Methods

Material throughout this chapter has been reproduced with permission from Arakelyan et al.28 This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial licence (CC-BY-NC) (https://creativecommons.org/licenses/by-nc/4.0/). This allows for the non-commercial reuse of the published paper so long as the published paper is fully attributed. The text below includes minor additions and formatting changes to the original text.

Some material in this chapter has also been reproduced from Jones et al. 2021.29 © 2021 International Union Against Tuberculosis and Lung Disease. Reprinted with permission.

Material throughout this chapter has also been reproduced with permission from Karat et al.30 This article is available under the Creative Commons CC-BY-NC-ND (https://creativecommons.org/licenses/by-nc-nd/4.0/) licence and permits non-commercial use of the work as published, without adaptation or alteration, provided the work is fully attributed.

Material throughout this chapter has also been reproduced with permission from Jones et al.31 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The text below includes minor additions and formatting changes to the original text.

Study context

In 2020, there were an estimated 7.3 episodes of TB disease per 100,000 population in England (equivalent to 4125 individuals).1 Almost three-quarters of those who developed TB disease were born outside the UK, with higher TB incidence rates seen in metropolitan areas. For example, in London, the 2020 incidence was 16.3 cases per 100,000 population, more than twice the national estimate.1 Additionally, the coronavirus disease (COVID) pandemic likely resulted in people with TB disease presenting less often to healthcare services and with a lower likelihood of being diagnosed, suggesting that these numbers underestimate the true burden of disease.1

In high-burden TB areas in the UK, outpatient treatment is delivered by multidisciplinary TB services, including doctors, nurses, social care workers, and DOT outreach workers. These services are designed to support the complex social and medical needs of TB patients through a ‘case management approach’, recommended by the Royal College of Nursing, to comprehensively follow-up confirmed or suspected cases.8 Each patient is assigned a case manager (CM), most commonly a specialist TB nurse, who is responsible for their care. At treatment initiation, CMs conduct a risk assessment to highlight specific medical needs the patient may have, such as complex TB disease, or social care needs, for example, housing or employment. This informs the CM’s judgement about whether enhanced case management (ECM) may be required. ECM includes a wide variety of support measures, usually titrated to the patient’s needs based on the judgement of the CM and other team members, and in consultation with the patient. Forms of adherence and other treatment support include supervision (e.g. DOT or VOT), increased contact and supportive person-time (e.g. home visits), and medication-organising aids (e.g. dosette boxes).

As part of standard TB care, all patients are reviewed by their CM at treatment initiation, week 2, and then usually monthly from months 1 to 6. Standard TB regimens in the UK for DS disease in adults are currently 6 months minimum in length, although treatment may last up to 2 years for patients with DR or complex disease.

Public health policies in the UK often differ between England, Scotland, Wales and Northern Ireland. For simplicity, we focus here on England, where most of the fieldwork for this study was conducted. England’s TB programme is co-ordinated by the UK Health Security Agency (formerly Public Health England) and NHS England (NHSE). National strategies and policies are devised and enacted in line with wider NHS regulations and priorities as well as the UK’s commitments to global programmes such as the WHO’s End TB Strategy.32 Strategic priorities for the TB programme in England at the time of this study, documented in the Collaborative Tuberculosis Strategy for England, 2015–2020,33 included improving access to services, improving treatment and care services, reducing DR TB, tackling TB in underserved populations, and strengthening surveillance and monitoring.

Research pathway

At the centre of this study was a multidisciplinary Intervention Development Group (IDG), made up of clinicians, researchers and other stakeholders including activists and patient advisors with experience of ATT. Intervention development was an iterative process: the group was active during 2019 and was presented with data from formative work conducted by our team.

Formative work included two scoping reviews of the literature describing the determinants of non/adherence to ATT; in-depth interviews with TB patients, caregivers and HCPs; and structured observations conducted in NHS TB clinics. Our iterative development approach allowed for new findings from formative work to be incorporated as they became available. The IDG interpreted and contextualised data, made decisions around how determinants were prioritised, and identified additional risk factors for consideration.

Guided by the IDG and the Medical Research Council (MRC)/National Institute for Health and Care Research (NIHR) framework for developing and evaluating complex interventions,34,35 a small core team of coinvestigators then built the intervention components using the Perceptions and Practicalities Approach (PaPA), a behavioural framework designed specifically to assist in the development of interventions to support informed treatment decisions and optimise adherence.36

The intervention package comprised:

1. a digital TB needs assessment (TNA) tool, to be administered by the CM at treatment initiation and at regular intervals until treatment completion
2. an interactive patient booklet to guide participants through treatment (the booklet was available in four languages and included a diary to record symptoms, adherence and side effects; information on a range of topics including emotional well-being and stigma; and contact details for local TB and social services)
3. two brief, animated videos (4 minutes each, titled ‘Helping you understand TB’ and ‘Questions about TB’), shown at enrolment and made available to view at home, intended to increase participants’ motivation and ability to take ATT
4. a detailed operations manual for CMs.

We tested the intervention package in a pilot study at four NHS TB service sites in London. This was designed as a cluster-randomised trial: at each study site, informed consent was sought from adults initiating TB treatment. Those who consented at the two intervention sites were provided with the intervention package. Those who consented at the two control/standard care sites were provided with the standard care package. The primary outcome of the trial was adherence to ATT (number of doses taken of a possible 168) among people enrolled at intervention versus control sites.

During the trial, we also assessed the feasibility of collecting behavioural and health economic data during routine clinical care. This was done primarily through questionnaires assessing perceptions, behaviour, quality of life and resource use. Finally, we performed a mixed-methods process evaluation to explore fidelity, implementation barriers and possible mechanisms of effect (Figure 1). This information was intended to contribute to the design of a potential large definitive study within the NHS adequately powered to formally evaluate the manualised intervention (or a modified version) for the management of TB disease.

FIGURE 1 Overview of research pathway. (Figure not reproduced here; see the original article.)

Formative work

Scoping review of quantitative literature

We performed a scoping review to explore determinants of non-adherence to ATT within high income, low TB incidence (HILI) settings such as the UK.

For this section of the work, scoping review methodology was selected as it provided both a broad overview and highlighted key gaps in evidence.37 We expected considerable study heterogeneity in the hits from our searches,38,39 including many different definitions and measurements of ATT adherence.40 The Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension Checklist for Scoping Reviews was used.38

Literature search

Five databases [MEDLINE, EMBASE, Web of Science, PsycInfo® (American Psychological Association, Washington, DC, USA), and Cumulative Index to Nursing and Allied Health Literature (CINAHL)] were searched in June 2019, using key terms related to TB, treatment, and adherence.

Search terms were mapped to the Population–Concept–Context (PCC) framework as recommended for scoping reviews.41 Identified studies were imported into EndNote [Clarivate Analytics (formerly Thomson Reuters), Philadelphia, PA, USA] and duplicates were removed. Any discrepancies were resolved through discussion. Reference lists of included studies were hand-searched to identify any further relevant studies. Eligibility criteria are listed in Appendix 1, Table 21. Included studies were English language, in peer-reviewed journals, reporting primary, observational, quantitative data on determinants of non-adherence to ATT, in countries classified as high income42 with low TB incidence rates (< 40 per 100,000 people) when the study was conducted.29 We included studies measuring adherence outcomes of both discontinuation (early cessation of treatment, including loss to follow-up) and suboptimal implementation (missing doses during treatment).43–45

Data extraction and synthesis

We included determinants if they were studied as primary exposures of interest or as potentially confounding factors and grouped them into demographic, clinical, health systems-related, and psychosocial categories. We rated the strength of evidence for extracted determinants using a proxy measure, based on the size and direction of the effect size (ES) estimate and statistical certainty. The four categories used to classify evidence for both predictive and protective determinants from strongest to weakest were:

• Category 1: Strongest: ES (ratio) ≥ 1.5, p-value ≤ 0.05
• Category 2: ES (ratio) ≥ 1.5, p-value > 0.05, small sample size (n < 154) i.e. study likely to be underpowered
• Category 3: ES (ratio) > 1.0 to < 1.5, p-value ≤ 0.05; and
• Category 4: Weakest: ES (ratio) > 1.0, p-value > 0.05.

