Op de Coul, Eline L. M. PhD*; Götz, Hannelore M. MD, MPH, PhD†; van Bergen, Jan E. A. M. MD, MPH, PhD‡; Fennema, Johannes S. A. MD, PhD§; Hoebe, Christian J. P. A. MD, SPH, PhD¶; Koekenbier, Rik H. MSc§; Pars, Lydia L. MSc‡; van Ravesteijn, Sander M. MSc†; van der Sande, Marianne A. B. MD, PhD*; van den Broek, Ingrid V. F. PhD*
The Chlamydia Screening Implementation (CSI) for 16- to 29 year-old residents in Amsterdam, Rotterdam, and South Limburg started in April 2008. This pilot implementation was designed to assess whether annual screening within a national program can reduce the population prevalence of Chlamydia trachomatis(Ct) and thus prevent serious complications such as pelvic inflammatory disease, ectopic pregnancies, and subfertility.
Achieving adequate levels of participation and capturing high-risk groups are cornerstones for an effective large-scale screening program. Effectiveness of population-based screening is usually negatively affected by high participation among persons at low risk for chlamydia.1 In this context, obtaining insight into screening coverage, demographic and behavioral characteristics of participants and nonparticipants, and self-selection for screening of people at risk is a vital first step in the evaluation of a screening program.
During the first CSI screening round from April 2008 to February 2009, 261,025 individuals aged 16 to 29 years were invited. In total, 1758 chlamydia infections were diagnosed (overall positivity rate, 4.2%) among participants in the first screening round; 1262 among women and 496 among men.1–3 Research on the acceptability of CSI among participants and nonparticipants showed that the Internet-based screening was highly acceptable to participants.2 Furthermore, almost 70% of the nonparticipants responding to a survey questionnaire reported justified reasons not to participate because of low (perceived) risk of infection, no sexual experience, or a recent sexually transmitted infection (STI) test. Twenty percent of the nonparticipants were less appreciative of the Internet usage, home testing, or posting samples.2
The crude participation rate in CSI, with the invited population as denominator, was 16% in the first screening round. Strong associations were observed between community risk level with both participation and Ct positivity. Residents from high-risk areas had low participation rates (13%) and high Ct positivity (7.4%) compared with residents from medium- and low-risk areas (participation, 20%; positivity, 3%–4%).1–3
The objective of the current study was to examine in more detail whether we reached the right populations for screening and to clarify if Ct positivity in high-risk areas was higher in general or because of a more efficient (self-) selection for screening. In this article, we recalculated the participation rates to the sexually active target population. Furthermore, we assessed the sociodemographic and sexual behavioral risk factors determining participation and Ct positivity in the first screening round, comparing the effect of individual and community-based risk factors.
MATERIALS AND METHODS
Details of the CSI program and the evaluation design are described elsewhere.1–3 In summary, the selective, population-based screening (using Internet and home sampling kits) was carried out in Amsterdam, Rotterdam, and South Limburg. All individuals aged 16 to 29 years in those areas received an invitation letter. In Amsterdam and Rotterdam, all sexually active persons could participate, whereas in South Limburg (low population density), eligibility for screening depended on the individual's score on expected Ct risk by using an Internet questionnaire. Persons with a Ct infection were referred to the general practitioner or public health service for treatment, and they were advised to have their current partner(s) treated simultaneously. The CSI Web site (available at: www.chlamydiatest.nl) contains information on the purpose of the screening and instructions on how to collect a (urine or vaginal swab) sample and return it to the laboratory.
People were asked to fill in an online questionnaire voluntarily after they replied whether to receive a test package. Data collection included age, gender, education level, sexual history, and symptoms. Additional questions were asked in the CSI acceptability and nonresponse study using email and paper questionnaires, respectively.2 Sociodemographic data were available for all invitees from population registers. Before screening, all city neighborhoods (geographic clusters) were classified as being at low-, medium-, or high community risk based on age, ethnic, and income profiles.3 In addition, a “status score” computed by the Netherlands Institute for Social Research (SCP, available at: www.scp.nl) was used as a proxy for socioeconomic status (SES). This score takes into account the average income per household in a given postcode area as well as the percentage of households with low income, without paid job, and with low education level. Categories of ethnic groups reflect the main group of immigrants in the Netherlands based on country of birth of the person and his/her parents from the population register. Education level was based on a person's reported level of completed or ongoing schooling.
