Young people’s preferences for the use of emerging technologies for asymptomatic regular chlamydia testing and management: a discrete choice experiment in England

Objective To undertake a comprehensive assessment of the strength of preferences among young people for attributes of emerging technologies for testing and treatment of asymptomatic chlamydia. Design Discrete choice experiment (DCE) with sequential mixed methods design. A staged approach to selection of attributes/levels included two literature reviews, focus groups with young people aged 16–24 years (n=21), experts’ review (n=13) and narrative synthesis. Cognitive testing was undertaken to pilot and adapt the initial questionnaire. Online national panel was used for final DCE survey to maximise generalisability. Analysis of questionnaire responses used multinomial logit models and included validity checks. Setting England. Participants 1230 young people aged 16–24 from a national online panel (completion rate 73%). Outcome measures ORs for service attributes in relation to reference levels. Results The strongest attribute influencing preferences was chlamydia test accuracy (OR 3.24, 95% CI 3.13 to 3.36), followed by time to result (OR 1.81, 95% CI 1.71 to 1.91). Respondents showed a preference for remote chlamydia testing options (self-testing, self-sampling and postal testing) over attendance at a testing location. For accessing treatment following a positive test result, there was a general preference for online (OR 1.21, 95% CI 1.15 to 1.28) versus traditional general practitioner (OR 1.18, 95% CI 1.12 to 1.24) or pharmacy (OR 1.15, 95% CI 1.10 to 1.22) over clinic services. For accessing a healthcare professional and receipt of antibiotics, there was little difference in preferences between options. Conclusions Both test accuracy and very short intervals between testing and results were important factors for young people when deciding whether to undergo a routine test for asymptomatic chlamydia, with test accuracy being more important. These findings should assist technology developers, policymakers, commissioners and service providers to optimise technology adoption in service redesign, although use of an online panel may limit generalisability of findings to other populations.

• To our knowledge, this is the first DCE study to examine the preferences of a large, nationally representative sample of young people for new approaches designed to improve screening for and treatment of chlamydia. • Robust sequential methods were used to select final DCE attributes and levels, including two literature reviews, focus groups with young people, and review by expert groups. • An online panel enabled access to the target population for screening, including young people with no personal experience of chlamydia testing and treatment. • A limitation of this work is that the literature reviews inevitably identified more potential attributes than could be included in the DCE.
• The use of an online panel may limit generalisability of findings to other populations and over-represent the acceptability of online care.
o n l y 5 A DCE requires respondents to choose between competing scenarios, e.g. service options, described in terms of their attributes (e.g. time to test result) and levels (e.g. 30min, 2hrs, 7 or 14 days) and compare these against at least one alternative scenario. Such studies are very useful because they allow a direct assessment of relative preferences for different existing and new service configurations or treatment approaches. 22 The aim of the present study was to undertake a comprehensive assessment of young people's preferences for emerging technologies for chlamydia testing and treatment services in the context of a "check-up test" when remote care could be medically appropriate. To our knowledge, no previous study has disentangled strength of preference for attributes associated with new and emerging options for testing (such as self-testing) or treatment (such as online prescription) of STIs. Because STI services are "open access" and not subject to gatekeeping by referral from a clinician, they are directly dependent on individuals' preferences. Early insight into the attributes that influence whether individuals would use a new testing or treatment pathway should therefore be helpful in informing product development, pathway redesign and the assumptions used in economic models for future chlamydia testing and management. 23 24 o n l y 6 Two literature reviews were undertaken to produce a comprehensive list of potential themes and factors that might influence young people's choices. These included: (i) a systematic review of the use of stated preference studies for STI testing and treatment (PROSPERO Reg: CRD42014014862); and (ii) a scoping review of other research exploring preferences and acceptability of STI testing and treatment services.

Focus Groups
Focus groups were run with young people aged 16-24 years (3-7 per group; total n=21) to identify which themes and factors young people consider important when choosing to test for STIs. Convenience sampling was used to identify participants. Focus group topic guides incorporated typical vignettes of situations individuals might encounter. Sessions were recorded, transcribed and thematic analysis was performed to produce a list of potential attributes and levels. 28

Expert Groups
Four expert groups were convened (n=13), including a range of professionals with expert knowledge of the service and technology context. Expert groups were asked to review the focus group findings in terms of clinical feasibility and practicality. 25 Narrative Synthesis Narrative synthesis 29 enabled outputs from the three prior stages to be synthesised for each potential attribute. This approach was adopted as it offered a clear approach to synthesis based on the following stages: (i) identification of a checklist of properties against which attributes can be considered; (ii) tabulation against the checklist; and (iii) conceptual mapping and triangulation against the checklist. The final synthesis highlighted a tension between young people's desire to be tested for a wide range of STIs irrespective of risk, versus clinical guidelines for selective testing of STIs based on population group prevalence and risk. The range of STIs presented in the DCE was consequently limited to Chlamydia. Because focus group findings indicated difficulty in understanding several dimensions of test performance, test accuracy was expressed in terms of the likelihood of a false negative result.

Questionnaire Design and Piloting
A generic pairwise choice with opt out question was selected for the questionnaire design. Respondents were presented with a series of choice sets for which there were 3 responses: 'option A', 'option B' or 'I would not test'. A sample choice set is illustrated in Figure 2. The questionnaire adopted a main effects design using full profiles (all attributes included in the study). Whilst some DCEs do include an attribute on economic costs, DCEs exploring preferences for STI testing services in England have generally excluded cost since the NHS provides treatment 'free at the point of delivery' 19 20 A cost attribute was therefore not included.
o n l y 8

Statistical Analyses
The multinomial logit (MNL) model developed by McFadden was used for analysing responses; this is recognised as the convention for threeresponse choice set studies ('option A', 'option B' or 'I would not test'). 25 32 Analysis utilised STATA13 SE with the method and code outlined by Ryan et al 22 Analyses presented odds ratios (ORs), 95% confidence interval (CIs) and coefficients for each attribute level. Variables within the model were all treated as categorical variables for the analysis. Attribute levels were specified using dummy coding, the preferred form of coding where ORs are to be calculated. Within the model, the levels that were dropped to form the reference levels reflected those aligned to a 'typical' sexual health clinic pathway (summarised in table 2). To test the internal validity of questionnaire responses, analyses compared full results against: (i) removal of responses where participants did not answer the repeated choice set consistently; (ii) removal of any respondents who took less than the minimum time (five minutes) observed in cognitive testing to complete the questionnaire; and (iii) removal of responses containing the opt out question data. Further tests for internal consistency and rationality were not included, since excluding responses on this basis may be viewed as an inappropriate imposition of rationality. 33 Demographic characteristics (gender, age and ethnicity) of respondents were compared with national Census data. 34 The influence of patient level characteristics (age, sex and STI testing history (yes/no)) on the likelihood of not choosing to test was examined. In addition to ex-ante planned sub-group analyses, if sufficient responses were received analysis was also planned to compare: (i) respondents who had, or had not, previously tested for an STI, and (ii) those who indicated their relationship status as 'single' versus those in a sexual relationship with one person.
Trade-off between accuracy and time to result was examined by considering the probability of uptake for tests with characteristics at the opposite ends of the spectrum i.e. 'lower accuracy (5% false negatives), faster time to result (30 min)' and 'higher accuracy (2% false negatives), longer time to result (14 days)'.

