Self-sampling kits to increase HIV testing among black Africans in the UK: the HAUS mixed-methods study

Stage 1

• Examine/evaluate barriers to, and facilitators of, provision, access and use of HIV SSKs by black African people, in primary care, pharmacies and community outreach settings.

• Determine appropriate SSK-based intervention models for different settings.

• Determine robust HIV result management pathways.

• Develop an intervention manual to enable intervention delivery.

Stage 2

• Determine the feasibility and acceptability of a provider-initiated, HIV SSK distribution intervention targeted at black African people in two settings:

  1. GP surgeries

  2. community-based organisations (CBOs).

Primary outcome

HIV SSK return rate.

Secondary outcomes

Point-of-delivery outcomes:

• acceptability of the targeted HIV SSK distribution

• acceptability and feasibility of the targeted SSK distribution among specified service providers.

Data collection outcomes:

• ability to record the number of people offered SSKs, accepting SSKs and returning SSKs

• feasibility of collecting correct contact details enabling follow-up, reminders and communication of results.

Pathway-to-care outcomes:

• proportion of those whose samples are reactive, who –

  • are informed of the results in person

  • attend for confirmatory testing at a NHS setting of their choice.

Overarching outcomes:

• cost per kit distributed and cost per HIV diagnosis per setting

• attrition rates

• confirmatory testing, proportion of those receiving a HIV-positive diagnosis and clinical stage at diagnosis

• feasibility and sensitivity of outcome measures (testing, behavioural and economic) for a definitive trial.

Search strategy and identification of studies

Ten electronic databases were searched using detailed search strategies:

1. OvidSP MEDLINE

2. OvidSP EMBASE Classic and EMBASE

3. OvidSP Global Health

4. OvidSP Social Policy and Practice

5. OvidSP PsycINFO

6. OvidSP Health Management Information Consortium

7. EBSCOhost Cumulative Index to Nursing and Allied Health Literature Plus with Full Text

8. The Cochrane Library

9. Web of Science Core Collection

10. Scopus.

The search strategy used for OvidSP MEDLINE is provided in Appendix 1.

Only studies written in English were included. The results were downloaded into a deduplicated database in EndNote 7 [Clarivate Analytics (formerly Thomson Reuters), Philadelphia, PA, USA]. The initial search was undertaken on 26 September 2014. Two further searches to update the database were undertaken on 17 April 2015 and 3 May 2016. Additional grey literature was retrieved from websites operated by the following organisations:

• Avert (www.avert.org)

• Terrence Higgins Trust (www.tht.org.uk)

• National AIDS Trust (www.nat.org.uk)

• Lambeth Council (www.lambeth.gov.uk/consultations/lambeth-southwark-lewisham-sexual-health-strategy-consultation)

• Naz Project London (http://naz.org.uk)

• Sexual Health Sheffield (www.sexualhealthsheffield.nhs.uk).

Inclusion and exclusion criteria

Only studies published since 1 January 2000 were included, as studies published earlier would be unlikely to reflect current technology or attitudes to HIV testing. Only studies conducted in the European Union/European Free Trade Agreement countries, North America, New Zealand or Australia were included, as studies conducted in other locations (particularly resource-poor settings) would probably have markedly different contexts and, thus, their results would not be applicable to the UK. Study populations that included lay groups, as well as health professionals, were included. Only studies that examined home/self-sampling tests for HIV were included as intervention studies. Studies without comparators were also included, as well as studies that compared home/self-sampling tests for HIV with routine service provision or other HIV testing interventions. Studies were included if they reported on any of the following outcomes:

• increase/decrease in the number of HIV tests

• proportion/number of confirmatory tests

• proportion/number of participants linked into care

• adverse events associated with HIV self-sampling

• proportion/number of false positives or failed tests

• increase/decrease in the reported history and frequency of taking HIV tests

• increase/decrease in the number and types of venue where HIV testing is offered.

Qualitative studies were included only if they reported one or both of the following:

• barriers to, or facilitators of, self-sampling reported by the general population

• barriers to, or facilitators of, self-sampling reported by service providers.

The following study designs were considered for inclusion:

• randomised or non-randomised controlled trials

• prospective or retrospective cohorts

• cross-sectional studies/prevalence studies

• pilot or feasibility studies

• qualitative studies (using in-depth interviews, FGDs and document analysis).

Studies that examined the use of, or views about, self-sampling for HIV in health-care workers were excluded, because the review focus was on uptake among testers, not service providers, as were all conference communications, because of insufficient detail and lack of peer review. Studies that focused solely on, or the outcomes of which were predominantly about, self-testing for HIV were also excluded at the study selection stage.

Study selection

Studies were selected using a two-stage screening approach. Reviewers Ibidun Fakoya and Esther Mugweni devised a checklist to independently screen titles and abstracts (see Appendix 1). When a consensus could not be reached about study inclusion, a third reviewer (FB) was consulted. Full-paper copies of the selected studies were screened and assessed independently by Ibidun Fakoya and Esther Mugweni, using a screening tool (see Appendix 1). Updated searches were screened using the same approach by Caroline Park, Thomas Hartney and Lisa McDaid. Inter-reviewer reliability scores of the different stages of the review were calculated using kappa in Microsoft Excel® 2010 (Microsoft Corporation, Redmond, WA, USA). The full-paper screening achieved a kappa score of 1.0, which indicates a high level of agreement between reviewers.

Data extraction, analysis and synthesis

Structured data extraction tools were developed to capture the required information from the included papers on study types, populations, SSK interventions and acceptability, feasibility and efficacy outcomes. Data were extracted by Caroline Park and checked by Thomas Hartney.

A meta-analysis was not conducted, as a result of the heterogeneity of the designs, methods, samples and outcomes of the included studies. There are a number of narrative approaches to data synthesis, including integrative synthesis, to primarily combine and summarise data, and interpretative synthesis that aims to generate new concepts and theory. 27 An integrative approach to summarising and presenting the data was appropriate to this review. The narrative synthesis is supported by tables in Results of the systematic literature review, which outline the key characteristics and findings of each included study, as relevant to the research questions. To reduce the risk of bias, the extracted data were first summarised by Lisa McDaid and then reviewed by Thomas Hartney. Disagreements in interpretation were resolved through discussion between the two authors.

Quality appraisal

The quality of the eligible papers was appraised by Thomas Hartney using the National Institute for Health and Care Excellence (NICE) quality appraisal checklist for quantitative papers. 28 Each quantitative paper was assigned a score for internal and external validity, from ++ (high quality) to – (poor quality). Each qualitative paper was assigned a single overall score. The appraisal process was validated by a second researcher (Lisa McDaid) assessing a sample of papers with a high level of agreement reached. Any papers that did not present self-sampling data separately from other forms of testing were excluded, but papers were not excluded on the basis of quality.

Non-specialist black African focus group discussions

Participants in the non-specialist FGD included members of the public who self-identified as being black African (n = 48). Three of the FGDs occurred in Greater Glasgow, and three occurred in Greater London. The participants were recruited via social media (n = 1) and African embassies in London (n = 6), as well as university student groups (n = 16), and CBOs (n = 24) in both Glasgow and London (missing data, n = 1). Participants were eligible if they self-identified as being black African and were aged ≥ 18 years. The sample was purposively selected sequentially during recruitment (with some interested individuals being set aside into a ‘pool’ of recruited participants in case they were needed at a latter stage) to ensure diversity of age, region of birth and HIV testing experience. Men were slightly over-represented in the sample, with 28 out of 48 male participants (58%), compared with 20 female participants (42%). The age ranged from 18 to 60 years. Participants were born in various regions of Africa, including East Africa (n = 17), Southern Africa (n = 10), West Africa (n = 10) and Central and North Africa (n = 3), and some were born in the UK, Europe or the USA (n = 7; missing data, n = 1). In order to ensure a balance of voices, one of the FGDs consisted only of people aged < 30 years (in London), another group consisted of men only (in London) and a further group consisted of people living with diagnosed HIV (in Glasgow). The other three FGDs were mixed in terms of participant gender, age and HIV testing experience (London and Glasgow). Nineteen participants had never tested for HIV. The black African non-specialist FGD participants were compensated £25 for participating in the discussion. Each FGD lasted between 1.5 and 2 hours, with an average of nine participants in each group (range of 7–11 participants). The FGDs were audio-recorded and transcribed verbatim.

Service provider focus group discussions and interviews

Six FGDs were conducted with service providers: three in Glasgow and three in London. Sequential purposive sampling (undertaken with the support of simple screening questions asked during the recruitment process) ensured the diversity of service providers from a range of professional backgrounds, all of whom provided HIV-related or other social services to black African people. Black African ethnicity was not a criterion for involvement in these FGDs. GPs were recruited via Clinical Research Networks (CRNs) in London and through established working relationships with members of the research team in both Glasgow and London. Community workers in both cities were recruited from organisations with extensive experience of delivering HIV prevention and care, as well as a range of other non-HIV-specific services to black African people. The research team approached pharmacies in areas with high numbers of African residents in Glasgow and London, with support from local pharmacy associations. Almost all specialist FGDs comprised those from diverse working backgrounds, in order to elicit contrasts within working and experiential contexts.

In total, 53 service providers participated in either a FGD or a supplementary interview. Those taking part in FGDs included HIV CBO staff (n = 15), pharmacists and pharmacy assistants (n = 9), general practitioners (n = 7), those who provided services to black African people (non-HIV focused) (n = 5), GP and specialist nurses (n = 3), African faith leaders (n = 3) and a health-care assistant (HCA; n = 1). The service provider FGD participants were offered reimbursement for their travel and their time given to the study. The level of reimbursement varied according to profession. The service provider FGDs lasted between 1.5 and 2 hours, with an average of seven participants in each group (range of 4–10 participants). These sizes fell within the ideal range to prompt discussion, while ensuring the participation of everyone in the group. 30

Following the FGDs, interviews with 10 highly specialised HIV service providers (including HIV clinicians, HIV service managers and service commissioners) in London (n = 3) and Glasgow (n = 7) were conducted to help check the acceptability of the intervention and procedures. This approach was undertaken to include diversity of voice in FGDs, accompanied by highly specialised expertise gained through interviews. Each interview lasted between 30 and 45 minutes. The FGDs and interviews were audio-recorded and transcribed verbatim.

Analysis of focus group discussions and interviews

An analysis of the qualitative data was undertaken using a ‘blended’ thematic approach drawing heavily on framework analysis. 31 NVivo version 10 (QSR International, Warrington, UK) was used to synthesise and code data within a thematic matrix to enable elucidation of conceptual associations. Both a priori concepts used in the development of the FGD topic guide, as well as emergent concepts arising from the data, informed the process of identifying the key thematic categories used in data coding. Two researchers devised an agreed coding frame, which was then used to index and chart the findings. The integrated model of behavioural prediction and change was used as the theoretical framework to assess attitudes, willingness and perceived behavioural control to use HIV SSKs. 32

Broad descriptive themes included the feasibility and accessibility of HIV testing, and existing knowledge and uptake of SSKs, and the practicalities of distribution emerged. Cross-cutting themes also surfaced, which influenced our analysis, particularly those concerned with trust and HIV-related stigma. The themes were then refined to devise a more detailed participant-led, inductive, thematic framework. Researcher team discussions and iterative analysis focused on the internal coherence and face validity of the resulting analytic structure.

Step 1: delineate key intervention components

The intervention development process began with identifying and conceptualising the diverse intervention components arising from a combination of existing SSK distribution practice and process-oriented data emerging from stage 1. The research team considered the sequential flow across social contexts, health professionals, SSK recipients and clinical governance procedures. In this way, the study team systematically considered the segmentation and flow of the intervention chain. This conceptual work also assisted in informing the topic guides for follow-up interviews with participants who agreed to take a SSK (regardless of whether or not they ultimately used it; see Appendix 4) and the choice of analytical approach for the intervention development work that followed.

Step 2: intervention barriers and facilitators, and relation to theoretical domains

Step 2 involved further consideration of the key intervention components identified in step 1 by utilising stage 1 data on barriers to, and facilitators of, the intervention. Appendix 5 provides an example of how the study team analysed the relevant data regarding the component ‘appearance and packaging of the HIV SSK’. Key barriers and facilitators were then mapped onto the theoretical domains framework (TDF). 34

The TDF is a metatheoretical framework that integrates key theoretical domains known to be important in understanding behaviour change across a range of populations. It provides a coherent way of organising explanations of why things do or do not happen in relation to either behaviour change or the implementation of particular intervention components. It enables insights into potential mechanisms of action for developing or optimising interventions. Table 1 illustrates the key domains of the TDF and provides a brief explanation of the content to which the particular domain refers.

Analysis and the mapping of barriers to, and facilitators of, the TDF domains were discussed within a single all-day event attended by the research team. Differences of opinion were resolved through consensus.

Step 3: identifying intervention components that could overcome barriers and enhance facilitators

In step 3, an ideal hypothetical intervention that minimised key barriers and amplified key facilitators was constructed. The behaviour change wheel33 was then used to structure the intended intervention, guided by the ideal intervention.

The behaviour change wheel33 links the domains of the TDF to the COM-B (capability, opportunity, motivation and behaviour) model of behaviour change36 (Figure 2). The COM-B model suggests that behaviour change is related to three key factors: (1) capability, (2) opportunity and (3) motivation. These three factors can be broken down into fine-tuned categories and, eventually, the TDF domains. Table 2 shows how the TDF domains relate to each COM-B component.

FIGURE 2.

The COM-B model. Adapted from Michie et al. 33 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is properly cited.

Adapted from Michie et al. 33 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is properly cited.

Step 4: viability of the intervention

The outcome of step 3 provided a range of potential ways in which the intervention could be structured that would reflect key mechanisms of action and reduce barriers to effective implementation. However, it was important to ensure that the resulting intervention was viable within busy service delivery contexts. Thus, the study team evaluated the intervention content with the APEASE (affordability, practicability, effectiveness and cost-effectiveness, acceptability, site-effects/safety and equity) criteria,32 assessing the viability of intervention function and behaviour change techniques for a real-world intervention implementable within the UK.

Strengths and limitations

Although limited in scope and scale, the range of methods used in stage 1 enabled the team to select a mix of data sources and analytical approaches in its systematic, theoretically driven approach to intervention development.

Findings of the formative stage 1 FGDs and interviews with specialist service providers and non-specialist members of the public, and the ensuing intervention development process, are presented in Chapters 4 and 5, respectively. The next chapter presents the methodology and results from the policy and systematic literature review.

