Article Text

Download PDFPDF

Studying complex interactions among determinants of healthcare-seeking behaviours: self-medication for sexually transmitted infection symptoms in female sex workers
  1. Gabriela B Gomez1,
  2. Pablo E Campos2,
  3. Clara Buendia2,
  4. Cesar P Carcamo2,
  5. Patricia J Garcia2,
  6. Patricia Segura3,
  7. William L Whittington4,
  8. James P Hughes5,
  9. Helen Ward1,
  10. Geoffrey P Garnett1,
  11. King K Holmes4,6,7
  1. 1Department of Infectious Disease Epidemiology, Imperial College London, London, UK
  2. 2Epidemiology, STI and HIV/AIDS Unit, Universidad Peruana Cayetano Heredia, Lima, Perú
  3. 3INMENSA, Lima, Perú
  4. 4Department of Medicine, University of Washington, Seattle, Washington, USA
  5. 5Department of Biostatistics, University of Washington, Seattle, Washington, USA
  6. 6Centre for AIDS and STD, Seattle, Washington, USA
  7. 7Department of Global Health, University of Washington, Seattle, Washington, USA
  1. Correspondence to Dr Gabriela B Gomez, Institute for Global Health, Imperial College London, South Kensington Campus, 15 Prince's Gardens, London SW7 1NA, UK; g.gomez{at}


Objectives To describe the frequency and determinants of self-medication for symptoms of sexually transmitted infections (STI) in a female sex worker (FSW) population. To present a methodology exploring the best predictors as well as the interactions between determinants of self-medication.

Methods A cross-sectional survey of 4153 FSW carried out in Peru. The prevalence of self-medication was estimated from the subsample of participants who had experienced symptoms of STI in the past 12 months (n=1601), and used successive logistic regression models to explore the determinants.

Results Self-medication prevalence for a reported symptomatic episode during the past 12 months was 32.1% (95% CI 29.8 to 34.6). It was negatively correlated with work in brothels (adjusted odds ratio (OR) 0.51, 95% CI 0.28 to 0.93; p=0.028) and awareness of STI services available for FSW (adjusted OR 0.49, 95% CI 0.29 to 0.81; p=0.006). Other determinants were organised at different levels of proximity to the outcome creating pathways leading to self-medication.

Conclusions The importance of the staggered analysis presented in this study resides in its potential to improve the understanding of associations between determinants and, consequently, the targeting of interventions. The awareness of STI services available for FSW increases access to health care, which in turn decreases self-medication. In addition, the sharing of information that takes place between brothel-based FSW was also related to a diminishing prevalence of self-medication. These two main predictors provide an opportunity for prevention programmes, in particular those designed to be led by peers.

  • Epidemiology
  • female sex worker
  • Peru
  • prostitution
  • self-medication
  • self-treatment
  • sexually transmitted infections
  • social epidemiology

Statistics from

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.

Self-medication is the choice to medicate oneself in response to symptoms without a clinician's diagnosis or prescription. It has consequences for the potential treatment success for the individual and the potential control of the spread of infection within the population.1–3 It is reported more frequently in patients with symptoms of sexually transmitted infections (STI) than in those with a general health problem.4–6 However, the prevalence of self-medication for STI symptoms varies broadly across populations and over time, ranging from 7.1% to 74.5% in earlier reports.7

The choice to self-medicate will be motivated by various factors related to the severity of symptoms, previous infections and treatment, information about the disease, availability of health care, income and cultural background. To be able to promote the best option of healthcare-seeking behaviour, it is important to understand why the choice to self-medicate is made, which are the main predictors, and how they relate to one another.

For this, we can explore individual and societal dimensions influencing self-medication and the expected direction of effects. A previous review found no clear determinants of self-medication at the population level, other than a lower rate for more recently published studies.7 Therefore, in this study we focused on individual-level determinants that can be explored in regression models. This statistical method will identify associations (causal and confounding), but our interpretation depends upon the creation and testing of hypotheses. In this study, we control for variables following a set of hypotheses to help distinguish between those measuring the same effect (an intervention targeted on one will also have an effect on the other), variables lying on the causal pathway (intervening over these would have an effect on self-medication), and confounding variables (intervening over these would have no influence on self-medication).