We calculated that a minimum of 154 participants would indicate an adequately powered sample size, using 90% power and 5% significance level, assuming 50% of individuals had the outcome among the unexposed, and a one-to-one ratio of exposed to unexposed or cases to controls. Where possible, determinants were classified based on ES in multivariable analyses.

Scoping review of qualitative literature

Our aim was to synthesise and critique the qualitative evidence on the mechanisms through which socioecological factors influence the experience of being on ATT in high income, low TB incidence countries. In our understanding of socioecological factors, we built on and expanded the model proposed by Munro et al.23

Research questions included:

• What socioecological factors affect individuals’ experience of ATT in HILI countries and how do these inter-relate?
• Through which mechanisms do these factors impact on individual adherence to ATT?
• What role does context have in mediating determinants and mechanisms of ATT adherence behaviour?

We drew on the multifactorial determinants of adherence to long-term medication, an approach originally proposed by WHO46 and elaborated further with work from researchers in sociobehavioural and ecological frameworks that seek to explain the web of proximate and distal influences on adherence behaviour.47

We followed the guidance for scoping reviews outlined by Arksey and O’Malley.48 Search strategies and extraction matrices were piloted and further refined to capture all the relevant information for the study.

Literature search and inclusion criteria

We searched six electronic databases [CINAHL, MEDLINE, PsycInfo, Applied Social Sciences Index and Abstracts (ASSIA), Web of Science and Scopus] in June 2019. The initial search included keywords related to TB, treatment and adherence. As this yielded primarily quantitative evidence, we included additional search terms specific to qualitative and mixed methods. Our final search strategy used the following keywords: ‘tuberculosis AND (adherence OR compliance OR concordance) AND treatment AND (qualitative OR ethnograph* OR anthropolog* OR sociol* OR phenomenol* OR narrative)’.

To capture a comprehensive range of qualitative studies relating to the experience of being and staying on ATT, we included studies from any discipline or theoretical tradition that used qualitative and/or mixed methods to describe (1) the experience of taking ATT or (2) factors influencing treatment non-adherence from the perspective of adult patients, care-givers, or healthcare providers (HCPs). We considered studies published in peer-reviewed journals from January 1990 onwards to reflect the relative increase in TB incidence in high-income countries that began in the early 1990s.28 We excluded studies that focused on children, latent TB, adherence to ATT in low- and middle-income countries and those published before 1990 in languages other than English. We did not use explicit definitions of adherence or non-adherence to screen studies for inclusion.

Study selection, data extraction and data synthesis

Following the removal of duplicates, three researchers screened titles and abstracts retrieved through database searches for relevance based on the selection criteria; a fourth researcher independently reviewed a random sample of 10% titles and abstracts to ensure consistency in screening. We sought out full texts and read them. For included articles, we examined lists of references and citations for additional relevant records.

We extracted the following data: study identifiers (study author, year of publication), settings, populations, aims, methods, theoretical approaches and factors shaping the experience of ATT, including specific reference to adherence behaviour. We used a narrative approach to categorise and synthesise the data according to the five socioecological domains listed above. We further examined and documented, where reported and relevant, the mechanisms through which identified factors influenced treatment behaviour.

Interviews with current and former patients, caregivers and healthcare professionals

We conducted interviews to elicit TB patients’ experiences of starting and staying on ATT as well as HCPs’ views on the barriers and facilitators to adherence. We interviewed adults who had been, or were still being treated, for TB, and individuals who had cared, or were caring, for an adult taking treatment for TB. We also invited HCPs who worked in one of four participating NHS Trusts [on the South coast of England, London (two sites), and Scotland] to take part.

Participants

We recruited participants from Barts Health NHS Trust, London; NHS Lothian, Edinburgh; Royal Free London NHS Foundation Trust, London; and University Hospital Southampton NHS Foundation Trust. Criteria for selection are provided in Table 1.

For HCPs, we aimed to interview a range of staff (specialist nurse, doctor, outreach worker and administrator). We gave priority to specialist TB nurses, as they most often act as CMs.

For patients, we attempted to interview roughly equal numbers from each trust, as well as trying to ensure representation of both genders and most affected ethnicities. We made special efforts to include patients without English as a first language. Adults who were currently taking or had recently completed treatment for TB (or other mycobacterial infection) were identified by HCPs in their TB service and asked to take part.

Interview approach

Interviews were conducted in confidential settings in hospitals or community health centres, often in consultation rooms that also contained a workstation, examination couch and sink. All interviews were conducted in person, in English, by one or two interviewers, and with the help of a topic guide that covered the following topics: pathways to TB care; treatment adherence issues; knowledge and perceptions of TB and treatment; adherence to ATT (experiences, enablers, and barriers, and incidence of and reasons for non-adherence); social support and structural and health systems issues (interactions with and potential obstacles to care). At various points during interviews, the lead interviewer confirmed with the participant what was being said through short summarisation.

Interviews were audio recorded [using an Olympus DS-9500 (Olympus Corporation, Tokyo, Japan)], securely encrypted and transferred to a password-protected computer. Hand-written notes were scanned and password protected, and the originals destroyed. Transcripts of interviews were used for analysis after being checked for comprehensiveness and completeness.

Analysis approach

In a stepwise manner, using an adapted framework method, two analysts cross-checked transcripts, identified themes, and piloted the coding system. First, we read transcripts several times and constructed narrative profiles of each patient that included accounts of how they presented themselves and their social networks; their TB diagnosis and early experiences of care; and their experiences of treatment. We paid attention to the unique trajectories of individuals and how demographic features such as gender, age, ethnicity, and social status impacted on pathways to and experiences of care, and the effects of TB and ATT on their lives. Second, using a conceptual framework of determinants of adherence derived from Munro et al.23 and our scoping review of qualitative literature,28 we categorised information from the profiles into a matrix to compare and contrast data on the five key themes (personal, social, structural, health systems and treatment-related) until theme saturation was reached. Third, we used the summarised information to create separate files for each of the five thematic areas. We supplemented retrieved data segments with information from the original transcripts and reviewed them through a constant comparison method to ensure that content and meaning of the themes were consistently applied. To examine how these multiple levels of determinants influence individuals’ adherence and the strength and direction of influence (positive or negative), we identified relevant subthemes, examined their occurrence and recurrence in the data, and mapped the relationships across these subthemes to illustrate direct (proximal) and indirect or mediating (distal) influences on adherence.

Clinic observations

In addition to the literature reviews and interviews, we also conducted clinic observations. These were performed, with patients’ consent, at the two sites in North Central London TB Network. Using field notes, we collected information on the delivery of the standard risk assessment, methods used to assess adherence and assign support and the education and information provided to patients. We collected additional data through observing patient appointments and discussing procedures with CMs, social care and outreach workers. Observations were conducted in parallel to the development of intervention content, to ensure that the feasibility of delivering emerging components in standard care was considered throughout the development phase.31

Intervention development

Findings from the formative research detailed above and the outputs of the intervention development process are technically ‘results’ of this study. However, it is important for the reader to have a clear understanding of the characteristics of the intervention, how it was expected to work and the context in which it was tested. As such, these details are provided below and not in the Results section.

Choice of theoretical framework

As described above, a variety of theoretical frameworks were encountered during the formative work. Each offered a different perspective on the ‘problem’ of adherence, and each improved our understanding of the wider (social, structural) influences on adherence behaviour and the mechanisms by which they might exert effects.

Important work by other authors included the framing by Munro et al. in a 2007 systematic review of qualitative research examining factors influencing adherence to ATT.23 Using meta-ethnography, the authors synthesised data from 44 articles and identified 8 major themes: organisation of treatment and care; interpretations of illness and wellness; the financial burden of treatment; knowledge, attitudes and beliefs about treatment; law and immigration; personal characteristics and adherence behaviour; side effects; and family, community and household support. The authors further stressed the importance of structural factors, such as poverty and gender discrimination. The authors’ illustration of interactions between the four major groups they identified (structural, social, personal and health service) is reproduced in Figure 2.

