Article Text

Download PDFPDF

Changes in HIV-related behaviours over time and associations with rates of HIV-related services coverage among female sex workers in Sichuan, China
  1. J T F Lau1,
  2. K C Choi1,
  3. H Y Tsui1,
  4. L Zhang2,
  5. J Zhang3,
  6. Y Lan3,
  7. Y Zhang4,
  8. N Wang5,
  9. F Cheng6,
  10. J Gu1
  1. 1
    Centre for Epidemiology & Biostatistics, School of Public Health, Faculty of Medicine, Chinese University of Hong Kong
  2. 2
    Sichuan Provincial Center for Disease Control and Prevention, China
  3. 3
    West China School of Public Health, Sichuan University, China
  4. 4
    Management Office of China-UK HIV/AIDS Prevention and Care Project, Beijing, China
  5. 5
    National Center for AIDS/STDs Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing, China
  6. 6
    China Country Office of Family Health International, China
  1. Professor J T F Lau, Centre for Epidemiology and Biostatistics, School of Public Health, Faculty of Medicine, The Chinese University of Hong Kong, 5/F., School of Public Health, Prince of Wales Hospital, Shatin, NT, Hong Kong; jlau{at}


Objectives: To investigate changes in HIV-related behaviours among female sex workers (FSW) and associations with services coverage rates.

Methods: Behavioural surveillance data from Sichuan, China, were analysed. A mapping exercise was conducted; FSW were recruited from randomly selected sex-work establishments in 19 sites in Sichuan, China, from 2003 (n = 7068), 2004 (n = 6875) and 2005 (n = 6833).

Results: Site variations were substantial. The random effect pooled AOR comparing the prevalence of condom use with regular sex partners, possession of condoms, HIV-related knowledge, HIV antibody testing and services coverage rates in 2005 versus 2003 ranged from 1.42 to 20.35. The 95% CI of these pooled AOR all excluded 1.0; hence rejecting the null hypothesis that such OR were not different from 1.0. Most of the AOR of these evaluative parameters (indicator of improvement) for the 19 individual sites in 2004 and 2005 (vs 2003) were significantly associated with coverage rates (Spearmen correlation coefficients  =  0.35 to 0.67, p<0.05).

Conclusions: Improvements were observed in relevant behaviours and coverage rates and the two were associated with each other.

Statistics from

In Asia1 2 and in China,3 the number of cases of heterosexual transmission of HIV has been increasing. Owing to the large population sizes of female sex workers (FSW) and their clients, even low prevalence of HIV among these groups would imply a large number of new infections. Around 6–11% of the adult males in Asian countries patronise FSW4 5 and there are around 4–6 million FSW in China.6 Furthermore, prevalence of inconsistent condom use among FSW is high in China7 and in Asia.8 Effective programmes to promote condom use among FSW are highly warranted.

Evaluation of the overall effectiveness of programmes targeting FSW at the community level has been reported in a number of countries9 10 and Behavioural Surveillance Surveys (BSS) has been used for the purpose.11 12 Such surveys have been implemented in countries such as the USA, Thailand and Vietnam.1315

In 2005, Sichuan ranked sixth among all Chinese provinces in the number of reported HIV cases.16 The HIV prevalence among FSW in some parts of Sichuan may be quite high—for example, 4% in Dazhou in 2004 (personal communication, Director of the Dazhou Center for Disease Control and Prevention (CDC), China, 18 July 2006). During 2000–2005, the China-UK HIV/AIDS Prevention and Care Project (China-UK Project) spent about 20 million pounds and served 54 000 people in Sichuan through 90 interventions. The Project also supported the development of BSS in Sichuan and Yunnan.17

Funding and technical support were provided by the UK’s Department for International Development and Family Health International (FHI). Intervention activities targeting FSW included mobilisation of community resources (for example, family planning workers and officials), seminars and training for FSW and keepers of entertainment establishments, publicity events and advertisements via different mass media, outreach education provided by trained peer educators, voluntary counselling and testing (VCT) services, free sexually transmitted disease checkups and support services, social marketing of condoms and so forth.

Using BSS data on FSW, this study compared prevalence of condom use and other HIV-related behaviours obtained from 2003–2005 in 19 China-UK Project sites in Sichuan to detect whether any significant improvements had occurred. The null hypothesis that there were no associations between service coverage rates and improvement observed in particular sites over time was also tested.


