Study Design

The study was conducted in a cohort of 253 female brothel-based sex workers in March 1994.  There are an estimated number of 1,200 brothel-based workers concentrated in 6 geographically defined localities in Singapore.   All 128 sex workers from a central locality in Singapore were assigned to an intervention programme.  This locality was identified to have high gonorrhoea rates and a mixed ethnic distribution of Chinese (80.6%) and  Malays 19.4%).  A locality that was situated about two kilometres from the intervention  locality and separated from it by two main roads was selected as a control group because it was   judged to be reasonably comparable to the intervention group in socio-economic class, number and type of clients and brothel establishment.  It was also the only other  locality with a mixed ethnic distribution of Chinese and Malays and was closest to the intervention group in gonorrhoea rates.  Diffusion of effects are unlikely to occur between the two localities because sex workers from different areas hardly interact due to the stigma and "competition" associated with their work.

The sample size was estimated to give a statistical   power of 80%, at alpha of 0.05 for detecting a 60% reduction of gonorrhoea from an estimated cumulative incidence of 20% in the five-month period prior to intervention to 8% in the five-month period after intervention.

The intervention, based on Green's PRECEDE framework (13), Bandura's social learning theory (14) and Ajzen and Fishbein's Theory of Reasoned Action (15) focused on developing the sex workers' negotiation skills and increasing their self-efficacy in always refusing sex without a condom if the negotiation fails; mobilizing support from brothels to support condom use; and intervening in their social network to change norms and effect behaviour.   Our previous studies on risk factors of STDs (16), knowledge and sexual behaviour (8) and condom negotiation (9) among sex workers were used to design messages on specific ways to negotiate condom use with different clients, alternatives to take in the event of client refusals, and the importance of peer cooperation so that one of them would lose their clients to others.  The messages also emphasized the need to use condoms with regular clients.  Environmental support was enhanced by getting brothel keepers to display posters and talk to clients on condom use and by instructing health  advisors from the public STD clinic to check on the brothels' compliance.

The intervention programme has been described in detail elsewhere (17).  In brief, it consisted of two two-hour small group sessions conducted by trained health advisors and the first author (Wong ML) in the public STD clinic.  Groups of about 16 sex workers were organized to provide a heterogeneous mix of sex workers with differing attitudes and skills in condom negotiation so as to get them to share their experiences and support each other under the guidance of a health facilitator.  Instructional methods included video presentations with local sex workers as actresses to demonstrate negotiation skills, role play, and peer group discussion of problems arising from their self-monitoring of condom use.  The experienced peers gave practical tips on how to deal with difficult clients.  Sex workers were also given stickers depicting "condoms must be used here" for display in their brothel-rooms to facilitate their negotiation task.  All gonorrhoea cases were given individual counseling.  Peer leaders were selected by the sex workers themselves to follow up on the sessions, given feed-back to health staff and act on problems encountered.  Three months later, a booster session was held with the   distribution of free condoms, comic scripts, pamphlets, and dissemination of congratulatory messages to all compliant and non-infected participants.  Sex workers who could get all their clients to use condoms were asked to share their skills with the others.  The control group did not receive any intervention.

Regular meetings were held  with peer leaders to get feedback and discuss problems encountered.  Some problems encountered were non-supportive brothel keepers who recommended clients to non-condom-using sex workers; acceptance by some sex workers of clients that were refused by their peers for not using condoms and problems of condom use such as slippage, breakage, pain, and clients' complaints of failure to ejaculate.   Prompt action was taken by the first author to provide individual counseling to non-condom-using sex workers that  were made known to her by the peer leaders. The stage of behaviour change (18) was applied to assess the sex worker's readiness to use condoms and identify their reasons for not using condoms so that advice  could be matched to their needs.  Sex workers who encountered difficult clients and experienced problems such as condom slippage and pain were given  specific tips that had been used effectively by their peers.  Gonorrhoea positives were also counseled.   Non-supportive brothel keepers were warned to support condom use and told of the benefits of a STD-free brothel.