Data Analysis of Participation and Positivity
Adjustment of Participation Rates to Sexually Active Population.
“Participant” was defined as a person who returned a sample to the laboratory. Participation rates were adjusted for the sexually active population by using information on age at the first sexual intercourse from 2 national sexual health surveys (16–24 years,4 25–29 years [Rutgers Nisso Group, Sexual health in the Netherlands, unpublished data, 2009). For each age-year group, the number of invitees expected to have had sexual intercourse at that age became the new denominator. The denominator was also corrected for undeliverable invitations (about 2% of mailing was undeliverable or “returned to sender”). Positivity rates were calculated as the number of positive Ct tests divided by the number of tests performed. Participation and positivity rates were stratified by gender. Confidence intervals for proportions (Wilson 95% CI) were calculated using Statistical Analysis System (version 9.2; SAS Institute, Inc, Cary, North Carolina).
Univariate and multivariate logistic regression analyses were conducted to investigate associations between participation and sociodemographic and sexual history variables (listed in Table 1) for Amsterdam and Rotterdam, for the overall sample and by gender . Data from South Limburg, where participant selection by risk criteria was applied, was excluded. The same analyses were conducted to study correlates of Ct positivity.
Variables showing an association of P< 0.20 (Wald test, univariate analysis) were included in the multivariate analyses. Highly correlated variables (Spearman correlation coefficients) were not included in the model together (strongest associated variables selected) or combined into new variables. Backward stepwise logistic regression analyses were performed, including variables with P< 0.05 for the likelihood ratio test.
Corrected Regression Analyses.
As questionnaire response rates varied across participant and nonparticipant groups, we took into account the potential impact of “nonrandom availability” of questionnaires by constructing regression models using different approaches . Odds ratios (ORs) for this model were calculated for both the groups with and without questionnaires. Both methods gave similar results, although the correlations were slightly weaker (smaller ORs, larger CIs, not shown) when these adjustments were made.
Individual and Community Risk Factors
Multilevel models were constructed to assess the effect of the cluster (neighborhood)-based invitations. In round 1, the number of clusters was 71 in Amsterdam and 52 in Rotterdam. Population sizes of the clusters varied from 451 to 4538 inhabitants in Amsterdam and 375 to 4374 in Rotterdam.3 The variable “cluster” was added as a second level of analysis in the multivariate logistic regression model. The method describes the amount of variation between cluster populations. We calculated a median odds ratio (MOR) for both the cluster effect on participation and positivity.5
Behavioral Risk Score.
A combined variable was constructed to indicate the individual level of sexual behavioral risk based on the following variables: number of partners in the past 6 months, age at first sexual intercourse, condom use at last sexual contact, concurrent partners, STI testing, and having symptoms of an STI. These variables were selected based on availability in CSI and the national sexual health survey (Rutgers Nisso Group, Sexual health in the Netherlands, unpublished data, 2009), and being a known determinant for Ct positivity.6,7 People could score a maximum of 7 points on this variable as follows—no partnerships (0), 1 steady partner (>1 year) (0), 1 steady partner (<1 year) (1), 1 casual partner (1), ≥2 partners (2); age at first sexual contact: ≤15 years (1), ≥16 years (0); condom use at last sexual contact: no (1), yes (0 points); concurrent partners: no (0), yes (1); ever tested for an STI: no (0), yes (1); symptoms of an STI: no (0), yes (1). The score was highly predictive for Ct positivity. Persons scoring 0 or 1 points were considered at low sexual risk (corresponding with an overall Ct positivity <3.5%), 2 points were considered at medium risk (Ct %, 3.5%–4.5%), and ≥3 points were considered at high risk (Ct %, 4.5%–11%). Scores were compared between participant subgroups (gender, ethnic groups, community risk areas) and the general population aged 16 to 29 years from the national survey. All statistical analyses were performed using SAS 9.2.
Participation Rate Among the Sexually Active Population
An estimated 20% of the invitees did not participate because they had never had sexual intercourse (Table 2) (Rutgers Nisso Group, Sexual health in the Netherlands, unpublished data, 2009).5 After adjusting to the sexual active target population, the overall participation rates reached nearly 20% (women, 25%; men, 13%; Table 2). The adjusted participation rate in the youngest group (aged 16–18 years) was 1.5 to 2.5 times higher than the crude participation rate, but it still remained lower than in people aged 19 years or above. Highest participation rates were seen in women aged 26 and 27 years: 28%.
