Objectives Decisions to use condoms are made within partnerships. We examined the associations between inconsistent or no condom use and individual and partnership characteristics. We also examined the relative importance of individual versus partnership factors.
Methods Cross-sectional study of heterosexual individuals enrolled from the sexually transmitted infections (STI) outpatient clinic in Amsterdam, the Netherlands, from May to August 2010. Participants completed a questionnaire about sexual behaviour with the last four partners in the preceding year. Participant and partnership factors associated with inconsistent or no condom use in steady and casual partnerships were identified.
Results 2144 individuals were included, reporting 6401 partnerships; 54.7% were female, the median age was 25 (IQR 22–30) years and 79.9% were Dutch. Inconsistent or no condom use occurred in 86.1% of 2387 steady partnerships and in 66.5% of 4014 casual partnerships. There was statistical evidence of associations between inconsistent condom use in steady partnerships and ethnic concordance, longer duration, higher number of sex acts, practising anal sex, and sex-related drug use. In casual partnerships, associations were found with having an older partner, ethnic concordance, longer duration, higher number of sex acts, anal sex, sex-related drug use, ongoing partnerships and concurrency. In multivariable models, partnership factors explained 50.9% of the variance in steady partnerships and 70.1% in casual partnerships compared with 10.5% and 15.4% respectively for individual factors.
Conclusions Among heterosexual STI clinic attendees in Amsterdam, partnership factors are more important factors related with inconsistent condom use than characteristics of the individual.
- Sexually transmitted diseases
- Unsafe sex
Statistics from Altmetric.com
Inconsistent condom use is common among heterosexuals1–3 and increases the risk of sexually transmitted infections (STI).4 The decision to use condoms is based on perceived risk and beliefs5 ,6 and also perceived willingness of the partner.1 Individuals who use condoms inconsistently have been reported to be of older age, non-white ethnicity and lower educational level, and to report an increased number of lifetime partners, younger age at first sex, more frequent sexual contact and drug use.1 ,7 ,8 In recent years, the importance of partnership characteristics has become increasingly recognised. For example, inconsistent condom use is associated with longer partnership duration3 and with partnerships in which women are older.9
Previous studies have tended to focus on either individual or partnership factors associated with condom use or its absence, but their relative importance is not known. In addition, while it is well established that condoms are more often used in new or casual than in longstanding partnerships, it is not known whether different factors are associated with condom use in steady partnerships and casual partnerships.3 ,10 The objectives of this study were to examine associations between inconsistent condom use in steady and casual heterosexual partnerships, and to examine the relative importance of individual versus partnership factors in steady and casual partnerships.
Materials and methods
Study population and setting
The study population was recruited from the STI clinic in Amsterdam, the Netherlands, where clients are tested and treated according to the clinic protocol.8 ,11 Between May and August 2010, heterosexuals were invited to participate in a cross-sectional study until approximately 1000 men and 1000 women were included if they visited the clinic for a new consultation, were at least 16 years old, could understand written Dutch or English and gave written informed consent. Men who had sex with men were excluded. The study was approved by the medical ethics committee of the Academic Medical Center.
Participants completed an online questionnaire at the clinic, asking about their sociodemographics and sexual behaviour. The questionnaire also included questions about sexual behaviour in self-defined steady or casual partnerships with up to four partners in the year preceding study participation and questions about partners’ demographics. No a priori definition of partnership status (ie, steady or casual) was provided to allow participants to subjectively determine partnership status.
The analyses were done at the partnership level: participants were represented with up to four partnerships. The outcome measure was inconsistent condom use within a partnership, defined as vaginal or anal sex during which condoms were not always or never used. The analysis was stratified by self-defined partner type (ie, steady or casual). Possible associated factors were selected a priori based on prior knowledge and examined in univariable and multivariable analyses; no further selection methods, such as backward or forward selections, were used. The selected participant factors were gender, age, ethnicity, education and number of partners in the preceding year. Education was categorised into lower (ie, similar to elementary school, high school or vocational training) and higher education (ie, similar to college or university). Ethnicity was based on country of birth and country of birth of the parents; if at least one parent was born in a country other than the Netherlands, the participant was regarded as belonging to a non-Dutch ethnic group. Ethnicity was then categorised into Dutch, Surinamese/Antillean or other.