Policy approaches and recommendations relevant to self-sampling kits

Most of the policy guidance documents yielded by this search were not specific to SSKs. The UK National HIV testing guidelines were produced by the British HIV Association (BHIVA), the British Association of Sexual Health and HIV (BASHH) and the British Infection Society in 2008, against the background of late HIV diagnosis and undiagnosed HIV status in the UK. 37 The guidelines advocated for the expansion of HIV testing services, including routine offering of HIV testing in GPs, in areas where the prevalence is higher than two per 1000 among 16- to 59-year-olds, to patients attending specified services, such as sexual health clinics or pregnancy termination services, and to patients who report practising high-risk behaviour and patients with indicator conditions. 37 Implementation of these guidelines was assessed using eight pilot projects in acute medical settings, emergency departments, primary care and community settings. 38 Findings from the pilot projects showed that the implementation of guidelines to expand HIV testing in the medical and community settings was both feasible and acceptable; HIV SSKs were successfully used in one of the pilot projects. A later review by Public Health England (PHE) on the evidence of the clinical effectiveness of HIV testing in medical and community settings noted that self-sampling could broaden the available testing options. 39 Other strategies have advocated for self-testing as an alternative option. 40 Indeed, the national response to HIV continues to evolve and, in April 2014, HIV STKs became legal in the UK. 41,42

With regard to policy specific to the black African community, NICE published specific guidance on increasing the uptake of HIV testing among black African people in 2011. 43 In 2014, NICE provided detailed recommendations for commissioners, including local authorities, Clinical Commissioning Groups and NHS England, on delivering HIV testing services. 44 NICE recommended that commissioners assess the local need for HIV testing for black African people and then develop a local HIV testing strategy with clear referral pathways, particularly for outreach point-of-care services. To address undiagnosed HIV and late diagnosis of HIV, NICE recommended that commissioners promote HIV testing, including through the use of modern HIV tests, and reduce barriers to HIV testing among black African people. In line with the 2008 guidelines mentioned above,37 NICE recommended that HIV testing should be offered by health professionals in primary and secondary care. Although SSKs were not specifically mentioned in these guidelines, SSKs have been commissioned by some local authorities as part of their HIV testing services. The NICE guidelines were updated in December 2016 and SSKs were considered a potentially innovative way of increasing the uptake of HIV testing among black African people, given that these kits may address known barriers to HIV testing in this risk group. 45 In the draft guidelines issued for consultation earlier this year, SSKs and STKs were endorsed as innovative ways of increasing uptake of HIV testing among black African people, given that they may potentially address the known barriers to HIV testing in this group. 45 Despite support within the policy documents for HIV SSKs as a means of increasing uptake of testing, evidence on the impact of SSKs on uptake compared with clinic-based testing was limited to one study.

Description of included studies

Table 3 presents the description of the 13 studies included in the review. 21,46–57 Of the included studies, nine were conducted in the USA21,46,47,50–55 and four were conducted in the UK. 48,49,56,57 Eight were cross-sectional surveys,47,49–55 three were prospective cohort studies,21,48,57 one was a qualitative study56 and one was a randomised controlled trial (RCT). 46 The total sample size across the papers was 15,816, with an average response rate of 78% (range 38–100%; information was not provided in two studies47,57). The majority of the studies included communities at high risk of HIV infection. Ten studies included MSM,21,47–53,56,57 three included people who inject drugs (PWID)46,47,50 and three included non-specified individuals or clinic populations who practised high-risk activities. 21,54,55 Only two studies focused on heterosexual populations who practised high-risk activities (both studies were based in the USA);47,50 all of the UK studies included only MSM. The only RCT in the included studies was conducted with PWID. 46 Most studies reported a predominantly white sample, although the sample in the RCT was 48% African American. 46 Among the studies that provided this information, the average age of participants ranged from 18 to 47 years.

The majority of the studies evaluated a dried blood spot SSK (n = 8),21,46,47,50,53–55,57 whereas five studies assessed an oral fluid test21,48,49,51,55 (two studies assessed both types of sampling21,55 and another two did not specify the type of test52,56).

The methods by which SSKs were distributed varied across the studies, with five studies including the option of kits being mailed out to participants21,48,49,51,54 and six studies requiring participants to pick up a SSK from a study site21 (including a pharmacy,47,54 a drug clinic,46 a public sex environment49 and a sauna;57 the remainder did not specify the method of SSK distribution50,52,53,55). Three studies offered participants a choice of both options21,49,54 and four studies did not specify how kits were distributed. 50,52,53,55 Nine studies required participants to return kits by post (the remainder did not specify the method of return or it was not applicable to the study type). 21,46,48,49,51,52,54,55,57 The qualitative study assessed a range of options with participants. 56

Quality appraisal

In accordance with the criteria used for both qualitative and quantitative studies, only a few high-quality papers were identified that related to the study outcomes. The majority of studies took the form of cross-sectional surveys47,49–55 (eight out of 13 papers), with only one RCT identified. 46

Only the RCT scored ‘++’ for both internal and external validity. 46 Two21,48 of the three prospective cohort studies21,48,57 scored ‘+’ or higher for both categories, as did six47,50–54 of the eight cross-sectional studies. 47,49–55 The qualitative study scored ‘+’ overall. 56 Two of the cross-sectional studies assessed the acceptability of a hypothetical offer of self-sampling and, overall, few of the studies directly compared the efficacy of different forms of testing. Those studies that scored poorly on both categories commonly featured a small sample size and/or deficient level of analysis.

Acceptability, feasibility and clinical effectiveness of self-sampling

All but one of the included studies50 reported on some measure of the acceptability of self-sampling (Table 4). In total, only five studies21,46,48,49,51 (three studies21,46,51 from the USA and two studies48,49 from the UK) reported on the distribution and return of SSKs. Within these, 1652 SSKs were distributed (range 80–716 SSKs) and 1373 participants returned a specimen (range 60–665 participants who returned a specimen). This suggests a median return rate of 77.5% (range 47.6–92.9%). The two UK studies48,49 distributed 80 and 126 HIV SSKs, with 62 (77.5%) and 60 (47.6%) returned, respectively. Both of these studies used oral fluid sampling, and no inadequate samples were reported. Twelve of the included studies did include (self-reported) data on the acceptability of SSKs to participants,21,46–49,51–57 but the measures used were inconsistent. Self-reported acceptability was generally high, with self-sampling reported to be broadly acceptable to participants (see Table 4). For example, Spielberg et al. 21 reported that 98% of participants agreed to take part in either oral fluid or dried blood spot bimonthly sampling in the future. Similarly, Fisher et al. 48 reported that 81% of MSM found (oral fluid) SSKs to be acceptable. However, Colfax et al. ,47 Wood et al. ,57 Skolnik et al. 54 and Sharma et al. 53 reported that few participants took up the offer of testing and/or it was the least preferred option among participants – it is notable that all of these studies used blood sampling. In terms of comparing the different types of tests, Spielberg et al. 21 found no difference in testing rates between dried blood spot and oral fluid testing, but this was the only study to compare the two methods. 21 Acceptability did not vary significantly by distribution method.

In terms of feasibility, only six studies included any documentation of how tests were completed, errors in testing, communication of results or linkage to care. Three of these studies used oral samples, two studies used blood samples and one study used both. There were few reports of errors in testing. Spielberg et al. 21 reported that 99% of both oral fluid and dried blood spot samples were adequate for testing, whereas Osmond et al. 51 reported two indeterminate results and 10 insufficient samples (out of 412 returned oral fluid samples) and Formby et al. 49 reported that seven (out of 126) equivocal oral fluid samples had to be retested (all tests gave negative results). Wood et al. 57 reported that three out of 30 tests of dried blood spot samples were not processed. Spielberg et al. 21 reported that staff members were concerned about the efficacy of telephone counselling and Osmond et al. 51 found that only half of participants tested telephoned for their results. Similarly, Bartholow46 reported that 22% of those tested did not report receiving their test results and Wood et al. 57 reported that 4 out of 30 test results were not communicated to participants. Fisher et al. 48 reported that 2 out of 62 required a reminder to return their sample. Formby et al. 49 also reported that oral fluid samples were delayed in getting to the laboratory, which required participants to be contacted and asked to resend their samples to ensure accuracy of results. Linkage to care was not assessed, because most studies (n = 9) had no reactive results. The studies that did have reactive results were unable to check on outcomes for linkage to care because of the features of their methodology.

Only six studies provided data on the clinical effectiveness of SSKs for HIV in increasing the uptake of HIV testing. In the one RCT included in the review, Bartholow46 reported that those in the self-sampling arm were twice as likely to have tested for HIV in the past month, but were not more likely to obtain their results than those in the clinic-based testing arm. In the USA, Colfax et al. 47 reported that the availability of SSKs had not increased testing rates among those not tested previously, nor had it significantly changed testing behaviour among those who do get tested. In the UK studies, Formby et al. 49 reported that SSKs offered an alternative means of testing, with 35% of participants having never tested for HIV before, while Fisher et al. 48 reported greater uptake of SSKs (62.5%) than clinic-based testing (37.5%). Formby et al. 49 concluded that the pilot successfully reached those not regularly engaging in HIV testing, including higher than expected numbers of bisexual men, men not otherwise tested in the last year and those not accessing existing sexual health services. Other studies provided data in support of this stance as well. For example, the only qualitative study included in the review noted that the use of SSKs could be a viable alternative to meet increasing demand for sexual health services, but to improve uptake, the method of service provision must be culturally sensitive and acceptable. 56 Wood et al. 57 concluded that including self-sampling in outreach settings could be effective in targeting MSM using sex on premises venues. Finally, five studies21,46,48,49,51 reported the HIV positivity rate, which was an average of 0.9% (12/1311). Two studies48,49 (both including MSM in the UK) reported a positivity rate of 0%.

Barriers and motivators to, and facilitators of, self-sampling for HIV

The final section of the review assessed the barriers and motivators to, and facilitators of, HIV self-sampling (Table 5). Key barriers included anxiety, concerns over the accuracy of testing, concerns about confidentiality, privacy and the lack of face-to-face counselling and fears about the difficulty or pain involved in collecting samples. On the other hand, one of the UK studies reported that there was no difference in uptake related to the importance of accuracy of results or willingness to wait for results. 48 Test reliability was reported to be a barrier in the other UK study. 49 In addition, Wayal et al. 56 reported potential barriers to uptake among British MSM, including preference for medical venues (which were perceived as discrete and appropriate, especially if symptomatic), fear that distribution in gay social venues could trivialise testing or promote stigma, concerns about the unreliability of the postal service for delivering samples and anxiety over waiting for the results.

Conversely, reported facilitators of SSKs were the availability of telephone (as opposed to face-to-face) counselling, and perceived anonymity, accuracy, convenience and ease of use. Finally, additional motivating factors that were reported to contribute to the acceptability of self-sampling were awareness of the seriousness of HIV56 and the benefits of regular/early testing,21,46 and agreement or awareness of being at risk of HIV. 21,46,53,56 Two studies also noted that a cash incentive could be a motivating factor to test. 51,52

Strengths and limitations

The studies included in this review were of relatively poor quality, with most data derived from cross-sectional studies and only one RCT included in the review. Most studies were conducted in the USA, which raises questions about the transferability of the findings to a UK context. Most of the studies, and all of the UK studies, were conducted with MSM, which again raises questions about the transferability of the findings to black African people not identifying as MSM in the UK. Furthermore, data on actual uptake and return of tests, the clinical effectiveness of self-sampling to increase the uptake of HIV testing and the clinical effectiveness of processes for linkage to care were largely absent, and this represents key knowledge gaps. The lack of standardised reporting of outcomes also made it difficult to compare findings across studies. Only one qualitative study was yielded by the search, despite the potential for such studies to inform the design and implementation of self-sampling for HIV interventions.

Recognition of risk as a precursor to HIV testing

In many cases, a crucial factor that determines whether or not an individual undertakes a HIV test is the recognition that they have faced an elevated risk of exposure to HIV (i.e. a motivation to test), along with having the capacity and opportunity (time and skills) to act on that concern. With the exception of interviews undertaken among HIV clinicians and some service providers in London, improved recognition of personal HIV risk and the need for awareness of the disproportionate impact of HIV among black African people in the UK were rarely linked with the need for improved HIV testing uptake in this population. The findings summarised in this chapter draw attention to the ways that HIV-related stigma and fear are inevitably linked to inaccurate risk perception, which in turn influences the motivation and capacity to test for HIV.

Perhaps unsurprisingly, the findings reveal a dissonance in perceptions of HIV risk between the epidemiological data on black African communities and the way that the risk is perceived by individuals within such communities. What we are referring to here specifically is the lack of awareness (and acceptance) of surveillance estimates that reveal that people of black African ethnicity account for almost one-third of adults with HIV in the UK. 1 Some participants felt that when potential targets were unaware of the scientific rationale for segmenting the population in relation to epidemiological categories, such approaches would be met with resistance:

I just see there being a bit of a disconnect, unless people are made to understand that this is, these are the clear [epidemiological] results that have been found, and this is why we’re, you know, suggesting to you. Until that gap is bridged, I just don’t. I just believe you’re going to be almost, you know, you’re just going to be met with, you know . . . ’Uh’?

London non-specialist group 3

At least some service provider participants felt that the dissonance that emerged between individual risk perception and community risk profiling was a challenge that may often be beyond the capacity of the service provider to bridge in ‘one-off’ encounters. These findings demonstrate that the diverse range of participants had pervasive concerns about the interpersonal and wider social implications of targeting HIV testing interventions for black African people in the UK.

Competing imperatives

The relatively low priority of HIV screening among people who had many more pressing needs to be met was frequently raised, as was a lack of time or funds to reach a testing site. In addition, some participants highlighted that there are those who lack the freedom or control in their lives to undertake a test while maintaining adequate levels of privacy to avoid HIV-related stigma. Qualms were also expressed about the extent to which confidentiality within health services could be trusted, which meant that the prospect of disclosure presented too much risk for vulnerable individuals.

Targeting HIV testing on the basis of actual or perceived ethnicity

A range of issues emerged among participants about the underlying inequalities that can impede the success of HIV testing interventions designed to target black African people. There were intense concerns about such offers being perceived to be driven by racism and discrimination among both service providers and non-specialist participants, awareness of the particular sensitivities that targeting could fuel antimigrant discourse and the threat that targeting would be perceived as divisive. Similarly, a concern was expressed that targeting all people considered by a service provider to be black African implied complicity with the homogenisation of highly diverse cultures and communities. Finally, there was discomfort with the assumption that targeting would probably rely on appearance and colour of skin (being ‘black enough’, as one specialised HIV health-care provider mentioned). Therefore, imbalances of power pervade accounts of black African people’s engagement with HIV testing to a considerable extent. Participants reflected on the way that such imbalances affected offers and uptake of HIV screening, and this is amplified rather than minimised when it comes to interventions that are designed to disproportionately benefit black African people.