Determinants for Participation and Positivity
Factors related to participation were studied in multivariate models. The models included demographic variables from population registers for all invitees (n = 261,025) and information on sexual history for persons with questionnaire data. For 31,466 persons, questionnaires on sexual history were available; 25,084 from participants (60%) and 6382 from sexually active nonparticipants (3%–4%). Women's response to the general questionnaire was higher than that of men (OR, 2.4 [2.3–2.5]). Also, older people (20–29 years old) replied more often than did younger ones (OR, 1.5 [1.4–1.6]), and migrant populations replied less often (ORs between 0.2 and 0.7).
Individual factors most strongly associated with lower participation model were male gender, non-Dutch origin, young age, lower education level, having a long-standing steady partnership, and no history and/or symptoms of STIs (Fig. 1). Women were about twice as likely to participate as men in all subgroups of age (range: female [F], 14%–24%; male [M], 6%–13%) and ethnicity (F, 18%–29%; M, 7%–14%), except for Turkish/Moroccan women (F, 5.7%; M, 5.4%). In general, Dutch invitees of both sexes had higher participation rates than non-Dutch invitees. Sexual risk determinants for participation were similar for men and women, except for having a history of STIs or clinical symptoms, which were only significant for women.
Of the community factors, living in a medium-/high community risk area or in an area with low SES was significantly associated with lower participation (Fig. 1).
Ct positivity was most strongly associated with younger age, non-Dutch ethnic background, steady partner of non-Dutch origin, high community risk level, and lower education. These factors were explored separately for men and women (Table 1).
The highest positivity rates were observed in Surinamese/Antillean men (9.7%) and women (8.4%) and women from sub-Saharan Africa (9.2%). Having a non-Dutch partner increased the risk of being diagnosed with a Ct infection, especially in women. Ct positivity was almost 10% among non-Dutch women with a non-Dutch steady partner.
Ethnic groups with the highest positivity rates participated slightly less (14%–16%) compared with the Dutch population (21%), in which the positivity rate was the lowest. Striking, however, were the low participation rates among Turkish/Moroccan men and women (both 5.6%), whereas Ct-positivity rates were average and higher than in Dutch nationals, especially in men: 4.0% (Table 1).
High Ct-positivity rates were also found in girls aged 16 to 19 years (8.0%). With increasing age, positivity rates became more similar for men and women (25–29 years, 3%).
Higher-educated people had a far more reduced likelihood of infection than those with medium or low education (high, 3%; low/medium, ≥9%). Further, participants in general were more often highly educated than nonparticipants (71% vs. 59% among those who filled in the questionnaire).
Multilevel Analysis: Cluster Effect
Invitations were cluster randomized (at neighborhood level) so that the screening can be delivered to people within social or sexual networks.3 In the multilevel model for participation, the MOR6 for cluster effect was 1.14 (1.11–1.16, P < 0.001). This suggests that the median heterogeneity between clusters can increase the individual odds of participation 1.14 times if a person moves from one cluster to another with a higher probability of participation. In the multilevel model for Ct positivity, the MOR for cluster effect was 1.03 and not significant. Therefore, geographic clusters were not independently associated with Ct positivity.
Individual and Community Risk Levels
A great difference in participation between high- and low-risk areas was observed for women and men: 16% versus 26% and 9% versus 13%, respectively. Participation in low-risk areas was higher than in high-risk areas in all ethnic groups. Ct-positivity rates were consistently higher in high-risk areas than in low-risk areas: 7.5% versus 3.2% for women and 7.2% versus 2.6% for men, respectively (Table 1). All ethnic subgroups (including Dutch) in high-risk areas had significantly higher Ct positivity than in medium- or low-risk areas.
We compared the individual level of behavioral sexual risk between CSI participants and the sexually active general population aged 16 to 29 years in highly urbanized areas (Table 3). CSI participants were more likely to have concurrent partnerships, symptoms of an STI, and more sexual partners in the previous 6 months compared with the general population. STI testing rates were higher for the latter group. The general population used condoms less often at their last sexual contact (with steady or casual partners), but they also had a steady partner more often.
