Identifying Critical Attributes and Attribute-Levels for a Discrete Choice Experiment on Oral PreExposure Prophylaxis (PrEP) Delivery Among Young People in Cape Town and Johannesburg, South Africa.

Background Oral Pre-Exposure Prophylaxis (PrEP) was approved for use in South Africa (SA) in 2017. The uptake and adherence of daily oral PrEP has been poor in high-risk populations in SA including adults and young people. We used a Discrete Choice Experiment (DCE) technique to identify critical attributes and attribute-levels to understand user preferences for daily and on-demand oral PrEP use among young people, including adolescents, in two South African cities. Methods Qualitative data were collected between September and November 2018 from 8 group discussions and 20 in-depth interviews with young people aged 13 to 24 years in Cape Town and Johannesburg. Participants were selected based on stratied (gender and age) purposive sampling. Using a semi-structured interview guide, interviewers discussed several attributes (dosing regimen, location, costs, side effects, and protection period) for PrEP access and use. Transcripts were coded to identify context-specic attributes and attribute-levels for delivering oral PrEP in South Africa. The adolescent community advisory board, expert and study team opinions were consulted for the nal DCE attributes and levels. Results We enrolled 74 participants: 51% (n=38/74) male, median age of 18.5 [IQR=16-21.25] years, 91% (n=67/74) identied as heterosexual and 49% (n=36/74) had not completed 12 th grade education. Using the qualitative data, we identied ve candidate attributes including (1) dosing regimen, (2) location to get PrEP, (3) cost, (4) route of administration and (5) frequency. After discussions with experts and the study team, we revised the DCE to include the following ve attributes and levels: dosing regime: daily, and on-demand PrEP; location: private pharmacy, public clinic, mobile clinic, ATM); cost: free-of-charge, R50 (~2GBP), R265 (~12GBP); side effects: nausea, headache, rash; and duration of protection: fulltime protection versus when PrEP is used).


