Annex E. An overview of HIV/AIDS epidemiology in Indonesia.

INTRODUCTION

This epidemiological overview focuses first on what was known about HIV, STI, and risk behavior in Indonesia in the mid-1990s. The overview then looks at how the situation has changed since that time, and looks for the impact of any interventions in Jakarta, an area in which several donors, including the Bank, intended to be active.

HIV AND STI IN INDONESIA IN 1995

Heterosexual Networking and STI Prevalence

The SAR describes high STI prevalence in the general population and in commercial sex settings, based largely on evidence from the port city of Surabaya. STI prevalence among nearly 1,900 sex workers surveyed in 1992-93 was 30 percent overall, with substantial differences between sex workers of different classes. (26)

Among women in the general population STI rates were far lower. A study of antenatal clinic attenders in a low-income areas of Surabaya and Jakarta conducted by the same team in 1992-93 found that 8.2 percent of women in Surabaya were infected with chlamydia, 4-5 percent in both cities had trichomonas infection, and under 1 percent in both cities were infected with either syphilis or gonorrhoea. (27) Several studies in other parts of Indonesia had similar findings. (28) Truckers, sailors and laborers in Indonesia were also tested. Prevalence ranged from 2.1 percent for syphilis to 8 percent for trichomonas. (29)

The SAR describes these truckers, sailors, and laborers as being "drawn from and arguably representative of the general population not only in Surabaya but elsewhere in urban Java." However, general population studies by the same authors recorded very much lower levels of sexual activity. For example, 77 percent of the sailors in the study population reported recent visits to sex workers, 10 times the highest rate reported by any sub-population in the population-based survey. (30) In a more recent household survey only

2.5 percent of the male population in West Java reported having sex with a sex worker in the last year. (31)

A population-based survey in East Java also found that non-commercial sexual networking was extremely low in the general population. Some 85 percent of single people in urban areas of East Java were not sexually active, and monogamy was the norm for sexually active men and women in both rural and urban areas. Just 16 percent of sexually active men (and 10 percent of all men) in urban areas reported having sex with more than one woman in their lifetime. That fell to 4.3 percent for sexually active men in peri-urban areas. Some 7 percent of urban women reported sex with more than one man in their lifetime, and in rural areas only 3 out of 692 women reported anything but abstinence or lifetime monogamy. (32) Close to 60 percent of the adult population in East Java and 55 percent nationally was classified as rural in the 2000 census. In the East Java city of Malang, a quarter of young men and one in 10 young women who were university students said they had ever had sex. The students' definition of "having sex" included masturbation and kissing and hugging. (33)

Given these findings, the assumptions about STI infection and sexual networking in the general population in the SAR appear high. This is important because a large part of the project benefits in the economic analysis centered on improving the treatment of STI as a means of reducing the transmission of HIV, and at least some of the service provision was focused on services reaching the general public, in particular women seen at MCH services. An overestimate of STI in the general population would lead to an overestimate of the likely spread of HIV, and also of the benefits of a program focusing on STI services.

Male-Male Sex

In the early 1990s, very little was known about men who have sex with men (MSM) in Indonesia. There was virtually no open gay scene, even in Jakarta, although foreign tourists created a demand for male sexual services in Bali.

The exception was among waria, transgendered males who frequently sell sex to heterosexually identified men throughout Indonesia. Waria sell both anal and oral sex; in anal sex they are frequently the receptive partner, but they also report selling insertive anal sex. Indonesian and foreign researchers had been studying risk behavior and HIV infection among waria in Jakarta since the early 1990s. Unprotected anal sex with multiple partners was the norm in this group; HIV infection rose from 0 in 1993 to 0.8 percent in 1995,34 putting it higher than prevalence among female sex workers.