Eight partnership factors were selected (ie, age difference, ethnic mixing, duration, number of sex acts, anal sex, drug use, whether the partnership was ongoing and concurrency). Age difference in years was calculated by subtracting partner's age from the participant's age. Drug use was defined as recreational use of XTC (3,4-methyleendioxymethamfetamine), GHB (γ-hydroxybutyric acid), (meth)-amphetamines, cocaine, poppers (amyl nitrate) or sildefanil shortly before or during sex. The questionnaire enquired about the month and year of first and last sex; it did not ask for the exact dates as it seemed unlikely participants would remember those accurately. Also, participants were allowed to skip these questions so as to avoid pure guessing. To calculate partnership duration, we assumed that partnerships started and ended on the 15th day of the month and subtracted the first date of sex from the last date. Partnerships were chronologically ordered by date of first sex to determine whether a previous partnership was ongoing at the time the current partnership was initiated. The data available allowed concurrency to be categorised as follows: (1) definitely concurrent, when the date of first sex within a partnership was more than a month before the end of the last sex date of the previous partnership; (2) unknown, when the date of first sex and the date of last sex of the previous partnership were in the same month; (3) and definitely non-concurrent, when the date of first sex was more than a month after the end of the last sex date of the previous partnership. We assumed that behaviour within a partnership is only determined by events or partnerships in the past or present, not by future partnerships. Because only information on a maximum of four partnerships was available, concurrency was missing for most first described partnerships (n=2071), except when it was the first lifetime partnership of a participant (n=73); in the latter case, the partnership was definitely non-concurrent. From 2045 non-first partnerships, the value of concurrency could not be estimated; this occurred if either the date of last sex of the previous partnership was missing or if the chronological order of all partnerships of a participant could not be determined.
Missing data for concurrency, drug use (n=7, 0.11%), anal sex (n=5, 0.08%), number of sex acts (n=3, 0.05%), ongoing partnership (n=2, 0.03%) and duration (n=1539, 24%) were imputed using multivariate imputation (MI) by chained equations.12 The MI prediction model included all variables that were selected for the main analysis and the variables ‘cohabiting’ and ‘number of lifetime partners’. Twenty imputed data sets were created. For every data set, a logistic regression model was fitted using generalised estimating equations to account for correlated outcomes within participants. Results from the 20 fits were pooled. Interactions between gender and other covariates were checked. Interaction terms were added if there was statistical evidence of interaction. Effects of continuous variables age, number of partners, age difference and partnership duration were modelled via restricted cubic splines13 with knots on the 2.5th, 25th, 50th, 75th and 97.5th centiles. Results were compared with results from complete case analysis (see online supplement). Diagnostic tests and sensitivity analysis were done to check the plausibility and robustness of the imputations (see online supplement).
To estimate the relative importance of participant factors and partnership factors, we used the scaled Brier score, a measure of explained variance (ie, between 0 and 100%).14 ,15 The Brier score is the squared difference between the predicted probability of the outcome and the observed outcome, averaged over the individuals. In the scaled Brier score, this number is scaled by its score under a non-informative model, that is, a model that includes an intercept but no predictors. We fitted the models for steady and casual partnerships separately for individual and partnership factors. Because the scaled Brier score does not account for clustering, we randomly selected one partnership per participant and estimated the score on a data set with one observation per participant.
During the study period, 5120 eligible clinic visitors were invited to participate, of whom 2200 (43.0%) gave informed consent. When compared with eligible but non-participating individuals, participants were more often of Dutch nationality (91.6% vs 82.9%) and more likely to have symptoms (34.8% vs 28.9%). Age, the number of partners in the previous 6 months and STI prevalences were comparable between participants and non-participants.
Participants were excluded from further analysis if neither vaginal nor anal sex occurred with any of the partners. Specific partnerships were excluded if neither vaginal nor anal sex was reported. In total, 2144 individuals reporting 6401 partnerships were included. They reported a median of 4 (IQR 2–6) partners in the preceding year; the median number of steady partners was 1 (IQR 1–2) and of casual partners 2 (IQR 1–5). Women represented 54.7% of the study population. They were younger than men (p<0.001), had more often received higher education (p<0.001) and reported fewer partners (p<0.001). Distributions according to ethnicity, condom use during vaginal sex and/or STI diagnoses were comparable.
Participants described 2387 steady and 4014 casual partnerships. On average, casual partnership duration was shorter than steady partnership duration, and the number of sex acts was higher in steady partnerships (table 1). Inconsistent condom use was reported in 2056 (86.1%) steady partnerships; in 962 (46.8%) of these partnerships, condoms were never used. Inconsistent condom use was reported in 2671 (66.5%) casual partnerships; in 1354 (50.7%) of these partnerships, condoms were never used. In univariable analysis, the OR for inconsistent condom use within steady partnerships compared with casual partnerships was 3.12 (95% CI 2.73 to 3.57).