Subjects and data collection

We analysed data collected within the Urban Community Randomised Trial for Prevention of STI (PREVEN),8 which evaluates the impact of a multicomponent STI/HIV preventive intervention in Peru. The intervention targeted female sex workers (FSW) and the general population of young adults. The general population arm of the intervention lasted 3 years (December 2003 to December 2006) and promoted healthcare-seeking for STI symptoms at either clinics or pharmacies. Pharmacy workers and clinicians were trained in the syndromic management of STI. Training began in July 2003 and was completed in late fall 2003, with an internet-based course for clinicians in 2004, and ongoing educational outreach for pharmacy workers. The FSW arm of the intervention lasted from July 2003 until December 2006. It included a mobile team visiting FSW at commercial sex venues in 8-week cycles to screen and treat for STI as well as to promote condom use and provide counselling. Counselling included information about STI, symptom recognition and services available for FSW to have regular STI check-ups. Outreach teams did not encourage FSW to go to pharmacies for treatment. The trial was evaluated through three prevalence surveys.

In December 2006, the final FSW survey was completed and provided the data for this analysis. The enrolment of approximately 200 FSW per city in 20 cities, following a time-location sampling, took place at sex work venues selected from an updated census of venues. After giving verbal informed consent and receiving STI/HIV counselling, a total of 4153 FSW agreed to participate (99.5% of those approached). A member of the local survey team administered the questionnaire face to face. It included sociodemographic, reproductive health, healthcare-seeking and sexual behavioural data. The participants then provided self-obtained vaginal swabs for nucleic acid amplification testing for Trichomonas vaginalis (TV), Chlamydia trachomatis (CT), Neisseria gonorrhoeae (NG) and blood for Treponema pallidum (TP) serology.

The PREVEN study obtained ethics approval from institutional review boards at the Universidad Peruana Cayetano Heredia, the University of Washington in Seattle and the US Navy Medical Research Center Detachment in Lima. This has been renewed every year.

Statistical methods

Data were analysed using Stata/SE 10.0. We classified participants as ‘self-medicated’ if they reported having had a symptomatic episode (vaginal discharge or genital sore or ulcer) in the past 12 months, and if the first action they took at the time upon noticing those symptoms was either seeking treatment from a private pharmacy or a traditional healer, or taking medicine they had at home or that a relative or friend gave them.

Continuous data were described by mean and standard deviation (SD), then grouped in categories and introduced as categorical variables into the models.9 We classified the department/province/district of birth into the three natural regions in Peru (coast, mountain and jungle).10 Income from the last client was calculated from data available for the last three sexual partners. Participants were considered as having a current STI if they tested positive for any of four curable STI: TV, CT, NG, or TP. A participant was considered to have active syphilis if she had a positive T pallidum haemagglutination particle agglutination assay positive and rapid plasma reagin reactive at 1:8 dilutions. Due to the nature of the intervention, we also added being recruited in an intervention city as a potential determinant of self-medication. We assessed associations between self-medication and its determinants using univariable and multivariable logistic regression models. Robust standard errors were calculated for all models to allow for intragroup correlations and the sample structure, FSW nested in cities.9 11 Variables significant at p<0.20 in the univariable analysis were retained for the multivariable models.11

Definition of determinants and logistic regression model building

There are many ways to generate a multivariable model with several possible sequences in which variables can be included or excluded. We chose to group our variables into three categories that reflect our initial assumptions. In doing so, our aim was to explore the associations within and between groups of characteristics to identify which variables are related in their effect on self-medication, and which ones predict it. At a population level, awareness of availability and access, as well as the perception of quality of health care play a role in determining self-medication. Therefore we grouped, at an individual level, variables related to sexual health and health care (diagnosis of a curable STI, knowledge of infections transmitted through sexual intercourse, awareness of STI services for FSW, and being recruited in an intervention city). General sociodemographic factors that might interact were grouped (age, education, region of origin, living alone and having an income separate from sex work), whereas sex work characteristics were considered a separate category (place of work in the past week, condom use at last sex by type of last sexual partner, number of sexual partners in the past week, age at first sex work, duration of sex work and income from last client).