This concern was further evidenced by a few non-specialist participants who raised concerns about HIV testing interventions in acute services that they perceived to be based on skin colour, rather than an individual’s actual risk of contracting HIV. It was argued that black African people had valid reasons to distrust health services as a result of previous experiences of racism, being patronised, not having the opportunity to give fully informed consent or being exposed to racial microaggressions in these settings:59

By the way, if you find yourself in an A&E [accident and emergency (department)], and if you’re black African, you will get a test anyway. So . . . [others laugh].

He is right.

They will just . . . They will just, you know, shove it into you, and when you ask a question, why didn’t you tell me, they go, oh, oh we are very sorry. But it’s always like that. It’s just not new, particularly if you’re in South London.

London non-specialist group 2

Concerns were also expressed about the difficulty of ensuring fully informed consent in the busy and emotionally heightened environment of acute care services. Other participants mused that they assumed that most blood samples drawn for routine purposes were already being screened for HIV, and that HIV screening was happening ‘behind closed doors’. In both types of discussions, non-specialists demonstrated little confidence in mainstream NHS providers’ judgement and communication strategies when it came to HIV.

Not all discussion about targeted approaches focused on their negative impact or repercussions. Some service providers noted that rationing limited resources made sense within a current climate of reduced public spending. As such, although targeting HIV testing to black African people was seen as problematic for a number of reasons, it was often counterbalanced by the epidemiological and practical need for such targeting:

I have trouble just targeting just black communities with that kit, for me it has to be universal for everyone I wouldn’t like to just target a specific population group.

This is screening! If the epidemiological studies that there’re high prevalence and new incidence rate in that race, in that particular community, that is really . . . really where resources should be.

London service provider group 2

Epidemiological evidence was presented as a neutral counterpoint to the barriers to targeting:

I think you can target in a way that’s honest . . . Because, for example, I think it’s very clear and honest to say, if you are from a high-prevalence area – if you’ve had unprotected sex in a high-prevalence area of HIV and then name where the high-prevalence areas are – you are at more risk of acquiring it.

London service provider interview 1

Furthermore, some service providers argued that the offer of a HIV test may be catalytic for individual testing decisions. From a provider perspective, acknowledging the consequences of not offering a HIV test to a black African service user or patient was considered as being one way to support providers in deciding to offer the test, even when they acknowledged that a challenging discussion could follow.

Finally, participants focused on reducing barriers to targeting through participant-led and culturally sensitive ‘approaches to enhancing targeting’. These included embedding the offer of a HIV test within a wider ‘bundle’ of targeted interventions to diffuse the specific stigma of HIV (e.g. within offers of targeted approaches to address sickle cell or bone marrow transplant donations or hypertension). A culturally sensitive approach to targeting that ensured the coproduction of targeting approaches with representatives of black African communities was also suggested, as was the introduction of such interventions by black African providers. These approaches were seen to reduce barriers to targeting by being particularly sensitive to the ethnic mix of those involved within the targeting interaction. Participants further noted that attention needs to be paid to the potential for health promotion or research materials to link HIV to black African people in the minds of the general public, and the harmful social outcomes that could result if such interventions were misconstrued by a wider audience.

Device practicability

Prior to taking part in this research, awareness of SSKs was not widespread among the non-specialist participants, and only a handful disclosed having used one in the past. A considerable proportion of participants initially assumed that results would be instant (i.e. self-testing), as there were national media reports about that technology being licensed just before the period of research. In addition, participants were invariably surprised by the volume of blood (400 µl) required for a sufficient TINY vial sample, and service providers in particular stressed routinely that they did not think that most members of the public would be able to produce a sufficient sample:

I mean, we do health checks and we take blood from the finger and our machines just been changed to take a much smaller sample, we have to take 40 microns [sic] of the blood, not a big amount which is why I kind of, I was a bit shocked at this. And just getting that amount of blood is actually sometimes quite traumatic for a person.

London service provider group 1

While fear of needles and blood was also discussed, a few pharmacists, nurses and general practitioners held the view that, physiologically, their black patients often struggled to produce finger-prick samples because of thickened skin on the fingertips.

Many of the service providers and non-specialist participants felt that the TINY vial kits would prove to be too complex for most people to use correctly. As one participant in a service provider group in Glasgow expressed:

I think that will be quite tricky. Certainly, I don’t think it’s one that you can tell them it’s that easy to do [. . .] maybe if the test was simpler.

Glasgow service provider group 3

Furthermore, a number of participants expressed concern that SSKs could be easily contaminated by users. Participants questioned the robustness of the technology and procedures on offer to the public, and also whether or not samples would be mismatched between two users. There was also an underlying worry that unskilled members of the public who used SSKs might introduce a risk to others (through contamination or spills) or increase the chance of invalid/inconclusive results in some way:

Can I be very honest? I don’t like this, and the reasons why I don’t like it is because it isn’t simple . . . It isn’t easy and, of course, this will not be popular. [. . .] By the time people put a jab and then put their hands and blood starts dripping, one, two, three, four, five, six, seven, eight, up to 20, I find it a bit . . . very, very cumbersome. It makes it very, very . . . It’s liable to a lot of mistakes. And so what do you then do?

London non-specialist group 2

In many of the discussions with service providers, they raised comparisons between these HIV SSKs and a range of kits for other conditions that are now designed for self-sampling. These comparisons highlighted that many such kits had not had great success (chlamydia and bowel screening kits in particular were perceived as underused and not cost-effective). In contrast, HIV service providers with experience in using dried blood spot kits for HIV self-sampling among MSM were encouraged by the benefits that self-sampling could bring to black African users.

HIV stigma and the need for privacy and discretion

HIV-related stigma was among the most pervasive concern emerging within these data, which supports extensive theoretical and empirical work in the field more broadly. 13,60,61 All participants were clear about the profound challenges that HIV-related stigma presents for prevention and testing interventions. The role of stigma was particularly evident in the concerns about potential isolation and resulting harm that could be experienced by SSK users. Indeed, it is this pervasive stigma that continues to make the promotion of HIV testing within this population so challenging. Stigma predicates against self-perception of risk and promotes profound fear about being seen to be accessing a HIV testing service, as a result of the social implications that may follow. It also makes providers uncomfortable in offering the test, because they too are aware of these social implications. Although SSKs were considered by many participants as having some scope in reducing these social risks of discovery, they were simultaneously understood to be a means of ‘keeping HIV underground’, providing cover to those who desire increasingly secretive means of confirming whether or not they are infected with HIV.

Closely connected to the matter of HIV-related stigma, and directly related to the feasibility and acceptability of the distribution of SSKs within the community, were the many facets of privacy that participants discussed. Although service providers may consider health providers and pharmacists to be trusted professionals who understand and adhere to data protection requirements, non-specialists were far less likely to share this view. Not only did they worry that presenting in such environments and requesting a HIV test might result in judgement and bias, it was also clear that a considerable proportion of these participants held deep-seated fears about who else may acquire access to their most personal health information as a result of such an interaction. In conjunction with this concern, other non-specialist participants pointed out that some potential users may lack power vis-à-vis their partner in order to independently use a SSK. One participant offered a scenario of a husband and wife, in which the former possessed control over the latter and prevented her from using the SSK, or in which the intention to use the SSK would lead to conflict in the household.

Among the many beneficial elements identified, privacy, discretion and the capacity to determine one’s own status in an environment of relative anonymity were regarded as considerable strengths, particularly among potential lay users of the kit. Participants felt that the ability to use a kit privately was an essential benefit for those who were frightened about attending a sexual health clinic, raising the issue in a clinical setting or seeking out a community organisation that provides HIV testing. To this extent, taking a sample in private was regarded as a means of avoiding the stigma that is heavily associated with those who seek out a HIV test:

[It’s] quite hard for some people to go and approach GPs or doctors to explain their situation. Like myself, I’ve been thinking about it. It’s been in my mind for a long time to do a test, because I’ve been hearing people, I’ve been watching this, I’ve been . . . you know what I mean, media’s talking about it, so I don’t even know my status, but when something like this came up, if it’s, like you said, I think it’s an opportunity for people like me to take the chance to do it.

Glasgow non-specialist group 3

But you’ve got to understand, most people, especially with location, where people live, they don’t have that time to go to the hospital and go through the whole process of getting a HIV test. And it’s not even a matter of going to the hospital. It’s a matter of I could just come home, I could do this and I could keep going with what I do in a normal day and then get my results, like they did [with] chlamydia tests.

London non-specialist group 1

Thus, for some, SSK distribution was regarded as a way to access HIV testing that was non-stigmatised, highly accessible and convenient, potentially increasing its appeal to those who had not considered testing in the past.

Choice and autonomy as both opportunity and risk

The findings described above should not be interpreted to mean that SSKs were perceived by research participants as a means of replacing traditional HIV testing and POCT. Instead, SSKs were described as a means to bolster the array of options on offer:

I like the idea of home sampling because it gives more choice, flexibility and opportunity for people to have an HIV test. So for example, if people are worried about confidentiality, they can do the test in the privacy of their own home. And even though they still have to send the result in to a lab, at least it’s not done through a third person, having to disclose their history and why they’re worried about HIV.

London service provider interview 1

Service providers and non-specialists alike were interested in the extent to which the SSK increased an individual’s autonomy over their health, their HIV testing options, and their decision-making around how, when and where to have a test. There was also an extent to which this autonomy introduced a sense of liberation around HIV, and a few participants (mainly service providers) mentioned that making SSKs available in public spaces for private use might help to tackle HIV-related stigma at a structural level. Not everyone agreed with this last point, however, as it was also mentioned that increasing the extent of privacy through the use of SSKs may serve to hide the issue of HIV even further away.

Chief among the concerns expressed about SSK distribution was that the kits appeared to circumvent the provision of sustained interpersonal engagement and support. In almost all groups and several interviews, participants voiced their concern that users would be alone while waiting to hear their results and, most crucially, when discovering the results. In a considerable proportion of groups and interviews, participants felt that the risk of suicide and self-harm was elevated among SSK users learning of a reactive result, because they would not be in the physical presence of a professional for this discussion. When self-testing for HIV was discussed, this concern was even greater, and the SSK was at least regarded as a means of better ensuring linkage to care than self-testing technologies. There was an assumption embedded within these exchanges that linkage to support and care services was far more assured with face-to-face testing services, and there was considerable doubt that this could always be achieved with SSKs.

To some extent, such comments reveal a fundamental concern about the loss of systems control that SSKs represent. This is ultimately a direct consequence of increased user autonomy. It is not surprising that this was a key tension that emerged among participants with regard to this technology, and it is of note that potential isolation being a key drawback of SSKs was raised by service providers and non-specialist participants in equal share.

Self-sampling kits and point-of-care testing

There was also considerable discussion about ensuring that HIV testing is accompanied by a talking intervention given by a skilled professional, enabling risk assessment and pre-test counselling (and, potentially, advice on the use of a SSK). To this extent, guidance suggesting that pre- and post-test discussions are not entirely necessary is not universally accepted by service providers or non-specialists. 37 In all focus groups, and several interviews, there were those who made it clear that their preferred testing pathway involved direct contact and a risk assessment discussion with a skilled service provider, followed up by immediate on-site POCT. Such individuals considered SSKs to be a ‘poor relation’ to POCT. One participant had the following query:

If you’re going to start offering it as an anonymous kind of thing that’s not connected to services, well, what’s the reason for doing that? What problem are you trying to solve by offering that?

Glasgow service provider interview 1

Those who thought testing should be delivered with interpersonal engagement said that it seemed peculiar to distribute SSKs when POCT afforded immediate sample and results collection, with full support and advice on offer.

Essentially, these findings highlight participants’ ambivalence to SSKs. On the one hand, participants expressed enthusiasm about the freedom and autonomy that this technology offers. However, they were also concerned that this comes at the sacrifice of immediate results, professional contact, support and clinically robust procedures. It was widely held by most participants that SSKs needed to be embedded in larger support interventions that involved discussions with skilled professionals.

Strengths and limitations

Although limited in scope and scale, the range of recruitment methods used for our FGDs helped to ensure that we had a diverse mix of non-specialist participants, most of whom were unlikely to have taken part in similar research previously. We started to approach data saturation by the time of the last focus groups and interviews, demonstrating that our use of purposive sampling for both specialist and non-specialist participants had helped to achieve a balanced range of perspectives and experiences. The decision to make some of the non-specialist groups more homogeneous (all male, younger, HIV-positive participants) was a benefit overall, as it helped to ensure careful consideration of distinct subject positions during both data collection and analysis. Although the decision to hold one-to-one interviews with HIV clinicians and other decision-makers ultimately resulted in divergent and multiple data sources for this phase, the option of mixing them in groups with other specialist providers was not only less feasible in terms of availability to attend, but it was also clear that, given the strongly divergent insights expressed by those in this sample (influenced by clinical experience and policy involvement), it was ultimately best to collect their data separately.

Ultimately, the use of these qualitative methods to collect data in two geographically distinct locations enabled the study team to (1) triangulate findings across study populations and geographic areas; (2) compare and contrast specialist/non-specialist perspectives on SSKs and their targeted distribution, and their rationales for these viewpoints; and (3) directly address these perspectives in the design of the intervention to be assessed in stage 2 of the study.

Conclusion

As the introduction of this new technology is meant to assist in circumventing the problems of low uptake of HIV testing among black African people, these findings remind implementers to be cautious about not introducing new problems while trying to address existing ones. The next chapter builds on the findings presented in this chapter to develop the intervention that was implemented in stage 2 of this study.

Setting and location

This is not a behavioural domain, but it is important to secure locations that maximise opportunities to reach African communities and settings that could facilitate trust in the intervention process. Furthermore, the limited amount of time with a potential user in some settings may prohibit intervention delivery on top of meeting the individual’s presenting needs.

The targeted offer of a HIV self-sampling kit

A central dimension of the intervention outline contained within the NIHR’s commissioned call was the specificity to black African people. It was imperative to reflect on which person or persons would be most appropriate to make a targeted approach. It was also important to consider which point in the path of the patient/client journey through routine services was the most appropriate time to make an offer of a SSK. It was equally vital to address risks about the potential implementation of the targeted offer to minimise harm and maximise uptake.

Key barriers in relation to the relevant TDF domain were:

• knowledge – lack of knowledge as to why they would target HIV testing at black African people (service provider)

• skills – lack of skills in initiating conversations about targeted self-sampling (service provider)

• social/professional identity – service providers may or may not see routine or universal testing as part of their day-to-day role (service provider)

• beliefs about capabilities – sense of low self-efficacy and capability concerning the targeted offer of a kit (service provider)

• beliefs about consequences – perceptions that a targeted offer will be perceived as racist and that this may compromise their relationship with a person they have targeted and other community members, and could lead to a damaged reputation (service provider)

• intentions – potential service users and providers may have prior intentions of testing through other routes (not self-sampling; both user and service provider)

• the environmental context and resources – population-level exposure to targeting of black African populations may lead to heightened stigma and disincentivise testing (both user and service provider)

• social influences – strong social norms concerning the avoidance of conduct that can be perceived as racist and discriminatory within a cultural context of xenophobia, racism, intergroup conflict, identity and social identity (service provider)

• emotions – service providers’ distress and negative affect will be a barrier to the targeted offer of a test (service provider).