RESULTS:
In total, 1,230 fully completed questionnaires were received, whilst a further 490 people commenced the questionnaire but did not complete it, providing a completion rate of 73%. Time to complete the 25 choice sets ranged from one minute 19 seconds to 30 minutes 19 seconds with a median time to completion of seven minutes 51 seconds. Demographic characteristics of respondents are summarised in the Supplementary Information File 2. Comparing the DCE sample to the characteristics of [16][17][18][19][20][21][22][23][24] year olds in England indicates that gender, age and ethnicity are broadly in line with national population demographics, with a slight under-representation of the [16][17][18] year age group. In terms of the o n l y 9 geographical residence of participants, the sample was broadly in line with the geographical distribution of [16][17][18][19][20][21][22][23][24] year olds identified in the ONS Mid-Year Population Estimates 2015 35 , with the exception of a lower proportion of respondents from the East of England.
Internal validity checks showed that, in comparison to the full dataset, all checks yielded very similar ORs with no change in the order of the strength of preference for a level within each attribute. Results of the internal validity checks are included in the Supplementary Information File 2 (Tables 2.1 -2.4).
The ORs for the full dataset and subgroups analysed are presented in Table 3. Additional DCE results, including the 95% CIs for all sub-group analysis are presented in Supplementary Information File 2.
Looking across all attributes and levels, participants expressed the strongest preference for accuracy of test result (OR 3.242). Table 3 shows differences in the strength of preference between males and females (OR 2.950 and 3.570, respectively), and between those who had previously tested or not tested (OR 3.000 and 3.482, respectively). Time to result was the attribute showing the next strongest preference across all subgroups. These results are consistent with the logical expectation that people will prefer higher accuracy and a shorter waiting time. Looking specifically at the trade-off between accuracy and time to result, Table 4 indicates that participants are willing to wait noticeably longer in order to have a test result with a lower chance of a false negative result.
When considering how to test, all subgroups demonstrated a preference for self-testing (OR 1.618). Testing via an outreach service in an educational/ work setting was found to be the least preferred option (OR 0.821). Results showed a consistent strength of preference for those options that do not involve direct contact with healthcare professionals (self-testing and postal self-sampling), compared with those that do. For consultation and treatment methods following a positive test result, there was a preference for non-sexual health clinic pathways with online consultation (OR 1.212), treatment via general practice (OR 1.183) and treatment via pharmacy (OR 1.158) preferred in the full dataset. At subgroup level (age, gender, previous testing history and relationship status), more variation was found in the preference order for this attribute, with the exception of the sexual health clinic which was consistently the least preferred option across all subgroups (see table 3 and supplementary information file 2). The full dataset shows that, for accessing a healthcare professional, there was no statistically significant preference for instant messenger or email access compared with accessing the professional face-to-face. Telephone access to a healthcare professional was the least preferred access option (OR 0.949), and the only statistically significant result when compared with face-to-face access (the referent). There was, similarly, very little difference in preferences for how young people might access antibiotics. o n l y 10

DISCUSSION:
Our findings indicate that, based on the levels included in the DCE questionnaire, participants were willing to wait considerably longer in order to have a test result with a lower chance of a false negative result. The conclusion for new test developers is that time to result is less important than accuracy, and that test users are unlikely to prefer a point-of-care test or self-test with lower accuracy than the tests currently available to them. There was a strong preference for remote access to testing, consultation, and antibiotic prescriptions, although for accessing a health professional there was no preference between online and face-to-face methods. This suggests a remote online pathway is acceptable to young people, as long as test performance remains equivalent.
In the various hypothetical situations presented, respondents showed a preference for chlamydia self-testing, self-sampling and postal testing over attendance at a testing location. For accessing treatment, a general preference was exhibited for online versus traditional GP or pharmacy services over clinic services. For receipt of antibiotics, there was little difference in preferences. We were also able to identify which attributes people may be willing to trade to maximise their utility. Looking at the trade-off between accuracy and time to result, we found that young people are willing to wait noticeably longer in order to have a test result with a lower chance of a false negative result, reinforcing the need for test equivalence.
The strengths of this DCE study include, firstly, the robust methods employed to select attributes and levels, which aligned with the recommendations of Coast et al, ensuring attributes are "manipulable in policy". 36 Secondly, the fact that participants were drawn from the general population targeted by the National Chlamydia Screening Programme, rather than from healthcare settings. This enabled us to access a demographically representative national sample of young people, including those who had not had previous contact with STI services. Finally, the use of an online panel enabled large scale data to be collected at a reasonable cost and allowed validity checks that would not otherwise have been possible with written questionnaire responses. The large sample size also allowed comparison between several subgroups identifying any differences based on age, gender, previous testing history and relationship category.
However, the study does have a few limitations. The first relates to the selection of attributes and levels. Whilst the selection process employed was very rigorous, this cannot detract from the fact that further attributes were identified, which might impact on individuals' choices. To mitigate against this, where such an attribute was excluded, information was provided in the survey background section to minimise respondents forming their own views on the impact of this attribute on the pathway.
Use of an online panel also provided accurate records of the time taken to complete the survey, permitting additional validity checks, which would not otherwise have been possible with a written questionnaire. On the other hand, use of an online panel did exclude populations who o n l y 11 do not have access to the internet, thereby potentially over-representing the acceptability of online care. However, given the extremely high proportion (97%) of [15][16][17][18][19][20][21][22][23][24] year-olds accessing the internet daily via a mobile device 37 and owning a smartphone 38 , it is evident that the vast majority of the target population can have access to online care pathways if they choose to do so. The question of whether these individuals have the degree of digital and health literacy needed for online self-sampling, testing and treatment was not explored in this study.
Identified barriers to accessing current sexual health services, such as embarrassment 6 ; access and convenience 7 ; stigma; 39 and privacy and anonymity concerns, 40 41 should be lower for an online service. Balanced against this, non-face-to-face access to testing and treatment services was valued less strongly than attributes such as test accuracy and time to result. In such cases a young person's preferences for remote versus face-to-face testing and treatment might be different.
Finally, it is difficult to compare the present results to other published DCEs because of key differences in focus. Miners et al (sample n=3,358) 19 and Llewellyn et al (sample n=233) 20 both only focussed on preferences within existing traditional service delivery models, and did not incorporate hypothetical future scenarios (e.g. point-of-care testing, self-testing or treatment via eHealth/mHealth solutions). Two other DCE studies, which did consider self-sampling at home for chlamydia screening 17 18 , sample size 174 & 126 respectively, both described sending the sample to a laboratory for analysis rather than a self-test. One of these studies 18 did identify a stronger preference for attendance at a family planning clinic, rather than self-sampling, but since participants were recruited from the waiting room of such a clinic they cannot be considered representative of the population targeted by the NCSP. A fifth study, which examined preferences for point-of-care testing, only surveyed clinicians undertaking STI testing (sample n=218), thus excluding the population targeted for chlamydia testing. 42 The only stated preference study that has considered patients' preferences for STI self-testing focussed on HIV (sample n=365). 43 The authors reported results in line with our findings, with respondents exhibiting a preference for tests which are accurate, timely and private/ anonymous. However, this study was undertaken in 2002 prior to smartphones and at a time when self-testing for HIV was still under development. Importantly for screening tests, only one DCE study to date 20 has sought to include non-service users (i.e. populations with no experience of STI testing). The authors identified a preference for testing for all STIs, in settings with healthcare professionals with specialist knowledge present, and for receipt of negative as well as positive results. This study used a convenience sample of 233 students from two universities which was unlikely to be geographically or socio-economically representative. All other studies have drawn their DCE samples from people who were either current service users, or attendees for other, linked services. Use of such samples represents a significant short-coming when considering the introduction of new digital technologies for asymptomatic chlamydia testing and management. Sample size is also generally smaller than the present study. o n l y 12 A number of questions remain which our research does not address. Firstly, recognising that the range of STIs included is an important consideration for young people in choosing to test 19 20 , further research is required to understand this better in the context of potential new pathways which incorporate other STIs (e.g. gonorrhoea). Secondly, the study highlights a number of methodological considerations where there is an absence of consensus that may warrant further exploration to improve consistency. These include the number of choice sets to include in a DCE and the use of repeated choice sets as an internal validity measure. Finally, given that cost was not included as an attribute in this study, it is not possible to provide an indication of willingness to pay, or the monetary benefits of potential service changes, from this DCE. 44 The present study is, to our knowledge, the first to present large scale, quantitative evidence of young people's preferences for attributes of new pathways and emerging technologies in the testing and treatment of chlamydia (based on a demographically representative national sample). The DCE methodology applied also produces a measure of the relative strength of preference between different attributes and levels, and potential trade-offs. This can provide useful evidence to technology developers, policy makers, commissioners and service providers. In particular, it provides a first insight into preferences for technologies currently under development, and those which might be available for use in the near future, compared to the features of existing products and services. This can indicate how young people may respond to changes in pathways and to the introduction of new technologies.