Background
The prevention of Human Immunode ciency Virus (HIV) in young people in South Africa remains a priority, rstly because it is a preventable chronic condition which would in turn negate the risk of associated physical and psychological co-morbidities, and secondly because of sustained high HIV prevalence and incidence among adolescents in both South Africa and sub-Saharan Africa (SSA). Moreover, interventions for HIV prevention within SSA have had limited effect, despite high levels of HIV awareness. The limited effect of HIV prevention interventions is largely attributable to poor adherence observed in major clinical trials [1][2][3][4][5][6][7].
Globally, pre-exposure prophylaxis (PrEP) shows e cacy in preventing HIV amongst men and women [8][9][10], with a daily PrEP regimen shown to be effective and associated with decreasing HIV incidence in heterosexual men and women, men who have sex with men (MSM), transgender women, and people who inject drugs (PWID) [2,9,[11][12][13]. This regimen has subsequently been recommended by the World Health Organization (WHO) for individuals at high risk for infection [2]. On-demand PrEP, which consists of four pericoital doses, represents a potentially more cost-effective approach [14]. However, the on-demand regimen is not yet recommended for populations other than MSM due to the lack of studies and the complexity of the regimen [15].
PrEP was made sequentially available in the public sector in South Africa in 2016 to sex workers, MSM, sero-discordant couples, and most recently to young women and adolescent girls [16]. However, there is still a lack of awareness and uptake of PrEP among young South Africans [17]. The expansion of PrEP is urgently required across SSA, but can only be optimised by understanding barriers to PrEP uptake in young people. Currently, there is no model of PrEP provision for adolescent boys who have been overlooked in HIV prevention research [15] but contribute to HIV incident infections. Emerging HIV epidemics in young men have been identi ed in urban communities around Soweto and Cape Town, South Africa [18]. There is concern that PrEP uptake amongst young men will be low given the disproportionally low PrEP uptake by young black men (10.1%) in the USA relative to their high HIV incidence [14]. This disparity between risk and PrEP uptake is no doubt multifactorial but identi es the need for young men to be included in studies, along with young and older women to improve PrEP uptake as part of HIV prevention messaging and services. Furthermore, empowering men and women from adolescence to use preventative methods such as PrEP will normalize the use of this strategy through their lives.
Discrete choice experiments (DCEs), also known as conjoint analysis, can direct researchers to user needs, and inform how to best formulate or implement a health intervention, as DCE allows researchers both to identify attributes [19] of an intervention that make it attractive and also to rank the importance of that attribute. In a DCE, interventions that are being investigated are deconstructed into their characteristics, which are referred to as attributes [20]. By allowing to infer important factors of individual behavior and predict decision-making under different conditions, the results from DCEs are critical in predicting the success of different approaches to implementing particular healthcare interventions in relevant communities or populations [21,22].  [25]. There has also been an increased interest in using DCEs to explore stakeholder preferences for healthcare interventions and changes in policy, as well as using them to support the prioritization, design and implementation of healthcare interventions [20]. The use of the DCE technique provides an opportunity to identify critical attributes and attribute-levels to understand user preferences for PrEP uptake among young people. The development of a DCE is a rigorous process of developing questions, conducting cognitive interviews, and modifying the questions as one goes along. The use of qualitative data allows using information from the target population to identify critical attributes and user preferences that are also context-speci c.
The Combined HIV Adolescent Prevention Study (CHAPS) aims to investigate the acceptability and feasibility of implementing daily and on-demand PrEP among adolescents in South Africa, Uganda and Zimbabwe and to determine an on-demand PrEP dosing schedule for insertive sex. As part of the social science component, we aimed to develop DCE choice sets to identify attributes and attribute levels for PrEP delivery among young people in South Africa.
The DCE technique can be simpli ed into four components: survey design, piloting and administration to participants, data analysis, and implement interventions for the relevant populations, or design further experiments [20,26]. In this paper we focus on the preliminary processes and data collected to design the DCE for survey admistration. Therefore, the aim of this paper is to describe the process of developing the DCE choice sets and the salient attributes and preferences for PrEP delivery in young people aged 13-24 years in two South African cities.

Study setting
Data were collected from September to November 2018 at the Perinatal HIV Research Unit (PHRU) in Johannesburg and the Desmond Tutu HIV Centre (DTHC) in Cape Town. Participants were recruited in informal peri-urban communities established as residential areas for black South Africans during apartheid, characterized by overcrowding, unemployment, with limited resources and service delivery [27].
The PHRU is situated at the Chris Hani Baragwanath Academic Hospital in Soweto, has conducted research in HIV prevention, care and treatment for more than 20 years as well as providing healthcare services for the population of Soweto. Soweto is southwest of Johannesburg city in the province of Gauteng, and is estimated to have a population of ~ 2.5-3 million in an area of 63 km 2 [28]. The prevalence of HIV in the Gauteng province is 15.2% among 15-49 year olds [29]. The second site, the DTHC is located in the Klipfontein/Mitchells plain sub-district, in the Southern part of the Cape Town Metro. Compared to other district municipalities within the Western Cape, the Cape Town Metro bares the highest population size and prevalence of HIV [29]. Both PHRU and DTHC have a well established Adolescent Community Advisory Boards (ACAB), who are deeply rooted in their respective communities.

Initial literature review
A literature search was conducted to start the attribute development process using the terms in Discrete choice experiment (DCE), Sub-saharan Africa (SSA), South Africa (SA), pre-exposure prophylaxis (PrEP), young people. The terms were used in a variety of combinations. While there has been a DCE-based study that included preferences concerning PrEP in South Africa [21], this study conducted in Johannesburg examined a variety of other HIV and pregnancy prevention methods along with PrEP, and focused only on South African women and adolesdcents girls in diverse settings. For our study, based on literature, important attributes for delivering PrEP to young people included the cost of PrEP, the location of PrEP dispensaries, and dosage frequency.
Procedures of the qualitative study for developing the DCE As part of the larger CHAPS study, qualitative data were collected between September to November 2018 through eight group discussions (GDs) and 20 in-depth interviews (IDIs) with young people aged 13 to 24 years; four GDs and 10 IDIs in Cape Town and Johannesburg, respectively. In total, 54 participants took part in GDs; 28 in Johannesburg and 26 in Cape Town and a total of 20 participated in IDIs.