Participant personal information collection

Recipients of SSKs must provide accurate personal information to enable their sample to be processed and the results returned to them. This is central to facilitate the later communication of a negative, unsuitable sample or reactive test results, and to secure an entry into HIV care for anyone who obtained a reactive result. Although providing this information is within the behavioural domain of the participant, the service provider who makes the targeted offer of the SSK may be instrumental in explaining or reassuring the participant of the need and safety of disclosing personal information within the context of the HIV test.

Key barriers in relation to the relevant TDF domain were:

• beliefs about consequences – personal details may not be confidential and could be seen by others or shared across agencies, with a negative impact on future health and social care (user).

Key facilitators in relation to the relevant TDF domain were:

• beliefs about consequences – clear information detailing exactly how personal information would be used and why it was required would reduce perceptions of negative future consequences (user)

• the environmental context and resources – using minimal information to enable the laboratory to contact the participant (user)

• the environmental context and resources – using online systems for completing study procedures and data collection (user).

Kit return

A further core component was the process of the participant returning the collected sample with sufficient information to enable processing and result communication. This element, again, represented a relatively complex self-managed behavioural domain. Having used the kit, the participant must properly enclose the sample along with an ID label and sample return form within the protective packaging and ensure that it is posted via the supplied postage-paid addressed envelope.

Key barriers in relation to the relevant TDF domain were:

• beliefs about consequences – perception that participants will not return the kit if they believe it will get lost or spoiled within the postal system (user)

• the environmental context and resources – Royal Mail (Royal Mail Group, London, UK) system would disincentivise kit return through widespread perception of distrust with the system (user)

• social influences – returning the kit to a community setting, for example, through faith leaders (as potential distributors), was linked to privacy concerns and disincentivising kit return (user)

• emotion – concerns about the affective impact of information if the testing process was not confidential (user).

Key facilitators in relation to the relevant TDF domain were:

• skills – previous competence with online tracking systems may facilitate kit return (user)

• beliefs about consequences – participants should be given a choice of how to return the kit to accommodate beliefs about consequences (user)

• intentions – intentions to test in situ at a distributor venue may enable uptake (user)

• memory and decision processes – having choices of ways of returning the kit was seen as facilitating test kit return (user)

• the environmental context and resources – postboxes are ubiquitous and can enable kit return with minimal effort (user)

• the environmental context and resources – medical facilities, such as GP surgeries, were trustworthy with management of samples (user).

Challenges to the development of an optimal self-sampling kit intervention

One of the key components of the HAUS intervention is the SSK itself. The type of specimen (saliva or blood) can have implications for its acceptability and for the accuracy of test results. The following section presents an overview of the UK regulations for medical devices for HIV testing and the resulting implications for development and viability of an optimal SSK intervention for HIV testing.

Discordance between acceptable self-sampling methods among study participants and UK regulations for HIV testing and diagnostics

The discordance between preferences of black African communities, service providers and UK regulations for HIV testing had several implications on development of an optimal intervention.

Aligning the study protocol with quality care and research governance framework

The best practices for provision of sexual health care in England are outlined in Standards for the Management of Sexually Transmitted Infections (STIs),66 developed by BASHH. These clinical recommendations also apply to Wales and Northern Ireland. Sexual health service standards for Scotland were published by NHS Quality Improvement Scotland in 2008. 67 Both of these standards provide guidance for achieving safe, high-quality services for management of sexually transmitted infections (STIs). According to these guidelines, development of clear care pathways to be utilised by all STI service providers both statutory and non-statutory is a vital component of STI management. Specialist GUM providers are advised to have an explicit leadership role for clinical expertise and clinical governance in the management of STIs. Provision of results as quickly as possible to service users, whether positive or negative, is emphasised for effective clinical management of infections and for user satisfaction. The recommended time period between consultation and receipt of results is ideally no more than 14 working days, taking account of the laboratory turnaround times. The guidelines emphasise that people accessing non-specialist or outreach services should receive the same standard of care as those accessing any other service for the testing and treatment of STIs. The Research Governance Framework for Health and Social Care68 outlines the responsibilities of researchers and care providers that apply to all research within the remit of Secretary of State. Ensuring confidentiality and quality of care provided to research participants is at the core of this framework.

In the context of SSKs for HIV testing, clear and robust pathways of care are needed for participants to send specimens for testing; delivery of test results; offering retests to those with indeterminate results and confirmatory tests to those with reactive results; and linking those who test positive for HIV to care. In order to align with the principles of the governance framework and given the sensitivities around HIV, the research team felt that it was important that an organisation with previous experience of managing HIV results, and ideally SSKs, was involved – this organisation would also need to be willing to assume the liability for the quality of care provided. Establishing collaboration with an organisation willing to accept the duty of care for all study participants, and not just those with reactive (positive) results, took considerable time and led to substantial delay. The reasons for the delay are described below.

The study team collaborated with TDL for the provision and processing of SSKs. A NHS trust initially agreed to take responsibility for the duty of care for all study participants. However, this arrangement was complicated by the fact that the team intended to use an ‘off-label’ assay to test for HIV. Although this trust had previously provided the same saliva-based SSKs using the GENSCREEN ULTRA Ag–Ab assay as part of outreach work to increase HIV testing in the gay community, they were unwilling to assume liability in the context of a research study. Eventually, the trust withdrew their support to fully partner the study and, in particular, to take responsibility of care for those participants who did not require direct personal contact. Cost pressures within the trust, leading to a reduction of staff as a result of the restructuring of commissioning of sexual health services, also made involvement problematic. Besides, for a NHS trust to assume liability for the assay, the assay would need to be supported by a performance evaluation. This was not feasible, as the trust did not house the required laboratory.

Similarly, other NHS trusts with SSK experience approached by the study team were unwilling to provide a duty of care to participants testing negative for HIV, because they would effectively have no contact at all with the individual (as the kits would be distributed in the community, processed by TDL and the patients informed of the results via text message sent by TDL). Eventually, UCLH agreed to be the ‘organisation providing care’ for all study participants; with laboratory services provided via a subcontract with TDL and a subcontract with the Central and North West London NHS Foundation Trust (CNWL) to deliver all results requiring direct contact with participants, given that UCLH did not have the prior experience or services in place to provide this. The process of establishing the clinical care pathways and necessary contracts took over 10 months, which adversely affected the ability to deliver the study on time and to scale.

Collaborating with general practices/community-based organisations

One of our study aims was to assess the feasibility of using existing services for a SSK intervention for a future Phase III evaluation. Based on results from stage 1, the study team decided to distribute HIV SSKs targeted at black African people in (GP) surgeries and CBOs working in HIV prevention in both London and Glasgow. Unfortunately, implementation of the HAUS study in Glasgow and collaborations with CBOs in London did not occur as intended.

Final intervention content

Table 8 illustrates the intervention functions and behaviour change techniques that were agreed across the team to become part of the HAUS intervention. These are the core components to be used within the intervention manual and within intervention training. As described in Step 4: viability of the intervention, these were derived after systematically having applied the APEASE criteria. 32 These criteria assess the viability of intervention function and behaviour change techniques for a real-world intervention implementable within the UK. In this way, the content of the HAUS intervention was finally agreed and drafted.

Study settings

The intervention was intended to be offered in GP surgeries and CBOs in both London and Greater Glasgow. Unfortunately, for the reasons described in Chapter 5, Collaborating with general practices/community-based organisations, implementation of stage 2 of the HAUS study in Glasgow did not occur.

Inclusion and exclusion criteria

To be eligible, potential participants needed to consider themselves to be ethnically black African, aged ≥ 18 years and able to provide informed consent. Potential participants were excluded if they were unable to read and understand English and/or could not provide a means of contact for result notification.

Statistical methods and sample size calculation

The sample size is based on the precision of the estimate of the return rate [e.g. the width of the confidence interval (CIs)]. The numbers were estimated using the well-known Wald approximation. As the primary measure is based on attrition (the return rate), no further allowance has been made for precision.

The HIV SSK projects, in the UK, (which primarily target MSM online) achieved return rates in the region of 60%. 20 Given that, in the UK, HIV testing rates tend to be lower in black African communities, the study team assumed a return rate of 50%. Distribution of a total of 600 kits per setting would enable a precision rate of at least ± 4.0% and distribution of a total of 380 kits per setting would enable a precision rate of at least ± 5.0%. Should the return rate be lower or higher with a fixed sample size, the size of the standard error would be smaller and the resulting CIs narrower.

More SSKs were to be distributed in London than in Glasgow, because the former has a larger population of black African people, and recruitment was limited in Glasgow to GP surgeries. We planned to distribute 600 SSKs in each site (GP surgeries and CBOs) in London and 380 in GP surgeries in Glasgow. Prior to the commencement of SSK distribution, it became clear that no SSKs would be distributed from Glasgow sites (for reasons outlined in Chapter 5, Collaborating with general practices/community-based organisations). Accordingly, the recruitment target was reduced from 1580 to 1200 participants.

Proportions were described using the number and percentage, and continuous data are described using the median and interquartile range. The relationship between baseline characteristics and enrolment is described using simple univariate logistic regression models. All analyses were conducted in SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

Self-sampling kit

The TINY collection device in conjunction with the ROCHE HIV Combi assay (Roche Diagnostics GmbH, Mannheim, Germany) was used to test for the presence of HIV Ab in participants who consented to take part in this study. The ROCHE HIV Combi is a fourth-generation assay, used for the detection of the HIV p24 Ag and Abs to HIV-1 (groups M and O) and HIV-2. It can detect HIV that was acquired > 4 weeks ago, and is CE marked. The ROCHE HIV Combi assay has a sensitivity of 100% (95% lower confidence limit 99.8%) and a specificity of 99.63% (95% lower confidence limit 99.42%). 70 This test requires collection of 400 µl of blood (obtained by pricking a finger with the supplied lancet) in a small collection tube. The test enables the specimen to be transported to a laboratory, by post, in a vial that contains preservatives that stabilise the sample for up to 21 days if temperatures are between 4 °C and 37 °C.

The package distributed to participants included (1) the TINY kit sample collection device (including self-retracting lancets); (2) a sample data form, which required three unique sample identifiers from the participants in order for the laboratory to be able to process the sample and to enable result notification; (3) the acceptability questionnaire (see Appendix 9); (4) a paid return envelope for sample and questionnaire return to the laboratory; and (5) the sample collection instruction sheet. The instruction sheet provided information on how to collect the sample; a link to a video demonstration (www.haus.org.uk); information on labelling the sample and postage; information on how the result would be communicated and why participants may be contacted; and a reminder about completion of the acceptability questionnaire. Both the instruction sheet and the video demonstration used black African imagery.

The Doctors Laboratory, a Clinical Pathology Accreditation provider, which takes part in NEQAS (National External Quality Assurance Scheme), supplied the kits, with each featuring a unique identifier. Sites were instructed to store the kits in a locked cupboard until distribution to avoid tampering and theft. Kit distribution to all sites was tracked by the project manager.

Screening procedures

Participant recruitment

An intervention script was provided to distributors in both GP surgeries and CBOs to introduce and initiate discussion about the study. This was supplemented by an aide-memoire in the form of key bullet points derived from the longer script. It was anticipated that this scripted interaction with potential participants would occur during routine consultations at GP surgeries or regular outreach activities undertaken by CBOs. Interested eligible potential participants were given the PIS, asked to read it, and given the opportunity to ask any questions. At this point potential participants were invited to enrol in the study and, if accepted, their eligibility was established.

If the participant declined, distributors were instructed to state, ‘That’s absolutely fine. We would really like to understand why people do not want to participate in the study. On this card are some of the common reasons why people say no’. They then showed the participant a card with nine reasons why people may not want to participate. The listed reasons were:

1. I have recently tested for HIV.

2. I do not believe I am at risk of HIV.

3. I would prefer to test elsewhere.

4. I would prefer not to use a SSK.

5. I would prefer not to know my HIV status.

6. I do not like being offered a HIV test just because I am black African.

7. I already know that I am HIV-positive.

8. I would prefer not to say.

9. Another reason.

If asked what was meant by the term ‘black African,’ distributors were instructed to clarify that this category included anyone who identified themselves as black African, whether they were migrants from Africa, African descendants or African nationals. A ‘frequently asked questions’ section was provided in the intervention manual to support distributors in fielding any questions that they may receive about HIV or the study process. If the potential participant was ineligible, the distributor thanked them for their interest, explaining that they did not meet the eligibility criteria, and suggested an alternative method of HIV testing.

Sample collection and analysis

Participants who self-collected a sample were also required to complete the sample form with three unique identifiers (initials, date of birth and a unique ID number) to enable the sample to be processed at the laboratory. The ID number was prepopulated on both the sample form and the acceptability questionnaire.

Follow-up of participants

Results communication and management

Anonymised results were communicated from TDL to the project manager at UCL on the same day that sample analysis occurred, via secure encrypted e-mail. There were three possible test results: (1) HIV not detected, (2) reactive sample or (3) insufficient sample (as a result of either underfilling of the TINY vial or gross haemolysis). UCLH provided a duty of care to all participants who returned a SSK, regardless of whether they were recruited at a GP surgery or a CBO. However, the duty of care was passed to CNWL for reactive results, insufficient samples and those without a mobile phone number. This arrangement arose because UCLH does not currently provide a self-sampling or HIV result notification service for patients. CNWL does, and its staff members have clinical expertise in the provision of HIV results. If a participant were to require confirmatory testing, the duty of care would pass to the health-care institution attended.

Purpose and scope of the evaluation

An understanding of what was implemented and how it was implemented is integral to explaining how an intervention works, in order to address questions of feasibility, or whether or not it was possible to deliver the intervention as planned in the chosen settings.

The principal concern of early process evaluation frameworks was capturing what was delivered in practice in order to avoid dismissal of sound intervention theories as a result of a failure to implement them effectively. 71 Most frameworks focus on the precise ‘form’ of delivery in terms of whether or not this represents fidelity to what was intended to be delivered, as well as measuring the reach of delivery (the dose). There is debate about whether or not adaptations in programme delivery, decided locally, enhance intervention clinical effectiveness or lead to poorer outcomes. 71,72

Advocates of strict fidelity73 argue that fidelity is essential if effective interventions are to be replicated, especially when an intervention’s ‘active ingredient’ may not be known. Advocates of local adaptation argue that interventions need to be tailored to local circumstances. 72 Durlak and DuPre74 propose a compromise, whereby the ‘core components’ of an intervention should be delivered in standard form, but less central intervention components or features can be modified to fit local needs. They present research that suggests that a balance between fidelity and adaptation is likely to be most effective, with the precise balance being dependent on the specific intervention.