FIGURE 2 Model of factors affecting adherence, from Munro et al., 2007. ↑↓ suggest a bi-directional relationship between factors. For example, health service interventions directed at patients are likely to influence patient adherence behaviour through the filter of ‘personal factors’. Similarly, patients’ interactions with health services are likely to be influenced by their knowledge, attitudes, and beliefs about treatment, as well as their interpretations of illness and wellness. Reproduced with permission from Munro et al.23 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

In 2017, Lin and Melendez-Torres developed a framework intended to reflect the experiences of migrants.49 This framework placed ‘relationships’ at the centre of a model of experiences relating to adherence, with ‘cumulative vulnerability’ (the compounding of difficulties faced by migrants across multiple systems), ‘acculturation’ (balancing one’s identity with relationships with society) and ‘interpretations of illness’ (individual understanding of the disease and the experience) exerting important effects on those relationships (Figure 3).

FIGURE 3 Theoretical model of factors affecting migrants’ experiences of ATT, developed by critical interpretative synthesis. HCW, healthcare worker. Reproduced from Lin and Melendez-Torres 2017.49 ©2017 John Wiley & Sons Ltd. Reprinted with permission. (Figure not reproduced here; see the original article.)

From data gathered during the scoping review and interviews, we developed two alternative framings. The first used qualitative data from reviewed literature to put forward a relational view of adherence, emphasising that adherence behaviour and the factors affecting it cannot be seen in isolation. It suggests that individuals’ motivation and capacity to take ATT must be understood within the complex and dynamic interplay across multiple socioecological levels, including social, health systems, and structural contexts. Our depiction of the relational dynamics of treatment-related behaviour among individuals on ATT is presented in Appendix 1, Figure 19.

The key mechanisms of influence described in our scoping review were access to TB-related information, prior experience of health services, and the balance of power, cultural expectations, and trust between HCPs and patients. Studies describing the influence of personal factors drew attention to the characteristics that led to suboptimal adherence in given contexts, highlighting the potentially important effects of health literacy, social marginalisation, isolation and comorbidities on care-seeking pathways and treatment-related behaviour.

The influence of structural factors on the livelihoods and social positions of people with TB was also discussed widely in the literature. Factors included health and social policies, funding for TB services, and socioeconomic conditions. Socio-economic marginality and limited social capital operated as mechanisms constraining individuals’ ability to effectively navigate TB services and ATT. Studies reporting on social factors illustrated how the experience of being a ‘TB patient’ was embedded within social structures rather than being merely an artefact of the health system. Finally, studies describing treatment-related barriers to ATT adherence showed that it was not only the commonly reported pill burden, lengthy duration, or side effects of medication that were problematic, but also their disruptive effects on everyday social and working lives, livelihood strategies, and routines.

Interviews with patients, caregivers, and HCPs conducted as part of the formative work for this study confirmed that many of the potential barriers to adherence described in the literature were also present in our context. Our interviews also illustrated how easily adherence could be disrupted, as shown in the quote below from the article published in 2021.

Changes [in circumstances] could occur slowly over the course of illness, for example, a progressive increase in the amount of alcohol consumed or a gradual reduction in mental wellbeing, or could be triggered by specific events or phenomena, such as severe side effects (e.g., leading to a sudden deterioration in function or hospital admission), pain, fatigue (such that someone who was previously able to work had to stop), social or physical isolation, episodes of stigmatisation or rejection, eviction, a loss of income, or setbacks in treatment. For example, one participant experienced worsening fatigue and side effects that meant he could no longer work; unemployment then created financial pressure for his family and led to the re-emergence of a previous alcohol habit, which strained relationships with his wife and children and worsened his previous mild cognitive impairment, culminating in him stopping ATT entirely. Reduced mental wellbeing, in particular, often put a strain on existing relationships, which in turn led to worsening anxiety, isolation, and difficulties maintaining a treatment-taking routine.30

We used data from interviews with patients and caregivers to map some of the key factors identified, their interactions, and the mechanisms through which they might affect adherence. Relationships between determinants and their potential effects on an individual’s ability to take ATT were mapped based on explicit descriptions by interviewees or through inference by researchers after analysis of all interviews. Determinants were categorised as ‘distal’ when their influence on adherence behaviour was indirect or mediated through intervening factors, or ‘proximal’ when they exerted a more direct effect on adherence behaviour. Sixteen ‘proximal’ characteristics were identified that had a more direct effect on individuals’ ability and motivation to take treatment, as illustrated in Appendix 1, Figure 20. ‘Positive’ characteristics (potentially increasing motivation and ability) included an ability to create and maintain a medication routine, the existence of at least one caring or close relationship, and regular employment and/or a supportive employer. ‘Negative’ characteristics (potentially reducing motivation and ability) included a ‘chaotic’ lifestyle, fragmented care, and severe treatment side effects.

After reviewing alternative frameworks, we decided to use the PaPA to guide intervention development.36 This was in large part because the PaPA was designed specifically to assist in the development of interventions to support informed treatment decisions and optimise adherence and is recognised as such within NICE guidance.50 The PaPA is based on a ‘no-blame’ approach to non-adherence, where patients are supported to make informed decisions about using medicines which are not based on misperceptions about the treatment or illness. The PaPA has underpinned the design of interventions delivered to a range of chronic illness groups.51–53 At the time of this study, the PaPA had not been applied to the design of adherence support for ATT.

The PaPA framework (Figure 4) suggests that adherence is best understood in terms of the interaction between an individual and a specific treatment. It addresses this by acknowledging that while adherence is influenced by a range of external (e.g. health systems) and internal factors (e.g. experience of symptoms), these ultimately influence adherence through the extent to which they affect the patients’ motivation and ability to take their treatment.

FIGURE 4 Perceptions and Practicalities model applied to adherence to ATT. (Figure not reproduced here; see the original article.)

Adherence support is therefore directed at the perceptions and practicalities that influence an individual’s motivation and ability to start and continue with treatment.36 Perceptual barriers include the beliefs that an individual has about their treatment and illness.54,55 Two key barriers are beliefs about how necessary a treatment is for a given illness (necessity beliefs), versus any concerns about taking this (concern beliefs), known as the Necessity-Concerns Framework (NCF).27 Therefore, the PaPA suggests that in addition to addressing the practicalities of taking treatment, adherence support should also promote a ‘common-sense fit’ between the problem (i.e. the illness) and the solution (i.e. the treatment) through increasing the perceived necessity for treatment and reducing concerns about taking it. The PaPA specifies three basic components that should be addressed in effective adherence support: (1) necessity beliefs: explain why treatment is necessary to maintain or protect health; (2) concerns: elicit and address patient concerns about treatment; and (3) practicalities: address practical factors to make treatment as easy as possible for the patient. An additional key element of the PaPA framework is its emphasis on tailoring adherence support to meet the needs of the individual. However, the ability to fully achieve this may be limited by the resources available within a given health system.

Intervention Development Group

An IDG was created and consulted to guide development and maintain the intervention’s relevance to HCPs and patients. The IDG included patient advisors with previous experience of ATT, a multidisciplinary team of health professionals with expertise in behavioural medicine, clinical trials, epidemiology, health services research, infectious disease, medical anthropology and respiratory medicine, and additional relevant stakeholders, such as HCPs and a representative from the UK TB Charity, TB Alert.

Formative research findings and analysis were presented to the IDG throughout development. The process was iterative; findings were incorporated as they became available. Among other things, the IDG helped prioritise the determinants to be targeted by the intervention and highlighted additional factors for consideration. The full IDG met three times between March and October 2019. Subgroups then worked to develop intervention components between November 2019 and September 2020; the full group approved the final products.

Constructing the intervention

The proposed intervention package comprised multiple components that could (a) identify patients most at risk of non-adherence; (b) discern salient modifiable barriers to adherence; and (c) tailor support mechanisms to meet individual needs by matching appropriate interventions to specific barriers.

Formative research influence on intervention

The risk factors for non-adherence identified in the literature reviews and patient and carer interview data identified a broader range of demographic, clinical, psychosocial, structural and health systems determinants than those listed on the standard care risk assessment form currently used in NHS TB services (Table 2).

The published quantitative literature contained few data on interventions to support adherence to ATT in HILI settings. Two previous RCTs using medical interventions were identified, of which only one was found to improve adherence compared to usual care.56,57 This provided health education counselling and incentives to patients initiating ATT in Los Angeles, USA in 1990.