Injecting Drug Users

Equally little was known about injecting drug users in Indonesia. A 1994 study in Surabaya estimated that there were at least 2,500 IDU in the city at that time. (35) Of 100 drug users surveyed, 41 were IDU, almost all male. Forty percent of them reported knowing more than 10 other IDU, and 63 percent reported sharing needles. Actual levels of sharing were probably higher because the question as asked did not encompass all types of previously used needles. All of the IDU reported multiple sex partners; one in five said they had used a condom sometime in their lives, but none used a condom at last sex. HIV prevalence was zero, and hepatitis B was 12 percent. In a paper published in 1994, Jalal and colleagues report evidence of drug injection in sex workers and others, concluding that "injecting drug use may be greatly underestimated." By the mid-1990s, Jakarta's only drug treatment hospital, Rumah Sakit Ketergantungan Obat (RSKO), was seeing 2,000 outpatients a year, and the proportion who were injectors was rising (RSKO, personal communication). The experience of several other countries, including Thailand, had shown that HIV prevalence among IDU can rise extremely quickly and can contribute to developing a critical mass of infection that can seed a wider heterosexual epidemic. However neither the Indonesian authorities nor their development partners expressed concern at the time over these early warning signs of growing drug injection. The projections used in the economic analysis of the SAR did not take drug injection into account, and no pilot prevention activities were planned for injectors.

HIV Surveillance Data

In Indonesia as in every developing country, HIV case reporting greatly understates the true magnitude of the epidemic. For this reason that the WHO and UN AIDS recommend sentinel surveillance for HIV--the anonymous testing of specimens collected from groups at risk during routine service provision. The material below summarizes the main findings from the national sero-surveillance system in the early 1990s.

In short, data from very active surveillance even in the high-risk groups found that HIV infection was extremely rare in Indonesia in the early to mid-1990s. Data from a very large number of blood donations suggest that, nationwide, a maximum of three Indonesian adults out of every 100,000 were infected with HIV. In the mid-1990s the Indonesian Red Cross, which conducts the screening, had no deferral criteria for high-risk individuals, and all positive blood samples were confirmed using Western Blot tests. In the five years from the start of surveillance in 1990 until 1995, HIV prevalence had not risen above 1 percent in any population in Indonesia. By comparison, HIV prevalence among sex workers in Thailand rose from 4 percent in 1989 to 30 percent by 1993.

THE EVOLUTION OF THE HIV EPIDEMIC IN INDONESIA SINCE THE BANK LOAN PROJECT

Data collection systems in Indonesia have improved greatly since the mid-1990s, when the Bank loan was prepared and appraised. This has increased our ability to understand the course of the HIV epidemic and the behaviors that drive it. HIV sentinel surveillance was expanded to include other sentinel groups such as JDU. Behavioral surveillance surveys (BSS) among high-risk sub-populations began in three cities in 1996. A further three cities were added in 1998, and the system was expanded to cover 16 cities in 13 provinces in 2002. One of the two areas in which the HSPMP and other donor projects were working, Jakarta, has been covered by this system since its inception.

Heterosexual Risk Behavior Continues Unchanged

The efforts of the government and its development partners to increase condom use in commercial sex and to improve the treatment of sexually transmitted infections appear to have had very little effect throughout the late 1990s. BSS data show that the proportion of sailors buying sex in Jakarta dipped during the economic crisis of 1998/99 but otherwise changed little, showing a slight but statistically significant downward trend from 57 percent in 1996 to 43 percent in 2002. Condom use among sex workers and their clients in Jakarta was constant or declined during this period (see Figure 1 of the main text) and STI prevalence actually rose among sex workers (Figure El).

Box E1. Projections of HIV Infection in Indonesia

Modeling HIV epidemics is notoriously difficult, especially in the
early days of an epidemic and where reliable data on the
behavioral parameters that drive an epidemic are not available.
However, projections are needed to demonstrate potential
infections averted, if a classic cost/benefit analysis of an
investment in HIV prevention is to be made.