What is Chlamydia?
Chlamydia is the most common sexually transmitted infection in England and the majority of infections are found in young people aged 15-24. Both men and women can get chlamydia, but most people with chlamydia have no symptoms and do not know they have an infection. The test for chlamydia is usually a urine sample for men and a vaginal swab for women. Once diagnosed, chlamydia can be treated with a single dose (2 or 4 tablets) of an antibiotic called Azithromycin.
In England there is a national chlamydia screening programme which recommends that people aged 16-24 are tested for chlamydia annually or when they change their sexual partner.

What happens if I don't get treatment?
If left undiagnosed chlamydia can lead to serious health problems.
In women this includes a condition called Pelvic Inflammatory Disease, which is an infection of the womb, fallopian tubes or ovaries. This can cause severe pain and in some cases (around 1 in 10) lead to infertility or ectopic pregnancy (where the pregnancy occurs outside of the womb in the fallopian tube). If you are pregnant and have chlamydia you can pass it on to your baby when you give birth and this can lead to the baby getting an eye infection or pneumonia.
In men chlamydia can lead to a condition called epididymitis, which causes soreness and swelling in the scrotum.  The following sections provide you with an explanation of the process of getting tested and treatment and the key terms used in the survey: The diagram below shows the stages up until you find out our result: The following sections provide more information on how you test, how long you wait for the result and the accuracy of the test.  This focuses on how you get your test, how the sample is taken and what happens to the  sample once you've taken it. There are six options in the questionnaire, an explanation of these is provided below:  Self-Test -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and interpret the result yourself (like a pregnancy test). This is different to the self-sample options below because you interpret the result yourself rather than it being interpreted by a healthcare professional.  Self-Sample and post off for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and send the sample in the freepost envelope to the laboratory for analysis. This is different to the self-test option above because the test is interpreted by a healthcare professional.  Self-Sample and take to pharmacy for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and then take the sample to a pharmacy for analysis. This is different to the self-test option above because the test is interpreted by a healthcare professional. Opening hours vary between pharmacies but many in towns and cities are open evenings and weekends as well as weekdays.  Self-Sample and take to your place of education/ workplace for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and take the sample to your place of education/ workplace for analysis by an outreach nurse. This is different to the selftest option above because the test is interpreted by a healthcare professional. Outreach services in education and workplaces are usually available one day a week.  Attend GP Practice, sample taken by a GP or Nurse -Book an appointment at your GP practice with a GP or practice nurse. Your sample will be taken by the healthcare professional you see.

How you Contact a Healthcare Professional
New technology means that there are now more options for completing your consultation to get treatment, or in the case of the online consultation, accessing a healthcare professional for advice. There are four options within the questionnaire:  Telephone -speak to a healthcare professional on the telephone.  Instant Messaging -have a discussion with a healthcare professional via an instant messaging service online  Email -have a discussion with a healthcare professional via email. During opening hours you will receive a response within 2 hours, and for emails sent out of hours you will receive a response the next day.  Face-to-face -have a face-to-face discussion with a healthcare professional.  Deliver to home address -provide your home address for your antibiotic to be posted to you. Your antibiotic will arrive within 1-2 working days.  Deliver to collection point -nominate an address other than your home address, for example a friend's address or a collection point for your antibiotic to be posted to. Your antibiotic will arrive within 1-2 working days.  Collect from Pharmacy -attend a pharmacy of your choice with your prescription and the pharmacist gives you the antibiotic tablets  Collect from Sexual Health Clinic -attend a sexual health clinic and the doctor or nurse gives you the antibiotic tablets In all of the choices available to you there are a number of things that you should assume are the same:  The service is provided to you free of charge although there may be a cost to you, for example in making phone calls, accessing the internet, or taking time off work to go to an appointment.  The healthcare professional who is providing your treatment is trained to be able to provide the service.  How you get your results is the same whichever option you choose, except for the self-test where the test will provide the result.  Your personal data is managed securely in whichever option you choose More choices for an option does not mean that it is more or less important, it just reflects that there are more choices within the scenarios you are being asked to consider.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45 46             State specific objectives, including any prespecified hypotheses (page 5)

Study design 4
Present key elements of study design early in the paper (page 5) Describe any efforts to address potential sources of bias (page 8)

Study size 10
Explain how the study size was arrived at (page 7)  *Give information separately for cases and controls in case-control studies and, if applicable, for exposed and unexposed groups in cohort and cross-sectional studies.

Conclusions:
Both test accuracy and very short intervals between testing and results were important factors for young people when deciding whether to undergo a routine test for asymptomatic chlamydia, with test accuracy being more important. These findings should assist technology developers, policy makers, commissioners and service providers to optimise technology adoption in service redesign, although use of an online panel may limit generalisability of findings to other populations.

Strengths and Limitations of the Study:
• To our knowledge, this is the first large-scale DCE study to examine the preferences of a geographically representative national sample of young people for emerging technologies designed to improve screening for and treatment of asymptomatic chlamydia. • Robust sequential methods were used to select final DCE attributes and levels, including two literature reviews, focus groups with young people, and review by expert groups. • An online panel enabled access to the population targeted for screening, including young people with no personal experience of chlamydia testing and treatment. • A limitation of this work is that the literature reviews inevitably identified more potential attributes than could be included in the DCE. • The use of an online panel may limit generalisability of findings to the small percentage of this population who do not currently access the internet, and therefore over-represent the acceptability of online care. Chlamydia is the most commonly reported sexually transmitted infection (STI) in England,3 with young people aged 15-24 accounting for 63% of diagnoses in 2016. 1 The estimated 4 annual cost of chlamydia treatment in 2015 was £249.8 million. 2 Undetected infections and 5 re-infections can lead to significant adverse health consequences, such as pelvic 6 inflammatory disease, ectopic pregnancy and infertility, which have an impact on National 7 Health Service (NHS) costs and health-related quality of life. 3 A National Chlamydia 8 Screening Programme (NCSP) was therefore rolled out in England for 16-24 year olds 9 between 2003 and 2008. 4 However, despite recent evidence of a reduction in chlamydia 10 prevalence in England for for 2000 to 2015, concurrent with large-scale population testing 5 , 11 and the fact that uncomplicated infection is easy to treat with oral antibiotics, uptake of 12 screening for chlamydia remains low. 6 Worryingly, there has also been a decline over the 13 last four years in the number of local authorities achieving the public health outcomes 14 framework indicator of 2,300 diagnoses per 100,000 population. 1  For many health services, digital technology is now widely regarded by policy makers as one 33 approach to improve access and reduce costs. 11 For chlamydia testing and treatment, a 34 range of digital options are available, and more technologies are likely to enter the market 35 within the next 3-5 years. 12 These include point-of-care tests 13 , online postal self-sampling, 36 eSexual health clinics 14 , apps and non-face-to-face consultation methods 15 16 , and self-tests 37 networked through mobile phones 17 . Such innovations provide an opportunity to redesign 38 current chlamydia screening services with the aim of improving testing and treatment 39 uptake. However, such service redesign should be based on a sound understanding of the 40 preferences of young people as service-users for specific attributes of such services. 41 42 Research specifically measuring young people's preferences nationally for chlamydia testing 43 and treatment service options is lacking. Even for STI testing services generally, relatively 44 few studies have used a discrete choice experiment (DCE) design to assess preferences in 45 the UK population. [18][19][20][21] In comparison with other preference elicitation methods, a DCE can 46 quantify the relative importance of different attributes that characterise a new or existing 47  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   5 product and/or service, identifying which attributes people prioritise and which they may be 48 willing to trade with the view to maximising their utility. 22 A DCE requires respondents to 49 choose between competing scenarios, e.g. service options, described in terms of a particular 50 attribute (e.g. time to test result) and a range of levels (e.g. 30min to 14 days) and to 51 compare these against an alternative scenario. DCE studies are very useful because they 52 allow a direct assessment of relative preferences for various existing and hypothetical new 53 service configurations or treatment approaches. 23 54 55 The aim of the present study was to undertake a comprehensive assessment of the 56 preferences of young people. targeted by the National Chlamydia Screening Programme, for 57 emerging technology options for testing and treatment in the context of a "check-up test" 58 where remote care could be medically appropriate. 14 To our knowledge, no previous study 59 has attempted to disentangle strength of preference for attributes associated with new and 60 emerging options for chlamydia screening (such as self-testing) or treatment (such as online 61 prescription). Because STI services are "open access" and not subject to gatekeeping by 62 referral from a clinician, the impact of such disruptive innovations will be directly dependent 63 on population preferences. Early insight into attributes that could influence whether 64 individuals are more likely to use a new testing or treatment pathway should therefore be 65 helpful in informing product development and pathway redesign for future chlamydia 66 screening service models. 24  This study was conducted using an exploratory sequential mixed methods design, adopting 72 recommended stages for undertaking a DCE 26 , as shown in Figure 1. In selecting methods to 73 design the questionnaire and conduct the DCE, reference was made to the International 74 Society for Pharmacoeconomics and Outcomes Research good practice checklist for conjoint 75 analysis. 27