The process evaluation was informed by Medical Research Council guidance75,76 and the wider implementation science literature. 77–79 It investigated not just implementation of the intervention (that is, fidelity, reach and acceptability), but also considered mechanisms of action (the impact in context) and potential future normalisation of the intervention. These three areas of exploration are described in fuller detail below.

1. Implementation:

   1. Were standardised intervention components implemented with fidelity of form?

   2. Where deviations from fidelity of form occur, do these reflect intentional adaptation (and, if so, with what motivation), unintentional drift or simple omission?

   3. What was the reach of the intervention?

   4. How did reach and acceptability vary across key subgroups of African people, and what contextual factors appear to affect this?

2. Mechanisms of action:

   1. Does the intervention increase access to HIV testing (opportunity)?

   2. Does the intervention provide an acceptable and viable option to establish HIV status (capability)?

   3. Does the intervention increase motivation to test for HIV?

   4. Are any potential harms suggested by qualitative data?

3. Normalisation:

   1. Is the intervention likely to be supported by key stakeholders in future funding and delivery of SSKs? Do they –

      • view the intervention as coherent?

      • commit to future participation?

      • commit to collectively take on the work arising from the intervention?

      • review progress in implementing the intervention?

To address these areas, the process evaluation collated information from the inception of stage 2. Figure 4 contains a flow chart of the data collection points relevant for the process evaluation. The following data were intended to be collected:

1. study diaries – the study team routinely recorded challenges and issues encountered in all aspects of the implementation of the feasibility study

2. training evaluation – distributors were asked to complete a before/during/after survey assessing the value of the training, the extent to which the training met its stated aims and their comfort and confidence in delivering the HAUS intervention

3. screening and recruitment logs – these data assisted in monitoring progress during recruitment and facilitated early intervention if support in recruitment was required

4. distributor weekly log – distributors were asked to complete a weekly log to monitor their recruitment progress and report any problems with recruitment (see Appendix 11); they were asked to rate how they felt about recruiting people to the HAUS study, their confidence level in targeting black African people and answering questions about the study, and whether or not they used techniques provided in the training session or study manual

5. site visits – up to two site visits were planned with each distributor, to support implementation of the intervention; detailed notes were taken during the visits to aid with understanding of the environments and circumstances in which the intervention was being delivered

6. data flow – the speed and success of data flow through all elements of stage 2 were monitored to assess intervention feasibility

7. support queries and responses – any contact with the study team by distributors or study participants was logged

8. site summaries – throughout the distribution period, data from distributors were collated and synthesised into a one-page intervention summary for each site, describing their key successes, challenges and experiences along the way; each summary was cross-checked by the project manager, and then e-mailed to each distributor in turn, for their consideration and comments

9. site close-down interviews – the site summaries were used as the basis for undertaking final face-to-face interviews with distributors at each site; focus groups were originally planned; however, as few individuals at each site undertook distribution (in many cases, only one individual took up distribution), one-to-one or small group interviews were a more feasible format; interviews lasted for approximately 30 minutes (range: 5–60 minutes); a semistructured topic guide (see Appendix 12) was used to structure these discussions; the focus of the interviews was to gauge acceptability of the intervention, and distributors’ experience of undertaking targeting, recruitment and consenting procedures, and fielding questions; the interviews were recorded when possible, and notes were taken throughout; annotations formed the basis of analysis, with recordings providing back-up when needed

10. qualitative interviews with study participants – the process evaluation also drew from the qualitative follow-up interviews with participants (see Follow-up of participants above).

FIGURE 4.

The HAUS study process evaluation plan. LA, local authority.

The raw and synthesised data were then coded and analysed, with a focus on eliciting themes relevant to the organising principles of feasibility (fidelity) and acceptability.

Fidelity

The HAUS intervention was designed to be implemented with highly standardised inputs and processes, with little or no site-specific tailoring. However, in practice, substantial local tailoring was observed. Therefore, our approach to evaluating implementation focuses both on fidelity of form and fidelity of function. Fidelity of form refers to the extent to which the intervention followed the standardised structures and processes developed for stage 2 of the HAUS study, as described in Chapter 6. This process evaluation examines the extent to which these standardised intervention components were implemented with fidelity of form. When we observed deviations from fidelity of form we assessed whether or not these were intentional adaptations (and, if so, what motivated them), unintentional drift or simple omission, and whether or not the adaptations ran with or against the logic of our theory of change (that behaviour is an outcome of three necessary conditions: capability, opportunity and motivation). This was a substantial issue in the HAUS study, as the intervention was delivered across 12 GPs and three African CBOs throughout London. We also examine fidelity of function, that is, the extent to which locally decided adaptations to the intervention processes were consistent with the overall theory of change.

Reach

Reach is the extent to which the target audience come into contact with the intervention. Moore et al. 75 recommend that process evaluations should include quantitative assessments of reach, in terms of, for example, proportions of the target audience who are aware of, and come into contact with, the intervention.

Assessing the reach of awareness of the HAUS intervention across the whole population of black African people in England and Scotland was not appropriate or feasible within the context of this study. Unfortunately, we were also unable to accurately determine what proportion of eligible persons in contact with the recruiters were actually exposed to the intervention. It was intended that the enrolment log would capture this information, but, in reality, these were often poorly completed. The data we have show that 349 people were offered the intervention, with 125 people agreeing to participate. In excess of 37,000 black African people are resident in the borough of Lambeth alone.

Normalisation

A further component of this process evaluation is to better understand the process by which interventions might best become embedded in distributors’ routine policies and systems of service delivery. This is often described in terms of ‘normalisation’ or ‘sustainability.’

There is a variety of frameworks that help evaluators assess intervention normalisation (e.g. the RE-AIM framework). 80,81 However, the study team chose to use the normalisation process theory. 77 The focus of this theory is to consider how implementers can enable the embedding of interventions within institutions and social contexts so that they are ultimately integrated. 77 The theory suggests that whether or not this happens depends on the following four ‘generative mechanisms’ that those working on implementation engage in individually and collectively:

1. coherence (how people make sense of a new practice)

2. cognitive participation (the willingness of people to sign up and commit to the new practice)

3. collective action (people’s ability to take on the work required of the practice)

4. reflexive monitoring (activity undertaken to monitor and review the practice).

Distributors

Within CBO settings, it was intended that community outreach workers would be the distributors of the SSKs, and this was what was reported from these distributors. Although the distributors within GP settings were intended to be PNs, during the training it became clear that this was not always going to be feasible. A range of GP locations told the project manager from the outset that other staff members in the surgery would be undertaking study recruitment and distribution, including: general practitioners, HCAs and practice managers. In all, 8 of the 12 practices used one or more of these alternate members of staff as recruiters and distributors for the study.

Discussions during site visits established that each practice took a different approach to managing staff workloads. This sometimes meant that HCAs were regarded by their senior colleagues as having more time to devote to research tasks than others, even if that was not actually the experience of HCAs. Alternatively, in a number of cases, general practitioners with a particular interest in sexual health took on responsibility for recruitment and distribution themselves. Though they were working under time pressures to stay within the typical 10-minute appointment schedule, they distributed the SSKs because they had a particular commitment to the issue. The following quotation is from a GP surgery where two general practitioners participated as distributors:

Why was it you two who did this?

Because we volunteered for it! We were the ones who were interested.

You got the e-mail, didn’t you?

Yeah, and I knew you would be keen.

Because we are doing a lot of gynae, and sexual health generally. So, it’s an area of interest.

In the main, we can conclude that in the majority of GP surgeries, it was not PNs (or not PNs alone) who recruited participants into the study and distributed kits. This action, while deviating from study intention, facilitated recruitment by increasing the opportunities in which it could be delivered, and, as such, should be considered a logical adaptation.

The intervention was designed to work alongside distributors’ existing commitments. However, it was reported to us by one general practitioner that staff members’ days off were used to attend HAUS training, and, in another instance, a general practitioner came into the surgery on her days off to undertake recruitment. Other distributors explained that it took a lot of additional time in the day (including overtime and loss of breaks) to fit in study recruitment, kit distribution and the associated research tasks.

Targeting

Within CBO outreach settings, it was intended that community workers would identify black African people whom they encountered in their routine outreach activities and locations to approach for study participation, and this is what was (in the main) reported by these distributors. In some cases, community workers also undertook special outreach events at settings, such as barbershops, in order to specifically recruit to the study.

Within GPs, it was intended that eligible individuals would primarily be identified from patient lists (using recorded ethnicity and/or country of birth data) at the start of the day, prior to the initiation of appointments. That way, a prepared enrolment log could be in front of each distributor, ready for when the patient came to them, and they could log the patient’s reason for refusal, or their agreement to take part on the enrolment log. This method was also intended to eliminate the need for practice staff to try to identify a patient’s ethnicity visually or to ask them their ethnicity prior to offering them the intervention.

Five GP surgeries informed us that there were usually or always insufficient data available to them to identify potential participants’ ethnicity in this way. As a result, practices resorted to a variety of alternative means of targeting black African people. Staff in at least four practices described using visual cues (both in the waiting room and in the consultation room) to identify potential participants. One practice, in particular, reported that this meant that many of those approached were ultimately found to be ineligible, as the majority were black Caribbean, rather than black African. Other tactics were also used to identify potential participants, in particular, knowledge of patient backgrounds, especially for those recruited in primary care. Having a range of practices to assess potential eligibility could cause confusion, as indicated by this comment:

Dr X asked the patient, who was happy to participate. I’m not sure if this particular patient fitted into the category though so was a little confused. The patient was born in the UK, so I could not remember whether this mattered or not.

Some practices did effectively use ethnicity data contained within their databases to identify some or all those eligible for participation. One practice designed a prompt to remind the distributor to introduce the study to come up on the practitioner’s screen when that patient was being seen. Another practice wanted to initiate such a system, but was ultimately unable to do so.

One GP used the ethnicity data that they held on patient lists to screen for eligibility and then contacted patients to invite them to make an appointment, so that the HAUS study could be introduced and discussed. This represents a considerable departure from the intention that the intervention was to take place alongside existing routine service delivery in GP settings when patients presented for other reasons.

Script to introduce the study

It was intended that distributors in CBO and GP settings would use the agreed script to present the study to eligible participants. The script and its importance were underlined at all training events, and laminated cards with the script wording were made available in each site manual.

When asked at follow-up about the extent to which they had adhered to the script, the staff at eight study sites said that they had not used the script consistently. Most said that they had used either the script or the PIS to become familiar with what information was to be communicated, and then they summarised it in their own words. One distributor remarked:

I used my own words, because it’s something I do on a daily basis. I just use my words, and the statistics are very important, they don’t lie. As well, I try to find out what part of Africa they are from.

Another distributor clarified that they did not follow the script word for word, preferring to memorise the key facts and state them in their own way.

It was felt that reading from a card could undermine confidence in the distributor, and one distributor said he was ‘not a robot’ and, therefore, needed to put things into his own words.

It was clear that, in most cases, the number of approaches made by each distributor in a single day or even a single week was too low to expect them to have committed the script to memory. Furthermore, at many GPs, study paperwork (including the site file) may not have been in the same room as the distributor, so they needed to raise the issue opportunistically and then find the relevant PIS and consent form afterwards. None of the CBO distributors reported adherence to the intervention script, as, once again, they had found it more comfortable to summarise and put things into their own words. Again, this can be regarded as a purposeful deviation from the intended intervention design, based on the practicalities of delivering it in busy settings and the professional expertise of health-care staff to put things in the most-appropriate language for a particular patient or situation.

When asked to confirm that they had all introduced the study in a way that made it clear that black African people were being targeted for distribution because of a higher prevalence of HIV in this population, all distributors agreed that they had made this clear in their tailored introductions to the study.

Recruitment process

In the intervention protocol, it was intended that one single distributor would take the participant through the entire recruitment process themselves, and that this would all happen within one 10-minute time period. Research paperwork (the consent form and the baseline questionnaire) was to be completed within this period, along with kit distribution and answering of any questions.

In the main, all CBO distributors attempted to adhere to this process, and all struggled to keep the required components to a 10-minute time frame. On at least one occasion during distribution at a barbershop, the barber first introduced the study to potential participants, and then sent interested participants to community workers at the back of the shop.

This practice of ‘funnelling’ potential study participants to other distributors who would then take them through the eligibility criteria and the study documentation was also evident in a few GPs. In these cases, a PN or a general practitioner would introduce the study, and then send interested individuals on to a designated colleague to complete the process. This was regarded as a means of extending reach, while utilising time and expertise efficiently. One distributor explained: ‘It was easy to do in surgery as another member of staff does the consent’.

Another variation, applied in a few practices, was for staff to break the recruiting process down into distinct stages, in order to avoid appointments backing up. For example, general practitioners at one surgery introduced the study during the appointment, and then asked interested individuals to go out into the waiting room to review and complete the study documentation on their own. These patients were then called in between appointments until all questions were answered and all documents completed. Again, this represents an opportunistic departure from intervention fidelity, as an attempt to undertake at least some recruitment in this setting.

At another practice, a HCA approached patients who looked to be black African and introduced the study to them in the waiting room (rather than in private as had been intended). Individuals were left with a copy of the PIS and invited to approach the HCA in her consultation room if they wanted to take part, and the rest of the process was carried out there. Finally, a HCA at a third practice said that he tended to introduce the study at the start of the consultation, and then, based on the patient’s response, gauged the remaining consultation time. If the patient seemed interested, the distributor tried to make sure he had time to go through all the stages of study recruitment by the end of the appointment. He felt that this tactic helped to prioritise the patients’ presenting health needs over the needs of the research study, while also ensuring that they did not leave the consultation without his having mentioned it.

Although in CBO settings it was suggested that the instruction video could be shown to potential participants as a part of the recruitment process, this option was not often utilised, because of time pressures or internet connectivity problems, or because it was felt that it could be ‘off-putting’ by at least one distributor.

Study procedures

It was intended that the introduction and recruitment components of the study, including sharing of the PIS and completion of the consent and baseline questionnaires, would take no more than 10 minutes. Feedback from almost all distributors indicated that it took considerably longer to get through the entire process, which, in the particular case of oversubscribed GP surgeries, caused major feasibility issues:

The only problem is the time, as no extra time has been allocated for this study while some minutes are needed to provide the information to patients as well as when patients asked questions to understand more about the study.

Difficult time wise to fit/ add into a 10 min[utes]. GP consultation when all the time has already been used up with the patient’s agenda/ symptoms.

