Discussion

Material throughout this chapter has 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. The text below includes minor additions and formatting changes to the original text.

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 summary

Curative treatment for TB takes months to years. The drug regimens used can cause adverse effects, and the stigma attached to the disease can make it hard for people to complete treatment as planned. Ensuring good adherence to medication is seen as an important contributor to a successful outcome. We designed a manualised intervention for use in a UK NHS setting that might improve drug adherence by identifying and then offering management solutions to important modifiable factors likely to influence treatment adherence.

The intervention was developed using the PaPA conceptual framework to integrate evidence from the published literature, interviews with people with TB, their carers/close family, and HCPs from centres across the UK. The approach was supported by an IDG, consisting of people with lived experience of TB, HCPs and researchers. This was then trialled in a cluster-randomised pilot and feasibility study in four London TB services, with two located in North-East and two in North-Central London TB Network. We also performed a process evaluation, assessed intervention cost, and measured the behavioural impact of the intervention compared to standard care for people being treated for TB disease, as well as the feasibility of collecting health resource and cost information.

Regarding the intervention, we found that:

• our TNA was effective at detecting people at higher risk of non-adherence, suggesting that it may be useful for improving the current risk assessment tool
• dosette boxes (as adherence support), and transport vouchers and referrals for housing and financial assistance (as social support) were the measures most often taken up by patients
• the MMB was unpopular with some participants, in part because of its large size
• although the intervention package was generally acceptable to staff, it often made consultations longer, which meant that fidelity to the manual was not always maintained
• the intervention produced short-term improvements in participants’ perceived need for ATT treatment, but these were not sustained over time; and
• we did not substantially improve adherence in the intervention group – this was possibly because adherence levels were already very high (~85%) and because our pilot trial was not powered to detect a difference.

Regarding our broader feasibility aims, we found that:

• It was feasible to
  • design, build and test a multicomponent intervention in this way
  • deliver and evaluate this intervention in an NHS setting, despite the disruption of COVID
  • collect data on costs and data for a future cost-effectiveness analysis (use of healthcare resources and quality of life) in this context.
• It was difficult to
  • recruit people who were at the highest risk of non-adherence
  • measure adherence accurately.

These findings are discussed in more detail below.

Findings and achievements

Findings

Our pilot study found that the TNA intervention appeared better at the detection of patients in need of adherence support. In comparison to a measure of the need for adherence support based on a series of ‘classical risk factors’, the TNA detected substantially more people: 6 of 6 (100%) individuals with a classical risk factor in the intervention arm were identified as requiring adherence, plus a further 15 without these factors. By comparison, the control arm only identified 1/5 (20%) of patients who had a classical risk factor for non-adherence, as well as 3 additional people without these factors.

The intervention also resulted in more people being offered adherence support. For example, at baseline, 21/36 (58.3%) and 4/43 (9.3%) of people were offered specific support measures in the intervention and standard care arms, respectively. Similar results were observed for social support, with 10/36 (27.8%) intervention and 3/43 (7.0%) standard care participants offered support.

While most participants identified as requiring support were picked up at baseline, the repeated use of either TNA or standard care assessments during treatment did detect people with new needs and risks for non-adherence. This was again more often in the intervention group, where for example the TNA recommended social support in 10/36 at baseline and then 19 times after that (individual participants could have more than one indication for support) while standard care identified 3/43 at baseline and a further three after this.

We observed greater overall acceptance of adherence and social support measures in the intervention arm. For example, at baseline, in the intervention and control groups, 11/21 (52.4%) and 2/4 (50%), respectively, accepted adherence support. Thus, while the absolute number in the intervention arm agreeing to adherence support was higher, the actual proportion of those accepting was similar – suggesting that more needs to be done ‘downstream’ if the increased risk detection offered by the TNA is to be translated into greater patient acceptance. In contrast, measures of social support were accepted by 7/10 (70%) in the intervention arm at baseline and 0/3 (0%) in the control. It should be noted that the study was not designed to determine whether those who accepted support then used it, or the benefit/s that did or did not result from this. The latter includes not just those relating to TB outcomes but also from the wider individual health and societal perspectives.