77
Patient and public involvement: Young people (with and without experience of STI services) 78 participated in focus groups to inform priorities included in the DCE; in cognitive testing to 79 finalise questionnaire design; and in questionnaire completion to identify preferences, as 80 described in the Acknowledgements. 81 82 Selection of Attributes and Levels 83 The attributes and associated levels were determined using a four-stage approach. 28

85
Literature Reviews 86 Two literature reviews were undertaken to produce a comprehensive list of potential 87 themes and factors that might influence young people's choices. A generic pairwise choice with opt out question was selected for the questionnaire design. 120 Respondents were presented with a series of choice sets for which there were 3 responses: 121 'option A', 'option B' or 'I would not test'. A sample choice set is illustrated in Figure 2. The 122 questionnaire adopted a main effects design using full profiles (all attributes included in the 123 study). Whilst some DCEs do include an attribute on economic costs, DCEs exploring 124 preferences for STI testing services in England have generally excluded cost since the NHS 125 provides treatment 'free at the point of delivery' 20 21 A cost attribute was therefore not 126 included. 127 128 The questionnaire was developed using SAS 9.4 software to ensure that the design was D-129 efficient. 27  There was a strong preference for remote access to testing, consultation, and antibiotic 279 prescriptions, although for accessing a health professional there was no preference 280 Behavioural factors, such as embarrassment 7 , stigma 41 and privacy and anonymity 324 concerns 42 43 are known to influence uptake of sexual health services, as well as structural 325 factors such as convenience and perceived barriers to access 8 ; many of these may be lower 326 for a non-face-to-face service. Balanced against this, access to online testing and treatment 327 which did consider self-sampling at home for chlamydia screening 18 19 , with sample sizes 174 337 and 126 respectively, both described sending the sample to a laboratory for analysis rather 338 than a self-test. One of these studies 19 did identify a stronger preference for attendance at a 339 family planning clinic, rather than self-sampling, but since participants were recruited from 340 the waiting room of such a clinic they cannot be considered representative of the general 341 population targeted by the NCSP. A fifth study, which examined preferences for point-of-342 care testing, only surveyed clinicians undertaking STI testing (sample n=218), thus excluding 343 the population actually targeted for chlamydia screening. 44 The only stated preference 344 study that has considered patients' preferences for STI self-testing focussed on HIV (sample 345 n=365). 45  The present study is, to our knowledge, the first to present a large scale, quantitative 371 analysis of young people's preferences for attributes of potential new pathways to deliver 372 testing and treatment of asymptomatic chlamydia, based on a nationally representative 373 population. The DCE methodology applied also produced a measure of the relative strength 374  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 Table 1 Supplementary Information File 1 (pages 3-7).    Please read through the following background information before completing the survey. It contains information about chlamydia and the terms that are used in the survey.

What is Chlamydia?
Chlamydia is the most common sexually transmitted infection in England and the majority of infections are found in young people aged 15-24. Both men and women can get chlamydia, but most people with chlamydia have no symptoms and do not know they have an infection. The test for chlamydia is usually a urine sample for men and a vaginal swab for women. Once diagnosed, chlamydia can be treated with a single dose (2 or 4 tablets) of an antibiotic called Azithromycin.
In England there is a national chlamydia screening programme which recommends that people aged 16-24 are tested for chlamydia annually or when they change their sexual partner.

What happens if I don't get treatment?
If left undiagnosed chlamydia can lead to serious health problems.
In women this includes a condition called Pelvic Inflammatory Disease, which is an infection of the womb, fallopian tubes or ovaries. This can cause severe pain and in some cases (around 1 in 10) lead to infertility or ectopic pregnancy (where the pregnancy occurs outside of the womb in the fallopian tube). If you are pregnant and have chlamydia you can pass it on to your baby when you give birth and this can lead to the baby getting an eye infection or pneumonia. In men chlamydia can lead to a condition called epididymitis, which causes soreness and swelling in the scrotum.

For further information
If you would like further information on chlamydia and how to prevent it, or any of the other conditions described please visit: http://www.nhs.uk/conditions/chlamydia/Pages/Introduction.aspx

About the Survey…
In completing the survey you will be presented with a number of choices for getting tested and treatment for chlamydia. Please consider each set of choices and indicate whether you prefer option A or option B or whether you would not test, that is, you would not choose option A or option B.
Each choice involves two different scenarios for testing and getting treatment and there are differences between how you test, how long you have to wait for the result, how accurate the test is, how you have a consultation to get treatment for chlamydia, how you access a healthcare professional to get that treatment and how you get your antibiotic.
The following sections provide you with an explanation of the process of getting tested and treatment and the key terms used in the survey: The diagram below shows the stages up until you find out our result: The following sections provide more information on how you test, how long you wait for the result and the accuracy of the test.

How you test for Chlamydia
This focuses on how you get your test, how the sample is taken and what happens to the sample once you've taken it. There are six options in the questionnaire, an explanation of these is provided below:  Self-Test -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and interpret the result yourself (like a pregnancy test). This is different to the self-sample options below because you interpret the result yourself rather than it being interpreted by a healthcare professional.  Self-Sample and post off for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and send the sample in the freepost envelope to the laboratory for analysis. This is different to the self-test option above because the test is interpreted by a healthcare professional.  Self-Sample and take to pharmacy for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and then take the sample to a pharmacy for analysis. This is different to the self-test option above because the test is interpreted by a healthcare professional. Opening hours vary between pharmacies but many in towns and cities are open evenings and weekends as well as weekdays.  Self-Sample and take to your place of education/ workplace for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and take the sample to your place of education/ workplace for analysis by an outreach nurse. This is different to the self-test option above because the test is interpreted by a healthcare professional. Outreach services in education and workplaces are usually available one day a week.  Attend GP Practice, sample taken by a GP or Nurse -Book an appointment at your GP practice with a GP or practice nurse. Your sample will be taken by the healthcare professional you see. We would also like you to think about how you would choose to get treatment if your test result is positive. The diagram below shows the stages to get treatment: The following sections provide more information on how you get your treatment, how you contact a healthcare professional and how you get your antibiotics.

How you Contact a Healthcare Professional
New technology means that there are now more options for completing your consultation to get treatment, or in the case of the online consultation, accessing a healthcare professional for advice. There are four options within the questionnaire:  Telephone -speak to a healthcare professional on the telephone.  Instant Messaging -have a discussion with a healthcare professional via an instant messaging service online  Email -have a discussion with a healthcare professional via email. During opening hours you will receive a response within 2 hours, and for emails sent out of hours you will receive a response the next day.  Face-to-face -have a face-to-face discussion with a healthcare professional.