The review of qualitative literature found 46 articles reporting on intervention studies, all from low- and middle-income countries. In these contexts, more successful interventions improved relationships and strengthened communication between HCPs and patients, reduced feelings of isolation, and were developed and implemented with buy-in from administering HCPs. Less successful interventions were those which increased costs and travel for patients, and increased or did not account for issues around stigma.

Formative research also gave us ideas for potential intervention components. In interviews, patients described experiencing ‘information overload’ at treatment initiation; this was supported by clinic observations reporting the large volume of verbal information delivered at that time. Patients expressed that brief videos about TB and ATT may be helpful resources. They also found that seeing imaging of their TB disease was helpful and motivated treatment adherence by providing reassurance about treatment efficacy.

Healthcare professional interviews revealed that experienced CMs delivered a comprehensive assessment by checking for risk factors beyond the items listed on the standard form. These included patient attitudes and motivation (perceived or stated) towards treatment, their existing social support, their understanding of TB and perceived need for treatment. Clinic observations supported this, and additionally found that adherence support was assigned based on the CMs’ judgement of the patient’s social and clinical needs. This suggested that CMs utilise a ‘mind-lines’ over ‘guidelines’ approach,58 whereby the strength and validity of the risk assessment and adherence support provided is based on experience and judgement, rather than the tool itself.

Formative research analysis and operationalisation of theoretical framework

Figure 5 shows how formative research findings were used to contextualise the theoretical framework and inform intervention content and components.

FIGURE 5 Conceptual map of the intervention development process. LMIC, low- and middle-income countries. Reproduced with permission from Jones et al.31 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

We identified targets for intervention content and components by mapping each determinant to the PaPA framework to identify the modifiable, perceptual, and practical barriers resulting from each risk factor. Our interpretation of these barriers was informed by the qualitative data (e.g. patient and carer interviews describing treatment experiences) and the expertise of the research team. Importantly, the relational approach developed through the qualitative research (scoping review and interviews) highlighted the complexity of the interlinked and dynamic factors influencing patients’ experiences, and helped to operationalise the PaPA framework and ensure the contextual relevance of intervention content to the UK TB setting.

In parallel, we considered potential components to deliver intervention content. The formative research highlighted limitations in the delivery of the risk assessment and assignment of adherence support. It also identified a need for additional patient resources to help manage/avoid ‘information overload’. With these potential components in mind, we developed draft content based on the perceptual and practical barriers identified in the theoretical mapping of determinants.

Decisions made by the Intervention Development Group

The IDG agreed that standardising and expanding the risk assessment could improve the identification of risk factors across all HCPs. The IDG also agreed that repeated needs assessments were necessary to account for the fluctuating social and structural issues that patients may face during their several-month treatment for TB. The IDG discussed ideas for feasible and scalable interventions to address the specific social and structural risk factors highlighted in the formative research (e.g. homelessness, substance misuse, mental health needs). Given the varied capacity of TB services and resources available, it was suggested that interventions should be site-dependent (e.g. referral to dedicated social care teams or signposting to relevant external services when these are not available). Additionally, the IDG suggested creating an electronic risk assessment to maximise usability and continuity of care.

The IDG agreed that flexible, standardised patient resources at treatment initiation would support patient motivation and ability to take treatment. Prior work from research team members was in line with findings from our formative research, demonstrating the utility of visual representations of disease and treatment efficacy (including animations) in modifying illness and treatment beliefs and improving adherence in conditions such as human immunodeficiency virus (HIV) where stigma and lifestyle factors are relevant.59–61 The IDG agreed that brief videos could be an effective patient resource. Researchers also suggested providing a patient booklet, as a supplementary resource, to address the full range of perceptual and practical barriers identified in the formative research.

In a final meeting, the IDG agreed upon the design of a draft intervention structure and components (see Figure 5). Subgroups of the IDG, including researchers, HCPs and patient advisors, met regularly to finalise intervention content, and circulated this to the wider IDG before production. Codevelopment with HCPs ensured that components would be complementary to standard care procedures delivered by CMs. Members of the IDG also attended meetings with third party production companies to provide feedback on initial iterations of the booklet and videos and to finalise materials.

Details of intervention and theory of change

Major components

Tuberculosis needs assessment

We designed the TNA to be an enhanced, structured risk assessment that would be administered by CMs at treatment-start and each subsequent appointment. The TNA was web-based [run in the Research Electronic Data Capture (REDCap) database]62,63 and completed using a digital tablet; it identified a comprehensive list of risk factors and provided CMs with standardised solutions to address the specific needs identified. The TNA did this using locally available resources, for example, by suggesting referrals or signposting to external services such as drug and alcohol services, housing or social care services, and/or addressing perceptual barriers using brief conversations. The TNA also highlighted when additional adherence support should be offered. Additional support included elements of ECM used in standard care (e.g. DOT, VOT, dosette boxes and increased support). Additionally, reminders could also be set on the electronic MMBs used to measure adherence.64

Case managers were asked to administer the TNA at the enrolment visit. The REDCap database was programmed to recommend specific ‘linked actions’ for each risk factor identified (see Appendix 1, Figure 21). At subsequent visits, CMs were asked to readminister the TNA. To save time, the database was programmed to display a summary of findings from previous assessments. We expected that this would also allow CMs and patients to review, in a structured way, issues that may have been raised at previous consultations.

When used in the manner described above, we expected the TNA to have a range of effects, all of which, we hypothesised, would eventually increase the chances of the participant maintaining adherence to ATT. These included improved detection (by the CM/health service) of people with risk factors for non-adherence, increasing risk awareness among clinicians and patients and allowing for early provision of (and/or planning for) appropriate tailored support measures; increased continuity of care [through repeated assessments by the CM and communication of risks and support measures to the general practitioner (GP)]; improved perceptions of TB and TB treatment; increased necessity beliefs around the need for treatment; reduced concerns about illness and treatment; and reduced practical barriers to maintaining adherence.

Patient videos

We created two brief (4-minute), animated videos to increase patient’s motivation and ability to take their ATT: Helping You Understand Tuberculosis (TB) and Questions About TB Treatment. The structure of the videos was based on the PaPA and the NCF,36 and the content of the videos informed by the qualitative formative research and relational approach. Specifically, the first video addressed step 1 of the PaPA, by targeting perceptions of TB, beliefs about the necessity of ATT, and providing a common-sense rationale for taking treatment. We used a visual representation of the mechanism of action of treatment to demonstrate how adherence relates to the control of TB infection and disease. The second video addressed steps 2 and 3 of the PaPA by addressing common concerns about ATT (e.g. side effects, pill burden) and providing practical advice to make treatment as easy as possible to take (e.g. habit formation and anchoring).

The videos were shown to patients on a digital tablet during their baseline session. Patients were also provided with web links to share and re-watch the videos. Translated videos were available in Bengali, Hindi and Romanian, to match the most common languages spoken by people treated for TB at study sites.

By providing easily accessible, evidence- and theory-based information that was intended to motivate and increase participants’ ability to take ATT, we expected the animated videos to make it easier for participants to discuss their TB (and treatment and care) with their family or other close contacts. In turn, we expected that this would reduce stigma and increase the likelihood of the participant receiving support from outside the health service. The videos were intended also to help change participants’ perceptions and improve their TB self-management by increasing awareness and helping them build routines.

Interactive patient booklet

An interactive patient booklet, TB and Me, was designed to guide patients through their treatment. The core content and activities in the booklet addressed the three steps of the PaPA; included material also covered the key social and structural issues highlighted in our formative research. Information topics included: what to expect when taking ATT, emotional well-being, addressing stigma, and side effects. Contact information for the TB service and local social services was also provided (see Appendix 1, Figure 22).

The booklet also contained interactive resources to help patients understand and implement a treatment routine, such as creating a personalised medication list and a ‘medication action plan’ (detailing when and how patients will take their treatment). CMs completed these resources with patients during their initial session. Patients were also encouraged to use a ‘medication and symptom diary’ to record adherence and side effects, which CMs could review at subsequent follow-up visits (see Appendix 1, Figure 23). Translated versions of the booklet were made available in Bengali, Hindi and Romanian (see Report Supplementary Materials 4, 5, and 6).