Ultimately, the direction of the effect is likely to depend upon what is perceived as the best option in terms of quality, feasibility, acceptability, cost and outcome. After identifying the predictors of self-medication, we tested the fit of the model including only the predictors of the data using the Hosmer–Lemeshow goodness-of-fit test.11


A total of 1601 FSW (38.5%) reported having had a vaginal discharge or a genital sore or ulcer in the past 12 months. The mean age for this group was 25.9 years (SD 7.0) with 76.6% being less than 30 years old. All but six (0.4%) were born in Peru, and most were educated to secondary school level (75.7%). Few were married (3.6%) and the majority reported not living with a partner (61.1%).

Just over half of the FSW (55.7%) worked in bars and nightclubs. The average income from the last client was S./59.1 (SD 62.7) equivalent to US$18.2 (SD 20.3), and 35.1% reported having an income separate from sex work. The mean age of start of sex work was 21.7 years (SD 4.9) with a mean duration of sex work of 36.4 months (SD 52.2). Only 6.7% of this sample had not heard of infections transmitted through sexual intercourse. However, 22.4% reported not being aware of STI services available for FSW. Finally, 20.7% tested positive for a curable STI (TV, CT, NG, or TP).

Of 1601 FSW who reported abnormal vaginal discharge or a genital sore or ulcers during the past year, 58 (3.9%) did nothing, 957 (63.9%) sought treatment from a public or private hospital or clinic or the PREVEN mobile team, and 481 (32.1%; 95% CI 29.8 to 34.6) reported self-medicating. Of these 481, 75.7% reported going to a pharmacy for medication, 13.9% obtained medication from friends, 8.5% from a traditional healer and 1.9% used medication available at home from previous episodes.

Of the FSW who reported having had a genital ulcer or sore or vaginal discharge in the past 12 months, approximately 44% reported having sex while symptomatic (n=641/1465). Those self-medicating were more likely to have sex while symptomatic (odds ratio (OR) 1.69, 95% CI 1.34 to 2.13; p<0.001). In addition, during this symptomatic episode, 90.5% reported using condoms with clients, while 46.6% reported doing so with partners. If they were self-medicating for their reported symptoms, they were significantly less likely to use condoms with their partners (OR 0.57, 95% 0.45 to 0.72; p<0.001) and somewhat less likely to use condoms with clients (OR 0.80, 95% CI 0.54 to 1.19; p=0.274).

Determinants of self-medication

Table 1 shows the determinants associated with self-medication in univariable analyses. Self-medication was significantly more common in younger FSW, those living alone and having an income separate from sex work. It was also higher for those having a current STI, and those not knowing about STI. It was lower in those FSW who were aware of the STI services available, were brothel-based, those with a higher number of new, occasional, or regular clients and with a longer duration in sex work.

Table 1

Determinants of self-medication in FSW: univariable analyses (n=1601)

In table 2, we present the adjusted OR for determinants included in seven multivariable models.

Table 2

Determinants of self-medication in FSW: adjusted OR and 95% CI from multivariable analyses

Figure 1 illustrates the determinants of self-medication and their interactions inferred from the comparison across multivariable models. The main predictors of self-medication are placed at the centre of the figure, the next level is determinants related in their effects with other groups of characteristics. In the last level we observe the determinants whose effect is predicted by variables within the same group.

Figure 1

Interactions among determinants of self-medication. Levels of proximity to self-medication are demarcated by circles in dashed lines. Grey areas define groups of characteristics: sex work-related characteristics; sexual health and healthcare-related characteristics and sociodemographic characteristics. Being brothel-based and the awareness of sexually transmitted infection (STI) services for female sex workers (FSW) were the main predictors of self-medication, and therefore they were positioned in the centre of the figure. The effect size is shown as adjusted odds ratio (aOR) with 95% CI in the boxes to the right of the figure. In the second level of proximity to self-medication, we observe the determinants that interact with characteristics in other groups. The arrows show the determinants' interactions with other groups. In the third level, we positioned the determinants whose effect is controlled by other determinants in the same group. N, number; N/O client, new or occasional client; SW, sex work.