In univariable analysis, low educational level was the only participant factor associated with inconsistent condom use (table 2). All partnership factors, except whether the partnership was ongoing, were associated with inconsistent condom use (table 2). In multivariable analysis, we found statistical evidence of an interaction between gender and ethnicity of the participant (p=0.012). Dutch women were more likely to report inconsistent condom use than Dutch men. Men of non-Dutch origin were most likely to report inconsistent condom use, but the opposite effect was found in women.
At the partnership level, most associations with inconsistent condom use remained in multivariable analysis. The associations with anal sex and number of sex acts were attenuated. There was no longer statistical evidence of an association for those in definite concurrent partnerships or age difference (table 2). The scaled Brier score of individual factors was 10.5%, and the explained variance of a model with only participant characteristics was 50.9%.
Of the individual participant characteristics, age was the only variable for which there was strong statistical evidence of an association with inconsistent condom use in univariable analysis. There was strong statistical evidence of associations between all partnership factors, except concurrency and inconsistent condom use (table 3). In multivariable analysis, the association with age of the individual was attenuated (p=0.097), but the trend was the same, with an increase in the adjusted OR (aOR) for inconsistent condom use for those aged 16–22. Low education increased the risk of inconsistent condom use (aOR 1.33; 95% CI 1.09 to 1.62). The likelihood of inconsistent condom use decreased when the number of partners of the individual participant increased from one to five, while from five partners onwards the aOR remained constant (overall effect p=0.020). All associations between partnership characteristics and inconsistent condom use remained. The strength of association with age difference was attenuated, but inconsistent condom use was still less likely when the partner was much older (p=0.022) (table 3). Within a casual partnership, men were generally older than women; median age difference reported by men was 2 (IQR 0–5) years, while the age difference reported by women was −1 (IQR −4 to 1) year (p<0.001), but the effect of age difference on inconsistent condom use did not differ significantly for men and women. Concordant ethnicity was associated with inconsistent condom use (aOR 1.33; 95% CI 1.14 to 1.56). Inconsistent condom use was also more common in longer partnerships (p=0.003). A higher number of sex acts in the preceding year, anal sex and sex-related drug use were associated with inconsistent condom use (p<0.001) (table 3). Inconsistent condom use was more often practised in casual partnerships that were still ongoing at the time of inclusion (aOR 1.42; 95% CI 1.10 to 1.82), while consistent condom use was more often practised in partnerships that started as concurrent partnerships (aOR 0.87; 95% CI 0.75 to 1.02). Individual-level factors explained 15.4% of the variance and partnership factors explained 70.1%.
In sensitivity analysis, aORs and CIs compared with those obtained after MI hardly changed. In complete case analysis, 710 steady partnerships and 1607 casual partnerships were included; inconsistent condom use was reported in 83.0% and 67.1%, respectively. Overall, CIs became wider, but the direction of the effects and the effect sizes did not change much from those based on MI data sets. The results of complete case and sensitivity analysis are shown in the online supplement.
Among heterosexual STI clinic attendees in Amsterdam, we found that partnership factors were more important factors of inconsistent condom use than individual factors. The strongest factors associated with inconsistent condom use in steady and casual partnerships were longer partnership duration, an increased number of sex acts with a partner and practising anal sex. Inconsistent condom use was also more often reported in partnerships with concordant ethnicity and in casual partnerships in which the reported partner was older.
The strengths of this study, conducted at the largest STI clinic in the Netherlands, are the large number of partnerships described by participants and detailed information that was collected. Information was available for up to four partners per participant and included many characteristics, enabling us to study partnership characteristics and mixing by age and ethnicity. We regard the study population as a good reflection of the total eligible clinic visitor population as there were no age differences or differences in sexual risk behaviour and STI prevalence between participants and non-participants. Participants were more often Dutch than non-participants, which was probably due to the fact that the questionnaire was only available in Dutch and English. STI clinic visitors are, however, a high-risk population, so extrapolation of the results to the general heterosexual population might not be possible.
We used a self-defined partnership definition (ie, steady or casual) to allow participants to subjectively determine partnership status. The advantage of using this subjective definition is that it is in accordance with the perception of the participants when deciding to use condoms or not. On the other hand, not everyone uses the same definition, which might introduce bias. For example, it is not inconceivable that women consider partnerships more often steady than men.