Booster group sessions were held after five months, one year and two years of intervention.  The sessions dealt mainly with discussing their problems and sharing tips to prevent condom rupture, leakage, failure to ejaculate, and pain from prolonged  condom use as well as reinforcing the benefits of condom use.   After one year when about three quarters were always refusing unprotected sex, we promoted the message that it was a norm to use condoms so that non-compliant sex workers were pressured to change their behaviour.  Compliant sex workers were also encouraged to influence and put pressure on their non-condom-using peers so that  their business would not be affected by loss of clients to the latter.

The same self-report questionnaire on sexual behaviour and condom use, available in Malay and Chinese, was administered twice, first at baseline and then five months after the intervention to both intervention and control groups.  Where there were illiterate respondents (15% of the total respondents), questions were read out to them and they were told to mark their responses.

Condom use was assessed quantitatively by asking the women how many of their non-spontaneous-condom-using client on an average out of ten (i) did they negotiate for condom use and (ii) used condoms following negotiation.  These questions were protested, used in our previous surveys and found to be well understood by the sex workers who could relate better to units of 10 rather than a percentage.  The total number of new cases of gonorrhoea that had occurred in the intervention and control groups during the 5-month period prior to the intervention and the 5-month period after the intervention were obtained from records of the sex worker's  routine biweekly endocervical culture tests.

The control group was no longer retained in the study after the 5-month follow-up for ethical and administrative reasons.  However, the intervention group was followed up for two years with the same questionnaire administered at one year and two years and their routine clinical records checked for gonorrhoea.   The first author also conducted in-depth interviews with 22 randomly selected sex workers to assess factors leading to sustained condom use.

The three primary outcome variables are (i) negotiation skill as measured by the self-reported average success rate in persuading clients to use condoms, (ii) behavioural change as measured by the percent always refusing vaginal sex without a condom and (iii) cumulative gonorrhoea incidence.  Gonorrhoea incidence was used as an objective indicator of the impact of intervention on condom use in view of the short incubation period and short duration of the  infection, high specificity and sensitivity of the diagnostic test, and rapid and effective response to treatment.  In contrast, other STD's such as chlamydia may become chronic and would not be influenced by condom use.  Cumulative gonorrhoea incidence for the 5-month period before and after the intervention programme is calculated   by dividing the total number of persons with new occurrences of gonorrhoea within the specified period i.e. either the 5-month period before  the intervention or the 5-month period after the intervention by the total number of sex workers who were   enrolled into the study.  The proportion is then expressed as a percentage.   For example, if 100 gonorrhoea-free sex workers were enrolled at the beginning of the 5-month period prior to the intervention programme, and 20 of them subsequently developed gonorrhoea within this period the pre-intervention cumulative gonorrhoea   incidence would be 20%.  Likewise the post-intervention cumulative gonorrhoea incidence would be 10% if 10 out of this cohort of 100 subjects developed gonorrhoea during the five months after the intervention programme.

The primary goal of the data analysis was to statistically determine the change in negotiation skills, safe sex behaviour and gonorrhoea incidence  from prior to intervention (baseline) to post intervention between the intervention and control groups.  The paired t-test was used to compare the success rate in persuading clients to use condoms between baseline and post-intervention separately for the control and intervention groups (Table 2).   Multiple covariance analysis was used to compare the mean follow-up success rate between control and intervention groups statistically  adjusting for baseline success   rate, race and age difference of the sex workers (19).  A modification of Cox regression model for cross sectional data was used to estimate the observed and adjusted rate ratios  of intervention to control group in behaviour outcomes of always refusing sex without a condom and cumulative gonorrhoea incidence.  Although the Cox regression model was  specifically developed for prospective cohort studies, it was recently used for cross-sectional studies (20).  In Table 2, for example, the gonorrhoea cumulative incidence rates of the intervention group at two points in time: the 5-month period before intervention and the 5-month period after intervention was computed and compared with the respective rates in the control group as rate ratios.  The post-intervention rate ratio was adjusted for race and baseline gonorrhoea rates using modification of Cox's model.  This approach yields the  cumulative incidence rate ratio as the effect measure rather than the odds ratio, if the logistic  model was used.  The odds ratio is less meaningful, and furthermore  it is not applicable in this situation as the outcome measures are not rare events.

