USAID and MEASURE Evalation, a project managed by the University of North Carolina at Chapel Hill in collaboration with ORC Macro, Tlane University, and John Snow, Inc. nder the Cooperative Agreement HRN-A The opinions expressed are those of the athors, and do not necessarily reflect the views of USAID.

2

3 Table of Contents Acknowledgements... iii Acronyms... v Exective Smmary... vii Chapter : Introdction.... Overview.... The Spread of HIV.... Stdy Objectives... 6 Chapter : Data and Methods Nationally Representative Data Sorces Spplemental Data Sorces... 8 Chapter : Backgrond Characteristics of Respondents Introdction Backgrond Characteristics... 9 Chapter 4: Levels and Trends in Sexal Behavior Introdction Sexal Behavior among Yoth Sexal Behavior Otside of Marriage Mltiple Sexal Partners Commercial Sex Condom Use... Chapter 5: Fertility Related Behavior Introdction Fertility Levels and Trends Family Planning Other Proximate Determinants of Fertility... 9 Chapter 6: Smmary Contry Smmaries Comparative Analysis of Sexal Behavior Comparative Analysis of Fertility... 6 References... 9 Appendix A: Tables... 4 Appendix B: Comparability of Uganda GPA and DHS Srveys... 7 i

4

5 Acknowledgements We wold like to acknowledge the many individals who contribted in varios ways to this report. The report was prepared with significant inpt from ABC Stdy Director Dog Kirby of ETR Associates; Connie Carrino, Shanti Conly, and David Stanton of USAID s Office of HIV/AIDS; Nomi Fchs, Jim Shelton, and Jeff Spieler of the Office of Poplation and Reprodctive Health; Linda Morse, Depty Assistant Administrator, and Anne Peterson, Assistant Administrator of the Brea for Global Health at USAID in Washington, D.C. Karen Stanecki of UNAIDS and Lara Heaton of the U.S. Censs Brea provided most of the data on HIV prevalence trends. This report also draws pon data analyses condcted by Silvia Alayon and Shelah Bloom of MEASURE Evalation and George Bicego formerly of MEASURE DHS+. Members of the ABC Stdy Technical Review Panel (Mohammed Abdllah, Saifddin Ahmed, Willard Cates, Tom Merrick, Carolyn Ryan, and Amy Tsi) reviewed an earlier draft and provided many sefl comments dring two Febrary 00 meetings of the stdy panel. This report also benefited from nmeros discssions and written exchanges with colleages who have thoght careflly abot these data, inclding Edward Green and Vinand Nantlya of the Harvard Center for Poplation and Development Stdies, Ties Boerma of WHO, Rand Stonebrner, Jimmy Whitworth of the London School of Hygiene & Tropical Medicine, Reginald Finger, and varios staff from USAID in Washington, D.C., and USAID missions, inclding Elise Ayers, Steve Hodgins, Patrick Osewe, and Michael Strong. Lastly, we wold like to express or gratitde for the contribtions of the many dedicated researchers in Cameroon, Kenya, Thailand, Uganda, and Zambia who have worked to help nderstand the spread of the epidemic and who have provided mch of the data that was sed or cited in this report. iii

6

7 Acronyms AIDS Acqired Immne Deficiency Syndrome ANC Antenatal Care ASFR Specific Fertility Rate BUCEN Brea of Censs CBoH Central Board of Health CPR Contraceptive Prevalence Rate DHS Demographic and Health Srvey GPA Global Programme on AIDS HIV Hman Immnodeficiency Virs MoH Ministry of Health NASCOP National AIDS and STD Control Programme NGO Nongovernmental Organization PSI Poplation Services International SBS Sexal Behavior Srvey STD Sexally Transmitted Disease TFR Fertility Rate UN United Nations UNAIDS United Nations Joint Programme on HIV/AIDS USAID United States ncy for International Development WHO World Health Organization WTFR Wanted Fertility Rate v

8

9 Exective Smmary The USAID-spported ABC Stdy examines how prevention behaviors may have affected HIV prevalence as well as fertility patterns in three contries where HIV prevalence declined dring the 990s (Uganda, Zambia, and Thailand) and in three contries where it appears not to have declined (Cameroon, Kenya, and Zimbabwe). This Phase I Report compares levels and trends of behavioral indicators sch as abstinence, age of sexal debt among yoth, faithflness in sexal relationships/redction in mltiple sexal partners, and condom se (especially with non-reglar partners). These types of behaviors are often referred to collectively as the ABCs (for Abstinence, Being faithfl, and Condom se) of HIV prevention. The ABC Stdy considers sch behaviors within the context of HIV prevalence trends in the six contries dring the 990s. Among the declining prevalence contries, the most dramatic redction occrred in Uganda, where prevalence among pregnant women in the capital city of Kampala, for example, fell from abot 0% in 99 to approximately 0% by the end of the decade. And while prevalence levels had generally been lower in other regions of the contry, similarly large declines occrred (althogh in 00, national prevalence may have increased slightly for the first time in a decade). Thailand also experienced a significant decrease in HIV prevalence, especially in high-risk grops, among whom the epidemic was concentrated (as opposed to sb-saharan Africa's typically more generalized epidemics). In addition, there are sggestions of a sbstantial decrease in prevalence among rban yoth in Zambia dring the 990s. Cameroon, Kenya, and Zimbabwe all experienced large increases of HIV prevalence in the 990s, althogh prevalence varied considerably among them, ranging from almost 0% in the western African contry of Cameroon to abot 5% in Kenya (where the epidemic may have recently begn to stabilize), to more than 0% in the sothern African nation of Zimbabwe. The report relies primarily on nationally representative srvey data, mainly from Demographic and Health Srveys (DHS). Phase II of the ABC Stdy will explore the varios contextal, sociocltral, and programmatic factors that appear to be associated with the behavioral trends described in this report. The final ABC Stdy report is de to be issed by late 004. The Phase I Report incldes the following findings and conclsions: In the two sb-saharan African contries that experienced declines in HIV prevalence dring the 990s (Uganda and Zambia), increases in all of the ABC behaviors (abstinence/delay of sexal debt, being faithfl/partner redction, and condom se with non-reglar partners) occrred. In Thailand, there were sbstantial redctions in commercial and other non-marital partnerships, as well as large increases in condom se for commercial sex. The same mltifaceted behavior change is not evident in the three contries that did not experience declines in HIV prevalence, where condom se increased, bt there was little reported change in abstinence or partner redction behaviors. In Uganda, the contry with the earliest and greatest redction in HIV prevalence, declines in nonreglar and mltiple sexal partnerships were greatest between the late 980s when the rate of new infections (HIV incidence) likely began to drop and the mid-990s. Levels of premarital sex also declined dring this period, along with an increase in the average age at sexal debt throghot the decade. Increases in condom se with non-reglar partners occrred throghot the 990s, althogh levels of condom se were not sbstantial ntil the middle of the decade. While the trends in sexal behavior within these contries are consistent with trends in HIV prevalence, the differences in patterns of sexal behavior across the five African contries are not sfficient to flly explain differences in HIV prevalence. Other factors sch as male circmcision (which is a norm in Cameroon, bt nearly absent in Zimbabwe) mst also be considered. vii