Dosette boxes were most often used as an adherence support measure. We could not determine if this was due to patient or TB service choice. Given that much of the trial was conducted during the COVID pandemic, it is possible that anxiety about physically interacting with others led to the greater acceptance of dosette boxes over DOT or VOT. Also, the MMB may have been regarded by participants as a form of support, meaning additional methods were not required.

The offer and uptake of measures of social support were harder to ascertain, as resource availability and practice varied between the two sites in the intervention arm and only one site routinely documented any acceptance to social support measures. A range of support was offered across the different months of treatment. In future studies it would be important to collect more information about how this was done and what then happened to those patients.

Finally, we did not see evidence that adherence was improved in the intervention arm of the study. This may be a ceiling effect as adherence was high in all study arms, with the percentage of 168 doses (i.e. 24 weeks of daily medication) taken exceeding 80% for both intervention and control.

Implementation, fidelity, adaptations, dose

The IMPACT intervention seemed implementable through group and individual training sessions. In practice, CMs often made adaptations to our suggested approach. Most frequently, this meant using their clinical judgement over the manualised procedures. As discussed previously, it was identified during formative work that some CMs effectively used a ‘mind-lines’ over guidelines approach to care,58 where assessing and addressing risk for non-adherence is based primarily on clinical experience and judgement. Indeed, the process evaluation revealed that less experienced CMs generally perceived the study intervention more positively – and in particular the TNA.

Observations suggested that the basic components of the intervention were delivered to all patients. Nurse checklists confirmed this at treatment baseline, though this decreased over time. Similarly, while all patients completed TNA at baseline, the rates were lower during follow-up. CMs also reported adapting the way they delivered the TNA and booklet.

Reach and acceptability

Of patients who were approached and eligible, 63% consented to participate in the study. Study demographics seemed to represent the general TB population in the UK, though identified social risk factors were lower in the sample compared to the general TB population. Mental health concerns were the most frequent social risk factor – highlighting the importance of targeting patient’s mental health concerns throughout their treatment for TB. Our study population generally had fewer ‘risk factors for non-adherence’ than might have been expected, including DR TB, even though this was the population we had in mind when designing the study. This may have been because these individuals did not wish to take part in a study given the time commitment, or they did not consider themselves to need additional support, or, at the time of the pilot, the London TB service saw fewer patients with DR TB than would usually be expected.

The intervention was highly acceptable to patients. All intervention elements were found to be helpful, including videos, medication diary, nurse phone calls, TNA, and the patient booklet. The mean ranking for each element was more than 70 on a scale of 100 being most helpful – indicating that they were considered very helpful. Ninety per cent of patients interviewed said the intervention materials changed the way they thought about their TB and its treatment, 79% indicated that the intervention made it easier for them to take their medication, and 84% would recommend the intervention to other patients starting treatment.

Psychological assessments

This pilot study aimed to assess the acceptability and feasibility of the IMPACT intervention. The estimates of effect for process variables, adherence-related perceptions and practicalities were collected to ascertain proof of concept if deemed feasible. It was not powered to assess the impact of the intervention on patients’ beliefs. However, there was an interesting signal where the intervention group saw greater improvements in the perceived need for treatment, compared to the control group. Changes in concerns about treatment, practical barriers to taking treatment, and the experience of side effects were no different across arms. This may be related to the small sample size as well as that both arms had high adherence rates throughout and felt positively about their treatment. The baseline data show that most patients had very positive attitudes towards their treatment for TB: 90% were convinced that they needed the treatment and only 24% reported high concerns, suggesting they were well aligned with their treatment. Nonetheless, some improvements in beliefs were seen, although these were short term, suggesting that an intervention ‘top up’ may be beneficial part-way through treatment. Pragmatically this might be during the transition from intensive to continuation phase (i.e. at 8 weeks), and/or 8 weeks into the continuation phase, as there was a drop-off in adherence and beliefs in the need for treatment around then. Of interest, we noted also a decrease in the completion of some TNA elements over time; it may be important, therefore, to provide further support to HCPs and CMs to ensure that the TNA is used as intended throughout treatment. This has been reported in other studies focusing on adherence and may be particularly important when implementing new approaches to treatment support and optimising their delivery throughout care.11,82