Sociodemographic variables

The effect of living alone as a determinant on self-medication is mediated by other sociodemographic determinants. We observe that it becomes not significant once we control for sociodemographic factors (model 1). The effect on self-medication of having an income separate from sex work is attenuated by variables such as sexual health and healthcare-related knowledge (model 3), sex work characteristics (model 5), or both (model 7) independently of other sociodemographic variables. Age was associated with a lower rate of self-medication independently of sex work characteristics or other sociodemographic variables but was associated with sexual health and healthcare-related variables.

Sexual health and healthcare-related variables

There seems to be no association between the intervention and self-medication. The intervention, which involved field visits to FSW every 2 months to talk about STI, provide STI care including bimonthly presumptive therapy with metronidazole for vaginal infection, and encourage attendance at specialised STI clinics, resulted as expected, in an increased knowledge of STI (OR 1.94, 95% CI 1.29 to 2.92; p<0.001) but not in awareness of the STI services available (OR 1.14, 95% CI 0.90 to 1.45; p=0.27). In turn, awareness of STI services remains an independent predictor of self-medication after controlling for all other characteristics (models 3 and 6).

Sex work-related variables

The associations of self-medication with having a higher number of new, occasional or regular clients, or with a higher duration in sex work became not significant after adjusting for all sex work determinants included in model 4. This was not altered when exploring sex work and sociodemographic (model 5) or sexual health and healthcare-related characteristics (model 6). Being brothel-based was independently associated with a lower rate of self-medication consistently in all models.

The multivariable logistic regression model, including only the two main predictors of self-medication (being brothel-based and awareness of STI services for FSW), fitted the data well according to the Hosmer–Lemeshow goodness-of-fit test (p=0.4004).


In this large study of FSW in 20 cities throughout Peru, we found that self-medication for STI symptoms in the past year was reported by 32%. This high prevalence of self-medication is in accordance with previous estimates.12–17 In Peru, the prevalence of self-medication for all health issues has been estimated at between 36% and 52%.18–21 For STI symptoms, Garcia et al22 reported a prevalence of 22.2% in a community sample of rural women (not FSW). Self-medication in FSW has also been reported to be common.13 23–26 However, in Peru, only one preliminary study investigated it and reported a similar level of 39.2%.27

We found self-medication to be negatively correlated with work in brothels and awareness of STI services for FSW. In Peru, these two determinants are associated with each other. Regular STI check-ups are required by brothel managers who request FSW to seek health care at local STI clinics, the Centro de Referencia de Enfermedades de Transmisión Sexual (CERETS). However, some FSW might choose not to access specific healthcare services to avoid being exposed or because it could result in withdrawal of their work permit by brothel management if an STI was found. Indeed, stigma in addition to lack of awareness of services available were the principal obstacles to accessing health care and treatment in previous reports from a FSW population in Côte d'Ivoire.16

Other studies have also reported an association between place of work and self-medication, with those FSW with a larger number of clients, such as freelance in the Philippines or brothel-based in Thailand, being more likely to use antibiotic prophylaxis or self-medication for STI symptoms.13 23 Indeed, Wong and Yilin17 found that FSW commonly share health information and practices among peers. Yet the self-medication and health advice offered was in general considered inappropriate and at times harmful. In Peru, another key factor influencing this association is a peer system promoting unusually extensive health services already in place.

Factors initially significantly associated with self-medication, such as age, living alone, other income besides sex work, number of clients, duration of sex work, current STI and knowledge of STI became not significant after adjustment. However, multivariable analyses suggested interactions among many of these variables. For instance, the place of work is also presumably associated with the income from sex work, the number of clients and with alternative sources of income separate from sex work. Increased duration in sex work could lead to more knowledge about STI and more awareness of where to access health care, thus being related to age and ultimately in their effect on self-medication.