10 While fertility declined in all five sb-saharan African contries between the late 980s and the late 990s, the more modest redctions in Uganda and Zambia sggest that the increases in age of sexal debt and declines in non-reglar partnerships that occrred there may have had little impact on fertility. Sexal Behaviors: Trends and Levels The report incldes the following observations on specific behavioral trends and levels: of sexal debt: In Uganda, the median age of sexal debt among both yong men and women increased by abot one year over the corse of the decade. Yong women and men in Zambia, especially in rban areas, also postponed sexal debt. The percentage of 5- to 9-yearold rban women reporting ever having sex dropped from 56% to 40% between the early and late 990s. For rban males, an even sharper decline from 67% to 4% took place between the middle and end of the decade. By contrast, the trend data for yong men and women in Zimbabwe and women in Cameroon or Kenya (which lack trend data for men) show little evidence of postponed sexal debt. Premarital sex: In Uganda, a large decline in sexal activity among nmarried 5- to 4-year-old females (from 5% to %) and males occrred dring the first half of the decade (althogh there was an increase among yong women to 7% dring the second half of the decade). Data from Zambia indicate a decline (from 9% to 6%) throghot the decade in the percentage of nmarried yong women reporting sex in the previos year. The percentage of nmarried yong men in Zambia reporting sex in the previos year dropped sharply from 6% to 4% in the latter half of the decade. In Kenya there was also a decline in sexal activity among nmarried men aged 0 4, bt the percentage reporting activity (56%) remained high. At the end of the decade, Cameroon had the highest rates of sexal activity reported by both yong nmarried men (58%) and women (5%). Non-reglar partnerships: Large declines in non-reglar (non-marital and non-cohabiting) partnerships among men and women occrred in the early 990s in Uganda and in the late 990s in Zambia. Sb-national Global Programme on AIDS (GPA) srvey data from Uganda in the mid- 990s indicate that many fewer men and women reported casal partners than in the late 980s (5% of men, down from 5%, and 6% of women, down from 6%). The declines in Zambia occrred among adolescents and adlts in both rban and rral settings. By the late 990s, only 9% of Zambian men reported a non-reglar partner, compared with 5% in the mid-990s. Data from Thailand in the early 990s also indicated there were large declines in the percentage of men reporting non-marital and commercial sex partners (from % to 0% for the latter between 990 and 99). Kenya saw only a small decline in non-reglar partnerships among men and virtally no decline among women. Trend data on non-reglar partnerships are lacking for Cameroon and Zimbabwe. Among the six contries, however, these two contries had the highest reported rates of sch partnerships at the end of the decade abot 60% of men in Cameroon and 40% of men in Zimbabwe reported a non-reglar partner in the previos year, compared to 5% in Kenya and less than 0% in Uganda and Zambia. Singles with mltiple sexal partners: In Uganda, a large decline from 54% to % (GPA data) occrred among single men reporting two or more partners in the previos year between viii