How you get your antibiotics
Once you've completed your consultation you need to get your antibiotic tablets. In this survey there are four options available to you:  Deliver to home address -provide your home address for your antibiotic to be posted to you. Your antibiotic will arrive within 1-2 working days.  Deliver to collection point -nominate an address other than your home address, for example a friend's address or a collection point for your antibiotic to be posted to. Your antibiotic will arrive within 1-2 working days.  Collect from Pharmacy -attend a pharmacy of your choice with your prescription and the pharmacist gives you the antibiotic tablets  Collect from Sexual Health Clinic -attend a sexual health clinic and the doctor or nurse gives you the antibiotic tablets In all of the choices available to you there are a number of things that you should assume are the same:  The service is provided to you free of charge although there may be a cost to you, for example in making phone calls, accessing the internet, or taking time off work to go to an appointment.  The healthcare professional who is providing your treatment is trained to be able to provide the service.  How you get your results is the same whichever option you choose, except for the self-test where the test will provide the result.  Your personal data is managed securely in whichever option you choose More choices for an option does not mean that it is more or less important, it just reflects that there are more choices within the scenarios you are being asked to consider.                  State specific objectives, including any prespecified hypotheses (page 5)

Study design 4
Present key elements of study design early in the paper (page 5-6)    Objective: To undertake a comprehensive assessment of the strength of preferences among young people for attributes of emerging technologies for testing and treatment of asymptomatic chlamydia.
Design: Discrete choice experiment (DCE) with sequential mixed methods design. A staged approach to selection of attributes/ levels included two literature reviews, focus groups with young people aged 16-24 years (n = 21), experts' review (n=13), and narrative synthesis. Cognitive testing was undertaken to pilot and adapt the initial questionnaire. Online national panel was used for final DCE survey to maximise generalisability. Analysis of questionnaire responses used multinomial logit models and included validity checks.
Outcome Measures: Odds Ratios (ORs) for service attributes in relation to reference levels.

Results:
The strongest attribute influencing preferences was chlamydia test accuracy (OR 3.24, 95% CI 3.13-3.36), followed by time to result (OR 1.81, 95% CI 1.71-1.91). Respondents showed a preference for remote chlamydia testing options (self-testing, self-sampling and postal testing) over attendance at a testing location. For accessing treatment following a positive test result, there was a general preference for online (OR 1.21, 95% CI 1.15-1.28) versus traditional GP (OR 1.18, 95% CI 1.12-1.24) or pharmacy (OR 1.15, 95% CI 1.10-1.22) over clinic services. For accessing a healthcare professional and receipt of antibiotics, there was little difference in preferences between options.

Conclusions:
Both test accuracy and very short intervals between testing and results were important factors for young people when deciding whether to undergo a routine test for asymptomatic chlamydia, with test accuracy being more important. These findings should assist technology developers, policy makers, commissioners and service providers to optimise technology adoption in service redesign, although use of an online panel may limit generalisability of findings to other populations.

Strengths and Limitations of the Study:
 To our knowledge, this is the first large-scale DCE study to examine the preferences of a geographically representative national sample of young people for emerging technologies designed to improve screening for and treatment of asymptomatic chlamydia.  Robust sequential methods were used to select final DCE attributes and levels, including two literature reviews, focus groups with young people, and review by expert groups.  An online panel enabled access to the population targeted for screening, including young people with no personal experience of chlamydia testing and treatment.  A limitation of this work is that the literature reviews inevitably identified more potential attributes than could be included in the DCE.  The use of an online panel may limit generalisability of findings to the very small percentage (3%) of 15-24 year-olds who do not currently access the internet, and therefore over-represent the acceptability of online care.

INTRODUCTION:
Chlamydia is the most commonly reported sexually transmitted infection (STI) in England, with young people aged 15-24 accounting for 63% of diagnoses in 2016. 1 The estimated annual cost of chlamydia treatment in 2015 was £249.8 million. 2 Undetected infections and re-infections can lead to significant adverse health consequences, such as pelvic inflammatory disease, ectopic pregnancy and infertility, which have an impact on National Health Service (NHS) costs and health-related quality of life. 3 A National Chlamydia Screening Programme (NCSP) was therefore rolled out in England for [16][17][18][19][20][21][22][23][24] year olds between 2003 and 2008. 4 However, despite recent evidence of a reduction in chlamydia prevalence in England for for 2000 to 2015, concurrent with large-scale population testing 5 , and the fact that uncomplicated infection is easy to treat with oral antibiotics, uptake of screening for chlamydia remains low. 6 Worryingly, there has also been a decline over the last four years in the number of local authorities achieving the public health outcomes framework indicator of 2,300 diagnoses per 100,000 population. 1 Barriers identified for young people accessing STI testing and treatment services include tangible service attributes, such as location of service, and personal or behavioural factors such as the stigma associated with attendance, embarrassment, fear of being recognised and privacy concerns. [7][8][9] A range of options are currently available for young people to access asymptomatic chlamydia testing and treatment services. For testing, options include attending a genito-urinary medicine (GUM) clinic, testing via primary care, e.g. pharmacies/general practitioners (GPs), or via internet testing services where a number of websites offer free self-sampling kits online with samples sent to laboratories for analysis and results communicated directly to patients. For individuals who screen positive, options for accessing antibiotic treatment include attending a GUM/sexual health clinic, GP practice or pharmacy. Even though a national screening programme exists in England, large geographical variations exist in testing coverage, from 16% of young people tested for chlamydia in the West Midlands region to 27% in London, and in the proportion of positive cases treated. 1 10 For many health services, digital technology is now widely regarded by policy makers as one approach to improve access and reduce costs. 11 For chlamydia testing and treatment, a range of digital options are available, and more technologies are likely to enter the market within the next 3-5 years. 12 These include point-of-care tests 13 , online postal self-sampling, eSexual health clinics 14 , apps and non-face-to-face consultation methods 15 16 , and self-tests networked through mobile phones 17 . Such innovations provide an opportunity to redesign current chlamydia screening services with the aim of improving testing and treatment uptake. However, such service redesign should be based on a sound understanding of the preferences of young people as service-users for specific attributes of such services.
Research specifically measuring young people's preferences nationally for chlamydia testing and treatment service options is lacking. Even for STI testing services generally, relatively few studies have used a discrete choice experiment (DCE) design to assess preferences in the UK population. [18][19][20][21] In comparison with other preference elicitation methods, a DCE can quantify the relative importance of different attributes that characterise a new or existing product and/or service, identifying which attributes people prioritise and which they may be willing to trade with the view to maximising their utility. 22   between competing scenarios, e.g. service options, described in terms of a particular attribute (e.g. time to test result) and a range of levels (e.g. 30min to 14 days) and to compare these against an alternative scenario. DCE studies are very useful because they allow a direct assessment of relative preferences for various existing and hypothetical new service configurations or treatment approaches. 23 The aim of the present study was to undertake a comprehensive assessment of the preferences of young people, targeted by the National Chlamydia Screening Programme, for emerging technology options for testing and treatment in the context of a "check-up test" where remote care could be medically appropriate. 14 To our knowledge, no previous study has attempted to disentangle strength of preference for attributes associated with new and emerging options for chlamydia screening (such as self-testing) or treatment (such as online prescription). Because STI services are "open access" and not subject to gatekeeping by referral from a clinician, the impact of such disruptive innovations will be directly dependent on population preferences. Early insight into attributes that could influence whether individuals are more likely to use a new testing or treatment pathway should therefore be helpful in informing product development and pathway redesign for future chlamydia screening service models. 24 25

METHODS:
This study was conducted using an exploratory sequential mixed methods design, adopting recommended stages for undertaking a DCE 26 , as shown in Figure 1. In selecting methods to design the questionnaire and conduct the DCE, reference was made to the International Society for Pharmacoeconomics and Outcomes Research good practice checklist for conjoint analysis. 27 Patient and public involvement: Young people (with and without experience of STI services) participated in focus groups to inform priorities included in the DCE; in cognitive testing to finalise questionnaire design; and in questionnaire completion to identify preferences, as described in the Acknowledgements.

Selection of Attributes and Levels
The attributes and associated levels were determined using a four-stage approach. 28 More detailed information is provided in Supplementary Information File 1.

Literature Reviews
Two literature reviews were undertaken to produce a comprehensive list of potential themes and factors that might influence young people's choices. These included: (i) a systematic review of the use of stated preference studies for STI testing and treatment (PROSPERO Reg: CRD42014014862); and (ii) a scoping review of other research exploring preferences and acceptability of STI testing and treatment services.