Study design

The methodology of the BSS for FSW in Sichuan has been documented in detail in a China-UK HIV/AIDS Project report.18 Another study documents the implementation of BSS in a particular site.12 BSS on FSW were conducted in the same 19 sites in Sichuan in 2003 (n = 7068), 2004 (n = 6875) and 2005 (n = 6833) using the same survey methodology.1820 A panel of international and national experts translated and modified the BSS questionnaire developed by the FHI, which has been used in some other Asian countries.14 Quality of data collection was monitored by some provincial, national and international panels of experts.

In the mapping exercises, peer workers identified geographic locations of sex work establishments, and field visits were made to estimate the number of workers present in these establishments. A number of such sex work establishments (ranging from 5–30 establishments per site) were randomly selected (probability proportional to number of sex workers). According to the national BSS guideline,21 a rough sample size of 400 per site per year was targeted. It was distributed to the individual establishments, roughly proportional to the estimated number of FSW working there. Convenience sampling method was used to recruit study respondents on site; face to face interviews using a structured questionnaire were administered in settings where privacy was ensured. Trained fieldworkers, who were medical staff of the CDC, provided assurance that participation was absolutely voluntary, anonymous and confidential. Verbal informed consent was obtained and the fieldworker signed a form pledging that he/she had explained the information clearly to the respondents. A cash incentive of RMB50 (about US$6) was given to the participant.


Information about respondents’ sociodemographic background, age the first time they had sex and duration of sex work were collected. Five condom use indicators (CB1 to CB5), three other HIV-related indicators (HB1 to HB5) and five indicators related to relevant prevention services received in the last 12 months (HS1 to HS5) were used in this study (see Box 1). These parameters were used in international BSS supported by Family Health International22 and were refined by an expert panel.

Statistical methods

Summary statistics of the percentage distributions of the eight evaluative parameters, CB1–CB5 and HB1–HB3, as well as the five service coverage parameters (HS1–HS5) measured in the 19 surveillance sites in the 2003 to 2005 surveys were presented. Odds ratios (OR) comparing data measured in 2004 and 2005 versus those measured in the baseline survey (2003) were derived for each of the 13 individual parameters using separate multivariate logistic regression models adjusting for relevant background characteristics. These adjusted odds ratios (AOR) are indicators of behavioural change over time. (Heterogeneity of the adjusted odds was assessed using Cochran’s Q test,23 see Appendix.) Random effects models are recommended in face of heterogeneity;24 hence, pooled AOR and their 95% confidence interval (CI) were estimated by the DerSimonian-Laird random effects method.25 The null hypotheses that whether these pooled AOR differed from 1.0 were tested by inspecting whether respective 95% CIs included 1.0.

The null hypothesis that the 19 sites’ adjusted odds ratios for CB1–CB5 and HB1–HB3 (comparing data of 2004 vs 2003 and 2005 vs 2003, which can be seen as indicators of behavioural changes) were not associated with the service coverage rate indicators HS1–HS5 for these sites (estimated by the area under the curve of the coverage rates of the respective years) was tested using Spearman correlation coefficients. The individual sites of particular years (2004 vs 2003 and 2005 vs 2003) were therefore used as units of analyses (n = 38).

All the statistical analyses were performed using SPSS 14.0 (SPSS Inc, Chicago, Illinois, USA) and the freeware package RevMan 4.2.26


Background characteristics

Of all respondents, 67.4% were of age 24 years or less: 81.3% attained junior high or lower level of education; 95.2% were of Han ethnicity; 44.4% worked in clubs or saunas or karaoke bars and 37.4% worked in “hair salons”; 58.5% had engaged in sex work for 1 year or less. Across the 3 years these characteristics were comparable (table 1).

Table 1 Background characteristics of respondents

Prevalence of condom use behaviours and HIV-related behaviours in 2005 and 2004 versus baseline data (2003)

The mean prevalence of condom use for the 19 sites in 2005 and 2004 versus those obtained from the baseline surveys (2003) were compared in table 2: condom use in the last episode of sex with her regular sex partner (CB1: 45.8% and 36.5% vs 38.1%), consistent condom use with her regular sex partner in the last month (CB2: 21.5% and 16.1% vs 15.9%), last episode condom use with client (CB3: 87.8% and 81.9% vs 81.2%), consistent condom use with clients in the last month (CB4: 52.7% and 40.0% vs 40.9%) and possession of condoms when being interviewed (CB5: 88.4% and 82.6% vs 75.5%).