In the early 1990s, two models were used to project the HIV
epidemic in Indonesia--EpiModel and iwgAIDS. EpiModel is a
curve-fitting program which was designed not to make forward
projections, but rather to compute, using information about the
natural history of HIV infection, the likely number of cumulative
HIV infections, AIDS cases and deaths implied by a given
level of current prevalence. In the opinion of leading
epidemiologists: "Models that use epidemic curves, such as
EpiModel,
should not be used in situations where extensive spread of HIV has
not been documented." (a)

EpiModel was used in Indonesia in 1994 to make projections
described in Jalal and others (1994). The paper predicted that
"it is
most likely that Indonesia will have half a million infections
within 4 years (on the basis of WHO/CDC EpiModel projections)."
The parameters of the model are not given, but one of the data
points through which the curve was drawn was an estimate of
current HIV infections nationwide of between 40,000 and 50,000
(attributed to a personal communication from G. Loth and M.
Linnan). This is 10 times the number of infections used by UNAIDS
in their Indonesian models, and 233 times the reported HIV
infections. It is over 3,400 times the number of infections found
in HIV surveillance among 40,000 sex workers in that year; the
sex workers tested represent around a seventh of all those
estimated to be working in Indonesia at the time. Even using
this data
point, the epidemic would have had to grow twice as fast in
Indonesia as it had in any other country to produce the numbers
infected predicted by Jalal, and others.

The other model used in making projections for Indonesia was
iwgAIDS. Developed by Stephen Seitz at the University of
Illinois, it models demographic processes, sexual mixing dynamics,
and the force of infection. This complex model was
structured on African epidemics, and requires very detailed
behavioral and biological inputs. In the opinion of a number of
modeling experts, it is ill suited to Asian epidemics, which are
concentrated in high-risk sub-populations, including injection
drug users and MSM. (a) Attempts to use the model in the data-rich
environment of Thailand were abandoned in the mid-1990s.

The use of the iwgAIDS model in the Indonesian context appears to
have been based on a belief that the HIV epidemic would
follow the same course in all countries. (b) In addition, the
Indonesian model appears to have been parameterized on Thai
data. (36)
In Thailand, HIV infection in direct sex workers escalated from 4
percent to 30 percent in just four years, while in Indonesia it
remained consistently below 1 percent even in the epidemiological
"hot spots" of Jakarta and Batam over the same period.
Survey data available in the mid-1990s give some clue as to why.
Consumption of commercial sex is around 10 times more
common in Thailand, while Thai direct sex workers serve 4 times as
many clients as their Indonesian colleagues. This strongly
suggests that the use of Thai parameters would yield unreliable
projections in the Indonesian context.

The model results in the economic analysis of the SAR predicted
that in the absence of the project, 17,500 people in Jakarta
would be living with AIDS and 105,500 would have died of AIDS by
2010, while 140,000 would be living with HIV. The
3,000 cumulative AIDS cases the model predicts for Jakarta in 1995
compares to fewer than 40 cases ever reported in the city
at that time. Some 8,500 people were projected to be living with
HIV in 1995--0.15 percent of the adult population. This
compares with HIV prevalence in close to 150,000 blood donors
recruited without deferral in Jakarta of 0.001 percent. By
2010, it was projected that HIV prevalence would escalate to 2.5
percent of the projected adult population.

The internal consistency of the model results is poor. In the
absence of war or other cataclysmic events which greatly
increase
adult mortality, the difference between cumulative HIV infections
and current HIV infections would be expected to be roughly
equal to the number of AIDS deaths. In the models published in the
project appraisal, there are as many non-AIDS deaths
among young HIV-infected adults as there are AIDS deaths. Such
dramatic internal inconsistencies are a strong signal that the
model outputs are unstable.

Because EpiModel, iwgAIDS and other existing packages fail
adequately to describe concentrated epidemics with a variety of
sub-populations at risk as they have developed in Asia, the UNAIDS
reference group on HIV estimates and projections
recommended the development of a more appropriate model. The Asian
Epidemic Model (AEM), developed by Wiwat
Peerapatanapokin and Tim Brown of the East West Center, is a
curve-fitting model with a significant number of behavioral
parameters, which builds a picture of the national epidemic from
sub-epidemics in specific sub-populations. It has been
parameterized for Jakarta using data available to the end of 2003,
and is used in generating some of the projections in this Annex.

(a.) MAP 1997. (b.) Linnan 1992.