The potential effects of the patient booklet were in large part dependent on the individual participant. CMs were required to review specific sections of the booklet at treatment initiation and during follow-up visits, including strategies for building and maintaining a medication routine. However, to get maximal benefit from this intervention component, participants would need to engage with the booklet outside of appointments and be willing to use it as a tool to help communicate with their family, close contacts, and/or CM. If used as intended, the booklet was expected to help reduce stigma, improve illness self-management, increase engagement with care, and facilitate better communication and relationship/s, any/all of which we expected to help improve adherence to ATT.

Support for case managers

A training manual provided to CMs included information on intervention development and theory, intervention components and guides on how to incorporate these into standard care (see Report Supplementary Material 3). Training was provided by two IMPACT researchers (ASKJ and PC), including an initial 90-minute presentation and question and answer session at each intervention site, and 1-hour individual sessions with each CM. CMs were also given checklists to follow intervention steps (see Appendix 1, Figure 24).

Minor components

We included structured phone calls as part of the intervention at weeks 1 and 6 of treatment. This was done to standardise the additional phone contact that some CMs reported having with patients. The purpose of the calls was to review patient progress with treatment and address any issues. Because our formative work suggested that imaging can motivate and reassure patients, CMs were also encouraged to share any available and appropriate imaging with patients (e.g. chest radiographs) at treatment initiation and completion. Finally, to tailor information provision to patient needs, at week 2, patients completed a short checklist to identify any booklet topics they would like to review with their CM.

Contextual modifiers

As described by some of the theoretical frameworks discussed above, numerous factors beyond the control of the individual and the health service may exert substantial influence on adherence. In addition, because adherence needs to be maintained for several months, it is also vulnerable to changes in circumstances, particularly if the change is drastic or unexpected.

Even without the massive disruption of the COVID pandemic, the effects of our intervention may have been modulated by the individual’s demographics: for example, a history of housing or immigration problems, incarceration or unemployment, any of which could increase stigma and social marginalisation, reducing access to care. Physiology, disease severity, comorbidities, other medications, current or previous issues with addiction, mental health history, experiences of stigma, and pre-existing strong beliefs about illness or medications are additional individual characteristics that could have affected how intervention components were received or understood, and/or limited their influence on adherence. Potential mechanisms include a higher likelihood of side effects, complex disease needing longer or more invasive investigation or treatment, and high levels of stress and anxiety, potentially leading to lower motivation to take treatment and other mental health problems. Finally, from a health service perspective, the character and severity of barriers to accessing care, the relationships between patients and their care team (particularly the CM), and the experience of ‘fragmented care’ (usually due to poor or absent communication between HCP teams) could all have affected adherence behaviour and/or reduced, negated, or distorted the effects of the intervention.

The logic model in Figure 6 provides an overview of the anticipated mechanisms of action and effects of the three major intervention components, indicating how their combined deployment through the channels described could lead to the desired outcome of improved adherence to ATT.

FIGURE 6 Logic model for the intervention package, showing channels targeted, and anticipated mechanisms and effects of each component. Note: Dotted lines indicate where anticipated effects are repeated between mechanisms. Reproduced with permission from Jones et al.31 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Figure 7 shows the common touchpoints in the treatment pathway where intervention components were to be delivered.

FIGURE 7 Timetable for delivery of intervention components over patient treatment pathway. a, Point of touch occurs outside of standard care treatment pathway. b, Month 6 is treatment completion for standard treatment. If treatment continues monthly, the TNA follow-up is completed at each session. ISQ, Information and Support Questionnaire. Reproduced with permission from Jones et al.31 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Pilot cluster-randomised controlled trial and feasibility study

The manualised intervention package described above (see Appendix 3, Figure 29) was evaluated within a pilot cluster-RCT (CRCT) and feasibility study, where it was compared to the standard care package of a needs assessment (see Appendix 3, Figure 28) plus offers of support. The evaluation included assessment of its feasibility in four domains: epidemiological, behavioural, health economic, and through a process evaluation.

Design

A CRCT design comparing the manualised intervention to the usual standard of care in four clinics treating TB in London (two each from North Central London TB Network and North East London TB Network). Within each network, one clinic was randomly allocated to the intervention and one to standard care. A CRCT design was selected because an individually randomised trial might suffer from substantial contamination. Specifically, the introduction of intervention components at a centre would likely influence how staff provide standard of care. Contamination through contact between individuals randomised to different arms would also be a concern. Since CRCT would be a natural design for any definitive trial we also felt it best to pilot this design to assess its feasibility.

Sites

The selected four TB clinics were large TB clinics in East and North London (Mile End and Newham clinics, from within Barts Health NHS Trust – North East London; Royal Free London NHS Foundation Trust and the Whittington Hospital – North Central London TB Network). Each clinic treats over 60 TB patients per year and manages a broad mix of patients, representative of the national demographic picture of contemporary TB. This includes established and newly arrived migrants, the homeless, those with mental health and physical comorbidities, people who misuse drugs or alcohol, and the immunosuppressed (through associated illness, including HIV, or medication). The clinics are also regional referral centres for the treatment of MDR TB.

Participants

Inclusion criteria

1. Age ≥ 18 years
2. Able to provide informed consent (see Report Supplementary Material 1)
3. Being started on treatment for TB disease affecting any part of the body

Exclusion criteria

1. Age < 18 years
2. Not able to provide informed consent
3. Already on treatment for TB disease
4. Not expected to live for the duration of the study (i.e. at least 6 months from starting treatment)

Target sample size

This was selected on a pragmatic basis to provide useful information about the ability to collect data and the uptake of the intervention. We did also consider power to demonstrate (at 95% significance level, two-sided testing) an increase in adherence due to the intervention (see published protocol),64 but these are only approximate and assume little clustering (intra-cluster correlation). The sample size of four clusters prohibits a formal analysis of intervention effect. The target sample size was 80 individuals, 20 per site. This sample size was intended to allow us to measure consent to enrolment for 80 individuals, data completeness for adherence and treatment outcomes for 80 individuals, data on acceptability and feasibility of the intervention package for around 40 individuals. We assumed that at least 26 individuals would be offered adherence support, providing data on its acceptability and feasibility.

Adherence monitoring

Multiple sources of data were used to monitor adherence: the Wisepill evriMED boxes; DOT, VOT, and ward records; side effect data; and finally, information on ‘pocket doses’ and intermittent box use. These are described below.

evriMED boxes measure patient’s adherence to a planned drug regimen by storing the drugs in a lidded container which, when opened (e.g. to remove the medication), records this activity. This measure is then used as a proxy for drug consumption. In this study, data on box openings were stored on the electronic recording module (inside the box). Data from the module were routinely downloaded by the research team when the participant attended their follow-up appointment (the research team reminded participants to bring their box to their appointments).

The box had to accommodate up to 1 month of medication, so its size had to reflect the maximum number of pills to be stored at any given time. In the UK, drugs need to be kept in their own packaging until consumed (the only exception to this is a dosette box which has daily, individual, pill containers that can be pre-filled by a HCP trained to do so, or individual drug blister packs made for patients by pharmacy). These requirements for medication storage meant that a custom evriMED box was designed for our study (Figure 8). The box was constructed from light but sturdy Correx polypropylene.

FIGURE 8 Medication monitoring box with lid closed and open. Note: A Styrofoam cup is placed next to the box to indicate the box size. With the lid open, the front recess used to contain the electronic recording module is visible. The action of opening the box triggers the module, though it has smart software that does not count quick open and shut activity as an actual box opening. (Figure not reproduced here; see the original article.)

Anecdotally, some participants reported not always storing the medication inside the box as intended. For example, some participants would remove medication from the evriMED boxes before they intended to take them, for example due to upcoming travel during which the evriMED boxes were left at home. In the analysis we have identified this as ‘pocket doses’ and other forms of intermittent box use.

As the boxes only record dates/times when they are opened, it is important to have an additional measure to demonstrate that the box is working on a given day. The evriMED boxes do this once a day by checking the internal circuits are working and recording the resulting ‘heartbeat’. These data are also recorded on the electronic module.

Where patients underwent DOT (i.e. in-person observation of medication consumption by a HCP) or VOT (similar, but not in real-time, i.e. asynchronous review of recorded video clip of patient taking their medication using the TB Find and Treat Team App), we collected these data.

Where a patient was hospitalised, we collected the ward record of their adherence from the ward drug chart. Information on side effects was also collected using the Side Effect Experience Questionnaire (SEEQ).