Table 2 Comparing prevalence of condom use behaviours, HIV-related knowledge/behaviour and service coverage rates in 2003 versus 2004 and 2005 surveys

The pooled AOR comparing 2004 and 2003 data ranged from 0.95 to 1.05 for CB1 to CB4 (all these respective 95% CI included 1.0) and was 1.81 for CB5 (all these 95% CI excluded 1.0, table 2). Similar pooled AOR comparing 2005 and 2003 data ranged from 1.42 to 3.29 for CB1 to CB5 and their respective 95% CI all excluded 1.0 (table 2).

Comparing 2005 and 2004 data versus the baseline data (2003) in the 19 sites, it is seen that on average 82.6% (2005) and 74% (2004) versus 65.8% (2003) of the respondents in the 19 sites gave at least four appropriate responses to the five HIV-related knowledge questions asked (HB1); 29.8% and 20.1% versus 15.0% had received VCT (HB2); 37.8% and 37% versus 35.1% self reported ever having some sexually transmitted disease (STD) symptoms (HB3) (table 2). The 95% CI of the pooled AOR summarising the individual OR of the 19 sites for HB1 and HB2 excluded 1.0 (1.63 and 1.32, respectively, in 2004; 2.94 and 2.43, respectively, in 2005); however, this was not the case for HB3 (table 2).

Service coverage rates

The baseline mean 2003 coverage rates of HS1–HS4 (HIV-related counselling services, general STD/HIV-related services, free condom/lubricant distribution, and free needle exchange services, respectively) for the 19 sites were 13.2%, 24.7%, 22.7% and 1.1%, which became 39.3%, 54.4%, 52.8% and 2.2% in 2004 and 58.0%, 75.8%, 71.1% and 3.7% in 2005 (table 2). The respondents of the 19 sites had on average received 0.94 types of the studied services in 2003 compared with 2.57 in 2004 and 3.50 in 2005. All the 95% CIs of the pooled AOR comparing service coverage rates in 2005 and 2004 versus 2003 for HS1 to HS3 excluded 1.0 (table 2).

Associations between changes in HIV-related behaviours (AOR) and service coverage rates

A total of 38 pairs of AOR and corresponding service coverage rates (19 for 2005 and 19 for 2004) were used to calculate Spearman correlation coefficients; hence, HS1 to HS5 may be considered as exposure variables whereas other variables may be considered as outcome variables. Except HB3, all the Spearman correlation coefficients between the AOR for the other seven evaluative parameters (CB1–CB5, HB1 and HB2) and relevant coverage rates (HS1–HS3 and HS5) were of statistical significance (ranged from 0.35–0.67, p<0.05,); however, HS4 was associated with CB1–CB3 but not CB4, CB5, HB1 and HB2. AOR related to HB3 were not associated with HS1–HS5. HIV-related knowledge (that is, HB1) was significantly associated with the AOR related to the other six evaluation parameters: CB1–CB5 and HB2 (Spearman correlation coefficients ranged from 0.32–0.71, p<0.05) (data not tabulated).


The baseline surveys reported prevalent unprotected sex with clients and with regular sex partners among the FSW studied. Relevant pooled AOR (2005 vs 2003) that were statistically significant from 1.0 (that is, the 95% CIs excluded 1.0) were observed in many though not all of the 19 study sites. Some improvements over condom use may therefore have occurred in these sites; however, the risk for HIV transmission through heterosexual commercial sex in 2005 was still substantial in Sichuan. The high prevalence of inconsistent condom use with clients could not be solely explained by non-availability of condoms as almost 80% of the respondents were carrying a condom when being interviewed in 2004 and 2005.

A growing number of respondents had received some HIV/STD prevention services.

Apparent improvements in evaluative parameters (AOR of CB and HB indicators) were also strongly associated with various service coverage rates and the number of types of services used. Improvements in HIV-related risk behaviours may at least partially be attributed to the improved service coverage rates.

Some studies used very similar BSS approaches to determine whether behavioural change has occurred in specific populations.27 28 These reports did not include detailed service utilisation data and were unable to correlate behavioural change with service coverage rates. A comprehensive set of parameters were used in this study, which therefore offers a new approach and adds to the literature of evaluating behavioural change in specific populations over time using BSS data.

Other observations were made. In 2005, only about 30% of all respondents on average had ever received VCT, which is an effective means of reducing unprotected sex;29 hence, there is much room for improvement. Some improvements in HIV-related knowledge were observed over time and the pooled AOR were correlated with condom use and other HIV-related behaviours. The significance of such correlations has, however, been mixed in the literature.30 Over the study period, the pooled AOR (2005 vs 2003) for ever having STD symptoms decreased in some sites but increased in others. Self reported, life long, STD prevalence data may not be sensitive enough to detect behavioural changes.