11 the late 980s and mid-990s. Between the mid- and late 990s in Zambia, the percentages of singles reporting mltiple sexal partners in the past year declined from 54% to 0% for men and also fell sbstantially among women (from 7% to 8%). Kenya may also have had a small decline in mltiple partnerships among singles, bt mltiple partnerships remained high (5%) among Kenyan men late in the 990s. At the end of the decade, levels of mltiple partnerships among singles were highest in Cameroon. This was tre for both single men (68%, compared with roghly 0% for single men in Uganda, Zambia, and Zimbabwe) and single women (0%, compared with 8% for Kenya, % for Zimbabwe, 8% for Zambia, and 6% for Uganda). Higher-level mltiple partnering among men: Regarding the important core transmitter grop of men reporting three or more non-reglar partners in the previos year, there was a very large decline in Uganda (in the GPA srveys, from 5% in 989 to % in 995). This figre remained low (%) at the end of the decade, in striking comparison with Cameroon (6%). Zambia also experienced a large decline from the mid-990s (from % to %), while there was little change in Kenya. No trend data are available for Cameroon and Zimbabwe, althogh the former had the highest percentage of men reporting three or more partners at the end of the decade. Zimbabwe, where factors sch as lack of circmcision and dry sex practices may contribte to HIV transmission, appears to be an anomalos case, with low levels of men (%) reporting three or more partners. It shold be noted, however, that some srveys have fond higher levels of mltiple partnership in Zimbabwe. Condom se with non-reglar partners: Uganda, Zambia, and Zimbabwe saw sbstantial increases in reported condom se with non-reglar partners in the late 990s. At the end of the decade, condom se at last sex with a non-reglar partner was highest for women and men in Zimbabwe (4% and 70%, respectively), followed by Uganda (8% and 59%). In Thailand, condom se with commercial sex workers also increased significantly among men paying for sex (from 6% to 7% between 990 and 99). Fertility-Related Findings The report also notes that fertility declined in all five sb-saharan African contries between the late 980s and late 990s, with steep declines occrring in Kenya, Zimbabwe, and Cameroon. The more modest redctions in Uganda and Zambia indicate that the increases in age of sexal debt and declines in non-reglar partnerships that occrred may have had little impact on fertility. While condom se with non-reglar partners increased in all contries, it remains very low (less than 4% in all contries) among reglar (marital or cohabiting) partners. The impact of condom se on fertility was ths likely to be minimal. At the end of the 990s, se of modern contraceptives was highest in Zimbabwe and Kenya (58% and %, respectively), moderate in Zambia and Uganda (% and 8%), and low in Cameroon (7%). The greatest increase in contraceptive se dring the 990s occrred in Uganda (p from % in 989), thogh it remained low among 5- to 4- year-old women. ix

12

13 Chapter : Introdction. Overview This report has been prepared as part of the ABC Stdy, a stdy jointly spported by the Offices of HIV/AIDS and Poplation/Reprodctive Health at USAID in Washington, D.C. The main objective of the stdy is to determine how and to what extent prevention behaviors may have affected HIV prevalence and fertility in three contries where HIV prevalence declined dring the 990s (Uganda, Zambia, and Thailand) and in three contries where there is little evidence of mch of a decline (Cameroon, Kenya, and Zimbabwe). The stdy is a joint effort between Harvard School of Pblic Health, U.S. Brea of the Censs (BUCEN), MEASURE Evalation, ETR Associates, and PSI/AIDSMark. The criteria for selection of the stdy contries was based pon the need to examine a manageable nmber of contries (maximm six), half of which actally had some evidence of a decrease in HIV prevalence (as opposed to stabilization of prevalence, which wold have entailed considering a very large nmber of potential candidates), in comparison with three other contries that had not experienced sch prevalence decline. The former were easily identified, as only a small handfl of developing contries have recorded an actal decrease in prevalence. The most widely acknowledged cases are Uganda and Thailand, and while in Zambia there has been only a relatively very small decline in national prevalence, a sbstantial decline appears to have occrred among rban yoth dring the 990s. The three other African contries were selected de to a combination of factors, mainly in order to inclde representativeness from each larger region of sb- Saharan Africa (one relatively high-prevalence contry from western, eastern, and sothern Africa) as well as contries for which reasonably adeqate seroprevalence and DHS behavioral data were available. Althogh ABC has typically been more of a programmatic emphasis in some high prevalence, generalized African epidemics, it was decided to also inclde the sotheast Asian contry of Thailand, in part for variety (regarding the different epidemiological and contextal factors involved in a more concentrated, low-prevalence epidemic). Figre. Median HIV prevalence in Uganda based on antenatal srveillance sites* % * Data have been interpolated for one-year gaps in site data However, most of the ABC Stdy's data analysis focses on behavioral and other comparisons across the five African contries. Under Phase I of the ABC Stdy, MEASURE Evalation was tasked with condcting a mlticontry analysis of HIV-related sexal behaviors and associated fertility patterns sing DHS and other national-level srvey data. BUCEN provided data on trends in HIV prevalence in these same contries. It is ths the objective of this report to present levels and trends in sexal behavior and fertility-related behavior in the context of the HIV epidemic in these six contries. The report draws on secondary analyses of nationally representative srvey data for five of the contries, while for the sixth, Thailand, available indicators are drawn from the pblished literatre. Where national-level data are lacking, the reslts are spplemented with findings from pblished research stdies. The strength of this analysis is that it draws pon nationally representative srvey data that were collected sing a standardized methodology and sing similar sampling designs. Since this is strictly a descriptive stdy, however, a casal association between trends or levels of sexal behavior and HIV prevalence within or across contries cannot be directly inferred. This stdy examines a nmber of indicators over time and across several contries, and most indicators are disaggregated by the backgrond characteristics of the respondents. As a reslt, the potential nmber of two-way comparisons is great and wold have made significance testing cmbersome. In addition, reslts

14 Figre. HIV Seroprevalence among 5- to 9-year-old antenatal women in Uganda: Figre. HIV Seroprevalence among 0- to 4-year-old antenatal women in Uganda: of significance testing wold have to be interpreted in light of the large variation in sample size for the different indicators, ths adding to the complexity of the report. For this reason, statistical testing is not inclded. When stdies referenced in this report have inclded statistical testing (as in the case of analysis of median age at sexal debt), statistical significance has been noted in the text. Phase II of the stdy, condcted by investigators and field interviewers from varios other research instittions, will explore the specific contextal, sociocltral, and programmatic factors that appear to be associated with the trends in sexal behavior described in this report.. The Spread of HIV While a comprehensive overview of HIV-prevalence trends for each of the stdy contries is beyond the scope of this report, a brief overview is presented here in order to pt the reslts into context. Six-Contry Smmary Uganda, Zambia, and Thailand saw some declines in HIV prevalence and likely incidence (the rate of new infections) dring the previos decade. Of the two sb-saharan African contries, both with generalized epidemics, the earliest and greatest declines occrred in Uganda. More recent and modest declines appear to have occrred in Zambia, with declines there most evident among yoth in rban areas. In Thailand, where the epidemic has been concentrated in high-risk poplations, srveillance data point to sbstantial declines in HIV