There was a considerable burden of data management and data tracking expected of each distributor. A few sites (particularly CBOs) found it difficult at first to keep a log of those who they had approached who had refused, partly because of the perception that the study team needed an age and other data for each person who had refused, as well as a reason for their refusal. Once the study team intervened to suggest that estimates and null values were acceptable entries in the enrolment logs, tracking of refusals improved.

Furthermore, it was essential that in all cases an identical kit ID number was attached to the paperwork an individual had completed, which matched with the kit ID number on their SSK. In at least one case, the SSK kit ID did not match that on the paperwork received by the study team.

There was a range of data flow that was essential to the smooth running of the research study, including electronic or faxed copies of consent and baseline forms and enrolment logs being sent to the project manager, and completion of weekly distributor logs to identify and respond to emergent challenges. Both sets of data were generally very slow to be returned to the study team. This introduced a number of unintended challenges, as the project manager was frequently not in possession of the consent form or contact details of a participant who had, for instance, sent back an insufficient sample, and who therefore needed to be called by telephone to be advised of their options (see Chapter 6, Results communication and management). In this way, the participants, Royal Mail and TDL were able to jointly act with far greater efficiency than the distributors and the study team in terms of relaying required information through to one another. These logjams occasionally meant that results could not be clearly communicated to individuals until the paperwork had caught up. Occasionally, these logjams took a number of weeks to resolve and, in a small number of cases, the paperwork never materialised, so the individuals had to be excluded from the study. All individuals who returned a sample were informed of their results regardless of whether or not they were subsequently included in study.

Connection to other HIV testing interventions

The HAUS study was not intended to be delivered (either as a follow-on or a precursor) with any other HIV testing interventions being undertaken by the distributing agencies. However, it was reported by one CBO that, following an event where the SSKs had been introduced, a group of interested people accompanied the individual back to the CBO office to undertake a POCT HIV test, because they regarded this testing option as more acceptable than the SSK on offer in the HAUS study.

Distributor support to take blood sample

The study was designed to assess the acceptability and feasibility of the targeting and use of HIV SSKs among black African people. However, two distributors (one CBO and one GP surgery) described assisting participants with the blood sample collection. In the case of the GP surgery, the practice manager decided that, as she found the kit so difficult to use, she would support every single participant she recruited, and reported that the entire time spent with each participant was between 30 and 40 minutes. The following excerpt is from an enrolment log:

I helped another patient today, and again the whole process took ages to complete. This was because the patient’s blood wouldn’t flow from her fingers, we used all the lancets in the box plus another 4, in the end the nurse came in to help but the patient’s blood seemed to be clotting the minute it started to flow. I’m not sure whether there is enough blood in the bottle to test. We really persevered and the patient said that if she was at home she would have discarded the whole thing. For me it’s becoming a very time consuming exercise.

Two further general practitioners reported that they had requests from participants to support them with the use of the kit, because of its complexity, but each of these distributors refused, because of a desire to maintain fidelity with the research protocol.

There were further data emanating from one CBO (and the participants whom it recruited) that during one distribution event in a barbershop, at least some participants took their blood samples together at the same time, with support from the distributor.

Location for sample return

There was one reported instance in which a sample was returned to a GP surgery rather than being posted by the participant to TDL via Royal Mail. Because the distributor in this practice was not working on the day the sample arrived, it was stored in the practice refrigerator until a few days later, when that staff member returned to work and was informed about the sample. The distributor then posted it.

Further data collected from distributors and participants at the CBO also led us to infer that, after this event at the barbershop, it is likely that the distributor posted the samples via Royal Mail, rather than the participants doing this themselves.

Communication with participants after sample return

There were a number of occasions when participants and distributors reported that a 2-week text reminder to return a sample was sent to people who had already returned their sample. This created some concern among participants about the potential for lost samples, and made some doubt the veracity of their results.

Although it was intended that a HA at CNWL would have access to the secure database to enable them to directly obtain contact information for those participants with reactive or insufficient results, this did not occur, as sites predominantly used paper forms, so there was delay in getting this information onto the database. Instead, the project manager notified the HAs directly when an insufficient sample had been received at TDL. Occasionally, because of missing or delayed consent and baseline data forms, there was a delay in access to sufficient information to contact the participant immediately. Furthermore, as a result of the HAs’ working context, there were occasionally further delays between notifying the HA team and communication of the result to a participant. The protocol does not state a time limit for these activities, but, in the training, distributors were told to inform participants that it could take up to 5 days for them to receive a result.

It was also reported that, in at least a few cases, a negative result was received by participants via text from TDL within 24–48 hours of a sample being posted, which appears to have caused concern among some participants that the result may not be trustworthy, as it had arrived more quickly than anticipated.

Descriptive results

The total number of study participants interviewed was 21, the median age was 40 years (range 18–67 years) and 12 participants were women. Of the 21 participants interviewed, nine received negative results, four sent samples that were unable to be processed (as a result of the samples being underfilled) and eight did not return their sample. Seventeen interviewees were recruited at GP surgeries and four interviewees were recruited at CBOs. Approximately half of the interviewees (n = 9) had previously had a HIV test. However, the majority were unaware that SSKs existed prior to participation in the study. Some interviewees had a notion that a kit could be ordered online, but did not have a more specific knowledge of either SSKs or STKs before enrolment.

Barriers to the acceptability of self-sampling kits

There were five inter-related themes that emerged as barriers to the acceptability of the SSKs. The first three all pertained specifically to the kit used (fear of needles, insufficient blood flow and issues with the vial). The last two meant that several interviewees submitted an underfilled sample, resulting in their being contacted by HAs, who explained that the interviewee’s sample could not be processed and described future testing options (see Results communication and management above). Those who experienced this problem reported several concerns that had an impact on the acceptability of the intervention. Finally, interviewees identified the stigma, fear and taboo surrounding HIV and HIV testing as a barrier, not only to SSKs, but to all methods of HIV testing. Each of these themes are discussed in greater detail below.

Facilitators of the acceptability of self-sampling kits

The convenience of the SSK emerged as a very salient theme in the data. The comprehensive kit was thought to have many positive features, including the lancets (perceived by many interviewees as non-threatening) and clear instructions – features that were frequently cited as making the SSK acceptable to use. In addition, participant trust in the SSK distributor facilitated SSK acceptance. Each of these facilitators is discussed in the sections below.

Acceptability and utility of training

All distributing organisations undertook mandatory training prior to the initiation of fieldwork. In total, 18 training sessions were delivered across London. These were delivered in the premises of 12 GP practices and three CBOs that participated, with the CBOs receiving two training sessions each. Some distributors asked for, and received, follow-up sessions, either to review key elements or to allow staff members who were unable to attend the first training session to catch up.

In total, 70 people attended the training sessions, following which 47 people (67%) voluntarily completed and returned an evaluation form anonymously. The participants identified their job roles as follows: professional health worker, including general practitioners, GP nurses, pharmacists, HCAs (n = 33); peer support worker (n = 4); CBO service manager (n = 8); and other (n = 2).

Training sessions typically lasted around 2 hours, and participants were asked to complete an evaluation form at the outset and following the training. These brief evaluations were primarily intended to be used iteratively to improve subsequent training sessions, but they also helped to assess the overall value that the training gave to intervention delivery.

Participants were asked to rate components of the training on a scale of 1 to 5, where 1 was ‘no benefit’ and 5 was ‘great benefit’ (Table 12). Each of these components received an average score of between 4 and 5, with the two highest-rated items being the overview of SSKs and study paperwork, as well as the clarification of study inclusion criteria. Some feedback demonstrated that a few participants were less confident about the benefit of the training to help them describe and model the use of the kit, and to explain to patients how their sample results would be delivered, but these views were very much in the minority.

The vast majority of participants (96%) agreed that the range of topics covered within the training was good. When asked ‘Did you get what you were looking for from the training event?’, one-third of respondents (n = 16) said that they had ‘completely’ got what they wanted, with half of respondents (n = 23) saying that they had ‘mostly’ got what they wanted. Four participants were ambivalent, with their responses to this question being ‘somewhat’. All those who responded to ‘The training has increased my understanding of the HAUS study and what is expected of me as a distributor’ agreed with this statement (n = 45; two responses were missing). Similarly, the majority of respondents (87%) reported that any questions or concerns they had at the start had been answered, with the remaining six people neither agreeing nor disagreeing.

Each individual was asked to summarise what they understood the HAUS study to be about and their role within it, both at the start and at the completion of training. Following the training, all participants gained a stronger understanding of the study aims and objectives, and clarity about their roles and responsibilities also became more pronounced. A range of other responses further reflected on the excitement participants felt at having a new way to encourage HIV testing, and there was a comment that this could be a helpful route to ensuring ongoing support for African people with diagnosed HIV.

The clear shift in participants’ responses indicates that participation helped to increase their knowledge about the study and their roles as recruiters. Indeed, several GP distributors at the subsequent exit interviews (see Acceptability of the HAUS intervention to primary care staff) indicated that, prior to the training, they were unaware of the epidemiology relating to HIV in the UK in general, and to HIV in black African communities in particular.

Acceptability of the HAUS intervention to primary care staff

Broadly speaking, the level of acceptability of the intervention to staff at GP surgeries remained relatively high, even after, in many cases, a relatively unsuccessful period of intervention implementation.

The only universal impediment to the acceptability of this intervention in primary care was the time it took to recruit to the study. In the context of 10-minute consultations, the research requirements of the study were the major impediment. Moreover, some felt that there was insufficient time to describe and execute the SSK intervention without causing a backlog of other patients. All the GP surgeries struggled to find sufficient time to implement the intervention routinely and most offered the intervention to far fewer patients than they had hoped, because of the pressure of time. Problems with the time it took to deliver the research requirements were exacerbated by concerns about the kit itself among some distributors, especially among those surgeries that allowed the patient to take the sample on-site. However, there did remain a degree of enthusiasm about the potential use of SSKs in GP settings as ‘add-ons’ for patients, especially where phlebotomy was based off-site or POCT was not perceived as feasible. In our closing interviews, a number of GP-based distributors talked about the way in which a SSK could benefit some of the people they see, and, in the main, it tended to be the research process that prevented timely distribution. There was even more support for SSKs that used an oral sample or a blood-based sample, provided that these would be simpler to use than the TINY vial (for African people and for all potential users).

The other key issue in the perceived acceptability of the intervention was targeting. The majority of recruiting staff found the concept of targeting black African people acceptable, at least during training, and relatively few reported adverse reactions from patients once the intervention was implemented, though some patients required an explanation as to why they were being targeted. One distributor noted that ‘the black community are more receptive and recognise [HIV] is an issue. Don’t take it personally, not a slight’.

Some surgery staff found the process of targeting difficult, usually because it was not feasible to establish the ethnicity of patients from patient databases, as these were often incomplete or contained erroneous data, or because staff did not have the time to check through patient notes. In the absence of pre-existing data, clinics adopted a range of strategies to identify potentially eligible patients that often varied between staff, even in the same clinic. In some settings, there was discomfort with the concept of targeting based on visible race, because this often led to a discussion about the intervention with patients who were not ultimately eligible. This was especially pertinent in areas where the majority of black patients were of Caribbean, rather than African, origin.

In up to half of the clinics, some (but often not all) distributors felt some unease with targeting black African patients. The concept of targeting black African people raised concerns about exacerbating both HIV-related stigma and xenophobia. In three clinics, the proposed targeting based on ethnicity was accepted during HAUS study training, but rejected once the distribution period began, resulting in little or no recruitment occurring at these sites. In all these sites, most recruiters seemed uncomfortable with the concept of targeting black African people largely for fear of causing offence.

In one clinic, after training, staff decided that discussing the study openly with patients or even asking patients whether or not they were black African could be considered offensive and give rise to accusations of racism. In other surgeries, there was some mild concern that targeting migrants and visible ethnic minorities might exacerbate racism and xenophobia. One distributor noted: ‘I am always a bit cautious about offering based on ethnicity because of what has been happening, in our country, and people feeling a bit uneasy about the colour of their skin and where they are from’.

Acceptability of the HAUS intervention to the staff of community-based organisations

The level of acceptability of the intervention to staff at CBOs remained high throughout the study, in spite of substantial recruitment difficulties. CBOs maintained that SSKs were a useful additional intervention for their choice of services when targeting black African people; they also felt that the SSKs were considered convenient by their clients, though HIV-related stigma clearly presented a barrier to some enrolment. Broadly speaking, SSKs remained acceptable to CBO staff in spite of evidence that SSKs were not necessarily a feasible add-on intervention to the outreach activities that formed a large part of their existing funded work. For all CBOs, pre-existing commitments to funded interventions meant that they had limited capacity to recruit to the HAUS intervention. CBO staff had varied perceptions as to the likelihood of acceptability among black African people as a result of HIV stigma, the time required to complete research processes and the process of using the kit itself.

Although the staff and volunteers at one of the CBOs struggled to implement the intervention, they reported that it ‘felt good’ to be able to offer SSKs as an option for their clients. Furthermore, they reported that participation in the HAUS intervention increased morale among staff, partly because ‘they had something concrete to give out to people’, but also because it allowed them to have a dialogue with clients about HIV testing. Staff at this CBO had hoped to integrate the SSK offer into the standard outreach work that they are funded to provide to black African people, but this proved difficult because the HAUS process was too time-consuming. The staff members reported trialling the intervention at three ‘events’ at barbershops in Lambeth, but they also reported that there was some resistance to their activities in these settings, with a high number of refusals, which they assumed was a consequence of a lack of privacy in these settings.

Staff from another CBO had a little more success with recruitment, but still found it difficult to implement the intervention. In spite of difficulties, the intervention remained acceptable to the organisation and its staff, and they reported fewer acceptability concerns for their clients than staff from the first CBO described above. They were comfortable with targeting African people (as they too were black African) and with the element of providing information required alongside the intervention, even though this could be time-consuming. Staff at both CBOs felt that participating in the HAUS intervention was a good experience, as it filled gaps in their existing range of services. For instance, even though they had a mobile testing unit, the HAUS intervention allowed them to distribute kits where the mobile unit could not go.

Staff at the third CBO expressed considerable enthusiasm about the HAUS study during initial contact and training. Like staff at the other participating CBOs, they felt that the HAUS SSKs would give them an important new tool to engage with service users, and would help to supplement their office-based point-of-care HIV testing service. Distributors felt that their service users had benefited from the intervention; although not many kits were distributed, they felt that awareness of testing had increased as a result of the discussions prompted by the offer of the SSK.

All CBOs reported that they approached many eligible people who refused to participate. CBOs reported two key barriers to participation that might be considered evidence that the intervention was not acceptable to the intended target audience. The primary acceptability problems identified by the CBOs were very similar to those identified by the GP surgeries, with the added complication that the CBOs’ recruitment tended to focus their recruitment efforts on outreach interventions, rather than selecting among their existing clients.