The study has some limitations. Onsite random sampling was not feasible. Although the interviews were strictly anonymous and privacy was ensured, reporting bias because of social desirability may exist as in all BSS studies. Associations rather than causal relationships were reported because this was not a cohort study. Data on the number of sex workers that refused to join the survey were not available and response rates were not determined. Since data on length of stay in the city were not measured, it is not clear how mobility affected exposure to services and the studied relationships between service coverage and outcomes. An assumption of independent samples across years was made in the data analyses. As part of an international collaborative project, the surveys were of high quality, having been conducted with careful planning, implementation, documentation and quality control. The project was fully supported by both the China CDC and the FHI.

In summary, the study demonstrates that BSS data could be used to investigate time trends of HIV-related behaviours and to evaluate the overall impact of various service programmes. The HIV-related prevention programmes of the China-UK Project have had an impact in promoting condom use among FSW in Sichuan, but the prevalence of unprotected sex among FSW remained high. These evaluations can be used for reporting to funding agents, for feedback to frontline workers, as well as for policy advocacy.

Key messages

  • Overall improvements in HIV-related knowledge and behaviours (for example, condom use with different types of sex partners, voluntary testing or counselling (VCT)) among female sex workers were detected in the 19 China-UK Project sites in Sichuan during 2003–2005 based on results of Behavioural Surveillance Surveys (BSS).

  • There was, however, still room for improvement as prevalence of inconsistent condom use with clients was still relatively high whereas prevalence of ever receiving VCT was relatively low in 2005.

  • Service coverage rates have been increasing over time in these sites and they are significantly associated with improvements in HIV-related knowledge and behaviours.

  • This paper illustrates how BSS data could be used to detect overall behavioural changes in specific populations and in multiple sites. The results should be useful for programming, policy formation and advocacy purposes.

Box 1: Description of studied parameters in three domains

Domain 1: Condom use evaluative indicators
  • CB1: whether a condom was used in the last episode of sex with her regular sex partner

  • CB2: whether condoms were consistently used with her regular sex partner in the last month

  • CB3: whether a condom was used in the last episode of sex with her sex work client

  • CB4: whether condoms were consistently used with sex work clients in the last month

  • CB5: whether possessing a condom(s) when being interviewed

Domain 2: Other HIV-related evaluative indicators
  • HB1: whether giving four or more appropriate responses to the five items asked of HIV-related knowledge, including “one could tell whether someone had contracted HIV by his/her appearance”, “HIV could be transmitted via dining with HIV-infected persons”, “HIV could be transmitted via handshaking with HIV-infected persons”, “HIV could be transmitted via mosquito bites” and “HIV positive pregnant women could transmit HIV to their infant”

  • HB2: whether ever received HIV voluntary counselling and testing (VCT)

  • HB3: self reported ever having STD symptoms

Domain 3: Service coverage rate indicators (ever received HIV-related prevention services in the last 12 months)
  • HS1: detailed individual-based HIV-related counselling services delivered by clinicians, outreach health educators or peer educators etc

  • HS2: general STD/HIV-related services, including lectures or workshops conducted in the Center for Diseas Control or sex service establishments such as karaoke bars or hair salons and distribution of information pamphlets by outreach workers in these settings

  • HS3: periodic free condom/lubricant distribution to female sex workers in different occasions

  • HS4: free needle exchange services that are run by trained injecting drug user peer educators who meet with their peers on a weekly basis to collect used needles and to distribute free new needles to them. Use of methadone maintenance treatment services on a daily basis in the methadone clinic at the cost of 10 Yuans (1.25 US$) per day

  • HS5: total number of types of the aforementioned services received


The authors would like to thank all respondents and staff of the China-UK HIV/AIDS Prevention and Care Project and CDC of Sichuan. Thanks are extended to Dr J H Kim for her help in early drafts of the manuscript. The study was supported by the UK’s Department for International Development, Family Health International, and the China-UK HIV/AIDS Prevention and Care Project.



  • Contributors: JTFL is responsible for the concept, write up and finalisation of the paper. KCC is responsible for data analyses. HYT and JG assisted in drafting and editing the paper. LZ, JZ, YL, YZ and FC were in charge of data collection and quality control of the study. NW took part in the concept, review and editing of the paper.

  • Funding: The study was supported by the UK’s Department for International Development, Family Health International, and the China-UK HIV/AIDS Prevention and Care Project.

  • Competing interests: None.

  • Ethics approval: Ethics approval was obtained from the National Center for Disease Control in China.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.