[FIGURE E1 OMITTED]

[FIGURE E2 OMITTED]

More than half of sex workers with an STD self-treat or seek no treatment. There was a spike in use of public sector services during the crisis years of the late 1990s, probably because these services were more affordable than the private doctors who had been the norm until then (Figure E3). More recently, only 1 in 10 of those reporting STD symptoms reported that they went to a public sector clinic for care in 2002, similar to levels of public service consultation before the East Asian crisis. Those using public services tend to choose hospitals--the proportion using MCH or Puskesmas services was consistently low in all years. Similarly, around half of the men reporting STD symptoms self-treat or seek no treatment, while the share using public services has fluctuated under 25 percent in all years, dropping to just 13 percent in 2002.

These data come from those who report symptoms. In the 1998 STI survey among sex workers, only 28 percent of those who tested positive for either chlamydia or gonorrhoea reported having experienced any recent STI symptoms. This means that the proportion of those infected who are not receiving medical attention is higher still.

These data suggest that STI strategies that focus on improving services in public sector facilities to which infected people are expected to self-refer have not been successful in Jakarta.

HIV Prevalence Among Sex Workers in Indonesia

Despite the failure of prevention programs to reduce risk behavior, HIV prevalence remained consistently low among sex workers in most areas of Indonesia throughout the 1990s. Because the supply of female sex workers is large, the turnover of women is high, and the frequency of commercial sex among clients is low, Indonesian sex workers have few partners in their commercial lives relative to their colleagues in, for example, Thailand. In addition, circumcision among men is almost universal. These factors combine greatly to limit the likelihood of a rapid rise in HIV infection among sex workers in the absence of any other "seed" factor. It is not possible to recalibrate the iwgAIDS model used for projections in the early 1990s with the data on sexual behavior now available, but the Asian Epidemic Model (AEM) can be used to make comparative projections. If one assumes, as the projections prepared for the SAR did, that injecting drug use will not assume a major role in the epidemic, uses parameters on sexual networking in the general population drawn from studies in East Java in the early 1990s, and adds parameters from the sex industry that only became available in behavioral surveillance from 1996, one can look at the expected development of the HIV epidemic in Jakarta.

[FIGURE E3 OMITTED]

That seed factor was provided by the HIV epidemic among IDU, discussed below. The interaction between drug users and sex workers was the probable cause of the rise in HIV infection among sex workers that began in around 2000--the year HIV prevalence first exceeded 5 percent among sex workers in any site. By 2003, HIV prevalence among sex workers exceeded 5 percent in a number of sites around Indonesia, but sites with no infections also remained common, sometimes in the same city. In Jakarta, HIV prevalence among sex workers ranged from 1.2 percent to 6.8 percent in different sites in 2003 surveillance.

Male-Male Sexual Risk Expands as Society Liberalizes

Surveillance among waria in Jakarta stopped in 1997, although the 6 percent prevalence recorded that year made this the most affected group in the country at that time. It was resumed in 2002. Condom use at last anal sex with a client rose slightly over that period, but consistent condom use remained low. The consequence was a rapid rise in HIV infection among waria, to 22 percent in 2002. (37)

The social liberalization that occurred in urban areas in the late 1990s saw the emergence of an increasingly active gay scene. A number of cities opened bars and nightclubs where men gather to meet other men, often potential sex partners. In addition, a male sex industry (distinct from the transgender sex industry) developed in many cities. In Jakarta in 2002, men provided sex services to male clients in around 70 massage parlors, and freelance sex workers also worked in a number of cruising areas. Behavioral surveillance among male sex workers and other men who have sex with men in Jakarta, Batam (Riau), and Surabaya recorded very high levels of unprotected anal sex, despite quite high levels of awareness about HIV. HIV surveillance was conducted only in Jakarta--some 3 percent of male-identified sex workers and other MSM were infected with HIV.

HIV Explodes among IDU

The most significant development in the HIV epidemic in Indonesia has been the explosive growth of HIV among injecting drug users. This group has been included in sentinel surveillance in Jakarta since 1996, but HIV was first found in 1999, by which time 16 percent of injectors were infected. Just two years later, close to half of injectors tested positive, and similar rates were recorded at treatment centers in West Java, as Figure E4 shows.