Response Rates

Initially, 253 sex workers participated in the baseline interviews with no refusals.  Of these seven subjects did not complete the study at the 5-month follow-up due to one death from the control group and six (two from the control group and four from the intervention group) quitting prostitution.  They were subsequently left out  of the analysis.  246 subjects were followed up at five months with 122 subjects in the control group and 124 in the intervention group.  96 (77.4%) of the intervention subjects completed the one-year follow-up  and 74 (59.7%) the two-year follow-up.  Of those lost to follow-up at two years, 33 (26.6%) had quit prostitution and 17 (18.7%) of those still working did not turn up for the interview.   There were not discernible differences in age (35.7 yr vs 37.7 yr); number of clients (9.0 vs 9.1) and proportion with no schooling (37.3% vs 38.8%) between those followed up and those lost to follow-up .  None of those still working defaulted in the routine gonorrhoea tests.

Both groups were in their late thirties and had similar educational level (Table 1).  They were also comparable in negotiation skills and behaviour of always refusing unprotected sex.   Although the intervention group had less Malays than the control group, the Chinese and Malays in the intervention group were comparable in factors associated with the outcome variables such as mean charge per client (US$ 16 vs US$ 14) and number of clients (8.9 vs 9.0).  They were also similar in prevalence of anal sex (<2%) and behaviour of always refusing sex without a condom (45.1% vs 43.5%).  Similarly the Chinese and Malay in the control group did not   show any discernible differences in sociodemographic characteristics and sexual behaviour.

Controlled Evaluation of the Intervention at Five Months

Table 2 shows the results of condom use negotiation, refusal of sex without a condom and cumulative gonorrhoea incidence in the intervention and control groups at 5 months after intervention.

The intervention group showed a considerable improvement in negotiation skills, with the average success rate in persuading clients to use condoms rising from 66.1% at baseline to 80.2% at 5 months post-intervention, a 14.1% difference.  This is in contrast to a small improvement of 3.2% for the control group.  The difference in improvement rates between the intervention and control groups, statistically adjusted for baseline   difference in success rates, race and age of the sex worker was 20.7% (p<0.001).

The proportion of sex workers who reported always refusing sex without a condom increased considerable by 20.8% from 44.4% at baseline to 65.2% post-intervention in the intervention group.  In contrast, the control group showed a decline of 5%, from 40.2% to 35.2%.  Sex workers exposed to the intervention were almost twice more likely than controls to refuse unprotected sex (observed rate ratio 1.85, 95% CI: 1.25-2.76).  Adjustment for baseline difference in rates of sex refusal, race and age, did not materially alter the rate ratio (adjusted rate ratio 1.90, 95% CI: 1.22-2.94).  In order not to lose a client, a high proportion of intervention subjects resorted to helping clients to masturbate (69.8% vs 39.6% in controls, p<0.001) and providing massage (40.9% vs 18.6% in controls).

The 5-month cumulative gonorrhoea incidence in the intervention group declined significantly by 77.1% from 10.%5 pre-intervention to 2.4% post-intervention, compared to a non-statistically significant decline of 37.6% from 19.7% to 12.3% in the  control group.   Before intervention, the gonorrhoea incidence of the intervention group was about half that of the control group (p=0.067).  After intervention, the intervention group showed only a 0.2 fold cumulative incidence of the control group (observed rate ratio 0.20, 95% CI: 0.06-0.68).  The rate ratio remained  unchanged after adjustment for race and baseline differences in gonorrhoea rates in the control and intervention groups (adjusted rate ratio: 0.21, 95% CI: 0.06-0.73), implying that the much   lower gonorrhoea  rate in the intervention group is unlikely to be confounded by the baseline difference in gonorrhoea rates.  None of the sex workers were HIV positive.

The success rate in getting clients to use condoms  increased to 79.1% at one year and 91.1% at two years post-intervention.  The proportion always refusing unprotected sex increased to 75% at one year and to almost all (93.1%) at two years (Figure 1).  Figure 2 shows the trend in the 5-month cumulative gonorrhoea incidence from 20 months before to 20 months after the intervention.  The decline over the preintervention period was slow and less marked compared to a steep decline in the first five-month period after the intervention.  The low incidence was maintained around 3% and this was followed by another dip to 1% after the reinforcement session at one year post-intervention.