Health economics

Our objectives were to assess the feasibility of capturing patient-reported HRQoL information from participants and relevant information on health and social care resource use (via a participant-completed modified CSRI plus TB service patient and other records). A further objective was to use some of this information to calculate the costs of delivering the TNA and standard needs assessment. Costs of needs assessment delivery using enhanced or standard methods varied substantially across time points due to variation in the amount of time TB nurses spent. Across time points, mean time taken to perform the needs assessment was 15–65 minutes among TNA recipients in the intervention arm and 10–47 minutes in the control arm. Mean costs for needs assessment were £13–£55 among TNA recipients in the intervention arm and £9–£40 in the control arm across time points. For patient-completed questionnaires, we achieved response rates over 70% on EQ-5D-5L and the modified CSRI at any time point. However, variables from records and CRFs contained more missing values.

In the systematic literature review conducted by Laurence et al.,83 provider-incurred and patient-incurred costs for TB treatment were assessed. Across 59 studies reviewed, hospital admissions constituted 74% of all provider-incurred costs for treating drug-susceptible TB and 63% in high-income countries. Our findings from the resource use information captured in this study were consistent as inpatient admissions and visits to A&E were the largest cost drivers in the whole group at baseline (71%), 2 months (33%), and 4 months (57%). In two US studies reviewed by Laurence et al.,83 medication costs for treating DS TB were US$311 and US$654. Costs for TB-related medications from our trial were comparable; with mean costs being £59–£376 across time points.

Process evaluation

Of the proposed mechanisms of action in the study logic model, our process evaluation suggested that the main intervention components increased motivation and ability for patients to take treatment. Specifically, the video was more likely to change patients’ perceptions of TB, whereas the booklet appeared to improve the practicalities for patients of taking treatment.

Observations revealed that the intervention components did influence the delivery of care provided by some CMs, such as the way that adherence or side effects were assessed and addressed. Several CMs also felt that the TNA was better able to detect risk factors for non-adherence, as asking repeatedly made patients more forthcoming with information. Two CMs felt that the TNA made them more likely to offer adherence support than they would do with usual care. In particular, it appeared that less experienced CMs found that the TNA helped formulate their patient’s care and plan. One possible explanation would be that the TNA enabled these CMs to detect and address the wider range of ‘informal risk factors’ that experienced CMs may additionally assess within their usual care, suggesting that the intervention may have achieved its aim of standardising these elements of patient care across the TB service.

The two CMs with the least experience in delivering the intervention felt that it limited their ability to establish rapport with patients. It may be that the perceived barriers imposed by the intervention subside with more understanding and experience of delivering the intervention and incorporating it into usual care procedures. Again, this highlights the importance of adequate training and providing ongoing support for staff and patients.

While there was some evidence that the intervention may have enhanced patients’ engagement with their care (e.g. by empowering them to adopt a more questioning attitude), this was balanced by a lack of evidence to suggest that the intervention improved the CM–patient relationship or continuity of care. It may be that these mechanisms are less likely to be addressed in a patient-focused intervention and could perhaps be targeted more effectively by health systems-level approaches.

Achievements

Our study gathered, synthesised, analysed, and published a large amount of information relevant to understanding and improving the experiences of people living with TB. We also designed, built and tested a complex intervention based on a theoretical behavioural psychology framework in an environment that, due to the global COVID pandemic, was much more turbulent than expected, and where all ‘non-essential’ NHS services were de-prioritised. We worked with NHS TB services that had many members of staff redeployed to direct COVID activity, yet we managed to recruit to the study and kept the NHS TB teams engaged with the project.

As discussed above, the study was not intended to elicit the effectiveness of the intervention at changing adherence. The data gathered, however, have been and will continue to be useful in improving our understanding of ATT non/adherence in the UK. Furthermore, although adherence was high among study participants, the intervention indicated the potential to influence beliefs about the necessity of TB treatment. We were also able to estimate the sensitivity of the TNA and gained insights into important actors, processes and relationships required for its implementation, as well as how we might maximise its benefit if taken forward (e.g. by building in regular support and training for both patients and staff during an anti-TB treatment course). Data from this study can also be used to improve understanding of how MMBs and other DATs might be used in the UK.