[FIGURE E4 OMITTED]

Until 1998, less than 1 percent of ADDS cases seen at Jakarta's Rumah Sakit Cipto Mangunkusomo, the reference hospital for HIV treatment, were IDU. In 1999 that shot to 9 percent and has continued to rise since. (38)

Indonesia's Prisons--Possible Breeding Grounds for HIV

The prison system has emerged as a special concern in Indonesia's HIV epidemic. HIV surveillance in prisons has shown very rapid rises in prevalence in prison populations--more than one prisoner in four is now infected with HIV in more than one province. An early assumption was that rising prevalence among prisoners merely reflected rising prevalence among IDU, since drug offenders make up a significant and growing proportion of the prison population. But recently the surveillance system has picked up evidence of transmission within prison. This is probably related to ongoing injection in prison, as well as anal sex between inmates. Since prison terms are typically short, with around a quarter of prisoners incarcerated for just three months, infection in jail carries the extra danger that a prisoner will be released while in a stage of high viraemia. If he resumes sexual activity or injecting on release, the likelihood of onward transmission is high.

Putting it Together: The Actual Course of the Epidemic

With the benefit of hindsight, greatly improved data sources, and more appropriate modeling software, it is possible to recreate with some accuracy the path of the epidemic in Jakarta, and to project it forward with greater confidence. The evidence of the surveillance system together with the results of this exercise show that HIV prevalence began to escalate significantly in the early 2000s, driven very largely by a new epidemic in IDU. The take-off point for the epidemic was about five years later than that predicted by iwgAIDS, which predicted an early and rapid rise despite the fact that it included only limited transmission among drug injectors.

[FIGURE E5 OMITTED]

Figure E5 compares the outputs of iwgAIDS projections made in 1995 and AEM projections fitted to prevalence recorded in all sub-populations at risk and in the general population until mid-2003. By 2010, current HIV prevalence was 50 percent lower than predicted by iwgAIDS and cumulative infections were one-third predicted levels. The similar levels of current HIV infection predicted by AEM by 2010 cannot be understood as a validation of the iwgAIDS projections, since that model did not foresee the contribution of injecting drug use to the sexual as well as the non-sexual epidemic. Because IDU in Jakarta are overwhelmingly young men and injection is a recent phenomenon, rates of sexual activity are high. A quarter of male IDU reported in behavioral surveillance in 2002 that they had had sex with a female sex worker in the last year. The sexual interaction between IDU and other high-risk groups has provided a "booster" effect that has created a critical mass of infection in the commercial sex sector and thus had a major impact on the unfolding epidemic. While only around one-third of infections projected by AEM for 2010 will be in IDU, most of the other two-thirds are sexual infections that originated at some stage with an infection transmitted by needle-sharing. As Figure E3 showed, had there been no HIV epidemic among drug injectors in Jakarta, a sexual epidemic would not have developed by the end of the first decade of this century.

Leaving aside models, there is little evidence from surveillance, case reporting, or special studies that HIV is spreading widely among people who do not engage in definable risk behaviors and their primary sex partners. The exception to this rule is the country's easternmost province of Papua, where estimates made at the district level at the end of 2003 suggest that HIV already exceeds 1 percent in the general population.

In late 2002, the MOH led an estimation process that began by estimating the size of each at-risk population in each province, estimating prevalence within each of those populations, and then aggregating into a national total. This process, validated by WHO and UNAIDS, estimated that 90,000-130,000 Indonesians were living with HIV at the end of 2002. (39) The process was repeated in early 2004, but at that time estimates for around a third of the epidemiologically more important provinces were made at the provincial level, using data collected at the district level. The resulting estimate was in the same range as the 2002 estimate, but the composition of populations was rather different. With improved data sources it was found that the number of clients of sex workers had previously been overestimated while infection rates in men who have sex with men were also too high. Infection rose among IDU and their sex partners.

(1.) The mission was conducted in connection with a case study of the effectiveness of the Bank's HIV/AIDS assistance to Indonesia (Elmendorf, Jensen and Pisani 2004), one of four country case studies for OED's evaluation of the Bank's assistance for HIV/AIDS control. This PPAR provides greater detail on the performance of the HIV/AIDS and STDs Prevention and Management Project, one element of that case study. The PPAR has been prepared by one member of the case study team (A. Edward Elmendorf), and takes into account findings from the case study, to which team members Eric Jensen and Elisabeth Pisani provided significant contributions.