When using a scoring variable based on the behavioral variables in Table 3 (mentioned in Methods), 32% of the participants were classified as high, 27% as medium, and 41% as being at low sexual risk. In the general population (in highly urbanized areas), this was slightly different: 25% classified as high, 25% as medium, and 50% as being at low sexual risk. This variable was studied in relation to Ct positivity for various participant subgroups (age, gender, ethnicity, and community risk level). Among men, the proportion of high-risk participants was slightly higher than that among women (33% vs. 30%). Turkish/Moroccan people showed lower levels of sexual risk (high risk, 27%; Dutch, 31%; Surinam/Antillean, 34%). Young people (aged 16–19 years) showed higher levels of sexual risk (39% high risk) compared with persons aged 20 to 24 (34%) and 25 to 29 (27%) years.
In high community risk clusters, the distribution (in %) of the categories medium and high individual sexual risk did not differ significantly from medium and low community risk areas (high: 33%, 30%, 33%, respectively; medium: 26% for all). Even though patterns of sexual risk of participants were fairly similar between these areas, Ct positivity varied to a high degree (Table 4). Residents of low-risk areas who reported high levels of sexual risk had significantly lower Ct positivity than residents with similar sexual risk in high-risk areas. Highest Ct positivity was observed for Surinamese/Antilleans in high community risk areas with high individual sexual risk: 19.5% (Dutch, 11.2%; Turkish/Moroccan, 9.8%).
This article describes who was reached in the first screening round of the CSI in the Netherlands, the largest online systematic chlamydia screening program to date. We showed that the overall participation rate remained low (under 20%), also after correction for sexual activity. Groups with a sociodemographic background related to a higher chlamydia risk (such as young people, ethnic minority groups, and people with lower education) were less likely to participate and more likely to test Ct positive. We also showed that living in a medium- or high community risk area was also associated with higher Ct positivity. This effect was present in all ethnic groups including Dutch nationals. Given that reported levels of sexual risk among participants from areas of different community risk were fairly similar, this suggests that Ct prevalence in high-risk areas is higher in general and is less likely the result of different (self-) selection processes of people living in these areas. Although the overall participation rate was low, self-selection for participation on behavioral characteristics (number of sexual partners, symptoms and history of STI) was observed, also when compared with the general population, and the relatively high positivity rates suggest that people at risk for infection have been reached to a certain extent.
To obtain further insight into interrelations of behavior and sociodemographic factors, we need to study outcomes over several screening rounds. Determinants for participation are expected to change after repeated invitations because motivations for uptake may depend on previous test results, characteristics of relationships, and dynamics of sexual networks. Aral et al.8 recently described that multilevel strategies are required to understand and prevent the spread of STI, as complex systems are underlying that may evolve over time. Therefore, strategies to control chlamydia spread or to promote screening uptake could benefit from insights in these (changing) systems. We explored the impact of sequential roll-out of invitations in neighborhoods to cover sexual networks in relation to community and individual factors. The cluster-based invitations had a small effect on participation, independent from individual determinants and community risk level. However, clusters showed no association with Ct positivity. Possibly, community risk areas and individual risks explained most variation in Ct positivity, yet it might also illustrate that sexual networks cross cluster area borders.
Most of our findings are in line with other international chlamydia screening programs or population surveys, showing that women and older age-groups (<24 years) have higher participation rates and that ethnic groups differ in their screening behavior.9–12 People who were less educated were less likely to participate in CSI, a finding that is also consistent with other programs and might be related with individual knowledge about chlamydia or lacking awareness of the asymptomatic nature of infections.13,14
A great benefit of our population-based program design is that sociodemographic data from population registers were available for all invitees.3 However, information on sexual behavior was only available for the group that voluntarily filled in questionnaires, and nonparticipants were underrepresented. We examined the impact of this limitation, which resulted in smaller associations between variables and the outcome of participation. We cannot rule out the possibility that determinant factors for participation had a similar influence on the person's motivation to fill in the online questionnaires.
Although behavioral data were lacking for subgroups in the participation model, the positivity model was less hampered by missing data because the availability of questionnaires was fairly good (60%) among the participants. Likely, this contributed to the generalizability and representativeness of these results. However, missing data for particular questions could still be a source of bias. The large number of questionnaires from participants (>24,000) also provided the opportunity for comparisons between subgroups. We showed that about a third of the participants had a high individual risk level and, likely, they participated because they perceived themselves to be “at risk.” However, a large group may have had a smaller risk, and particularly in the first screening round, they could have participated out of curiosity or “just to be sure.” These patterns might change in the subsequent screening rounds of CSI.