Outcomes of interest

Adherence

The primary outcome of the trial was adherence (doses taken) over 24 weeks (168 days) from the start of treatment. The data sources were cleaned in the following order: data from the evriMED boxes; then from DOT, VOT and ward records; side-effect data; and finally recorded pocket doses and intermittent box use. Dose taking was measured from the date of treatment initiation, although when side effects meant the regimen was completely restarted, we altered the treatment initiation date to match (this was usually during the first few weeks of treatment).

We initially took the box opening data from the evriMED boxes as our starting point for cleaning the adherence data (see Appendix 1, Figure 25). For days where the box was not opened, we checked for a heartbeat (to indicate that the box was working normally). If a heartbeat were not present, we checked to see if alternative data sources provided useful information. We also over-wrote the box data with the alternative data sources, where required, for example, if a patient were hospitalised in the middle of treatment, we used their ward data rather than their box data.

With the alternative data sources slotted into a patient’s timeline of treatment, each day could be coded as: dose taken, dose not taken, and data missing. This was done in two ways (see Appendix 1, Table 22), one for the main analysis and one for the sensitivity analyses. The proportion of doses taken was calculated in both cases. Both variables had their denominators adjusted for dates of withdrawal and if the patient completed treatment earlier than 168 days because the final appointment was earlier than this time point.

Two additional variables were also created for sensitivity analyses. These allowed for the exclusion of individuals when (1) it was either known that they chose not to use the evriMED boxes for part of their treatment or we knew that the patients had pocket doses but were unsure when this had happened (17 patients; ‘intermittent box use’), or (2) we were missing ward dates and were thus unable to pinpoint where particular periods of (non-)adherence occurred (two patients).

Finally, an additional variable recorded adherence up to the end of treatment for individuals on regimens longer than 6 months. This was coded as per the primary outcome adherence variable.

Adherence and social support offered and accepted at different time points

In both arms of the trial, adherence support measures were offered to patients identified as needing such support. It was recorded as (a) whether adherence support measures were offered (yes/no/unknown), (b) whether any adherence support measures were accepted (yes/no/unknown) and (c) which measures were accepted (yes/no/unknown).

These offers and the responses were made at the following time points: at baseline, 2 weeks, 1 month, and then monthly until the end of treatment. The forms of adherence support were:

• Dosette (to help organise medication and make it clear which pills are taken on which day). This was done with a clear plastic cover front which needed to be punctured to get the medication for a given day. Dosette boxes were usually filled by TB nurses.
• Blisterpack (where medication was made up specifically by pharmacy for a patient).
• DOT.
• evriMED box reminders (i.e. the inbuilt alarm in the box was set to sound daily).
• VOT.

In both arms of the trial, social support measures were also offered to patients identified as needing such support. It was recorded as (a) whether social support measures were offered (yes/no/unknown), (b) whether any social support measures were accepted (yes/no/unknown) and (c) which measures were accepted (yes/no/unknown), specifically in the two intervention sites. Social support options offered at different sites were dependent on what was available within the TB service at that site. The biggest difference between the two intervention sites was that one had a social care team and the other did not.

These offers and the responses were made at the following time points: at baseline, 2 weeks, 1 month and then monthly until the end of treatment. The forms of social support that were available at intervention sites are listed in Appendix 1.

12-month treatment outcome

Outcomes 12 months after diagnosis were collected from the London and National TB registers and coded as follows:

• Treatment completed.
• Lost to follow-up.
• Still on treatment.
• Not yet 12 months since diagnosis.

Other variables

Participants’ demographic and clinical details were gathered at baseline. Variables collected and coded are listed in Appendix 1.

Withdrawal

Three variables were available for withdrawal: whether the patient withdrew (‘Yes’, ‘No’), date of withdrawal and reason for withdrawal (medical reasons, patient choice or other). The month of withdrawal was calculated using the date of withdrawal and the date of the baseline appointment. For analyses of the adherence data and the patient flow, withdrawal data were removed from the date of withdrawal inclusive. For the other analyses (e.g. psychological assessments and health economics), data on the date of withdrawal were retained.

Late exclusions

Patients recruited into the trial who were later found not to meet the inclusion criteria were classed as late exclusions.

Analysis approach

The epidemiological analyses broadly sought to assess the performance of the intervention at improving adherence, including by triggering the deployment of adherence-promoting support measures, as follows.

Primary objective

How does treatment adherence (percentage of prescribed doses taken) differ between intervention and standard care arms of the trial, assessed at 6 months (168 doses) from the start of treatment?

Secondary objectives

• What is the percentage of patients consenting to the study from those approached?
• Are there demographic differences between patients who decline participation and those included in the study?
• What are the proportion of patients withdrawing during the study and the reasons for withdrawal?
• What is the completeness of data available for measures of adherence?
• What are the patterns of adherence (initiation, implementation and discontinuation) overall and by study arm?
• Did adherence patterns differ among the people who were identified as needing and not needing adherence support?
• What is the proportion of patients who were offered and accepted adherence support in both the intervention and control arms, at baseline and monthly follow-up appointments?
• What is the proportion of patients completing and still on treatment at 6 months, overall and by arm?

Data cleaning and analyses were undertaken using Stata 17 software (StataCorp 2021, StataCorp LP, College Station, TX, USA), with some figures constructed in Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA). A flow chart was made detailing the number of patients approached, enrolled into the study, withdrawn from the study, and late exclusions. Late exclusions were removed from the analyses. Basic demographic details of non-consenters were compared with those who consented. Reasons for non-consent were described. Baseline characteristics of those who withdrew were compared to those of the overall consenting population. The reasons for and timing of withdrawal were examined. The baseline characteristics of the individuals who consented were described and compared between the trial arms. The proportion of days for which adherence data were missing was reported.

To answer the pilot’s primary objective, the user-written ‘clan’ command in Stata was used to provide a ‘cluster summary’ analysis.65 Our primary effect estimate was a risk difference, calculated from the arithmetic means of the cluster-level adherence proportions in each arm (t-test on the cluster proportions, stratified by TB network). This was the unadjusted model. (We also used the arithmetic means of the cluster-level adherence proportions to calculate overall adherence for everyone in the pilot.) Next, the model was re-run, adjusting for age, ethnicity and gender. Sensitivity analyses were conducted on unadjusted and adjusted estimates. Three were planned: one using the alternatively coded adherence variable, one excluding intermittent box users, and one excluding patients for whom it was not possible to determine when they were admitted to a hospital ward.

Lasagna plots were constructed that documented the adherence patterns of patients in the trial during the first 6 months of treatment. These were for the control and intervention arms of the trial across the main analysis and the three sensitivity analyses. The relative distribution of discontinuation (early cessation of treatment) and suboptimal implementation (intermittent missed doses) was also examined graphically.6

To examine how well the intervention identified patients in need of adherence support, we cross-tabulated adherence support offered at baseline against the variable charting if patients had one or more classic risk factors for non-adherence. The offer of adherence support, its acceptance, and the types of support used were also graphed. This process was repeated for social support.

Treatment completion at 12 months, overall and by study arm, was tabulated.

Psychological assessments

The primary aim of this activity was to assess the impact of the intervention on beliefs about TB medications. The study also aimed to assess the change over time in representations of TB, practical barriers to adherence and side effect experiences between the intervention and control arms by conducting ‘cluster summary’ analyses.65

Participants in both arms were asked to complete questionnaires during their treatment which focused on illness perceptions, treatment perceptions and the practicalities of taking treatment. These were administered in line with the schedule in Appendix 1, Table 23.

Illness perceptions

The Brief Illness Perception Questionnaire (Brief IPQ)66 is a validated tool to assess patient’s perceptions of their illness (in this case ‘TB’). We used the eight, Likert-scale items of the Brief IPQ assess patient perceptions of the consequences, timeline (acute or chronic), identity (symptoms associated with the illness) and controllability (personal and treatment) of TB, as well as how emotionally affected they are by TB and how well they understand their illness (coherence). We added an additional item which asked patients how concerned they were about passing TB on to others, to measure perceptions of transmissibility. Items are answered on a 11-point scale with relevant anchors for each item.