15 Figre.4 HIV Seroprevalence for antenatal women by age in Chelstone, Zambia: % Figre.5 HIV Seroprevalence for antenatal women by age in Macha, Zambia: % Figre.6 HIV Seroprevalence for women by age in Lsaka, Zambia: % prevalence (and in all likelihood incidence) in military recrits, commercial sex workers and their clients, and STD clinic attendees, althogh rates among intravenos drg sers have remained high. Of the remaining three contries in this report (Cameroon, Kenya, and Zimbabwe), all experienced increases in HIV prevalence dring the 990s, thogh the magnitde of the epidemic varied by contry. The most recent srveillance data sggest, however, that the epidemic may be stabilizing in Kenya and perhaps Zimbabwe, whereas Cameroon has experienced a large increase in prevalence in recent years. Uganda Uganda was one of the first contries in sb- Saharan Africa to experience the epidemic spread of HIV. The first AIDS cases were identified in Uganda in 98 in the Rakai district on the shores of Lake Victoria (Serwadda, 985). The nmber of HIV infections increased rapidly throghot the contry, and by 988 Uganda had one of the highest rates of HIV infection in Africa. By 990, HIV prevalence in major rban areas was as high as % among pregnant women attending antenatal services at sentinel sites (hereafter referred to as antenatal women). After a decade of increasing nmbers of new cases, the spread of the epidemic began to wane in the early 990s. Since 99, there is evidence of a consistent decline in HIV prevalence in antenatal women (Figre.). As seen in Figres. and., these declines occrred most dramatically among antenatal women aged 5 9 and 0 4 years indicating that the nmber of new infections was also likely on the decline (MoH, 00). Data from a poplation-based cohort stdy also fond that declines in HIV prevalence occrred among yong women aged 9 and 0 4 years and among men aged 0 4 and 5 9 dring the 990s (Kamali 000). More importantly, a more recent analysis that incldes data throgh 999 shows a significant redction in HIV incidence in this stdy poplation with a decline from 8.0 per 000 person-years in 990 to 5. per 000 in 999 (Mblaiteye, 00). There is some indication, however, that HIV incidence was mch higher in the 980s and that declines in HIV prevalence in the 990s largely reflect declines in HIV incidence that occrred dring the late 980s (Low-Beer, 00). Evidence from sentinel srveillance of yonger antenatal

16 women and the poplation-based cohort indicates that incidence is likely to have contined to decline, thogh at a mch slower rate, throgh perhaps 997. The most recent prevalence data for yong antenatal women sggests, however, that the rate of new infections may perhaps be increasing somewhat for the first time in a decade. Among all antenatal women, national prevalence increased from 6.% in 000 to 6.5% in 00 (MoH 00). Zambia In Zambia, AIDS was first detected dring the mid- 980s. Althogh initially concentrated in rban areas, it qickly spread to semi-rban and rral areas (Fylkesnes, 00). In 998, prevalence among 5- to 49-year-olds in Zambia was estimated at 0% (MoH and CBoH, 999). An examination of data from antenatal women attending HIV sentinel srveillance sites between 990 and 998 shows a relative stability of HIV prevalence among women living otside major rban areas, while there was evidently a decline in HIV prevalence among rban residents between 994 and 998. Among 5- to 9-year-olds, the overall pictre between 994 and 998 appears to be that of a general decline in HIV prevalence (see Figres.4.6). A poplation-based srvey also showed a decline in prevalence between 996 and 999 for men and women in most age grops bt with a significant decline only among yong women (aged 5 9 in rban areas, 5 4 in rral areas) with the sharpest declines occrring among women in rban areas (Fylkesnes, 00). A nationally representative serosrvey in fond HIV prevalence was 5.6% among adlts of reprodctive age (CSO et al., 00). Figre.7 HIV Seroprevalence for antenatal women by region, Thailand: % Figre.8 HIV Seroprevalence for military conscripts, Thailand: % Figre.9 HIV Seroprevalence for antenatal women 5 4 in Yaonde, Cameroon: 99 and 998 % Thailand In Thailand the epidemic began relatively slowly in the mid-980s, and AIDS cases were detected primarily among men who had sex with men. This is different from the epidemic in sb-saharan Africa, where the disease is primarily spread throgh heterosexal sex. By the late 980s, HIV infection rates among injected-drg sers began to rise rapidly, and high levels of HIV infection were fond among brothel-based female sex workers and among men seeking services at STD clinics. In addition, in many areas of the contry, HIV infection was increasingly 4

17 Figre.0 HIV Seroprevalence for antenatal women by age in Nairobi, Kenya: % Figre. HIV Seroprevalence for antenatal women in Harare, Zimbabwe: % Figre. HIV Seroprevalence for antenatal women in Zimbabwe: % fond among women seeking antenatal care (UNAIDS, 998a). Figre.7 shows the rise in HIV prevalence in antenatal women dring the early 990s with a leveling off in the late 990s at abot %. Sentinel srveillance of key high-risk grops shows a decline in HIV infection among military conscripts starting arond 99, a proxy grop for yong men in the general poplation (Figre.8). In addition, HIV prevalence declined among female sex workers and male STD clinic patients between the mid- and late 990s, thogh rates among intravenos drg sers contined to remain high (0 40%) (UNAIDS, 00b). Cameroon Cameroon experienced steady increases in HIV prevalence beginning in the early 990s. Althogh the level of HIV infection remained lower than in many contries in eastern and sothern Africa, it now has one of the highest rates in West Africa. Median HIV infection among antenatal women in Cameroon's two main cities (Yaondé and Doala) rose from % in 990 to 5% in 996 and then p to 9% in 000. Among antenatal women aged 5 4 years in Yaondé, HIV prevalence increased from less than % in 99 to more than 5% in 998 (Figre.9). Kenya Kenya, thogh somewhat less affected than Zimbabwe and Zambia, has seen HIV prevalence among pregnant women steadily increase from abot 5% in 99 to % in 999 (NASCOP, 999). While data from sentinel sites across the contry show increases in HIV prevalence dring the 990s, HIV prevalence in antenatal women less than 0 years of age in Nairobi may have declined somewhat between 99 and 997 (Figre.0). Similar trends are not seen, however, among women aged 0 4 years. Preliminary data for 00 recently released by the MoH indicate that HIV prevalence among antenatal women in Nairobi declined from 6.5% in 000 to 4.4% in 00 (with even greater declines in yonger women), perhaps an indication that the epidemic was stabilizing. 5