Contribution to existing knowledge and comparison with other studies

This is the first study of its kind in an NHS setting, and one of the first in the global north, high-income, low TB incidence area. Systematic reviews of work in other healthcare settings have reported inconsistent results of DATs or intervention packages based around DATs improving adherence. What in part distinguishes this study from most others is that our focus was on developing an intervention that addressed the underlying psychosocial factors promoting non-adherence, and we used the digital technology to measure adherence rather than specifically encourage it. We believe that this is important, as using DATs alone (even in programmes where non-adherence triggers an increasing level of provider input)84 does not necessarily have the effect that might be expected from the amount of service and HCP input required to provide differentiated support. Nonetheless, the studies that have suggested DATs could improve treatment outcomes used a DAT together with a person-centred strategy, depending on how adherent with therapy participants appeared to be when receiving a baseline standard of care.85,86

A recent mixed-methods systematic review of psychosocial support interventions to improve treatment outcomes for people with TB (by Maynard et al.)87 identified 23 studies, the majority of which were in low- or middle-income settings with a high TB incidence (none from the UK, Western Europe, or the USA). Ensuring access to care was regarded as key, while financial support was also seen to be important. The authors concluded that for psychosocial support to be successful it needed to act as the basis for other levels of support relevant to the individual rather than be standalone. This was something we sought to do with the various components of our intervention. Recent studies that adopt a similar approach to us are yet to report, though again are exclusively in high TB incidence settings.88

Strengths and weaknesses

The pilot study was inevitably limited by its pragmatic design. Attempting to conduct a cluster-randomised trial within an active clinical service made us vulnerable to changes in the operating environment, and those resulting from the pandemic were of a magnitude we could not have anticipated when planning the study. This included both the impact on wider society (e.g. restrictions on movement and social gatherings, pressure on health services), changes at clinics (e.g. staffing, competing priorities), and changes in individuals’ mental state and behaviour (e.g. psychological distress, risk perception, health-seeking). These events also made it challenging to carry out study activities, not least because the pandemic also affected members of the study team. Of interest, the purpose-built MMB was popular during the COVID lockdown and much less so once social restrictions were lifted. This may be due in part to the size of the box itself (roughly similar to a shoebox), as in the UK there is a requirement for medication that is not packaged specifically for a study must be given to patients in the original medication packaging. We wanted the study to be as close to real life care as possible (and hence used standard medication) and did not have the resources to obtain pre-packaged trial medication. Patients were therefore using a container that stored all their anti-TB medication but was neither small nor discrete, and hence more likely to be noticed by others once travel and meeting-up were permitted. This, as so much in TB, speaks to the stigma and shame that remains associated with the condition, and has been reported elsewhere in studies using medication monitors.89

Methodological and other strengths and weaknesses of the various study components are discussed further below.

Formative work

Our scoping review of the quantitative literature was limited by: (1) the considerable heterogeneity in measurements and definitions of non-adherence across studies, which hindered comparability and the conclusions that could be drawn; (2) the omission, from most reports, of information around context required to understand mixed findings for determinants; (3) the inclusion of only English language studies, which may have biased our results; (4) given that we used scoping review methodology, the absence of a quality assessment for included studies; and (5) the inclusion of all data on reported determinants, whether measured as primary exposures of interest or as potential confounding factors – some of these may have been subject to bias, in which case our ability to understand the relationships between ‘determinants’ (exposures) and non/adherence (outcomes) would have been impaired.

Our scoping review of the qualitative literature started from the premise that ‘adherence’ is primarily a clinically driven outcome that should be seen within the broader frame of treatment experience; hence, we were unable to advocate for a precise, standardised way of assessing and measuring ‘adherence’. However, this approach allowed us to show that individuals’ treatment-related behaviour occurs within a complex interplay of factors and processes, functions at multiple socio-ecological levels, and is shaped by individuals’ evolving social and structural circumstances. No quality assessment was performed, and quality control was limited to the inclusion of peer-reviewed publications. Additionally, only papers written in English were included, which may have led to selection bias.