Study population, sampling and recruitment process
Fieldworkers used a convenience sampling strategy to recruit eligible men and women 13 -24 years, who were HIV negative on a rapid HIV test (Soweto used the ABON HIV 1/2/O Tri-Line Human Immunode ciency Virus Rapid Test Device and Cape Town used the Alere Determine test) and reported sexual activity in the past three months. Most participants in our study had moderate to no prior knowledge about PrEP. In Soweto, participants for IDIs and GDs were actively recruited through community organisations, taxi ranks and outside premises of public meeting places such as shopping malls and community parks. In Cape Town, participants were recruited through local schools, shopping malls, the library, word of mouth referrals and walk in at the research clinic.

Data collection procedures
IDIs and GDs were strati ed by age and gender across both sites, but the GDs also included mixed genders in Cape Town. Prior to the interviews, participants completed a short ve-minute sociodemographic questionnaire [Additional File 1]. GDs and IDIs were facilitated by experienced and trained researchers uent in English and local languages used (e.g. IsiZulu, IsiSesotho and IsiXhosa). They were conducted in secure and private locations at both sites. During the GDs, two researchers were present, one served as the main facilitator and the other as a note taker and co-facilitator. A semi-structured guide including open ended questions and relevant probes was used to facilitate and guide the discussions during the IDIs and GDs [Additional File 2]. Participants were allocated a participant identi cation number, which was used while completing the demographic questionnaire and during the GDs and IDIs. Participants used pseudonyms during all discussions to maintain anonymity. The IDIs and GDs lasted 36 minutes on and 100 minutes on average, respectively. All IDIs and GDs were audio-recorded, transcribed verbatim, and later translated to English.

Measures
The brief socio-demographic questionnaire included questions on gender, age, race, primary home language, sexual orientation, and education level.
Trained qualitative researchers used a semi-structured interview guide to conduct the GDs and IDIs. The guide covered topics related to identify several attributes for PrEP access and use, including: choice of dosing regimen (i.e. daily versus on-demand PrEP), alternative routes of PrEP administration (i.e. implant and injection), dispensing location (pharmacy, clinic, adolescent/youth centre, doctors o ce), and time to be spent in obtaining PrEP.

Ethical approval
This study was approved by the Human Research Ethics Committees of the University of the Witwatersrand and University of Cape Town. Written study informed consent or assent was obtained from all participants. In Soweto, parents/ legal guardians provided written informed consent for participants younger than 18 years. Participants and parent/legal guardian signed separate forms for audio-recording. Cape Town had a parental waiver and did not require parental consent for participants under 13-17 years. Therefore, for Cape Town, participants provided written consent regardless of age. Participants were reimbursed R100 (~ 5GBP) in Johannesburg and R120 (~6GBP) in Cape Town respectively for their study participation.

