Introduction Behavioural surveys can provide insight into the attitudes and context associated with risk of sexually transmitted infections (STIs), but interpretation is hampered by theirreliance on self-reported STI history rather than confirmed diagnoses. We aimed to determine the feasibility and acceptability of linking clinic patients’ online survey data on STI risk factors with the national surveillance dataset on STI diagnoses (GUMCADv2), which contains electronic patient record (EPR) data routinely submitted to Public Health England by sexual health clinics.
Methods Between May and September 2016, attendeesat 16 sexual health clinicsacross Englandwereinvited to completean onlinesurveyon knowledge, attitudes, and behaviours around STI risk, using a clinic tablet or personal device (e.g. smartphone). Participants were given a unique study identifier (ID) to type into the survey, and provided consent to participate and to data linkage. Screening questions routed eligible participants, ≥15 years old and sexually active in the past year, to the full survey. Recruiting clinic staff kept record of study IDs and corresponding patient IDs. We linked surveysto GUMCADv2 with deterministic and probabilistic methods, using the recorded ID numbers as well as age, gender, and clinic attendance date. We examined recruitment and linkage success for a target of 7500 eligible attendees, and used univariable logistic regression and Z-test for proportions to assess selection bias.
Results 6283 clinic attendees agreed to take part in the study, of whom73.6% (4626) completed the survey. 95.9% (4437) of survey respondents were eligible; 59.2% of our recruitment target. Survey completionamong those agreeing to participate was higher in the 9 clinics that recruited ≥50% of their target than those that did not (84.9% vs 53.0%). Survey completion was also higher in those who agreed to fill in the survey in clinic than in those who agreed to do soat home (87.3% vs 16.8%). 91.2% (4048/4437) of participants consented to data linkage; of these 83.7% (3391) could be linked to GUMCADv2. Consent did not differ by age or gender, but washigher among men who have sex with men than heterosexual men (95.5% vs. 88.4%; p<0.01), and lower among Black Caribbean than white participants (87.1% vs 93.8%, p<0.01). Compared to all clinic attendees at study sites in the study period, a higher proportion of participants were <25 years (30.8% vs 37.1%,p<0.001), were black Caribbean (8.1% vs 14.5%, p<0.001), and had a same-day sexual health screen (70.2% vs 77.4%, p<0.001) or a chlamydia/gonorrhoea diagnosis (6.4% vs 7.7%p<0.01) recorded on GUMCADv2.
Conclusion Online behavioural surveys with data linkageto EPRsare highly acceptable to sexual health clinic attendees and technically feasible, and could improve the knowledgebase needed to tailor STI risk reduction efforts. However participants with higher risk profiles may be oversampled. Poor recruitment in some clinics was likely due to limited availability of clinic staff to facilitate and encourage use ofclinic tablets