Our interviews were conducted in healthcare facilities by a medical doctor, which may have influenced the information volunteered by participants, though this was not explicitly discussed with participants, the interviewer was not involved in routine care provision, and one other non-clinician was present for most interviews. Despite reassurances of anonymity, participants who were still receiving care may have moderated their criticisms of the health service out of fear of repercussions. Individuals with DR-TB are likely under-represented in our data, though many of the issues we discuss are relevant to them, given the longer and more arduous treatment regimens. Interviews were conducted in English, and some migrant experiences will not have been captured; however, many of our participants had at least limited English and several described challenges around language and communication. Qualitative data were not linked to empirical estimates of adherence in any of the patients interviewed and the designation of factors as ‘key determinants’ as well as the relationships between factors depended on self-report by participants and interpretation by authors. A strength of our approach was that participants were recruited from four NHS Trusts in three cities, reflecting some of the variation in care models and resources available in different parts of the country. In addition, participants were not recruited from any particular risk group, and their common experiences may be closer to those of the ‘average’ person treated for TB in the UK.30

Intervention development process

Through coproduction with a multidisciplinary IDG, our development process ensured that intervention components would be compatible with UK TB services. From the patient perspective, this aligned the accessibility and relevance of intervention content to patient needs. For HCPs, this ensured the components were deliverable by CMs and applicable to current care practices. Furthermore, although the IMPACT intervention was piloted in London, the intervention was designed to be relevant to other UK services and HILI settings. The IDG included researchers and patient representatives from across the UK. The formative research reviewed literature across similar HILI settings and captured perspectives across three UK sites (Southampton, London, and Edinburgh), with differing levels of TB incidence and models of care. Including stakeholders throughout the design process is likely to help to reduce implementation barriers when incorporating the intervention into standard care, should that be recommended.

A further strength of our development process was the use of varied research methods to obtain evidence to inform intervention content. In practice, this meant that the various pieces of formative work were conducted in parallel, such that findings were incorporated into intervention design as they became available, rather than the results of one piece of research sequentially informing the design of the next. Using varied research methods also enhanced the development process: for example, clinic observations enabled a more rapid understanding of the context of current TB care, which provided useful initial insights into the relevance of emerging intervention components.31

Pilot trial

We did not collect data on whether the additional adherence and social support measures offered and accepted were actually used, and therefore cannot say whether the intervention, which detected more health and social care needs, actually led to better patient outcomes. In addition, we do not have data on whether the people identified as needing adherence support were the ones who most needed it. Given that we had few participants with MDR-TB or who had particularly complex social, psychological, or medical backgrounds, it seems likely that the population for whom the intervention was designed were relatively small in the pilot study. Why this should have been may in part reflect the local TB populations at the time of the study (i.e. during COVID) who were actively seeking health care, as well as the reality that those we really wanted to target are not only less likely to come to clinic but also perhaps more likely to decline to take part in a study that will take more time than a standard consultation, and could suggest to them that they are being closely monitored when using the MMB.

Although the MMB provided us with relatively granular data on individual adherence patterns, it also created some potential problems. First, providing boxes to participants in both arms of the pilot trial may have artificially boosted adherence overall, raising the baseline and making it more difficult to detect a meaningful difference between arms. Second, as indicated earlier, their large size may have discouraged some patients, and/or worsened adherence by increasing stigma. Finally, the ‘counting mechanism’ on the box used box opening to indicate medication being taken by the patient as directed. We have no guarantee that this final step happened, beyond assessing participant’s longer-term response to anti-TB treatment (to date, we are unaware of any trial participants returning to their service with relapse).

Psychological assessments

The tools used to assess beliefs about TB illness, its treatment, and adherence were reliable, as demonstrated by their psychometric evaluations (e.g. BMQ).67 However, the length of the questionnaires was a reported concern for the first five participants, leading to protocol amendments for the use of shorter versions. This change may have reduced the depth of information gathered and led to the exclusion of baseline assessments from our analyses, as the modifications were made during the trial. As some participants completed the full version while others completed the shorter version at baseline, it was not appropriate to compare the results across different questionnaire lengths for that time point. Additionally, the MARS adherence questionnaire was difficult for some participants to understand due to their limited English, which led to its exclusion from our analyses. Therefore, future work should focus on reducing questionnaire burden and providing those used in a greater number of languages particularly for individuals from minority ethnic groups, to ensure their engagement (which was something we did for the TNA booklet which we translated into Hindi and Romanian as well as English).