Data analysis
Qualitative data analysis A framework analysis approach was used to analyse the data [30]. Transcripts were coded to identify context-speci c attributes and attribute-levels for delivering oral PrEP in South Africa. Initial coding was conducted by two trained qualitative research team members at each site. Initially, the researchers read through four transcripts (two GDs and two IDIs) to gain an understanding of the data to develop an Excel spreadsheet codebook. Each transcript was then individually coded using a line by line technique to assign codes to text. A second analyst also coded the same transcripts using the line by line technique to identify codes. Both analysts then met to discuss codes identi ed and differences were discussed. If there was a disagreement among codes, this was discussed until a common agreement was reached Codes were then categorized into themes and sub-themes by both researchers. Themes identi ed were captured on an Excel spreadsheet and a comprehensive codebook was developed. The codebook was shared with the rest of the team from both sites for nal review and input.
Quantitative data analysis Descriptive statistics and frequencies were used to analyse socio-demographic characteristics using Statistical Package for Social Sciences (SPSS) Version 25.0 [31].
Qualitative analysis of attribute identi cation Table 1 illustrates the full list of all attributes and attribute-levels, which were identi ed in consensus among the two qualitative research analysts. Initially, eight candidate attributes with respective attribute levels were identi ed including (1) Frequency of administration, (2) side effects, (3) cost, (4) location of PrEP dissemination, (5) person responsible for dispensing, (6) duration of PrEP use, (7) pill intake, and (8) alternative routes of administration. Attributes and attribute-levels were identi ed from the transcripts and prominent participants' quotes were directly extracted to illustrate each attribute and attribute-level (Table 1). Injection 20 Frequency of administration: The majority preferred a daily PrEP dosing regimen over an on-demand PrEP dosing regimen, citing unplanned sexual intercourse as one of the main reasons. Substance use was also identi ed as a factor that leads to unplanned sex.
"The reason is because you don't choose when you will have sex just like Esihle mentioned that things such as Nandos, drugs and alcohol and the next thing you are kissing leading to sex." (GD, Females, 18-24 years, Soweto) "Because I don't know when I'm going to have sex. I don't plan to have sex it just happens." (IDI, Female, 13-17 years, Cape Town) There were however some participants who preferred to take PrEP on demand stating that they might not remember to take a pill daily and and that it would be easier to take PrEP only when they know that they were going to engage in sexual intercourse.
"I don't see the reason why I should take it [PrEP] when I know I am not going to have sex, I am not going to be at risk." (IDI, Male, 13-16 years, Cape Town) "How am I going to remember to take the pill so I would rather take it [PrEP] as an on demand." (GD, Females, 18-24 years, Soweto) Side effects: Some participants spoke of side effects from personal experience and some from information shared by people using PrEP. Participants stated multiple side effects, including, headaches, fatigue and "sleepiness", that could be experienced when taking PrEP. They expressed that some side effects might only be experienced for a short period, such as nausea, vomiting, stomach aches and diarrhea, especially when the body is not used to the drug. However, some participants preferred a young over a older health care provider to dispense PrEP.
"I would prefer them [HIV counselors] because they are educated to counsel you rst and make sure you are okay before you take something you don't know. Nurses can explain but they don't explain as well as a counsellor. A counsellor listens to you and you both listen to each other whereas a nurse just tells you what to do and they don't even want to know about your concerns." (GD, Females, 18-24 years, Soweto) "A person for young people who is able to reach out to young people, someone friendly. So it may happen that she handles such department. She has to be young person so that we girls are able to talk to her and boys can have a male to talk to." (GD, Males and Females, 13-24 years, Cape Town) PrEP dosing: Participants discussed different dosing periods for taking PrEP. These differed from the timeframe to take PrEP to be protected for the next sexual intercourse, to the time period PrEP should be taken once started on it.
"I would prefer if they take it [PrEP] once a day but for a lifetime." (IDI, Male, 18-24 years, Soweto) "I may maybe say I would take it [PrEP] ve days before [sex]. If I know that on that weekend I am going to be seen by my partner, something like that."(IDI, Male, 13-17 years, Cape Town) "It [PrEP] should be ten years just like the loop so that we can see how well people take to it. Like this thing that they put in your arm." (GD, Females, 18-24 years, Soweto) Pill intake: Overall, participants stated that they would be willing to take one pill or two pills per dose.
However, the majority preferred to only take one pill, especially when taking it over a longer period of time.
If taking two pills, participants would prefer to take one in the morning and the other in the evening.
"One a day…people will get bored in taking two or three pills." (IDI, Female, 13-17 years, Cape Towm) "It is easy to take one pill because if you take more than 1 pill the more you get discouraged to take it because you will be like so many pills why? Why do I have to take pills 3 times a day." (IDI, Female, 18-24 years, Soweto) Alternative route of administration: During the GDs and IDIs, alternative routes of administration were discussed. Participants reported they would chose an implant or injectable as it would only need to be administered once and would not require returning to the clinic as frequently compared to using oral PrEP.
"Since I am a forgetful person, I wish PrEP can be injectable just like contraception." (GD, Males and Females, 13-24 years, Cape Town) "Implant because injection people are scared of pain whenever we talking about injection comes pain in our heads but implant it is something which you do once off maybe then it last for a certain period which is intended for it to last." (IDI, Male, 18-24 years, Soweto)