18 Zimbabwe Finally, Zimbabwe has been extremely hard hit by the epidemic. The first cases of AIDS were identified in 985, and by 990 median HIV prevalence at sentinel antenatal clinic sites was estimated to be 9% in rban areas and % in rral areas. Since this time, HIV prevalence has contined to increase, and by 000 HIV prevalence among antenatal women across all sentinel sites was estimated to be 6% compared to 7% in 997 (BUCEN, 000). Data from sentinel sites in Harare indicate that seroprevalence among ANC women in the capital city peaked in the mid-990s and then stabilized throgh 000 (Figre.). Recent data from the srveillance system for 00 indicate that there may be a decline beginning in HIV prevalence among yong antenatal women (Figre.).. Stdy Objectives In generalized epidemics, sch as those in mch of sb-saharan Africa, HIV prevention programs among the general poplation have promoted delaying the onset of sexal debt among yoth, mtal monogamy, and condom se. However, the mix or emphasis of prevention messages has varied from contry to contry. In contries where the epidemic has been concentrated among high-risk poplations (sch as commercial sex workers), behavior-change approaches have sally been more targeted to these grops. Jst as the levels and trends of the epidemic in the six contries inclded in this report have varied, so have the programmatic responses. As sch, the behavioral response to HIV prevention efforts on a wide scale is likely to be mixed, as are the reslting effects on HIV incidence. In addition to changes that may be in response to HIV prevention programs, changes in sexal behavior may also occr as a direct response to HIV infection and reslting mortality. People may change their behavior de to fear of infection and death even in the absence of any formal HIV prevention program. This phase of the stdy aims to assess whether trends in sexal behavior are consistent with trends in HIV prevalence (and incidence where possible) in these six contries. Phase II will explore the broad programmatic, societal, and other contextal factors potentially associated with sexal behavior trends in these same six contries. In addition, changes in fertility may reslt from changes in sexal behavior in response to the epidemic, as well as to the impact of the epidemic on fecndity, mortality, and orphanhood. The relationship between AIDS and fertility is complex, and AIDS may have both depressing and increasing effects on fertility rates (Gregson, 994; Ntozi, 00; Gregson, Zaba, et al., 00; United Nations (UN), 00). HIV/AIDS and the adoption of HIV prevention behaviors can affect fertility patterns, and fertility can affect the risk of HIV infection. Many of the proximate determinants of fertility and HIV infection are similar, sch as sexal exposre within or otside marriage, contraceptive se, breastfeeding, and pathological sterility from sexally transmitted infections (Bongaarts, 978). In addition, changes in socio-economic, cltral, and other contextal factors (i.e., political stability and poplation policy) are also major determinants of reprodctive behavior as well as HIV infection. It is, therefore, very difficlt to discern the effects of AIDS on fertility among the many competing factors. While a comprehensive overview of fertility and fertility-related behaviors in the context of the epidemic is warranted, it is beyond the scope of this report. Therefore, with regard to fertility, this report aims to look specifically at whether changes in sexal behavior are consistent with changes in fertility-related behaviors and otcomes. 6

19 Chapter : Data and Methods. Nationally Representative Data Sorces While the analysis draws primarily on data from the DHS, other national-level poplation-based srveys are inclded when the srvey samples and indicators are thoght to be comparable. Becase of differences in the content and scope of the different srveys, and becase the introdction of detailed sexal behavior qestions in the DHS did not occr ntil the mid-990s, trend data for many indicators are available for the second half of the 990s only. Table. lists the nationally representative srveys inclded in the secondary data analysis. The srveys are groped by time period with those srveys condcted between 988 and 99 falling into the late 980s/early 990s, those condcted between 994 and 996 falling into the mid-990s, and those condcted between 998 and 000 falling into the late 990s. The srveys are nationally representative (with minor exclsions) and inclde women aged 5 49 and men aged An exception to the age range for the srvey is the 99 Kenya DHS where only men aged 0 54 were inclded. While data for women are available from the late 980s on, data for men are not generally available ntil the srveys condcted in the mid-990s. At the time of this report, the Zambia DHS data set was not yet available for secondary analysis. Slight differences in srvey implementation between the Zambia Sexal Behavior Srvey (SBS) and the Zambia DHS may reslt in some differences in indicator vales. A preliminary assessment of some of the data indicates that sing the 000 SBS to represent the late 990s in Zambia instead of the DHS may lead to somewhat of an overestimate of the decline in higher-risk sexal behavior and, to a smaller degree, of the increase in condom se. The differences are most apparent for men. While the overall pattern of behavior change reported here is similar to that of the DHS, the magnitde of the decreases shold be interpreted with some cation. While the fll data set was not available, a limited nmber of fertility indicators have been drawn from the Zambia DHS preliminary report as these data were not available in the ZSBS 000. While a comparative analysis of the five sb-saharan African contries inclded in this report was possible de to the availability of the DHS and other data, the same was not tre for Thailand. Althogh there is a wealth of data on sexal behavior in Thailand, most is obtained from specific poplations sch as commercial sex workers, their clients, military conscripts, and varios other occpational grops. These data are valable for tracking behavior change in specific grops, bt less so when making comparisons across contries. Two srveys condcted in the first half of the decade, the Srvey of Partner Relations and Risk of HIV Infection in Thailand in Table. Nationally representative srveys a for each contry sed in the analysis Cameroon Kenya Uganda Zambia Zimbabwe Late 980s/Early 990s (988 99) Year of Srvey Nmber of Women,87 7,50 4,70 7,060 4,0 Nmber of Men 84 b Year of Srvey Timing of Srveys Mid-990s ( ) Nmber of Women -- 7,540 7,070 8,0 6,8 Nmber of Men --,6 c,996,849,4 Year of Srvey d 999 Late 990s (998 00) Nmber of Women 5,50 7,88 7,46,79 5,907 Nmber of Men,56,407,96,55,609 a Unless otherwise noted, srveys were condcted as part of the Demographic and Health Srveys program. b In the 99 Cameroon DHS, only married men were srveyed, and no data on sexal behavior were obtained. c In the Kenya 99 DHS, only men aged 0 54 were srveyed. d Data for Zambia in 000 were obtained from the Zambia Sexal Behavior Srvey. While data from the Zambia DHS were not available at the time of this report, a limited nmber of fertility-related indicators were drawn from the preliminary report. 7