The most important study weakness was missing data, especially in variables measuring duration and timing of partnerships. Missing data arose because questions about first and last sex dates with each partner were not obligatory. To prevent introduction of bias we used MI.22 Based on the results of diagnostic tests and sensitivity analyses, we are confident that MI did not introduce bias. The definition of the primary outcome, inconsistent condom use, might be considered a limitation. In approximately 50% of the partnerships with inconsistent condom use, condoms were never used during sexual contact, while the remaining 50% had not always used condoms. We used this definition because one unprotected sex act can be enough to transmit an STI. This definition has been used in other studies23 ,24 and provides insight into risk behaviour during the course of the partnership. Other studies examined condom use at first and/or last sex,3 ,7 ,25 which reduces errors resulting from recall bias. Recall bias might have affected our analysis because questions were asked about self-reported behaviour in the year preceding participation and about partners’ details, which might have been forgotten. It is unknown to what extent participants reported their partners’ characteristics accurately, especially in very short partnerships.
We included the four most recent partnerships in the preceding year, while 46.5% of the participants had more partners in the period than they described in detail, of whom 81.1% had just one extra partner. This sampling strategy might cause bias because information of the other partnerships is missing and longer partnerships are more likely to be included than shorter partnerships.7
To estimate the contribution of individual and partnership factors, we calculated scaled Brier scores using the same data set that was used in model construction. This may lead to overoptimism in model performance. However, the difference in explained variation between individual-level factors and partnership factors was very large and conclusions will not be affected.
This study shows that stratification by partner type (ie, steady or casual) was justified in views of the strong associations with patterns of condom use, partnership duration and the number of sex acts. We found that within casual partnerships men were generally older than women and that condom use in partnerships in which the partner was older was more often inconsistent. This association was only found in casual partnerships and not in steady partnership, suggesting that differences become less important when familiarity between partners increases. This effect was already known for men,9 and we found the same effect for women.
Inconsistent condom use was more common in partnerships in which anal sex was practised. There might be various reasons for heterosexuals to have anal sex. It might be used as strategy to preserve female virginity; it prevents unwanted pregnancy; and it might be more adventurous and thus be a marker of risky behaviour. We also showed that in a population that was predominantly Dutch, concordant ethnicity within a partnership was associated with inconsistent condom use. This association has been shown for ethnic minorities in Amsterdam, but not yet for Dutch individuals.23 ,24 In the Netherlands, Surinamese and Antillean individuals have a higher chlamydia risk than Dutch individuals.26 ,27 Ethnic mixing patterns might help to maintain differences in chlamydia prevalence, especially because inconsistent condom use was more common in assortative partnerships.28
For future research, we suggest that ideally studies include both members of a partnership to gain additional insights into ‘social’ structures, for example, mixing patterns by age and ethnic group, which contribute to decision making about condom use. Increased understanding of individual and partnership factors associated with inconsistent condom use could contribute to more tailored and effective STI prevention campaigns to promote condom use.
In conclusion, we have shown that partnership factors, such as duration, the number of sex acts and mixing by age and ethnicity, are more important factors associated with inconsistent condom use than characteristics of the individual participants and are important to consider when studying sexual risk behaviour in heterosexuals.
Partnership factors are more important factors associated with inconsistent condom use than characteristics of the individual.
Partnership factors, such as duration, the number of sex acts and mixing by age and ethnicity, are important factors related with inconsistent condom use in heterosexual partnerships.
In future research, more emphasis on partnership characteristics is needed.
Studies that include both members of a partnership are recommended to gain additional insights into social structures (eg, mixing patterns).
We would like to thank the participants and M. van Rooijen for the data management and N. Wensink for data collection. We also thank two anonymous reviewers and the editors for their useful comments. This study was part of the ‘Academic Collaborative of the Public Health Service of the City of Amsterdam (GGD) and the Academic Medical Center at the University of Amsterdam (AMC); network for infectious diseases’.
Abstract in Dutch
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Abstract in Dutch - Online abstract
Short summary: Among heterosexual STI clinic attendees in Amsterdam, the Netherlands, absent or inconsistent condom use is associated with partnership characteristics. The associations between partnership factors and condom use were stronger than the associations between individual factors and condom use.
Correction notice A sentence under result section in Abstract has been corrected since published Online First.
Handling editor Jackie A Cassell
Contributors AM, MH, RG, TH, MK and MSvdL contributed to the design of the study. AM, RG and MSvdL performed the analyses. All authors contributed to the interpretation of the results, writing the manuscript and gave intellectual feedback. All authors have contributed to, have seen and approved the final submitted version of the manuscript.
Funding This work was supported by the Netherlands Organisation for Health and Research (ZonMw) (grant number 71150001).
Competing interests None.
Ethics approval Medical ethics committee of the Academic Medical Center in Amsterdam, the Netherlands.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement At present the data used in this analysis are not freely available because further analyses are underway.
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.