Our results illustrate the challenge to ensure that those needed to be captured most in a chlamydia screening program actually do participate. It is a complex process including individual and community factors that are also interdependent. As these factors might change over time, they should be monitored in next screening rounds. The results can contribute to the development of more (cost-) effective screening modalities, such as more targeted attention for high risk communities or personal risk scores to motivate participation.
The advisory committee: T. Bakkenist, P. van den Broek, T. Coenen, S. de Gouw, P. Bindels, B. Hoebee, J. Land, S. Morré, R. van Riel, M. Kivi, M. van der Sande, E. Steyerberg; RIVM collaborators and advisers on data analyses: Gerda Doornbos, Matty Meijer, Hendriek van Boshuizen, Katie Greenland, Marlieke de Kraker, Amanja Haasnoot, Anita Suijkerbuijk; RNG-WPF: Floor Bakker.
1. van Bergen JE, Fennema JS, van den Broek IV, et al.. Rationale, design, and results of the first screening round of a comprehensive, register-based, Chlamydia screening implementation programme in the Netherlands. BMC Infect Dis 2010; 10: 293.
2. Greenland KE, Op de Coul EL, Bergen van JE, et al.. Acceptability of the internet-based Chlamydia Screening Implementation in the Netherlands and insights into non-response. Sex Transm Dis 2011; 6: 467–474.
3. van den Broek IV, Hoebe CJ, van Bergen JE, et al.. Evaluation design of a systematic, selective, Internet-based Chlamydia Screening Implementation in the Netherlands, 2008–2010: Implications of first results for the analysis. BMC Infect Dis 2010; 10: 89.
4. de Graaf H, Meijer S, Poelman J, et al. Sex Under 25 [in Dutch]. Delft, the Netherlands; Eburon Publishers, Rutgers Nisso Groep, 2005.
5. Merlo J, Chaix B, Ohlsson H, et al.. A Brief conceptual tutorial of multilevel analysis in social epidemiology: Using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health 2006; 60: 290–297.
6. van Bergen JE, Götz HM, Richardus JH, et al.. Prevalence of urogenital Chlamydia trachomatis increases significantly with level of urbanization and suggests targeted screening approaches: Results from the first national population based study in the Netherlands. Sex Transm Infect 2005; 81: 17–23.
7. Valkengoed IG, Morré SA, van den Brule JC, et al.. Low diagnostic accuracy of selective screening criteria for asymptomatic Chlamydia trachomatis infections in the general population. Sex Trans Infect 2000; 76: 375–380.
8. Aral SO, Leichliter JS, Blanchard JF. Overview: The role of emergent properties of complex systems in the epidemiology and prevention of sexually transmitted infections including HIV infection. Sex Transm Infect 2010; 86 (suppl 3): 1–3.
9. Novak DP, Karlsson RB. Simplifying chlamydia testing: An innovative Chlamydia trachomatistesting approach using the internet and a home sampling strategy: Population based study. Sex Transm Infect 2006; 82: 142–147.
10. Goulet V, Barbeyracde B, Raherison S, et al.. Prevalence of Chlamydia trachomatis: Results from the first national population-based survey in France. Sex Transm Infect 2010; 86: 263–270.
11. Vaughan D, O'Connell E, Cormican M, et al.. “Pee-in-a-Pot”: Acceptability and uptake of on-site chlamydia screening in a student population in the Republic of Ireland. BMC Infect Dis 2010; 10: 325.
12. Colliers A, Verster A, Van Puyenbroeck K, et al.. Screening Belgian university students for Chlamydia trachomatisinfection: A feasibility study. Int J Adolesc Med Health 2009; 21: 343–346.
13. Lorimer K, Hart GJ. Knowledge of Chlamydia trachomatisamong men and women approached to participate in community-based screening, Scotland, UK. BMC Public Health 2010; 10: 794.
14. Langille DB, Proudfoot K, Rigby J, et al.. A pilot project for chlamydia screening in adolescent females using self-testing: Characteristics of participants and non-participants. Can J Public Health 2008; 99: 117–120.