Focus Groups
Focus groups were run with young people aged 16-24 years (4 groups, 3-7 per group; total n=21) to identify which themes and factors young people consider important when choosing to test for STIs. Convenience sampling was used to identify participants. Focus group topic guides incorporated typical vignettes of situations individuals might encounter. Sessions were recorded, transcribed and thematic analysis was performed to produce a list of potential attributes and levels. 29

Expert Groups
Four expert groups were convened (n=13), including a range of professionals with expert knowledge of the service and technology context. Expert groups were asked to review the focus group findings in terms of clinical feasibility and practicality. 26 Narrative Synthesis Narrative synthesis 30 enabled outputs from the three prior stages to be synthesised for each potential attribute. This approach was adopted as it offered a clear approach to synthesis based on the following stages: (i) identification of a checklist of properties against which attributes can be considered; (ii) tabulation against the checklist; and (iii) conceptual mapping and triangulation against the checklist. The final synthesis highlighted a tension between young people's desire to be tested for a wide range of STIs irrespective of risk, versus clinical guidelines for selective testing of STIs based on population group prevalence and risk. The range of STIs presented in the DCE was consequently limited to Chlamydia. Because focus group findings indicated difficulty in understanding several dimensions of test performance, test accuracy was expressed in terms of the likelihood of a false negative result.

Questionnaire Design and Piloting
A generic pairwise choice with opt out question was selected for the questionnaire design. Respondents were presented with a series of choice sets for which there were 3 responses: 'option A', 'option B' or 'I would not test'. A sample choice set is illustrated in Figure 2. The questionnaire adopted a main effects design using full profiles (all attributes included in the study). Whilst some DCEs do include an attribute on economic costs, DCEs exploring preferences for STI testing services in England have generally excluded cost since the NHS provides treatment 'free at the point of delivery' 20 21 A cost attribute was therefore not included.
The questionnaire was developed using SAS 9.4 software to ensure that the design was Defficient. 27 A full factorial design was ruled out in favour of a fractional factorial design because a full factorial design would have contained 3,072 possible alternatives, which would have been unmanageable in practice for individuals to complete or for a blocked questionnaire format to handle 31 . The smallest 100% efficient design that could be created included 48 choice sets. These were blocked (halved) into two questionnaires each with 24 choice sets using SAS JMP Pro 9.2.0. Choice set 1 was repeated as choice set 25 in each questionnaire to provide an internal validity check.
Three rounds of cognitive testing (n=9) were undertaken to check respondents' comprehension of information when making choices. Cognitive testing, undertaken based on two questionnaires of 24 sets, confirmed that a study based on two such questionnaires was acceptable to participants. Some modifications were made to levels where reasons for choice selection demonstrated that one level (e.g. 8 in 100 false negative) dominated the reason for selection. Cognitive testing also identified that implausible combinations (e.g. a postal test providing a result in two hours) impacted on completion of the task, whilst unlikely (but feasible) combinations did not. Only implausible combinations were therefore excluded from the choice sets. The D-Efficiency of the design 27 calculated by SAS JMP Pro 9.2.0, prior to the removal of implausible combinations, was 98.06. The final combination selected for the questionnaire was the one which created no duplicate choices and which provided an equal balance of the number of choice sets containing overlap between questionnaires.
Respondents were asked to provide socio-demographic information, including their age, gender, ethnicity, region of residence, sexual preference, relationship status and whether they had previously been tested for an STI. An introduction to the DCE questionnaire provided background information and an explanation of the attributes and levels. The introduction was included in the cognitive testing rounds to check comprehension within the target age range. As a result, the text was modified and diagrams added to illustrate the chlamydia testing pathways. The final phase of testing demonstrated that the number of choice sets was acceptable to respondents. The use of an online panel also provided completion time data to support the internal validity checks, and enabled an accurate record of time taken to complete the survey.

Final Attributes and Levels Selected
The six attributes and 24 levels selected for inclusion in the final DCE questionnaire are shown in table 1. More detailed information used to explain these attributes and levels is provided in Supplementary Information File 2 (pages 3-7).

Participants and Recruitment Procedure
Participants were drawn from the general population rather than from health care settings, thereby providing access to young people aged 16-24 years who had not previously been tested for an STI. An online national panel (YouthSight) 32 was used to maximise geographical reach and generalisability. Ex-ante sub-group analysis was planned by three age bands (16-18, 19-21, 22-24) and by gender (male, female). Where sub-group analysis is planned, it is recommended that there should be a minimum of 200 respondents in each sub-group, so a sample size of 1,200 was required in our study, 600 per questionnaire. 33 Consent was obtained online prior to questionnaire completion. Participants were offered a small reimbursement of one point (equivalent to £1) for completion of a survey of up to 20 minutes in length. Points could be exchanged for shopping vouchers.

Statistical Analyses
The multinomial logit (MNL) model developed by McFadden was used for analysing responses; this is recognised as the convention for three-response choice set studies ('option A', 'option B' or 'I would not test'). 26 34 Analysis utilised STATA13 SE with the method and code outlined by Ryan et al 23 Analyses presented odds ratios (ORs), 95% confidence interval (CIs) and coefficients for each attribute level. Variables within the model were all treated as categorical variables for the analysis. Attribute levels were specified using dummy coding, the preferred form of coding where ORs are to be calculated. Within the model, the levels that were dropped to form the reference levels reflected those aligned to a 'typical' sexual health clinic pathway (summarised in table 2). To test the internal validity of questionnaire responses, analyses compared full results against: (i) removal of responses where participants did not answer the repeated choice set consistently; (ii) removal of any respondents who took F o r p e e r r e v i e w o n l y 8 less than the minimum time (five minutes) observed in cognitive testing to complete the questionnaire; and (iii) removal of responses containing the opt out question data. Further tests for internal consistency and rationality were not included, since excluding responses on this basis may be viewed as an inappropriate imposition of rationality. 35 Demographic characteristics (gender, age and ethnicity) of respondents were compared with national Census data. 36 The influence of patient level characteristics (age, sex and STI testing history (yes/no)) on the likelihood of not choosing to test was examined. In addition to exante planned sub-group analyses, if sufficient responses were received analysis was also planned to compare: (i) respondents who had, or had not, previously tested for an STI, and (ii) those who indicated their relationship status as 'single' versus those in a sexual relationship with one person.
Trade-off between accuracy and time to result was examined by considering the probability of uptake for tests with characteristics at the opposite ends of the spectrum i.e. 'lower accuracy (5% false negatives), faster time to result (30 min)' and 'higher accuracy (2% false negatives), longer time to result (14 days)'.

RESULTS:
In total, 1,230 fully completed questionnaires were received, the platform analytics showed that 460 people had opened the questionnaire but did not complete it, providing a completion rate of 73%. No further information was available on the demographics of the 460 non-responders nor any information on the point at which they chose to exit the survey, so these people could not be included in our analyses. Time to complete the 25 choice sets ranged from one minute 19 seconds to 30  OR values for the full dataset and subgroups analysed are presented in table 3 with reference levels (1.00). Analysis of the trade-off between accuracy and time to result is presented in  There are also differences in the strength of preference between males and females for this attribute (OR 2.951, 95% CI 2.807-3.101 and 3.570, 95% CI 3.396-3.753, respectively), and between those who had previously tested or not tested (OR 3.000, 95% CI 2.820-3.191 and 3.482, 95% CI 3.331-3.640, respectively). Time to result was the attribute showing the next strongest preference across all subgroups (OR 1.806, 95% CI 1.711-1.906). These results are consistent with the logical expectation that people will prefer higher accuracy and a shorter waiting time. Looking specifically at the trade-off between accuracy and time to result, table 4 indicates that participants are willing to wait noticeably longer in order to have a test result with a lower chance of a false negative result.
When considering how to test, all subgroups demonstrated a preference for self-testing (OR 1.618, 95% CI 1.514-1.729) over attendance at a sexual health clinic. Testing via an outreach service in an educational/ work setting was found to be the least preferred option (OR 0.821, 95% CI 0.773-0.872). Respondents showed a consistent strength of preference for those options that do not involve direct contact with healthcare professionals (self-testing and postal self-sampling), compared with those that do. For consultation and treatment following a positive test result, there was a preference for non-sexual health clinic pathways with online consultation to access treatment (OR 1.212, 95% CI 1.150-1.277), treatment via general practice (OR 1.183, 95% CI 1.123-1.246) and treatment via pharmacy (OR 1.158, 95% CI 1.100-1.220) preferred in the full dataset. At subgroup level (age, gender, previous testing history and relationship status), more variation was found in the preference order for this attribute, with the exception of the sexual health clinic which was consistently the least preferred option across all subgroups (see table 3 and Supplementary Information File 3, Part  3). Finally, the full dataset shows that, if someone wants to access a healthcare professional, there was no statistically significant preference for instant messenger or email access compared with accessing the professional face-to-face. Telephone access to a healthcare professional was the least preferred access option (OR 0.949, 95% CI 0.903-0.998), and the only statistically significant result when compared with face-to-face access (the reference level). There was, similarly, no statistically significant difference in preferences for how young people might access antibiotics apart from a slight preference for the pharmacy versus sexual health clinic (OR 1.075, 95% CI 1.018-1.134).