Reasons given by the sex workers for the sustained condom use were that they had become more skillful in persuading clients and it had become a norm among them to use condoms.  Their confidence was reinforced when former clients returned and agreed to use condoms with no loss of clients.  The negotiation task was further facilitated by the display of posters depicting "100% condom use: condoms must be used here" as the clients perceived condom use to be mandatory.  Regular reminders from health staff and brothel keepers helped to reinforce their positive behaviour.  In addition, they had learnt techniques to prevent condom slippage, breakage and pain.  Self-reported condom breakage has decreased from 44.8% at one year to 20.3% at two years.

The interpretation of the effects of the intervention should be considered in the light of its limitations.   First, we used a non-randomized matched design that might threaten internal validity.  It was not feasible nor appropriate however to randomize brothels because their close proximity within localities would lead to diffusion of intervention effects across  brothels.  We also wanted to create an environment supportive of condom use and encourage cross-brothel discussion in the same locality.

Another concern was the baseline difference between the intervention and control subjects in race and gonorrhoea  incidence.  Actual field conditions and the limited number of comparable localities did not permit us to find a control group that could be matched on all the outcomes and factors associated with them.  The selected control group was the closest  possible to the intervention group in gonorrhoea  rates and race distribution.  Other than these two differences, the two groups were comparable in the other outcome measures and variables most likely to influence outcomes such as age, socio-economic class, and educational level.  In addition, differences in racial distribution are unlikely to confound results as race was not found to be associated with behaviour outcomes and gonorrhoea incidence.  Although we used statistical procedures to adjust outcome measurements for known pre-existing group differences, we are unable to adjust for unknown intergroup differences, that could have led to the observed improvements in the intervention group.   We, thus, selected behavioural outcome indicators specific to the intervention such as negotiation skills and problems of condom use such as breakage and slippage.

Third, intervention subjects might overreport condom use or desirable behaviours.  These biases were minimized by using  self-administered questionnaire assuring confidentiality and anonymity, and explaining the purpose of the survey.

Five months after the intervention, the intervention group showed considerable improvement in negotiation skills, and self-reported behaviour of refusing unprotected sex that was corroborated by a decline in gonorrhoea incidence.   The observed improvement in outcomes in the intervention group could be attributed to the effect of the intervention, baseline differences between the groups or other   unknown confounders.  However, adjustment for known baseline  differences in race, negotiation skills and  gonorrhoea incidence did not  affect the observed improvement in the intervention group, indicating that non equivalence of the groups is unlikely  to be solely responsible for the change.  The intervention group also showed  a more marked statistical improvement in negotiation skills and   increase in refusal of unprotected  sex in contrast to the control group which showed negligible changes.  Intervention subjects were also more likely to use alternatives like massage and masturbation in the event of negotiation failures.  As these outcome variables were specifically related to the intervention, it provides further evidence of the plausible effect of the intervention.

At the two-year follow-up, negotiation skills and refusals of unprotected  sex increased to almost 100%.  This was corroborated by the maintained low gonorrhoea  incidence with another dip after the reinforcement session.  The absence of a control group after the 5-month follow-up makes it difficult to interpret whether the sustained effects were due to the intervention or effects such as historical trends, ongoing public education, instrumentation or maturation.  To reduce the possibility  of making false conclusions, we used a time series design to allow for a more detailed  analysis of the trend of gonorrhoea incidence.  The overall decline in gonorrhoea incidence over the 20 month post-intervention period was much greater than the 20 month pre-intervention period.  The sustained greater decline after intervention is unlikely to be solely due to public education or maturation effects as we would then expect the same   degree in decline over the pre-intervention period.  There was no new public campaign during the post-intervention period.  Moreover, the sex workers attributed their sustained condom use to the skills  learnt from the intervention and the posters provided to them.