20 990 (Sittitrai 99) and the Srvey on the Effectiveness of AIDS Media on Behavior and Vales in 99 (Thongthai 995), provide nationallevel data that can be sed for comparative prposes. As these data were not available for secondary analysis, and becase some of the indicators differed in their calclations from others in the report, the few indicators available from these srveys are presented in separate tables. Data are also presented, when available, in the figres.. Spplemental Data Sorces In order to better nderstand changes in sexal behavior that may have occrred in the late 980s and early 990s, there has been a strong interest in condcting secondary analysis of the srveys carried ot with spport of WHO s Global Programme on AIDS (GPA). While the DHSs are valable sorces of data on sexal behavior from the mid- and late 990s, many of the DHSs condcted in the late 980s and early 990s did not inclde a male sample, nor did they collect information on sexal behavior beyond whether or not the respondent was sexally active. A direct comparison of the GPA data (when available) from the late 980s with DHS data from the mid- to late 990s has the potential to allow for a more thorogh assessment of changes in sexal behavior throghot the decade. For the contries inclded in this report, only Kenya and Uganda had GPA srveys with broad geographic coverage. The Kenya GPA data have not been inclded de to nmeros irreglarities in the data. The Uganda 989 and 995 GPA srveys were sbnational and have a strong rban bias (approximately 0% of the sample in 989 was from Kampala as compared to 6% in the DHS condcted the same year). However, becase these data make an important contribtion to the nderstanding of behavior change in Uganda in the late 980s and early 990s when the main decline in HIV incidence was likely to have occrred, a secondary analysis of the GPA data was condcted for this report. However, reslts are not compared directly with the nationally representative DHS data. A discssion of the Uganda GPA srveys is inclded in Appendix B at the end of this report. 8

21 Chapter : Backgrond Characteristics of Respondents. Introdction As factors sch as edcational attainment and work stats (a proxy for socioeconomic stats) are important determinants of reprodctive behavior, a smmary of levels and trends in these characteristics is presented here. In addition, becase of the importance of the interaction between male circmcision, STDs, and sexal behavior in the transmission of HIV, this chapter also incldes data on levels of male circmcision.. Backgrond Characteristics Edcation Edcation is an important determinant of individal behavior. Higher edcational attainment may be associated with lower fertility and even lead to positive changes in sexal behavior throgh varios mechanisms, sch as access to better knowledge for informed reprodctive choices, adoption of small family norms, se of modern contraception, and delayed ages at first birth and first marriage. Extra years of schooling may also lead to an increase in exposre to premarital sex becase of delayed marriage. For men, edcation may reslt in higher socioeconomic stats and a resorce gateway that can be sed to access commercial or casal sex. Table. presents levels and trends in edcational attainment among women and men dring the 990s. Reslts for women are also presented in Figre.. Overall, edcational attainment for women increased in all contries throghot the decade. The percent of women reporting they had not attended school was higher in the late 980s/early 990s than at the end of the decade. By the end of the 990s, these percentages had sbstantially declined by between 7% and 6% in all contries. A concomitant increase in the percent of women with secondary or higher edcation was also witnessed in all contries. The increase was mch greater in Zimbabwe (%) than in Uganda, Zambia, Kenya, or Cameroon, where increases ranged between 6% and 9%. Overall, Ugandan women lagged behind somewhat in secondary edcation throghot the 990s. Althogh Zambia and Zimbabwe witnessed some decline in the percentage of women with primary edcation, this was primarily de to a shift toward more women attaining secondary or higher edcation. Sex differentials in edcation are evident across contries and srveys (see Table.). Throghot the decade, the percent of men with no edcation was less than women, althogh the gap between women and men was narrow. Similar to women, the percent of men reporting secondary or higher edcation increased in all contries throghot the 990s. The percent of men with higher edcation almost dobled in Cameroon between 99 and 998 (from 4% to 46%), and in Zambia, 4% of men reported having secondary or higher level of edcation in 996 compared to 44% in Figre. Trends in edcational attainment among women Percent 9