Finally, it has been suggested that self-medication might influence condom use with clients.2 12 We found that even though self-medicating FSW in our sample were more likely to have sex while symptomatic and were less likely to use condoms with stable partners, they were not less likely to use condoms with clients. These results are in agreement with those of Todd et al15 who found, in a sample of injection drug users in Uzbekistan, that women were not less likely to use condoms if they self-medicated for STI symptoms.

Limitations of the study

We explored associations between self-medication determinants by using a succession of regression models. This staggered analysis has advantages and limitations. Whereas its importance resides in its potential to improve the understanding of risk pathways and, ultimately, the targeting of interventions, we might lose insights of associations within subgroups. However, a classic stepwise approach will not give us any insights into intermediate interactions between determinants, and the testing of each association across all determinants might put research studies at risk of multiple testing.

As with other studies that rely on self-reported behaviours, there may be a recall bias. We tried to limit this by asking about actions taken during the last symptomatic episode, and by limiting the recall period to a year. The participants were interviewed face to face, which might lead to a social desirability bias. Indeed, the finding that awareness of STI services was associated with self-medication might indicate some of this bias. It might be more acceptable to explain self-medicating by a lack of awareness of the services available. Nonetheless, the prevalence of reported self-medication was high, indicating that many FSW reported this behaviour openly. Further qualitative research would be appropriate to investigate in detail the extent of social desirability bias in these answers as well as motivations and practices of self-medication.

Although assessment of self-medication was not the primary focus of the PREVEN trial, we observed missing data for this measure only in 6.5% (n=1497/1601) of records. We did not look into self-medication use as prophylaxis. This was because self-medication questions were only asked following the report of a symptomatic event. This aspect of self-mediation is relevant in the context of sex work and has been reported frequently, in particular in Asia.12 13 23 In Peru, a study by Paris et al28 reported that 41% of FSW from the Amazonian region self-medicated with antibiotics, often to reduce their STI risk. However, the authors did not specify if the antibiotics were used for prophylaxis or in response to symptoms.

During this study, a social marketing campaign directed at the general population promoted healthcare-seeking for STI symptoms at either clinics or pharmacies, which had undergone extensive training in STI management in half of the cities. FSW in intervention cities were also advised on STI symptom recognition and the STI services available. Because FSW in these cities are also part of the general population, there is a possibility of contamination. However, this seems unlikely because after controlling for intervention, awareness of STI services available for FSW remained a predictor of self-medication but the PREVEN intervention itself was not associated with it.

Relatively simple interventions for sex workers, including disseminating information, providing access to health care and promoting condoms, have been found to be effective in many different settings.29 30 This study supports the importance of such basic healthcare interventions while highlighting the key role of a peer-led element in their dissemination strategy. The implementation of simple but effective interventions will improve the health of sex workers and contribute to a better control of STI.

Key messages

  • A high prevalence of medication use from the community was observed in a FSW population in Peru.

  • Awareness of STI services available for FSW increases access to health care, which in turn decreases self-medication.

  • Communication in brothels between FSW and managers is related to a diminishing prevalence of self-medication and to awareness of STI services.

  • Understanding the predictors of self-medication is key to the effective design of programmes to reach out and improve practices of treatment providers outside clinics.


The authors would like to thank all the participants in this study and the PREVEN study team for their efforts. Data in this manuscript were collected by the PREVEN Study.



  • Funding The PREVEN study is funded by the Wellcome Trust Foundation, (#GR-078835), National Institutes of Health CIPRA (#AI-053218) and NIH/NIAID UW STI/Topical Microbicide Cooperative Research Center (#AI-031448). GBG was funded by the Medical Research Council, UK.

  • Competing interests HW is co-editor of the journal Sexually Transmitted Infections. All other authors declare no conflicts of interest.

  • Patient consent Obtained.

  • Ethics approval The PREVEN study obtained ethics approval from institutional review boards at the Universidad Peruana Cayetano Heredia, the University of Washington in Seattle and the US Navy Medical Research Center Detachment in Lima, which has been renewed every year.

  • Provenance and peer review Not commissioned; externally peer reviewed.