Although the staff at CBOs had no initial qualms about targeting only black African people, they subsequently reported that some people approached were hostile about being targeted and even aggressive. African men were reported as especially challenging to engage. Even though staff at all three CBOs had experience of recruiting African people to other HIV prevention interventions, they all expressed surprise that there was still so much stigma associated with HIV among the wider community of African people in London. They felt disappointment that the intervention was limited to black African people, as it disqualified other black groups from participation. This issue was a bit off-putting to some black people approached, as they would indicate interest in the SSK, only to be turned away as a result of their black Caribbean status. This issue became especially problematic when the SSK was offered to a group of people, some of whom were eligible for participation and some of whom were not. One CBO also reflected that the usual users of their office-based HIV testing service were highly self-selecting, in that they were already primed to consider HIV testing, which was not the case in the wider community. The extent of disengagement from HIV, framed in hostile or suspicious responses, was described as being difficult to bear during kit distribution. Some staff felt that ignorance of HIV (including beliefs that HIV was a conspiracy or could be cured with prayer) was a barrier to using a SSK, but this is likely to be a disincentive to all types of HIV testing.

The second major barrier to recruitment was the amount of time it took to explain the study, fill out the paperwork and go over the kit contents. All CBOs intended to distribute SSKs in outreach sessions already scheduled for other funded projects, but this was not always feasible, as it took much more time than they expected to explain the study to potential participants, especially if they showed the video to potential testers.

As in the GP surgeries, CBO distributors reported problems with the actual intervention, including the amount of blood needed to fill the vial, the requirement to provide personal details to receive the result and the preference of some people to know the outcome of the test immediately. Some distributors reported that potential participants felt that they should be paid to participate in the research study.

Population

Sex-specific prevalence rates of HIV, including diagnosed HIV and estimated undiagnosed HIV, were used to determine the baseline number of HIV infections in our model population. 1 Data from PHE also provided information on the proportion of individuals with early- and late-stage HIV, and the proportion receiving antiretroviral therapy (ART)1 (see Table 13).

Model structure

A patient-level simulation was developed to assess the cost-effectiveness of SSKs among black African people in the UK compared with current practice. The model was developed using published data and results from the HAUS trial to predict individual participants’ transitions, costs and health outcomes. The model was created in Microsoft Excel 2010 in accordance with methodological recommendations for evaluations of new health-care technologies and interventions. 84,85

The same hypothetical cohort of 8000 patients was tested under two different HIV screening arms, similar to the trial: (1) intervention (SSK) or (2) comparator (current practice) (Figure 7).

FIGURE 7.

Decision tree of patient movements through the model.

The patient-level model is a lifetime patient-level simulation with the following three components:

1. the simulation of HIV disease progression

2. the probability of HIV transmission from HIV-positive to HIV-negative patients

3. the clinical effectiveness of SSKs in increasing the uptake of HIV testing and reducing the number of undiagnosed and untreated infections of HIV in black African people in the UK.

The model consists of cycles of 3 months’ duration. In each cycle, patients are in one of six health states or two death states (Figure 8). All patients enter the model alive and in one of the six health states. There is a 10-cycle run-in period to stabilise the model, during which no outcomes are measured. After the 10th cycle, patients who are not dead and not in the state ‘diagnosed HIV’ are offered a SSK when they are in the SSK screening arm.

FIGURE 8.

Model structure for health states.

During each cycle, patients can remain in their current health state or move to a new one, based on the following rules.

• Only sexually active patients can move from being HIV-negative to being newly HIV positive.

• Newly HIV-positive patients move directly into ‘HIV undiagnosed’ in the next cycle; patients can only move from ‘HIV undiagnosed’ to ‘HIV diagnosed’ if:

  • they are in the SSK arm, and accept and return a SSK test

  • it is done through current screening practices in both arms.

• Once a patient is ‘HIV diagnosed’, they have a probability of either receiving ART or not receiving ART; in both regimes, patients who are diagnosed as being HIV-positive receive ART, in line with HIV best-practice guidelines, whereby all patients start ART following diagnosis, regardless of CD4 count. 86

• Diagnosed and undiagnosed HIV patients have a 3-month probability per cycle of either HIV-related causes or death from other causes; HIV-negative patients have a risk of death from other causes only. This was taken from Office for National Statistics’ (ONS) life tables. 82

• All patients continue to cycle through the model until they reach an absorbing state of ‘dead’.

In the primary analysis, the SSK screening tool was only offered at one time point (baseline). A secondary analysis evaluated offering SSK screening tools at multiple time points (annually).

Clinical effectiveness of screening tools

Our model assessed two HIV screening tools: (1) SSKs and current practice, and (2) current practice only.

The clinical effectiveness of screening tools is dependent on the percentage of:

• HIV-negative and undiagnosed HIV patients who are given a HIV SSK (acceptance rate)

• patients given a SSK who return the SSK (return rate)

• returned tests that provide a complete result

• patients with undiagnosed HIV who are diagnosed as being HIV-positive after using the screening tool (dependent on the sensitivity of the screening tool used).

The SSK screening tool employed in this study, was a blood test sampling kit, with a sensitivity and specificity of 0.999 (95% CI 0.999 to 1.000) and 0.992 (95% CI 0.982 to 0.998), respectively. The percentage of patients who are given a SSK and who return the SSK was based on data from the HAUS trial. A range of values of acceptance and return rates were tested within sensitivity tests of the model.

HIV screening tools in current practice settings, such as sexual health clinics, GP clinics and antenatal services, are typically blood tests. The sensitivity and specificity of current practice screening tools were assumed to be the same as those for SSKs. The probability of testing was taken from the National Survey of Sexual Attitudes and Lifestyles (Natsal)-3, according to which the 5-year rate of HIV testing in black African men and women was 43.9% (95% CI 30.3% to 58.6%) and 46.1% (95% CI 35.6% to 57.0%), respectively. The probability of having a test in the past 5 years was based on HAUS trial data in a sensitivity test.

All individuals in the model who are diagnosed using a SSK receive confirmatory testing in current practice settings. In addition, all those who are not diagnosed as being HIV positive, and who are sexually active, could have additional testing in current practice settings, as per current practice testing rates.

Transition probabilities

In our model, events were assumed to have occurred when the patient-specific probability of an event was greater than a random number generated in Microsoft Excel. For example, if an individual who was HIV negative had a probability of HIV in a cycle of 0.4%, and the random number generated was 0.001 (0.1%), the individual would move to the ‘new HIV’ state. If, however, the random number generated was 0.553 (55.3%), the patient would stay in the HIV-negative health state. Events for patients were carried over cycles and, hence, the model has memory of past health states for each patient.

The probability of being infected with HIV was determined by four factors:

1. the transmission rate of HIV to HIV-negative patients for HIV-positive patients receiving ART87

2. the transmission rate of HIV to HIV-negative patients for HIV-positive patients not receiving ART87

3. the prevalence of HIV in the model population during that cycle

4. the probability that HIV-negative patients are sexually active; this was taken from the African Health & Sex Survey 2013–14: Headline Findings,26 which reported that 84.2% of men and 73% of women had a sexual partner in the previous year. A sensitivity analysis was conducted using the percentage of sexually active patients reported in the HAUS study.

A weighted average risk of HIV infection was applied to all patients who were HIV negative and sexually active in the model.

Patients who were HIV positive progressed through the stages of HIV (stage I, stage II, stage III and stage IV: death from HIV). At baseline, individuals infected with HIV were proportioned to the stage of disease. The proportions of individuals in the early and late stages of the disease were equal to the proportions of patients in each group reported by PHE in 201588 (see Table 14). Individuals who were free of HIV at baseline, and subsequently infected with HIV during the model, entered stage I of the disease. Progression through the stages of having HIV was determined by changes in CD4 count, whereby stages I, II and III were designated by a CD4 count of > 499, 350–499 and < 350 cells/mm³, respectively. Baseline CD4 count for newly infected patients and change in CD4 count were calculated using the algorithms reported in an economic model of lifetime outcomes and costs of HIV in MSM in the UK. 83 Patients who were HIV positive, receiving ART and adherent with medication, changed viral load and CD4 count in line with the algorithm set out in Nakagawa et al. ,83 based on treatment strategy and adherence.

The probability of dying from HIV was applied to all HIV-positive patients. Risk was stratified by sex and CD4 count, whereby risk increased 10-fold for those in the late stages of HIV (i.e. a CD4 count of < 350 cells/mm³). One-year probabilities were converted to 3-month rates, and then probabilities. 89

The probability of dying from non-HIV related causes was calculated from the ONS 2015 data,82 according to which death from non-HIV-related causes was equal to total mortality minus HIV-related mortality (International Classification of Diseases, Tenth Edition codes B20 to B24, as per the ONS guidelines). As mortality is higher in those with HIV, the probability of death from non-HIV-related causes was increased by 50%. 83 The risk of death from non-HIV-related causes was age and sex specific. We conducted a sensitivity analysis that increased the probability of death by 100% for non-HIV-related causes.

Clinical effectiveness of antiretroviral therapy

The clinical effectiveness of ART was applied to the transmission probability of HIV, whereby patients receiving ART were at a lower risk of transmitting HIV than those who were not. All patients who were newly diagnosed with HIV were assumed to be prescribed ART in line with current guidelines. Adherence with ART was based on the rate of adherence from PHE HIV statistics88 (see Table 14).

Costs

The costs included in our model were the cost of the HIV screening tool and the cost of HIV (Table 14). The cost of the HIV screening tool included the cost of SSKs and/or current practice screening tools, the cost of confirmatory testing, any additional testing and pre- and post-test counselling.

The cost of the SSK was calculated from the HAUS study and includes the cost of the test kit itself, as well as the time taken to deliver the kit. The cost of the current practice screening tool was calculated as £9.33. 38 In addition, pre- and post-test counselling was valued at £46.45 per test.

The cost of HIV was extracted from an economic analysis of early access to HIV services. 92 The cost of HIV included the cost of hospital inpatient, outpatient and day ward services, determined by the mean number of days spent in each service. The cost of HIV was stratified by CD4 count and the presence of ART. For undiagnosed HIV, the cost was estimated to be equivalent to individuals with early-stage HIV (CD4 count of > 200 cells/mm³) and who were not receiving ART, as these individuals are likely to have more illnesses/visits to the doctor than HIV-negative individuals.

All costs were inflated to 2015 values, using relevant national price index conversion rates. 94

Outcomes

The outcomes assessed in this analysis included societal and individual-level outcomes. Societal outcomes included the number of HIV prevented, while individual-level outcomes included the impact of HIV on mortality and morbidity.

The impact on mortality and morbidity was evaluated using QALYs, as recommended by NICE in the UK. 95 QALYs are calculated by multiplying a preference-based value of a health state (a utility score) by the amount of time spent in that health state, whereby 1 represents perfect health and 0 represents death. All individuals who were HIV negative were assumed to be in a ‘healthy’ state with a utility score of 0.824. If an individual was infected with HIV, a utility decrement was applied. 93 This was applied at the time of infection, as well as every cycle thereafter until the individual died. The utility values were taken from a cross-sectional study of health-related quality of life of people infected with HIV compared with the general population. Utility decrements were dependent on viral load, CD4 count and the use of, and adherence to, ART. As individuals progressed through disease stages, and/or the use of ART changed, the utility decrement applied was updated accordingly.

Cost-effectiveness was evaluated as the mean incremental cost per QALY gained of 8000 patients over a lifetime horizon. 96 The incremental cost-effectiveness ratio (ICER) for the HAUS study is defined as the total discounted cost of the intervention (SSKs) minus the total discounted cost of current practice, divided by the total discounted QALYs of the intervention minus the total discounted QALYs of current practice. An ICER of £30,000 per QALY is generally reported as the maximum value at which NICE is likely to approve the implementation of a new technology. However, Claxton et al. 97 have calculated that it currently costs £12,936 to generate an additional QALY in the English NHS. 97 As a result, it has been argued that this value may represent a more appropriate cost-effectiveness threshold to use.

The random numbers in the model meant that each run of the model would give a different result. As such, we took the average of each parameter in the ICER from 10 runs of the model. A Monte Carlo Error from the 10 runs of the model is reported to provide an estimation of the precision of 10 runs of the model.

All future benefits (QALYs) and costs were discounted at 3.5% per annum, converted to a 3-month rate. 95

Sensitivity analyses

We conducted a number of sensitivity analyses. The analyses were conducted in line with the Decision Support Unit guidance. 98

Barriers to, and facilitators of, the use of self-sampling kits

The first objective of stage 1 was to clarify barriers to, and facilitators of, provision, access and use of HIV SSKs by black African people in community settings. The policy review revealed a facilitative environment for STKs and SSKs for HIV, with the use of these kits being viewed as an innovative method of increasing uptake. Guidelines have shown that HIV testing in medical and community settings is both feasible and acceptable, with SSKs suggested as a route to broaden testing options. Evidence on the use and clinical effectiveness of SSKs, however, was scant – especially for black African groups.

The data yielded through interviews and FGDs echoed many of these reported barriers to, and facilitators of, SSK use. The privacy and convenience afforded by SSKs were viewed as major strengths, as they negated the need to attend a sexual health clinic and potential stigma. Service providers and non-specialists alike felt that SSKs increased individual autonomy and testing opportunities. However, both the FGD participants and the interviewees voiced concern over the volume of blood required to provide a TINY vial sample. Service providers in particular voiced doubts over the ability of members of the public to produce a sufficient sample. Participants also pointed out that fear of needles and blood, and the complexity of using the kit, could discourage some people from using the SSKs. Another concern was the potential isolation that could be experienced by users of the SSKs.

Many participants mentioned the stigmatising association of HIV with ‘sexual immorality’ and promiscuity (an association that many FGD participants and interviewees described as being heightened within black African communities) as a barrier to testing among black African people. SSKs, however, helped to mitigate the risk of others discovering that one was testing for HIV, as a result of the privacy afforded; however, as described above, what was regarded as the ‘benefit’ of privacy by some was regarded by others as carrying a risk of increased isolation for those receiving news of a reactive result.

The difference between individual- and community-level risk perception was also viewed as a barrier, with participants stating that those targeted may not appreciate the epidemiological data used as a justification to target black African people exclusively for the intervention. Participants recognised that, because of pervading HIV-related stigma and perceptions of being at a low risk of acquiring HIV, an opportunistic provider-led offer of a SSK was more likely to increase HIV testing than a user-led model. The participants warned that offers of SSKs exclusively to black African people may be perceived as racist and as fuelling anti-immigrant rhetoric.

Determining appropriate self-sampling kit-based intervention models for different settings

As participants in FGDs and interviewees recognised that black African people were already accessing primary care and CBOs, the addition of SSKs into their service options was viewed as a convenient extension that would immediately increase testing opportunities. There was almost universal approval of CBOs working with African communities and HIV prevention to be involved in distribution, as their staff members were considered to possess the expertise required to target black African people in a sensitive manner. CBOs, however, would not be extending their outreach activities, and HIV testing was already widely integrated into their services; therefore, there was some concern from the research team that this would not significantly increase HIV testing opportunities for the wider African community.