22 Work Stats Employment is closely linked to earnings and is reflective of a person s level of edcation and wealth. Table. shows that dring the 990s, there was a steady increase in the percent of working women in Cameroon, Kenya, Uganda, and Zimbabwe, thogh there may have been a small decline in Zambia. While there are fewer trend data for men, it does appear that men were more likely than women to work for pay, bt differences were narrower in the late 990s than in the early 990s. Male Circmcision While not traditionally considered a backgrond characteristic, male circmcision may be an important factor affecting HIV transmission and acqisition in a poplation. There is sbstantial evidence sggesting a protective effect of male circmcision on HIV acqisition by men. Circmcision is believed to redce the risk of lcerative STDs (Bailey, 00), which are thoght to enhance the transmission and acqisition of HIV. Circmcision may also protect against HIV directly as the foreskin appears to be a highly vlnerable point of entry to HIV and other pathogens (Patterson, 00; USAID, 00). A meta-analysis of stdies examining the relationship between male circmcision and the risk for HIV infection among males in sb-saharan Africa conclded that ncircmcised men are more than twice as likely as circmcised men to be HIV-infected. The effect was even stronger among men at high risk for HIV than among men in the general poplation (Weiss, 000; USAID, 00). Only two of the srveys inclded in this report obtained data on male circmcision. The 989 Uganda GPA srvey fond that 5% of men srveyed were circmcised, with higher levels of circmcision in rban areas than in rral areas (0% compared to %). This was a slightly higher rate of circmcision than reported elsewhere in Uganda. In the Rakai district in sothwestern Uganda, a poplation-based stdy fond that 7% of men were circmcised (Gray, 000). Differences in rates of circmcision between these srveys may reflect the differences in poplations srveyed and their cltral practices or in methods of data collection. Both samples, however, had a similar percentage of men who were Mslim (% of men in the Rakai stdy and % in GPA), and these figres are similar to the national figres reported in the 995 Uganda DHS (0%). The Zambia Sexal Behavior Srvey (ZSBS) 000 reported that 7% of men in Zambia were circmcised, with rates of male circmcision of 5% in rban and 8% in rral areas. For the other contries inclded in this report, rates of circmcision in Cameroon and Kenya were reported to exceed 80%, while they were below 0% in Zimbabwe and Thailand (Halperin, 999). In the mlti-center stdy, a srvey of men in Yaonde, Cameroon, fond that almost all men (99%) were circmcised (Avert, 00). While the rate of circmcision in Kenya appears to be roghly 80%, the same stdy fond that in Kism in western Kenya, only 8% of men were circmcised. The prevalence of HIV in Kism has been estimated to be as great as 5% in 997 (thogh it has since declined to 9% in 999), a rate mch higher than that fond in other parts of Kenya, where most men are circmcised (BUCEN, 00). 0

23 Chapter 4: Levels and Trends in Sexal Behavior 4. Introdction The aim of this chapter is to present national-level trends in sexal behavior, inclding condom se, in five sb-saharan contries (Cameroon, Kenya, Uganda, Zambia, and Zimbabwe), as well as in Thailand. Differences in sexal behavior across contries are explored, as are trends within contries. For prposes of presentation of the data in the tables, the contries are organized into two grops. The first grop incldes the contries that have experienced declines in HIV prevalence dring the 990s; the second grop incldes the contries that have not yet experienced a clear decline. Becase the greatest declines in HIV incidence in Uganda are believed to have occrred in the late 980s, special attention is paid to this time period for Uganda. As nationally representative srveys provide only a limited amont of data on changes in sexal behavior for this time period, reslts from the 989 and 995 GPA srveys are presented. Also inclded are findings from research stdies from the early 990s, where sexal behavior was assessed in the same poplation at two or more points in time. Before presenting the findings, the limitations of selfreported sexal behavior data need to be acknowledged. It is possible that respondents may nderreport sexal activity, and it is believed that this nderreporting is particlarly problematic among yoth and yong women. In addition, as an AIDS epidemic matres and the level of AIDS awareness increases, srvey respondents may be more likely to give socially desirable answers (sch as having sed a condom) and less likely to disclose more socially nacceptable behavior (sch as premarital sexal activity and mltiple sexal partners). This tendency may be even more prononced in highly edcated societies sch as Zimbabwe. Figre 4. Median age at first sex among women aged 5 4 years Figre 4. Median age at first sex among men aged 5 4 years 4. Sexal Behavior Among Yoth Sexal Debt A major component of HIV prevention programs is encoraging yong people to delay sexal debt. Two standard indicators to measre the onset of sexal activity are considered. Table 4. presents levels and trends in the percent of yong people aged 5 9 who had ever had sex by residence. Table 4. presents the median age of sexal debt among 5- to 4-year-olds and was calclated as the age at which one half of yoth aged 5 4 reported having had sex. Median age at first sex is also presented in Figre 4. for women and in Figre 4. for men. In both Uganda and Zambia, there was a clear trend towards delayed age of sexal debt among yoth. The sexal debt among girls in Uganda increased from 6.5 to 7. years dring the 0-year period, while for boys it increased from 7.6 to 8. years in the second half of the decade. An analysis of the sexal debt data that sed both crrent stats and retrospective data (ths allowing trends before the time of the srvey to be assessed) fond that there was abot a one year increase in age of sexal debt for yong men and women in Uganda over a 0-

24 Figre 4. Percent of women aged 5 4 years who had never been married Figre 4.4 Percent of men aged 5 4 years who had never been married year period (Zaba, 00). Similarly, the percent of yoth aged 5 9 reporting ever having had sex also decreased. A delay in the onset of sexal activity is most evident among rban women in the first part of the 990s, while a delay in the onset of sexal activity really began dring the second half of the decade for rral women. Between the late 980s and mid-990s, the percent of women aged 5 9 who had sex declined from 66% to 59% in rban areas with no change in rral areas. Declines of abot 5% and 0%, respectively, were seen among yong rban and rral women in the second half of the 990s. The Uganda GPA data also show a decline in sexal activity among both yong women and men between 989 and 995 (Tables 4.4 and 4.5). It is notable that the decline in sexal activity among women based on the GPA data is mch greater than that based on the DHS data, indicating that the changes in sexal behavior based on GPA estimates may overstate national-level estimates, even when reslts are disaggregated by residence. In addition to real changes in sexal activity, differences between the estimates from the GPA srveys may reflect differences in poplations srveyed as well as in srvey implementation. In Zambia, the data also indicate that both yong women and yong men are postponing sexal debt.