Expert opinion
The study team consulted two experts experienced in designing and implementing DCEs in South Africa.
Through an iterative process of discussions with these experts four of the initial eight candidate attributes were retained with three attribute-levels, each, except for the dosing regimen attribute, which consisted of two attribute-levels, daily and on-demand PrEP. Table 1 indicates whether the attribute identi ed during the qualitative analysis process was retained after the discussion phase with the experts (last column). The discussion was aimed at selecting the most relevant attributes with a maximum of three attribute-levels each. This was also important to ensure the DCE incorporates a manageable amount of attributes and level options to ensure that participants were not overwhelmed with the number of choice sets within the DCE. Consultation with other local team members ensured the language used for the attributes and attribute-levels was understandable and relevant within the South African context. For the location of PrEP dissemination it was deemed relevant to include a third attribute level, mobile clinic, as this has been a feasible approach to provide integrated health care, counselling services and dispensing of medication in underserved communities in Cape Town [32]. The nal attributes and attribute-levels for the DCE included: dosing (daily and on-demand PrEP), location (private pharmacy, public clinic, mobile clinic), cost (free-of-charge, ZAR50 (~2GBP), ZAR200 (~9GBP)), and (4) side effects (nausea, headache, rash). After the iterative revision process, SAS was used to develop a statistically e cient and balances fractional factorial design of the DCE. Using SAS Enterprise Guide 7.15 (SAS Institute Inc., Cary, NC, USA) and the attributes identi ed in the qualitative study, the maximum possible number of attribute combinations was 2 3 × 3 1 [33]. A smaller factorial design was further developed using the D-E ciency criteria reducing the choice sets to a feasible number of 6. The nal DCE was constructed with a total of six binary choice sets with two scenario descriptions and graphics of the attribute-levels and participants can chose between Choice A and Choice B (Figure 1).

Pilot testing of the DCE with Community Advisory Board Members
To ensure a community engaged approach to the DCE development, the ACAB of the PHRU rati ed the DCE choice sets as developmentally and contextually appropriate [20]. ACAB members regarded the number of choice sets as manageable, easy to understand, user-friendly and easy to follow. In addition, ACAB members agreed the illustrating graphics included in the DCE choice sets were complementary, relevant and context-speci c for the attributes and related levels.