Participants advised that targeted offers needed to be couched in clear terms, using epidemiological evidence, and in a way that would avoid the perception that offers were being made because of racist or xenophobic sentiment. It was deemed of utmost importance that distributors were regarded as trustworthy, knowledgeable and non-judgemental in order to foster a realistic degree of reflection about HIV risk among those to whom they encouraged SSK use; it was felt that staff in GP settings and outreach teams in CBOs possessed all of these qualities. Ultimately, there was a universal view among participants that SSK distributors needed to actively resist HIV-related stigma, rather than potentially reinforce it.

The formative work revealed concerns about the limited amount of time available to providers to initiate discussions and gain informed consent, especially in GP surgeries. Embedding the intervention into routine practice was considered to be a critical component by the research team if the intervention was to be sustainable. Therefore, specific appointments for intervention delivery should not be part of the model.

Prior to intervention development, it was apparent that, ideally, a saliva-based test or a more user-friendly SSK would be preferable to the TINY vial. Unfortunately, despite this knowledge, and as described in Step 4: viability of the intervention, we were unable to proceed with an alternative option, as we could not identify a service provider that was willing to assume the liability for using an off-licence product in the context of a research project. At the time of writing, the TINY vial was the only CE-approved HIV SSK available in the UK. An alternative option would have been to formally test a saliva-based SSK against the TINY vial; however, this would have represented a significant shift in the original research question and design, which was not considered an appropriate option for the HAUS study.

The final intervention model for both settings utilised a theoretically informed scripted conversation in conjunction with distribution of a SSK during a routine encounter. The script was designed to assist providers in overcoming their anxieties about targeting black African communities (addressing the issue of capability), while also motivating users and increasing opportunities for HIV testing; it also ensured that the message was consistent, in order to enable intervention fidelity. The information supplied with the kit (including a web link to a video demonstration) and via the provider was designed to further increase the capability and motivation of service users. The study team were not able to deliver the intervention with what they believed would be the optimum SSK, owing to governance and regulation restrictions. A copy of the intervention and training manuals are available on request.

Determining robust HIV result management pathways

The formative work highlighted a disconnect between the accepted current practice of delivery of results by text message or telephone call, and the relatively widespread concern that SSK users may be isolated and at risk of self-harm when receiving test results. The use of SSKs was perceived to reduce opportunities for HIV prevention, with a lack of recognition that there is no strong evidence that routine pre- or post-test counselling subsequently affects the behaviour of those who receive a negative test result. Participants also warned that a profusion of HIV testing interventions could lead to a disjointed and confusing service landscape.

The clinical pathways within the HAUS study were all subject to clinical governance regulations and designed to ensure that best practice would be delivered. The research team felt that it was important that an organisation with experience and expertise in managing HIV results, and ideally SSKs, were involved to ensure appropriate and sensitive delivery of the test results, including the offering of retests to those with indeterminate results and confirmatory tests to those with reactive results, and linking those who tested positive for HIV to care.

The process of establishing these clinical pathways highlighted the uncertainty and confusion over where the duty of care and liability should lie for a SSK intervention delivered in community settings by non-NHS trust employees.

The provision of results as quickly as possible to service users, whether positive or negative, is emphasised throughout clinical standards, and the recommended time period for receipt of results is ideally within 48 hours, and no more than 14 working days. The process evaluation demonstrated that, although this was achieved for everyone testing negative, for those with insufficient samples, there was an occasional delay while information from recruiting sites was awaited. The FGDs highlighted significant concerns regarding confidentiality and privacy among potential users; this led us to request only initials (rather than a full name) with the returning sample, which, in hindsight, was an error, as it meant that we could not make contact with participants until the additional information was received from the recruitment site. All negative results were delivered by short message service (SMS) via an automated system, and this worked well. No reactive results were received, so the processes for confirmatory testing and linkage to care were not tested. There was also concern that people may not test if their general practitioner was automatically informed of the result. As a consequence, we obtained specific consent to notify general practitioners of the test results; 98.9% of participants recruited through primary care consented to this, suggesting that this concern was unfounded.

Establishing the feasibility and acceptability of interventions for providers and service users

The primary objective of stage 2 was to determine the feasibility and acceptability of a provider-initiated, HIV SSK intervention targeted at black African people in two settings: GP surgeries and CBOs. Our findings indicate that, although many aspects of the intervention were acceptable, scale-up of the intervention to a Phase III evaluation is not feasible. All sites reported that the time it took to offer the kit and complete the research-relevant forms (consent, baseline questionnaire and enrolment log) was a major impediment to the feasibility and acceptability of the trial. With hindsight, a pilot of stage 2 in a GP surgery and a CBO might have identified some of these issues; however, project timelines and resources did not enable this. Consent at an individual level was sought to enable the collection of baseline data, tracking of data and follow-up interviews. Consent at a group (GP or CBO) level, and limiting data collection to service evaluation, may have proved more feasible and better reflected how the intervention would be implemented in practice.

Although interviewed participants reported that the targeting of black African people was acceptable, this is in contrast to the large proportion of respondents who indicated on the acceptability questionnaire that it was not acceptable to offer a HIV test based on a patient’s ethnicity. Many distributors, especially at GP surgeries, also continued to express unease at targeting black African patients only, despite the provision of training and a script designed to specifically overcome these barriers. This unease arose from concerns about exacerbating HIV-related stigma and xenophobia.

Our findings highlighted that community providers, as well as many black African people, would prefer to use SSKs that involve the collection of an oral specimen as opposed to a blood specimen. However, at the time of writing, in the UK, there are no CE-marked HIV testing assays that can be used on saliva. Not being able to proceed with a saliva-based option had a direct impact on the study feasibility. It meant that the SSK took longer to distribute, as there was more to explain and greater motivation required for a blood-based SSK; it was no longer possible to demonstrate kit use in the field, and the more complicated procedure to collect the sample was likely to have influenced the return rate and to be responsible for the extremely high rate (16.9%) of insufficient samples. This figure is remarkably similar to that reported by the national HIV self-sampling service among their black African users to date (Tim Alston, personal communication). This suggests that an alternative to the TINY vial is required if HIV SSKs are to be an effective means to increasing the uptake of HIV testing in black African communities.

Evaluating the clinical effectiveness of self-sampling for HIV in increasing the uptake of HIV testing by black African people

As a result of the failure of the study to recruit adequate numbers to the study, we were unable to evaluate the clinical effectiveness of SSKs in increasing the uptake of HIV testing in black African people.

The fact that 12 of the 65 participants who returned a kit reported having never previously tested, and that most kit returners tested because of the opportunistic offer of the test, suggests that if distribution of HIV SSKs in these settings were rolled out at scale as part of the range of available HIV testing options, there would be the potential for increasing the uptake of HIV testing. However, at a structural level, these services do not have time to ‘bolt on’ a SSK intervention (or indeed anything else), unless there is a strong benefit or incentive to do so.

Determining the cost-effectiveness of distributing the self-sampling kits among black African people over other screening methods

The model of a SSK dispensed to black African people in GP surgeries or in CBOs presented in Chapter 10 suggests that SSKs may be a cost-effective way to identify new infections of HIV, as SSKs resulted in more QALYs for less cost.

We showed that it was feasible to collect data as part of the HAUS trial in order to update values in the model. However, there were challenges in collecting data on the time taken to explain SSKs to participants. Although there were limited data available on the time taken, this proved to be inconsequential to the model, given that SSKs could cost as much as £50 per test and still be cost-effective.

The estimation of the uptake rates of SSKs taken from the HAUS study and used in the model may not reflect real life. The uptake rate for SSKs might be higher, as patients may have been put off by the research process. It is also possible that the use of an alternative SSK, which is easier to use, may increase return rates and reduce the number of incomplete results. It is likely that these improvements would improve the cost-effectiveness of the SSK.

Further work required includes running the model and the probabilistic sensitivity analysis more times, whereby values in the model are varied within given CIs and distributions to provide an estimate of the probability that SSKs are cost-effective compared with current practice. The extra value of perfect information would provide a monetary estimate of the value of additional research to provide more certainty of the result. However, the computational load required to run this analysis is significant, potentially taking days to run, and was not possible as part of this study.

Monitoring the ability to trace participants with reactive results, confirmatory testing and linkage into specialist care

No reactive results were obtained, so we were unable to assess the ability to trace participants with reactive results through the process of confirmatory testing and linkage to care.

Determining the cost per person of distributed kits and the cost per HIV diagnosis per setting

Data on the cost per kit distributed were collected as part of the trial. This was dependent, however, on who distributed the SSK. If a conservative upper estimate was made of a PN taking 5 minutes to explain the kit, the cost of staff time was £4. The economic model did not determine cost per setting as originally planned. It is unlikely that community workers would be paid more than a PN; thus, the cost-effectiveness could be expected to be greater in CBO settings, unless the time taken to explain the kit differed by setting. A limitation of the model is that the time to specifically target the individual and explain the kit could not be determined, as, within the study, the research components were so enmeshed with this process.

The cost of the SSK itself was £3.24 if not used, and a returned kit cost £15 (as this included the cost of a retest for incomplete tests). In a lifetime patient-level model of 8000 black African people in the UK, the cost per detected infection of HIV using a SSK was £6000. However, SSKs resulted in more QALYs for a total lower cost once the cost of treating HIV was included, dominating current practice. The Monte Carlo CI of 10 runs of the model suggested a significant increase in QALYs, but not cost.

Assessing the feasibility of collecting data for a lifetime cost-effectiveness model alongside the potential Phase III evaluation

Limited data were able to be collected to inform the cost-effectiveness model, because of a need to reduce the time taken for patients to complete questionnaires as part of the baseline assessment. The values that were collected as part of the questionnaire that could inform the model (the percentage of sexually active patients and the proportion of people tested in the past 5 years) tallied with the best-available data in the literature, suggesting that data from the literature are likely to be adequate and a better source of information. We were able to obtain better estimates of the proportion of people who agreed to use a SSK and returned the SSK to inform the model, although real-life scenarios, rather than trials, are likely to provide more realistic data.

Assessing the feasibility and, if appropriate, the optimal trial design for a future Phase III evaluation

Based on the findings of the HAUS study, in particular the challenges with recruitment in both GP and via CBOs, a future Phase III evaluation was not considered feasible by the study team.

Inclusion criteria: quantitative studies

• Population: any lay population in any country.

• Intervention: SSKs for the detection of HIV.

• Comparison: any other HIV testing method.

• Outcomes: completion of a SSK for HIV; participant-reported acceptability; adverse events; first-time testers; test/strategy; linkage; accuracy; and cost-effectiveness.

• Study design: any randomised or non-randomised evaluation or observational design.

• Publication date: 2000 to present.

• Language: English.

Inclusion criteria: qualitative studies

Studies in any population group in any country describing, summarising or analysing the experiences and perceptions of any population group about SSKs for the detection of HIV. This will include studies of experiences of, and responses to, personal completion of SSKs for HIV detection and also perceptions and opinions of potential use. Any text-based qualitative research methodology will be included: interviews, focus groups and ethnography. Studies published from 2000 until the time of searching will be included.

Screening

Titles and abstracts of studies retrieved using the search strategy and those from additional sources will be screened by two researchers to identify studies that potentially meet the inclusion criteria outlined in Inclusion criteria: qualitative studies. Studies excluded at this stage will be excluded as ‘not relevant to SSKs for HIV’ and will be counted to allow completion of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. At this stage, both quantitative and qualitative studies of SSKs for HIV will be retained together. When an abstract is not available and the title cannot be used to assess relevance, the article will be retained for full-text screening.

The full text of these potentially eligible studies will be retrieved and assessed for eligibility. All the PDF files will be incorporated into the bibliographic database, which will be delivered to the rest of the team. A more detailed coding scheme for exclusion reasons will be developed and recorded for completion of the PRISMA flow diagram. Eligible quantitative and qualitative studies will be separated at this stage and treated separately in the PRISMA flow diagram.

At each stage, the two researchers (EM and Gemma Philips) will:

• complete a 5% sample of records in parallel to check the consistency and feasibility of inclusion/exclusion criteria application

• split the remaining records and screen individually, with a 10% sample being double-screened.

Any disagreement between the two researchers (EM and GP) over the eligibility of particular studies will be resolved through discussion and, when needed, a third researcher will be consulted (FB and/or CD?). For all the excluded references, the exclusion reason will be recorded in an EndNote/Microsoft Excel spreadsheet.

The researchers will contact the investigators for clarification when eligibility cannot be determined from the published study, and for any new studies or references that may not be published yet, and to check that the search has captured important articles in the field. We will review backward and forward citation of Web of Knowledge for the publications finally identified as relevant after the screening process.

Full-paper screening

Search strategy

1. exp HIV Infections/ (238,345)

2. exp HIV/ (86,013)

3. hiv.ti,ab. (225,871)

4. ‘hiv1’.ti,ab. (710)

5. ‘hiv2’.ti,ab. (149)

6. ‘hiv type 1’.ti,ab. (3784)

7. ‘hiv type 2’.ti,ab. (186)

8. human immunodeficiency virus.ti,ab. (69,068)

9. human immunedeficiency virus.ti,ab. (4)

10. human immuno-deficiency virus.ti,ab. (186)

11. human immune-deficiency virus.ti,ab. (245)

12. (human immun* adj3 deficiency virus).ti,ab. (435)

13. acquired immunodeficiency syndrome.ti,ab. (14,483)

14. acquired immunedeficiency syndrome.ti,ab. (10)

15. acquired immuno-deficiency syndrome.ti,ab. (94)

16. acquired immune-deficiency syndrome.ti,ab. (4904)

17. (acquired immun* adj3 deficiency syndrome).ti,ab. (5037)

18. Sexually Transmitted Diseases, Viral/ (1207)

19. or/1-18 (313,426)

20. (sample adj1 collect*).ti,ab. (5944)

21. home dried blood spot.ti,ab. (0)

22. (alternative adj3 test*).ti,ab. (3862)

23. (option* adj1 test*).ti,ab. (270)

24. ((Home* or self* or mail*) adj3 (collection* or sampl* or specimen* or test* or kit)).ti,ab. (10,969)

25. or/20-24 (20,907)

26. 19 and 25 (1003)

27. limit 26 to (english language and yr=‘2000 -Current’) (673)

Community-based organisations

General practice

PHE

Dr Anthony Nardone.

Terrence Higgins Trust and King’s College Hospital NHS Foundation Trust

Dr Michael Brady.

Pharmacist

Mr Alistair Murray.

Positively UK

Ms Rebecca Mbewe.