Health economics

We showed that EQ-5D-5L data can be collected from this patient group with good response rates. We also showed that our modified CSRI was acceptable to participants, and that it yielded useable and relevant information for a future full trial with full cost-effectiveness analysis. There were limitations in the data collected from site records as these were not as complete as the participant-completed questionnaires. In future work, therefore, we would need to consider how we ensure that sufficient useable information can be collected. There were also limitations to our calculation of the intervention cost in each arm. First, our calculations for needs assessment delivery were limited by data availability on session length. The findings could have over- or underestimated the true costs of the intervention as the variable for session length contained considerable missing values. Anecdotal evidence from study team members suggested that it was challenging to record the session length in a busy clinical setting. Our calculations for TB-related medication costs were also limited by the availability of record-based data.

Process evaluation

Our sample size was small and our interviews were restricted to participants who could speak English, limiting the generalisability of our findings. Restrictions on in-person interviews during COVID meant that a video link was required for some interviews. As a result, we had to reconsider our interview strategy and accept that we would obtain fewer data than planned.

Take-home messages

1. We developed and piloted an intervention to detect and manage potential non-adherence among NHS patients treated for TB.
2. Intervention components included a needs assessment, treatment monitoring, signposting of appropriate resources, an interactive booklet and treatment diary, and educational videos discussing perceptions about TB and the practicalities of taking treatment.
3. In the pilot trial, the needs assessment was sensitive in detecting factors that may reduce adherence; however, a similar proportion of people in both study arms declined offers of support to mitigate these.
4. Adherence, recorded by a MMB, was high in both study arms.
5. The intervention was acceptable to patients and staff, though took longer than standard care to administer at each visit. If considered for routine use, in its current form it will require ongoing staff support and training during a patient’s treatment.
6. We used a large number of assessments to evaluate the intervention’s impact, and this may have altered its effects.
7. Our pilot study sample did not include many people who were most likely to non-adhere.
8. This is the first study of its kind in an NHS setting: it shows that it is possible to work closely with NHS services during a significant external event (COVID) and collect good-quality clinical and health-economic data.

Remaining key uncertainties

The pilot study was limited to answering its primary and secondary objectives. The remaining key uncertainties are summarised below as questions:

• How can we include the people we believe might benefit most from the intervention? Would this require consent to be obtained at a TB service-level rather than from the people with TB attending that service?
• Did we deploy/get agreement to deploy the wrong social/adherence support?
• Do we need to use all components of our TNA, or could we remove some and still retain benefit?
• If the TNA is helpful, what does this look like? Is it higher chance of cure? Or less risk of post-TB complications? Or might the benefits even be outside of TB, and reflect that by receiving support for identified social care issues the patient is able to maintain longer term health and well-being?
• Although the sites look balanced in terms of demographic criteria, were they, in fact, not balanced for the key things that lead to non-adherence? Or was there something else about the way the different sites performed?
• Why was the acceptance of support measures so low?
• Did we improve people’s experience of care, rather than their adherence – i.e. was our mechanism of action as intended?
• Did the pandemic boost adherence?
• How might we redesign the intervention to optimise its functionality within a clinical TB service?

Challenges, lessons, learning and implications for planning and decision-making

Pilot trial

The cluster-randomised pilot trial included in this study was conducted from October 2020 to September 2022 and was therefore heavily impacted by the global COVID pandemic and the resulting disruption to health services. A full exploration of the effects of the pandemic on TB programmes, the wider health service, health-related behaviour, and population health is beyond the scope of this report. Instead, we refer the reader to the published accounts by other groups.90 Important insights from the literature include the disproportionate burden of disease, mortality and economic hardship seen among deprived communities and ethnic minorities; this was due in part to the exacerbation of pre-existing inequalities which, in some areas, was made worse by lockdowns and other government measures.91 Studies have also reported effects on risk perception and care seeking,92 health and illness beliefs,93 community health services,94 HCP mental health,95 and implications for global TB control.96,97 Specific major events that are likely to have influenced both the running of the study and the interpretation of our findings include national lockdowns (March–May 2020; November 2020; January–March 2021; Figure 18), policies around mask-wearing and occupancy of shared spaces, and task-shifting and other reorganisation of health services to meet the sudden increase in demand.

FIGURE 18 Timeline (2018–23) showing selected study events and UK government COVID measures. Adapted with permission from Walker et al.98 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.)