(2.) National surveillance records, MOH, and U.S. Census Bureau HIV database.

(3.) Republic of Indonesia, MOH tables contained in a January 1997 Bank memorandum.

(4.) Linnan 1992.

(5.) Water supply and sanitation in low-income communities is also included in the health portfolio.

(6.) Jalal and others 1994.

(7.) Jalal and others 1994.

(8.) World Bank 1994.

(9.) The Japanese grant provided for financing of an institutional specialist, but the individual was not hired.

(10.) As with most other Bank-financed projects, the presentation of the project development objectives in the Loan Agreement was the only statement of objectives explicitly agreed between the borrower and the Bank.

(11.) The ICR treats the PMU as a project unit, and reflects the continuing confusion on this issue that persisted during HSPMP implementation. Since closure of the Bank loan, the PMU appears to have disappeared, in practice if not in form.

(12.) Some other donor-financed HIV/AIDS projects also suffered from the financial crisis and the changing context for HIV/AIDS work in Indonesia. A UNDP project was particularly affected (Janssen and Purwaningsih 2003)

(13.) The term 'Directly Transmitted Diseases' is used in the Indonesian English translation of decrees and other working documents.

(14.) The Bank responded to a non-objection request for a procurement award with an estimate that the recommended bidder's prices were 50 percent to 350 percent higher than international market prices. According to the ICR, the letter found that collusive practices between bidders and authorized medical equipment suppliers appear artificially to raise equipment prices, and recommended that supplies be procured through the UN.

(15.) Jansen and Purwaningsih 2003.

(16.) The spike in condom use reported by sex workers but not clients in 1999 may be related to the economic crisis, which reduced demand for commercial sex among Indonesian men. The women working in the north Jakarta port area covered by this surveillance are likely to have had proportionately more foreign clients in 1999 than in other years, and BSS data show that foreign clients are more likely to use condoms than Indonesian clients.

(17.) World Bank 2003b.

(18.) World Bank 2004.

(19.) Source: World Bank 1996 (SAR); no baseline or target values were given.

(20.) Source: World Bank 1996, Annex 4.

(21.) Source: World Bank 2000b (ICR).

(22.) Source : World Bank 1996.

(23.) Source : World Bank 1996, Annex 4.

(24.) Source: World Bank 2000b.

(25.) This annex is based on analyses by Elizabeth Pisani for the OED case study of the Bank's HIV/AIDS assistance to Indonesia (Elmendorf, Jensen, and Pisani 2004).

(26.) Joesoef and others 1997.

(27.) Joesoef and others 1996.

(28.) For a comprehensive summary of study findings, see ROI/MOH and WHO 2000.

(29.) Linnan 1995.

(30.) Kambodji and others 1995. 4.7 percent of urban males reported ever having bought sex; that percentage fell to 2.5 percent for periurban men and less than 1 percent for rural men.

(31.) Riono 2001.

(32.) Kambodji and others 1995.

(33.) Kambodji and Linnan 1995a.

(34.) Lubis and others 1997.

(35.) Kambodji and Linnan 1995b.

(36.) The parameters have not been published and repeated requests to the authors of the models for information about input parameters have gone unanswered. Two epidemiologists who were asked to contribute data recall that most of the input parameters used derived from Thai data.

(37.) Pisani and others 2004.

(38.) Sedyangingshih-Mahamit and Gunawan 2000.

(39.) Republic of Indonesia 2003.

Table E2.HIV prevalence data from national surveillance, 1990-1994/5

Year      Population           Number     Number       HIV
                               tested    positive   Prevalence
                                                    (percent)

1990      Female sex workers    4,420       0           0
1991/92   Female sex workers   20,293       0           0
1992/93   Female sex workers   38,444       0           0
1993/94   Female sex workers   52,870       3         0.006
1994/5    Female sex workers   39,790       12         0.03
1993/94   Pregnant women         572        0           0
1994/5    Pregnant women        4,750       0           0
1992/93   Blood donors         533,865      8         0.002
1993/94   Blood donors         705,345      5         0.001
1994/5    Blood donors         748,813      17        0.002