25 There was a small decline in the percentage of 5- to 9-year-old females who reported ever having sex (from 6% in the early 990s to 56% in the late 990s). The decline for boys was greater (65% in the mid-990s to 44% in the late 990s) and occrred over a shorter period of time. Median age at first sex for yong men also increased from 6. to 8. years over the five-year period (this is likely to be somewhat of an overestimate), thogh there was little change in median age at first sex for girls. A srvival analysis sing both crrent stats and retrospective data did, however, show a statistically significant increase in age at first sex for women (of abot half a year) and a small bt insignificant increase for yong men (Zaba, 00). Among yong women, changes in age of sexal debt occrred only in the rban areas with little change in rral areas. Figre 4.5 Percent of women aged 0 4 years who were first married by exact age 0 years Figre 4.6 Percent of men aged 0 4 years who were first married by exact age 0 years * Figres for Cameroon are for married men only. There is little evidence of a postponement of sexal debt among yong men and women in Zimbabwe and among women in Cameroon or Kenya (there were no trend data available for men). However, a srvival analysis of the Kenya data that sed both crrent stats as well as retrospective data (and therefore cold assess change over time) fond a small bt insignificant increase in age of sexal debt for women and men in Kenya dring the 990s (Zaba, 00). Despite the increase of age of sexal debt in Uganda and Zambia, women still experienced an earlier age of sexal debt than their conterparts in Kenya and Zimbabwe. On the other hand, yong men in Uganda and Zambia initiated sexal activity at an older age than their conterparts in Cameroon and Kenya. The late age of sexal debt reported among yong men and women in Zimbabwe is particlarly srprising given the high HIV prevalence. Yet Cameroon, with the earliest onset of sexal debt for girls and the second yongest for boys, had mch lower rates of HIV infection. In general, yong women in the 990s in Uganda, Zambia, and Cameroon became sexally active abot one year earlier than did yong men. Kenya shows the opposite pattern, while there is very little gender difference in Zimbabwe. at First Marriage Delay of sexal initiation may potentially reslt in a concomitant delay in the age at which people enter into first marriage. Other factors sch as increased years of schooling and rbanization may also reslt in later marriage. Table 4. and Figres 4. and 4.4 show the percentage of never-married women and men according to crrent age. Of all the contries in this report, the greatest changes in the percentage of never married yong women over the past decade occrred in Uganda (Figre 4.). From the late 980s to mid-990s, the percentage of never-married women aged 5 9 years declined, indicating a move toward earlier marriage. This was followed by a sbstantial increase in the percentage of never-married yong women in the second half of the decade, indicating a move away from marriage in this yong age grop. There was, however, only a marginal increase in

26 women aged 0 4 years who had never been married. It is plasible that this shift toward marriage among 5- to 9-yearolds in the early part of the decade was a reslt of the fear of AIDS associated with premarital sex. While the reasons for the trend away from marriage in yong women in the second half of the decade are nclear, this is consistent with the increase in premarital sex (described in the next section) as fewer sexally active women in this age grop were married. Figre 4.7 Percent of yong women who were sexally active and married by age, 989 and 000, Uganda Figres 4.5 and 4.6 and Table 4.4 present the percentage of women and men aged 0 4 years who were first married by the exact age of 0 years. The data show that in Uganda, Zambia, and Zimbabwe there was little change in the proportion of women aged 0 4 years who were first married by 0 years of age. In Cameroon and Kenya, however, there seems to have been a delay in age at first marriage over the past decade. Yong women in Kenya and Zimbabwe married later than their conterparts in Uganda, Zambia, and Cameroon. Abot three qarters of yong women in Uganda and almost two-thirds in Zambia and Cameroon were first married by age 0 across all srveys. As there has been increases in ages of sexal debt in Uganda and Zambia, the reslt is a shorter period of time in which premarital sexal activity can occr and a redced likelihood of exposre to HIV throgh sex with non-marital partners. The narrowing gap between sexal debt and marriage is illstrated for yong Ugandan women in Figre 4.7. Men married, on average, later than did women. Althogh trend data for men are more limited than they are for women, there is a clear indication that the proportion of yong men who were first married by age 0 in Uganda and Cameroon (based on married men only) declined dring the 990s (Figre 4.6). There was little change in Kenya and Zimbabwe, and trend data are not yet available for Zambia. Althogh the age of sexal debt for men increased dring the 990s in Uganda, the concomitant increase in age at first marriage means 4 Figre 4.8 Premarital sex: Percent of never-married women aged 5 4 who had sex in the previos year that there wold have been a similar period of time between when men become sexally active and marry for the first time, dring which premarital sex occrs. Premarital Sex The percent of never-married 5- to 4-year-olds who had sex in the previos months by residence is presented in Table 4.5. This indicator reflects age of sexal debt as well as sexal activity among single adolescents who are already sexally active. Consistent with changes in age of sexal debt, there was a positive change for yong women in Uganda and Zambia (Figre 4.8). The proportion of single 5- to 4-year-old women who report having had sex in the previos year declined in Uganda and Zambia in the first part of the decade. Contining declines in Zambia are seen among yong women,