DISCUSSION:
Our findings indicate that, based on the levels included in the DCE questionnaire, young people are willing to wait in order to have a chlamydia screening test result with a lower chance of a false negative result. The conclusion for new test developers is that time to result is less important than accuracy, and that test users are unlikely to prefer a point-of-care or self-screening test with lower accuracy than the tests currently available to them. There was a strong preference for remote access to testing, consultation, and antibiotic prescriptions, although for accessing a health professional there was no preference between online and face-to-face methods. This suggests a remote online pathway is acceptable to young people, as long as test performance remains equivalent.
In the various hypothetical situations presented, respondents showed a preference for chlamydia self-testing, self-sampling and postal testing over attendance at a testing location. For accessing treatment, a general preference was exhibited for online versus traditional GP or pharmacy services over clinic services. For receipt of antibiotics, there was little difference in preferences. We were also able to identify which attributes people may be willing to trade to maximise their utility. Looking at the trade-off between accuracy and time to result, we found that young people are willing to wait noticeably longer in order to have a test result with a lower chance of a false negative result, reinforcing the need for test equivalence.
The strengths of this DCE study include, firstly, the robust methods employed to select attributes and levels, which aligned with the recommendations of Coast et al, ensuring attributes are "manipulable in policy". 38 Secondly, the fact that participants were drawn from the general population targeted by the National Chlamydia Screening Programme, rather than from healthcare settings. This enabled us to access a demographically and geographically representative national sample of young people, including those who have had no previous contact with STI services. Finally, the use of an online panel enabled large scale data to be collected at a reasonable cost and allowed validity checks that would not otherwise have been possible with postal questionnaire responses. The large sample size also allows comparison between several subgroups to explore differences based on age, gender, previous testing history and relationship category.
However, the study does have a few limitations. The first relates to the selection of attributes and levels. Whilst the selection process employed was very rigorous, this cannot detract from the fact that further attributes were identified which might impact on individuals' choices. To mitigate against this, where such an attribute was excluded, information was provided in the survey background section to minimise respondents forming their own views on the impact of this attribute on the pathway.
Use of an online panel also provided accurate records of the time taken to complete the survey, and permitted additional validity checks which would not otherwise have been possible with a written questionnaire. On the other hand, use of an online panel excluded young people who do not have access to the internet, thereby potentially over-representing the acceptability of online care. However, given the extremely high proportion (97%) of 15-24 year-olds accessing the internet daily via a mobile device 39 and owning a smartphone 40 , it is evident that the vast majority of the target population could access online care pathways if they choose to do so. The question of whether these young people have the degree of digital and health literacy needed for online testing and treatment was not explored in this study. Finally, information was unavailable on young people who opened but decided not to complete the questionnaire, so systematic comparison with those who responded was not possible.
Behavioural factors, such as embarrassment 7 , stigma 41 and privacy and anonymity concerns 42 43 are known to influence uptake of sexual health services, as well as structural factors such as convenience and perceived barriers to access 8 ; many of these may be lower for a non-faceto-face service. Balanced against this, access to online testing and treatment is valued less strongly than attributes such as test accuracy and time to result. In individual cases a young person's preference for remote versus face-to-face testing and treatment might differ.   21 both only focussed on preferences within existing traditional service delivery models, and did not incorporate any hypothetical future scenarios (e.g. point-of-care testing, self-testing or treatment via eHealth/mHealth solutions). Two other DCE studies, which did consider self-sampling at home for chlamydia screening 18 19 , with sample sizes 174 and 126 respectively, both described sending the sample to a laboratory for analysis rather than a selftest. One of these studies 19 did identify a stronger preference for attendance at a family planning clinic, rather than self-sampling, but since participants were recruited from the waiting room of such a clinic they cannot be considered representative of the general population targeted by the NCSP. A fifth study, which examined preferences for point-of-care testing, only surveyed clinicians undertaking STI testing (sample n=218), thus excluding the population actually targeted for chlamydia screening. 44 The only stated preference study that has considered patients' preferences for STI self-testing focussed on HIV (sample n=365). 45 The authors reported results in line with our findings, with respondents exhibiting a preference for tests which are accurate, timely and private/ anonymous. However, this study was undertaken in 2002 prior to smartphones and at a time when self-testing for HIV was still under development. Importantly for screening tests, only one DCE study to date 21 has sought to include non-service users (i.e. populations with no experience of STI testing). The authors identified a preference for testing for all STIs, in settings with healthcare professionals with specialist knowledge present, and for receipt of negative as well as positive results. This study used a convenience sample of 233 students from two universities which was unlikely to be geographically or socio-economically representative. All other studies have drawn their DCE samples from people who were either current service users, or attendees for other, linked services. Use of such samples represents a significant short-coming when considering the introduction of new technologies for asymptomatic chlamydia testing and management. Most sample sizes were also smaller than in the present study.
A number of questions remain which our research does not address. Firstly, recognising that the range of STIs included is an important consideration for young people in choosing to test 20 21 , further research is required to understand this better in the context of potential new screening pathways which incorporate other STIs (e.g. gonorrhoea). Secondly, our study highlights a number of methodological considerations where there is an absence of consensus that may warrant further exploration to improve consistency. These include the number of choice sets to include in a DCE and the use of repeated choice sets as an internal validity measure. Finally, given that cost was not included as an attribute in this study, it is not possible to provide an indication of willingness to pay, or the monetary benefits of potential service changes, from this DCE. 46 The present study is, to our knowledge, the first to present a large scale, quantitative analysis of young people's preferences for attributes of potential new pathways to deliver testing and treatment of asymptomatic chlamydia, based on a nationally representative population. The DCE methodology applied also produced a measure of the relative strength of preference between different attributes and levels, and potential trade-offs. This can provide useful evidence to technology developers, policy makers, commissioners and service providers. In particular, it provides a first insight into preferences for the type of technologies currently under development, and those which might be available for use in the near future, compared to the features of existing products and services. This can indicate how young people may respond to changes in pathways and to the introduction of new technologies. Within the context of current UK sexual health policy and commissioning of sexual health services 12 , this DCE provides supportive evidence for the policy direction of remote chlamydia testing and treatment. However, whilst young people overall expressed a stronger preference for attributes such as self-test, online consultation etc, a small proportion still preferred existing pathways. This suggests that, in order to maximise benefit, face-to-face services should continue to be available in addition to any online screening and treatment service. This will ensure that services are inclusive and accessible irrespective of digital/ health literacy, while recognising that people's needs and preferences may change depending on their personal circumstances.

Supplementary File 1 -Selection of Attributes and Levels
Four research stages were used to determine the attributes and associated levels included in the Discrete Choice Experiment.

Inclusion/ Exclusion Criteria
The inclusion criteria were identified as:  any stated preference study within the scope of STI testing and treatment services. This included but was not limited to products (e.g. tests, drugs, condoms, microbicides) and services (e.g. screening and screening programmes, and service providers e.g. GPs, CaSH clinics and GUM clinics);  There was no date limiter, with all published studies included to end of 2014.
Exclusion criteria included studies:  not related to humans;  not published in English;  from outside of the OECD High Income Countries;  not related to the diagnosis or treatment of STIs (e.g. vaccinations).