Discussion
We explored PrEP delivery preferences among young people from a peri-urban setting in a multisite context in South Africa to develop a DCE. There are few methodological literature that describes the process to systematically infer attributes and its levels for a DCE, which is considered an important technique to better understand and provide guidance on user preferences for interventions and products [20,22]. Previous research in South Africa and SSA that used a DCE focused on hypothetical HIV prevention methods [41] , HIV testing, HIV and family planning services [21]. One multinational study conducted in seven countries among key user groups aged 16 -³41 years to elicit acceptability of PrEP and factors to determine uptake [42].
Even though DCEs are increasingly used in the eld of sexual and reproductive health, there is limited research on young people's PrEP preferences in developing countries. Young people are considered a high risk group of acquiring HIV infection and important bene ciaries in accessing PrEP, yet PrEP uptake is low amongst South African youth [17]. Amongst a sample of 772 young people aged 16-24 years in the Eastern Cape, South Africs, only 1.7% had used PrEP [17]. Once young people initiate on PrEP, a substantial proportion do not persist with it [43]. Our documentation of the DCE methodology provided a unique opportunity for targeted investigation on how to best implement PrEP amongst this population to support uptake in resource-limited settings.
During the qualitative interviews, young people expressed an initial preference for daily over on-demand PrEP. Using daily PrEP would ensure protection for spontaneous or unanticipated sexual activities. Young people are often less likely to anticipate sex, which also includes communication and control within relationships [25]. This may be particularly true for young women, as they often have less decision making power and be less likely able to negotiate when sex occurs [44]. This might impact on chosing daily versus on-demand PrEP. In addition, young people are at high risk of non-adherence to medications, including contraception, ART or other chronic medications [45]. While it may be more di cult for young people adhering to a strict xed pre-intercourse and/or post-intercourse PrEP dosing regimen, it may also impact on optimal adherence to a daily PrEP regimen.
Participants in our study discussed the tradeoffs between locations to access PrEP with costs and con dentiality. Participants opted for accessing PrEP at their local clinics, mainly because services would be free. Others preferred to access PrEP at a pharmacy, for ease of access, privacy and con dentiality.
Stigma and discrimination are commonly known barriers to accessing public health care facilities and lead for young people to less likely engage in sexual and reproductive health care, including HIV services [46,47]. Therefore, alternative locations for accessing PrEP might promote PrEP uptake among young people and support privacy and con dentiality, which was an important aspect for young people in our sample when getting PrEP. During a South African based demonstration project amongst adolescent girls and young women, PrEP initiation was high in youth-orientated, family planning and mobile outreach clinics [43]. Offering PrEP at different types of locations could be a suitable delivery model for young people. This may mitiage additional barriers, such as transportation, long distance to dispensing sites, and healthcare provider attitutes [48,49]. In the South African setting, an integrated reproductive health care delivery model to access PrEP is feasbile amongst adolescent girls and young women [43].
Participants in our study discussed their willingness to use alternative modalities for PrEP administration including injectables and implants. Current research is underway that tests the effectivenss of long-acting injectable formulations and intra-vaginal rings and potential alternatives to oral PrEP [50]. Further evaluations are required that focus on the willingness and perceptions of alternative application methods of PrEP amongst young people.
To better ensure that the nal DCE was re ective of youth preferences we were able to incorporate a community engaged approach. The ACABs were actively involved in reviewing the choice sets for content, language, graphics and layout. This phase afforded the research team the opportunity to test the DCE choice sets supported response e ciency and content validity within the DCE [51,52]. A youth engaged approach has been shown to work well in other settings, such as interventions for mental health, substance use and HIV prevention interventions [53,54].
Next steps are to deliver the nal DCE in a larger CHAPS survey and then to assess the preferences of young people for PrEP delivery in two South African cities, Soweto and Cape Town. It is anticipated that the results will provide important insights on how to align the implementation of PrEP to be most accessible and bene cial for young people, who present a key population in the HIV prevention response.

Limitations
A few participants had prior PrEP experience, but the larger sample had limited to no prior exposure to PrEP. To avoid a potential bias based on those with PrEP experience, we included recruitment to locations outside of the clinic setting. Many participants that discussed topics related to PrEP access and use, drew from experience with other medication or from reports of others who took PrEP or where exposed to PrEP before. This was particularly true when they discussed the location to access PrEP and possible side effects. Limited prior knowledge on PrEP might have in uenced participant's responses. However, this was largely mitigated by the fact that discussions during the phase with experts con rmed the relevancy of the attributes and attribute-level identi ed during the qualitative interview phase.

Conclusions
There is limited methodological literature describing the step-by-step process of developing a DCE for PrEP in adolescents, especially in resource-constrained countries. Using a systematic and robust approach, we use a developed and designed a DCE with attributes and levels for PrEP uptake that can be tested in young people in SA. We provide the process followed for the DCE technique to understand user preferences for daily and on-demand oral PrEP among young people in South Africa. Availability of data and materials

List Of Abbreviations
The qualitative datasets generated and/or analysed during the current study are not publicly available to maintain participant privacy and con dentiality, but are available from the corresponding author on reasonable request.