Of the people approached, 63.1% consented to participate, though there was a somewhat skewed recruitment population consenting compared to those who did not. 18.8% (15/80) people who enrolled later withdrew, half for personal reasons and slightly over a quarter for medical reasons. This feels like a large proportion, though anecdotes suggest that many non-medical withdrawals were because of the number of assessments performed as part of the study, rather than the intervention itself. Future work can learn from this and choose more judiciously what questionnaires should and should not be used.

The MMB did create problems for some patients. Other adherence measures such as pill counts or urine testing, for example, for isoniazid, were deemed impractical in this setting. Future work will need to consider carefully how we measure adherence more discreetly without losing accuracy.

Health-related quality-of-life utility scores in context

Overall, it was feasible to calculate utility scores using EQ-5D-5L responses that were collected in the trial, and they broadly agreed with values for similar populations found in the literature. For example, a recent systematic review by Park et al.77 included two studies in their meta-analysis that had used the EQ-5D-3L (UK value set), which is the precursor to the 5L questionnaire in our study (and is thought to suffer more from ceiling effects due to having only three response levels per question rather than five). They found utility scores for treated TB of 0.88 and 0.91, and for active TB of 0.43 and 0.70. In addition to utility scores, Park et al. meta-analysed four studies that collected EQ-VAS scores and reported a mean increase of 21.60 after ATT. The mean (SD) EQ-VAS score of 79.7 (18.2) in the overall group here at 6 months aligns with the range of mean EQ-VAS scores Park et al. noted for treated TB groups (73.9–96.1).77 Similarly, our mean (SD) baseline EQ-VAS score of 66.1 (20.9) is similar to Park et al.’s active TB groups (54.5–80).77

Potential improvement of intervention

We delivered the intervention to all patients at each visit in the intervention arm when on treatment. This was to account for any change in life circumstances that may have arisen during the study – and indeed 10–15% of participants in the intervention arm did require further adherence support after the initial assessment. However, this made all visits longer than those in the standard care arm, and towards the end of a patient’s treatment there was a (perhaps inevitable) drop off in how it was delivered by CMs. So, were all components needed at each clinic visit? And if not, what could be removed and when? The process evaluation informed us that the specific intervention components were all popular, though each appeared to have a somewhat different role for patients. Also, less experienced CMs were more comfortable delivering the intervention throughout the study. Therefore, instead of changing the intervention, there is an argument for using a screening triage at each patient visit that can indicate the level of assessment required. This could, for example, be based on the previous 2–4 weeks’ adherence data, which could be available to clinicians and patients and enable a stepwise, graded response based on the individual’s contemporary adherence patterns. For example, if there were no apparent issues with adherence at baseline, the patient would be screened again at week 8 (when moving from intensive to continuation phase). If no issues were detected at week 8, the assessment would not be repeated or may be repeated at week 16 of a 24-week regimen; if, however, the risk of non-adherence appeared higher at week 8 than at baseline, screening would be repeated more frequently. In the pilot we did not provide this feedback to CMs. In a larger study it would be reasonable to do this for both CMs and patients. While it is rarely done in a TB service, in areas such as HIV medicine it is common, and patients’ plasma HIV load results are often reviewed jointly by the patient and healthcare team, and suggestions made to improve or reinforce drug adherence. Including patients in decision making could be further enhanced by using incentives such as travel or food vouchers. It should also be noted that shortening the duration of treatment (as recently reported for both DS and DR TB) may also improve medication adherence given the generally fewer number of medications prescribed, and so reduce the treatment fatigue probably experienced by both patients and staff after a few months.99,100

Our results suggest that completing the EQ-5D-5L and modified CSRI was not burdensome for TB patients during treatment. Further research should be conducted in a definitive trial to evaluate whether TNA improves quality of life and explore which HRQoL domains would benefit the most from the intervention. Database design can be fine-tuned to enhance data collection.

Unanticipated pathways

The medication box used to measure adherence in the pilot study affected many patients’ motivation and ability to take their treatment. This seemed more common in the standard care arm, although it was noted by patients in the intervention arm as well. This is perhaps not surprising given that measures which provide a degree of order (as well as dispensing the drugs as prescribed such as a dosette box) are routinely used to assist with treatment.

In future trials where adherence is an outcome measure, the influence of monitoring created by measures of adherence should be considered and attempts made to try and reduce these where possible, plus minimise any stigma that exists around people taking treatment for TB and others’ knowledge of this.