Slightly more than one fifth (26.6%) were no longer in prostitution at two years.  We did not find any discernible differences in socio-demographic characteristics among those followed up and those lost to follow-up.   Thus selection bias from attrition is unlikely to be gross.

The intervention had increased negotiation skills and condom use and the effects were corroborated by a decline in gonorrhoea incidence.  The programme's success was due to comprehensive planning that considers behavioural, socio-cultural and environmental factors; the use  of quantitative and qualitative methods to identify sex workers' needs; multiple health education methods aimed at changing behaviour;  and involvement of sexworkers in designing messages.  Factors leading to the sustained condom use after two years were: close monitoring and prompt action  on problems encountered; involvement of sex workers in identifying solutions; and reinforcement sessions focusing on condom use as a norm.  This project can be adapted for use among brothel-based sex workers in Southeast Asian communities with comparable socio-cultural environments by applying the above-mentioned factors.

Our project may not be applicable to street-based sex   workers whose characteristics and working conditions differ.  We started by studying brothel-based sex workers to increase our understanding of the nature of commercial sex, and hopefully used this as an entry point to gain access to the less easily defined group of street-based sex workers.  We believe that certain principles used in our  project can be  adapted to increase condom use among street-based workers.  These include the application of behavioural theories  and involvement of sex workers in planning.

References

Armitage P, Berry B Statistical Methods in Medical Research, 2nd edn.  Balckwell: Oxford, 1987.

Asamoah-Adu A, Weir S, Pappoe M, Kanlisi N, Neequaye A, Lamptey P.  Evaluation of a targeted AIDS prevention to increase condom  use among prostitutes in Ghana.  AIDS 1994; 8: 239-246.

Campbell CA.  Prostitution, AIDS, and preventive health behaviour.  Soc Sc Med 1991; 32:1367-1377.

Capocaccia R, Mariotti S.  Evaluation of a risk factor variation in relation to their baseline values in a controlled preventive trial.  J Chron Dis 1982; 35:509-520.

Chew SK, Snodgrass I, Monteiro EHA.  HIV infection in Singapore : The second wave.  Epidemiol. News Bulleting, Singapore.  1992; XVIII:25-27.

Green L W, Kreuter MW.  Health Promotion Planning, An Education and Environmental Approach.  Palo Alto: Mayfield, 1990.

Hanenberg RS, Rojanapithayakorn W, Kunasol P, Sokal DC.   Impact of Thailand's HIV control programme as indicated by the decline of sexually transmitted diseases.  Lancet 1994; 344:243-245.

Lee J.  Analysis of covariance by the SAS GLM procedure.  Comput Biol Med 1987; 17:221-225.

Lee J.  A note on the comparison of group means based on repeated measurements of same subjects.  J. Chron Dis 1980; 33: 673-675.

Lee J.  Odds ratio or relative risk for cross-sectional data?  Int J Epidemiol 1994; 23: 201-203.

Roumeliotou A, Papautsakis G, Kallinikos G, Papaevangelou G.  Effectiveness of condom use in preventing HIV infection in prostitutes.  Lancet 1988; 1249.

Swaddiwudhipong W, Nguntra P, Chaovakiratipong, Koonchote S, Lerdlukanavonge P, Chandoun C.  Effect of health education and condom promotion on behavioural change among low socioeconomic prostitutes in Mae Sot, Tak.  Thailand.   Southeast Asian J Trop Med Pub Health 1990; 21:453-457.

Wong ML, Chan R, Koh D, Wong CM.  Theory and Action for Effective Condom Promotion: illustrations from a behaviour intervention project for sex workers in Singapore.   Int'l  Quarterly of Community Health Education.  1995; 15: 405-421.

Wong ML, Tan TC, Ho ML, Lim JY; Lim R, Wan S, Chang R.   Knowledge and sexual behaviour related to HIV and AIDS of female sex workers in Singapore.  Health Education Journal 1994; 53:155-162.

Wong ML, Tan TC, Ho ML, Lim JY, Wan S, Chan R.   Factors associated with sexually transmitted diseases among sex workers in Singapore.  Int J of STD and AIDS 1992; 3: 332-337.

Top