Search Strategy
The following databases were searched on 28 April 2014 to identify studies published to the end of 2013, the saved database searches were re-run in April 2015 to search for any studies meeting the inclusion criteria published between 1 January and 31 December 2014 and no further studies were identified:

Literature Review 2 -Preferences for and acceptability of mainstream sexual health services 1.3.1. Objective
The objective of this literature review was to identify which factors might influence individuals' decisions to access testing and treatment services for STIs.

Inclusion/ Exclusion Criteria
The inclusion criteria were identified as:  any study which indicates individuals' preferences or acceptability of STI testing and/ or treatment services;  studies published between Jan 2004-Sept 2014;  conference abstracts, where the abstract enabled the extraction of information on study focus and key findings.
The date range for the literature review was selected to limit the volume of results identified to the last 10 years. This recognised the fact that the previous literature review only identified one stated preference study which met the inclusion criteria pre-2004, and that none of the studies identified included relevant new technologies e.g. internet or smartphone based services.
The exclusion criteria were identified as any study:  not published in English  not related to humans  not related to preferences for sexual health services  from outside of the OECD High Income Country List  not related directly to testing and/ or treatment provision e.g. drug characteristics, health promotion interventions  not offering a perspective provided by service user/ potential service user e.g. clinician  focused on non-mainstream service provision e.g. STI testing in A&E, dedicated service provision for specific high risk groups such as men recently released from prison, sex workers, injecting drug users.

Search Strategy
The key search terms included were:  Sexually transmitted infections, sexually transmitted diseases, sexual health  Test, treatment, service  Patient preference, acceptability, choice, uptake, access.
Individual STIs (e.g. Chlamydia) were not included in the search strategy, as it was determined that attributes and levels could potentially be identified from any STI/ sexual health product or service.
A smaller selection of databases was chosen for this scoping review, with the databases selected being those which had generated a high return rate of relevant articles in the first literature review and encompassing journals where research relating to nursing, AHPs and psychology is published. The three databases searched were:  Medline  CINAHL  PsycINFO.

. Sampling
Convenience sampling was ultimately chosen for this research due to the challenges with accessing the population and being able to undertake purposive sampling within the time constraints of this phase of the research.

Data Management & Analysis
Thematic analysis was selected as the core approach to analyse the focus group data recognising that this is one of the three 'main methods' for analysing focus group data and is the method most closely aligned to that used to design the focus group.

Objectives
The objectives of the expert groups were to consider whether:  The potential attributes met best practice requirements as defined by ISPOR Conjoint Analysis working group (Bridges et

Methods
Expert groups were run as focus groups to enable discussion around a series of key questions. The key issues raised by each of the expert groups were summarised by attribute using a matrix. The expert groups were not recorded, fully transcribed or analysed thematically as their purpose was to build on the outputs of the focus groups and capture the salient points for consideration, which was achievable through the matrix coding.

Objectives
The purpose of the focus and expert groups was to inform the selection of attributes and levels for the DCE. Achieving a balance between attributes of importance to the study population and attributes which will deliver impact to services and technology developers was the key consideration in selection. This led to the decision that a formal ranking or consensus method should not be used with either the focus or expert groups as a combined view of both was required. Instead, the approach adopted was a narrative synthesis developed for implementation reviews to enable

Methods
The approach taken in the narrative synthesis is outlined below, adapted from the key elements of synthesis identified by Pope and colleagues (2007):

Element of Synthesis Approach Taken Developing a Theoretical Model
Identification of a checklist of properties against which attributes can be considered to inform the selection process

Developing a Preliminary Synthesis
Tabulation against the checklist -to create a matrix of checklist criteria against attributes to enable visual comparison of perspective Exploring Relationships in the Data

Conceptual mapping and triangulation against the checklist Assessing the Robustness of the Synthesis Product
Critical reflection on the synthesis process Please read through the following background information before completing the survey. It contains information about chlamydia and the terms that are used in the survey.

What is Chlamydia?
Chlamydia is the most common sexually transmitted infection in England and the majority of infections are found in young people aged 15-24. Both men and women can get chlamydia, but most people with chlamydia have no symptoms and do not know they have an infection. The test for chlamydia is usually a urine sample for men and a vaginal swab for women. Once diagnosed, chlamydia can be treated with a single dose (2 or 4 tablets) of an antibiotic called Azithromycin.
In England there is a national chlamydia screening programme which recommends that people aged 16-24 are tested for chlamydia annually or when they change their sexual partner.

What happens if I don't get treatment?
If left undiagnosed chlamydia can lead to serious health problems.
In women this includes a condition called Pelvic Inflammatory Disease, which is an infection of the womb, fallopian tubes or ovaries. This can cause severe pain and in some cases (around 1 in 10) lead to infertility or ectopic pregnancy (where the pregnancy occurs outside of the womb in the fallopian tube). If you are pregnant and have chlamydia you can pass it on to your baby when you give birth and this can lead to the baby getting an eye infection or pneumonia. In men chlamydia can lead to a condition called epididymitis, which causes soreness and swelling in the scrotum.

For further information
If you would like further information on chlamydia and how to prevent it, or any of the other conditions described please visit: http://www.nhs.uk/conditions/chlamydia/Pages/Introduction.aspx

About the Survey…
In completing the survey you will be presented with a number of choices for getting tested and treatment for chlamydia. Please consider each set of choices and indicate whether you prefer option A or option B or whether you would not test, that is, you would not choose option A or option B.
Each choice involves two different scenarios for testing and getting treatment and there are differences between how you test, how long you have to wait for the result, how accurate the test is, how you have a consultation to get treatment for chlamydia, how you access a healthcare professional to get that treatment and how you get your antibiotic.
The following sections provide you with an explanation of the process of getting tested and treatment and the key terms used in the survey: The diagram below shows the stages up until you find out our result:  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   3 The following sections provide more information on how you test, how long you wait for the result and the accuracy of the test.

How you test for Chlamydia
This focuses on how you get your test, how the sample is taken and what happens to the sample once you've taken it. There are six options in the questionnaire, an explanation of these is provided below:  Self-Test -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and interpret the result yourself (like a pregnancy test). This is different to the self-sample options below because you interpret the result yourself rather than it being interpreted by a healthcare professional.  Self-Sample and post off for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and send the sample in the freepost envelope to the laboratory for analysis. This is different to the self-test option above because the test is interpreted by a healthcare professional.  Self-Sample and take to pharmacy for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and then take the sample to a pharmacy for analysis. This is different to the self-test option above because the test is interpreted by a  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   4 healthcare professional. Opening hours vary between pharmacies but many in towns and cities are open evenings and weekends as well as weekdays.  Self-Sample and take to your place of education/ workplace for analysis -Order a test kit online or collect one from a community location e.g. pharmacy or supermarket, provide the sample yourself and take the sample to your place of education/ workplace for analysis by an outreach nurse. This is different to the self-test option above because the test is interpreted by a healthcare professional. Outreach services in education and workplaces are usually available one day a week.  Attend GP Practice, sample taken by a GP or Nurse -Book an appointment at your GP practice with a GP or practice nurse. Your sample will be taken by the healthcare professional you see.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  The following sections provide more information on how you get your treatment, how you contact a healthcare professional and how you get your antibiotics.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

How you Contact a Healthcare Professional
New technology means that there are now more options for completing your consultation to get treatment, or in the case of the online consultation, accessing a healthcare professional for advice. There are four options within the questionnaire:  Telephone -speak to a healthcare professional on the telephone.  Instant Messaging -have a discussion with a healthcare professional via an instant messaging service online  Email -have a discussion with a healthcare professional via email. During opening hours you will receive a response within 2 hours, and for emails sent out of hours you will receive a response the next day.  Face-to-face -have a face-to-face discussion with a healthcare professional.

Study design 4
Present key elements of study design early in the paper (page 5-6) Describe any efforts to address potential sources of bias (page 8)