Treatment perceptions

The Beliefs about Medicines Questionnaire (BMQ; © Prof Rob Horne)67 is a validated questionnaire that assesses beliefs related to medicines prescribed for a specific illness (BMQ-specific) and general beliefs about medicines as a class of treatment (BMQ-general). The BMQ-specific has two subscales, assessing beliefs related to the perceived necessity of treatment and concerns about treatment (in this case, ‘TB medicines’). The BMQ-general includes three subscales which measure perceptions of general harm from taking medicines, the overuse of medicines by practitioners, and the benefit of medicines to patients. During the pilot trial, the full measure of the BMQ-general was adapted into a brief profiler as a result of feedback regarding the impact of questionnaire burden upon data collection. The adapted BMQ-general short profiler consisted of three items, with one each assessing harm (‘Medicines do more harm than good’), overuse (‘Doctors prescribe too many medicines’) and benefit beliefs (‘Medicines are usually beneficial’), with possible responses ranging on a 5-point scale from ‘strongly disagree (1)’ to ‘strongly agree (5)’. The BMQ-specific full version included 17 items, with 12 assessing treatment concerns (‘How worried are you about taking TB medicines?’) and 5 evaluating necessity beliefs (‘How important are our TB medicines to you?’), answered on a 5-point scale, from ‘strongly disagree (1)’ to ‘strongly agree (5)’ (higher scores indicate higher necessity beliefs and concerns). Finally, a Necessity–Concerns Differentials (NCD) score was calculated by subtracting a participant’s necessity score from their concerns score to create a numerical indicator ranging from −4 to +4 that captured each person’s view on their personal need for TB treatment in relation to their concerns. Positive NCD scores would indicate higher necessity belief and lower concerns, while negative scores would be the opposite.

The Perceived Sensitivity to Medicines (PSM; © Prof Rob Horne)27 scale is a validated tool which assesses patient’s perceptions regarding their personal sensitivity to the effects of medicines. As above, the PSM was also adapted during the study to a one-item measure profiler (‘I am more sensitive to medicines than most people’) answered on the same response scale as the BMQ-general.

Perceptions of side effects were measured using an adapted version of the SEEQ (© Prof Rob Horne). This questionnaire has previously been used to measure side effect attribution in patient’s taking highly active antiretroviral therapy to treat HIV infection.68 This questionnaire asks patients to report whether they are experiencing a range of symptoms (‘yes’ or ‘no’) as attributed to their TB medicines. The 15 symptoms include non-specific symptoms which may occur as adverse effects of TB medicines (headache, fatigue, stomach upsets, nausea, vomiting, diarrhoea, chills/fever, dizziness), and specific side effects of TB medicines (joint pain, skin rashes and itching, pins and needles, bodily fluid discolouration, yellowing of the skin, visual disturbances, skin irritation or peeling of the mouth).

Treatment practicalities

Three items from the Medication Practical Barriers to Adherence Questionnaire (MPRAQ; © Prof Rob Horne)69 were used to assess practical barriers patients face in taking their treatment for TB. These items assessed how patients fit their medicines into their lives (‘I am sometimes too busy to take TB medicines’), remember to take their medicines (‘It is difficult to remember to take TB medicines’), and are able to physically take their medicines (‘TB medicines are very difficult to swallow’). Items were answered on the same response scale as the BMQ-general short profiler.

Analysis approach

Descriptive- and cluster-level analyses were implemented utilising the ‘clan’ package of Stata (Version 17; StataCorp 2021; see pilot trial epidemiological analysis approach). The cluster-level t-tests were run on patients’ beliefs about TB (BMQ-specific) to compare the over-time change from baseline to 2 weeks, 2 weeks to 3 months and 3 months to 6 months between arms. Further, representations of TB medication (BMQ-general), practical barriers to adherence (MPRAQ) and SEEQ compared the change-over-time from 2 weeks to 3 months, 3 months to 6 months.

Health economics

The overall aim of the health economics aspect of this study was to assess the feasibility of collecting appropriate data that enable us to perform a full cost-effectiveness analysis using the manualised intervention’s TNA compared to the standard needs assessment in an NHS patient population in a definitive trial. This future RCT would aim to calculate the incremental cost-effectiveness ratio as the difference in costs divided by the difference in quality-adjusted life-years (QALYs) of using TNA compared to the standard needs assessment, over the study time horizon, from a health and social care perspective.

Data collected in the feasibility study included the time spent delivering the TNA, time spent delivering the standard needs assessment, quality of life and utilisation of health and social care resources. The objectives included calculating the level of data completeness for the quality of life and resource use information collected, estimating the costs of providing the enhanced and standard needs assessment, and assessing the feasibility of calculating utility scores and resource use costs by applying standard tariffs and unit costs.

Cost of interventions

The study intervention was the delivery of a manualised package including TNA, interactive patient booklet and educational videos related to treatment for TB. Costs for the TNA were calculated by capturing how much staff time was required in the intervention arm, which was recorded in case report forms (CRFs) as the session time with a TB nurse, and applying standard unit costs. The amount of staff time required to deliver the corresponding standard needs assessment was captured similarly in the control arm. Descriptive statistics to describe the resource use and corresponding calculated costs for the TNA and standard needs assessment were reported and no formal comparisons made. CRFs were designed to collect this information at every study time point: at baseline, 2 weeks, then monthly from 1 to 6 months.

Resource use

We collected health and social care resource utilisation data using the modified Client Service Receipt Inventory (CSRI)70 and healthcare records to assess the feasibility of capturing data for generating cost information from the perspective of the NHS and Personal Social Services (PSS). This included information on visits with a TB nurse, social worker or support worker, hospital outpatient appointments and procedures, hospital admissions and visits to accident and emergency (A&E), use of primary and community care services, and DOT or VOT sessions. We additionally captured participants’ out-of-pocket costs, although this is outside the NHS + PSS perspective, as missed opportunities to control TB have cost implications not only for the NHS, but also for patients and associated members of the public. The modified CSRI also captured referrals to other social services, although information on these referrals was not designed to be costed. As the study did not follow up the outcome of external referrals, downstream actions ensuing from the referrals could not be tracked or costed. The modified CSRI was administered to participants at baseline, and at 2, 4 and 6 months, asking each time for the previous 2 months’ worth of information. Prescription data on TB-related medications were retrieved from medical records by research nurses.

Health-related quality of life

In the UK, NICE recommends that QALYs are calculated using utility scores derived from responses to the EQ-5D, a five-item, self-rated instrument covering mobility, self-care, usual activities, pain/discomfort and anxiety/depression. We collected EuroQol-5 Dimensions, five-level version (EQ-5D-5L) responses from participants at baseline, 2, 4 and 6 months, to assess the feasibility of capturing this information for generating utility scores that can be used to calculate numbers of QALYs per participant in each arm, using area under the curve methods and controlling for baseline values.71,72

Data completeness

Response rates for the patient-completed EQ-5D-5L and modified CSRI questionnaires were calculated from variables recording questionnaire completion. The number of respondents who were recorded as having answered the questionnaire at each time point was divided by the number of eligible patients at that time point to derive the response rate. Percentages for missing responses were obtained from dividing the number of missing responses in a variable by the total number of eligible patients at a certain time point.

Costs of intervention

The costs were calculated from the perspective of NHS and PSS and reported in 2020–1 Great British pounds sterling. Time spent by a Grade 7 hospital nurse (e.g. with job title ‘Nurse advanced/team manager’) delivering the intervention or control was costed using the hourly wage in Unit Costs of Health and Social Care published by the Personal Social Services Research Unit.73 Descriptive statistics were reported for the duration and costs associated with needs assessment delivered during a TB nurse session in TNA recipients and by randomisation arm. The amount of missing data was also reported.

Resource use

Information on services used was costed using nationally published sources including Unit Costs of Health and Social Care 2021,73 National Schedule of NHS Costs 2019–20, (NHS, 2021) and NHS Agenda for change-pay rates (NHS, 2021). Costs from 2020–21 were used, where available. Where this was not available, costs from 2019–20 were uprated to 2020–21 values using the NHS cost inflation index.73 For services where only costs per episode/visit were available, these were costed as frequency of service use multiplied by the cost per visit. If an hourly rate was available and the amount of time for service use was captured in the study, costs were calculated as length of appointment multiplied by the hourly rate.

Medication costs

Information captured on TB-related medications was costed using the online British National Formulary.74 Where multiple prices were available for different manufacturers and pack sizes, we used the lowest price and smallest pack size that was relevant to the prescription regimen. The number of packs of medication prescribed to a patient at each time point (baseline, 2 weeks, then monthly from 1 to 5 months) was calculated via the following formula:

number of units per dose × frequency × number of days in prescription.

Pack sizes were rounded up and then multiplied by the price per pack to derive the cost for each individual medication in the treatment regimen. Then, we summed up the costs of individual medications to arrive at the total TB-related medication costs for each patient. We only included costs for medications prescribed between baseline and 5-month study visits. Medications prescribed at the 6-month study visit and thereafter would be taken after the 6-month study time frame, which is outside the scope of our costing. Descriptive statistics were reported for resource use and other cost-related items at each time point, both overall and by randomisation arm, including the mean and standard deviation (SD).

Quality of life

Responses collected from EQ-5D-5L at each time point were converted into utility scores using the Value Set for England.75,76 Descriptive statistics were reported for utility scores and EQ-VAS responses at each time point, both overall and by randomisation arm, including the mean and SD. EQ-visual analogue scale (VAS) responses are not typically reported for cost–utility analyses but these are often used in the TB treatment literature as a quality-of-life measure.77

Analysis approach

Analyses were descriptive. Data captured from EQ-5D-5L, modified CSRI, and records were cleaned and analysed in Stata 17 (StataCorp, 2021) with some figures constructed in Microsoft Excel.

Process evaluation

The aim of the process evaluation was to understand how the IMPACT intervention may affect patient adherence to treatment. Specific objectives were to:

• explore the implementation, feasibility, fidelity and adaptations, dose and reach of the intervention
• explore the contextual factors which may influence the implementation of the intervention
• explore the potential mechanisms of impact through which the intervention may influence adherence to treatment for TB
• understand how the intervention might be refined in further development work.

Design

Data were collected from a variety of sources using a range of methods, including:

• Patient self-report questionnaires
• CRFs
• Data from intervention components
• CM checklists for intervention sessions
• Interviews with patients
• Interviews with CMs
• Clinic observations.

Recruitment

Patients in the pilot study were asked if they would like to participate in an additional evaluation study, to have a researcher observe one of their CM- (or other nurse-) led appointments and/or to be interviewed about their treatment for TB. Any English-speaking participants were eligible. Patients were approached to consent to the evaluation study either at recruitment or retrospectively during their participation, by a study research nurse as per the pilot study recruitment procedures.

Case managers at all study sites were asked to consent to the evaluation study if one of their patients was enrolled in the evaluation study and had an eligible appointment to be observed (see Report Supplementary Material 2). Additionally, all CMs at intervention sites were also approached to consent to an interview about their experience of the intervention.

Measures

Patient self-report questionnaires

Data from self-report questionnaires completed by participants in the main pilot trial were used for the process evaluation analysis. The questionnaires are described above (see Appendix 1, Table 23 for the questionnaire schedule).

Intervention feedback. Intervention arm patients completed an intervention feedback questionnaire at 6 months, asking about how helpful they found different components of the intervention (from ‘0 – not at all helpful’ to ‘10 – extremely helpful’), including the TNA, booklet, videos, and phone calls. Patients were also asked how helpful they found the interactive resources of the booklet (medication diary and medication action plan), and whether they showed their booklet to others. Patients were further asked whether they watched the videos in their own time (and if so, how many times), and whether these were shared with others. Patients could provide any additional comments about the intervention components in an open-text box. Patients were asked general questions about whether the intervention changed the way they thought about their treatment, made it easier to take their treatment, and whether they would recommend the intervention materials to other patients starting treatment. The response options for these items were ‘Yes’, ‘Somewhat’, and ‘No’, with a follow-up open-text box asking patients to explain their answer in a few words. Open-text boxes also asked patients to state what aspect of the intervention materials they found most helpful in supporting their treatment, and whether they had any additional comments to make about the intervention overall.

Nurse checklists for intervention sessions

Case managers were asked to complete a checklist at each appointment to measure fidelity by assessing whether each of the intervention steps was completed. The checklist also contained open-text boxes to provide information on any alternations made or other comments. CMs were also asked to record the approximate duration of the consultation.

Case report form

A CRF was completed by the study research nurse to collect clinical and trial data on all participants in the pilot study. Data captured on the CRF for patients in both study arms relevant to the process evaluation analysis included appointment length, psychosocial risk factors identified by the standard care needs assessment/TNA, adherence support assigned, and any referrals/support provided for social risk factors. Finally, we recorded whether the TNA was completed at each relevant appointment.

Data from intervention components

Data collected through use of some of the intervention components were also analysed. This included the summary of the TNA items selected and the use of the information and support questionnaire.

Observations

Observations of CM-led appointments were conducted at four study sites with patients who consented to the evaluation study between March and September 2021. A study researcher (ASKJ) observed any face-to-face baseline or follow-up appointments along the standard treatment pathway. Data were collected using structured forms. Observations at intervention sites focused on the delivery of the intervention components, including any deviations to the intervention. Observations during standard care delivery captured information on standard care practices, to allow comparisons between arms.

Interviews

Structured interviews were conducted from March to September 2021 with patients and CMs who consented to the evaluation study.

Patients from all four study sites were purposively interviewed at two stages of treatment (early interviews ~1 month from treatment initiation, late interviews > 3 months into treatment), to understand effects and use of the intervention across the treatment timeline, and to compare treatment experiences between the study arms. Responses were manually recorded onto a structured form during the interview by the interviewer. Interviews lasted approximately 45 minutes and were all conducted via telephone except for one face-to-face interview. Telephone became the preferred method as data collection took place during COVID restrictions. All interviews were conducted in English by a member of the study team (MD and ASKJ).

Case managers at intervention sites were also interviewed to provide information on the implementation of the intervention into standard care practices and potential mechanisms of action. Responses were also recorded onto a structured form as described above. Interviews lasted approximately 45 minutes. All interviews were conducted by telephone except for one face-to-face interview (per CM preference). An overview of activities and their timing is provided in Appendix 1, Table 24.

Analysis

Quantitative data were analysed using the Statistical Package for the Social Sciences (SPSS; SPSS Inc., Chicago, IL, USA) version 27 and Stata version 17 (StataCorp 2021).

Data from the intervention feedback questionnaire were analysed using descriptive statistics and frequencies. Data from open-text boxes were coded and categorised, with frequencies reported where applicable. The nurse checklist was analysed using descriptive statistics and frequencies. Any open-text boxes were coded and categorised, with frequencies reported where applicable. Descriptive statistics and a cluster-level t-test were used to analyse differences in appointment length (baseline and average follow-up appointment) between groups. The frequency of TNAs conducted at eligible appointments was reported and compared using a chi-squared test for independence between intervention sites. The frequency of items selected on TNA summaries and the information and support questionnaire was also reported.

Frequencies of closed-ended responses on the interviews and observations were reported. Data from open-text boxes on the structured form were analysed to identify codes and themes across the observations, and patient and CM interviews, using the steps provided in the thematic analysis approach.78 Coding was completed using the software package NVivo (QSR International Pty Ltd., 2020).

Codes were developed both through inductive and deductive analysis, informed by the topics of the interview and observation guides (e.g. intervention components) and by iterative analysis of the data. Two members of the research team (ASKJ and MD) coded 50% of the data each. To reduce bias, the coders also overlapped and double-coded 10% of the files. The coders met to discuss initial approaches and codes identified after coding the first four files each. After consensus was reached on an approach, the remaining files were coded. The coders met again to discuss each code and highlight those which needed refinement or merging across the data sets to reach consensus. This led to the first iteration of the coding framework. At this stage, one coder (ASKJ) created conceptual maps of the coding framework and presented these to the wider process evaluation analysis team (MCIL, RH, MD, KK), who agreed upon the emerging framework and analysis approach. The two coders reviewed the codes of all files again, to ensure fit against the coding framework. The conceptual maps of the coding framework were then reviewed to identify themes based on the study logic model (see Figure 6), and the process evaluation key functions (context, implementation, mechanisms of impact) described in the 2014 UK MRC guidance on process evaluations of complex interventions.79