Globally, stigma associated with mental, neurological and substance use disorders remains rampant and continues to have an impact on the health and quality of life of people with these disorders.Reference Markowitz1–Reference Clement, Schauman, Graham, Maggioni, Evans-Lacko and Bezborodovs3 Reduction of stigma in mental illness was a key priority in the World Health Organization's 2013–2020 Mental Health Action Plan. Evidence on implementation and effectiveness of stigma reduction campaigns in low- and middle-income countries (LMICs), however, is poor and, until recently,Reference Potts and Henderson4 was absent for regions such as Africa.Reference Corrigan, Morris, Michaels, Rafacz and Rüsch5 Evidence is urgently needed in LMICs, which bear a disproportionately higher burden of these disordersReference Murray, Aravkin, Zheng, Abbafati, Abbas and Abbasi-Kangevari6 and where outcomes are worse owing to additional factors such as under-equipped mental health systems.Reference Lora, Hanna and Chisholm7
The few anti-stigma interventions for mental illness conducted in LMICs have either focused on educational approaches only or have augmented educational approaches with video or in-person contact-based approaches.Reference Clay, Eaton, Gronholm, Semrau and Votruba8 However, little is published about the effectiveness of participatory approaches for stigma reduction involving people with lived experience of mental illness and the target audience in creating and disseminating anti-stigma campaigns. This gap may explain the mixed results on effectiveness of these campaignsReference Hanisch, Twomey, Szeto, Birner, Nowak and Sabariego9 and the resulting low impact on indicators of well-being such as improved socioeconomic positions.
The use of participatory digital methods to implement stigma-reduction interventions in mental illness is gaining momentum among health advocates in both low-income and high-income settings. For example, ‘photovoice’, a method that uses photography to address challenges facing marginalised people, has been used in LMICs to address the impact of climate change on mental health.Reference MacFarlane, Shakya, Berry and Kohrt10 Use of films in a method called participatory video is effective in improving the understanding of mental illnesses.Reference Whitley, Sitter, Adamson and Carmichael11 In participatory video, a group of marginalised people come together to script, film and produce contextually relevant educational videos about an issue that affects them. The workgroup has total control over the entire editorial process and a facilitator only guides the team on technical aspects, such as use of equipment or production software, but not on the theme and content of the films.Reference Sitter12 For people with mental illness, taking part in participatory videos creates opportunities to share their experience and address stereotypes such as narratives that link mental illness with violence and crime. They also provide opportunities for changing the narrative to more positive experiences such as journeys to recovery, as was the case in one recent participatory video project.Reference Carmichael, Adamson, Sitter and Whitley13
Digital interventions to reduce stigma surrounding mental illness have been successfully implemented in LMICsReference Potts and Henderson4 and the use of digital means of communication has grown exponentially globally. Participatory video has been successfully used in Kilifi, Kenya to engage communities about health researchReference Davies, High, Mwangome, Hanlin and Jones14 but not as a stigma reduction strategy. Our present study had two aims: (a) to evaluate the effectiveness of a participatory video project in lowering stigma against people with mental, neurological and substance use disorders; and (b) to examine the sociodemographic correlates of stigma in Kilifi, Kenya. We used knowledge, attitudes and behaviour to conceptualise stigma and used validated toolsReference Bitta, Baariu, Fondo, Kariuki, Lennox and Newton15 to measure the impact of the participatory video intervention on stigma scores immediately after the video exercises and 4 months later.
Method
Design
Between August 2020 and February 2022, we used a repeated measures design to measure the effectiveness of participatory video in lowering mental illness-related stigma by comparing baseline scores and scores at two follow-up time points: one immediately after the intervention, that is approximately 2 h after the baseline scores, and the second 4 months after the intervention. These time points are referred to here as baseline, time point 2 and time point 3 respectively. The study population were adults >18 years from a general population in Kilifi, Kenya. Written informed consent was obtained from all participants.
This was part of a larger mental illness stigma reduction campaign dubbed Difu Simo (‘Breaking Free’), which was established in Kilifi County in 2019 (difusimo.org/#home). Difu Simo conducted mass awareness campaigns in the community within Kilifi. The campaigns involved personal stories from people with lived experience of mental illness, dialogue sessions between traditional healers and biomedical practitioners, question-and-answer sessions from the audience and performances of songs, poems and dances which had messages about mental health and mental illness. Therefore, some members of the target audience for the participatory video intervention presented in this study may have participated in these mass campaigns before, during or after the participatory video intervention.
Ethical approval
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human subjects/patients were approved by the Scientific and Ethics Review Unit under protocol KEMRI/SERU/CGMR-C/167/3933.
Study setting
This study was conducted in Kilifi County, which is located along Kenya's coast of the Indian Ocean. Kilifi covers an area of approximately 12 246 km2 and has a population of approximately 1.5 million, with 54% of the population aged between 15 and 64 years. Most of the county is rural (74%) and the main economic activities are agriculture, fishing, tourism and small-scale trade. Christianity is the most common religion (68%) and Kiswahili is Kenya's lingua franca.16 The burden of mental, neurological and substance use disorders such as depression,Reference Nyongesa, Mwangi, Wanjala, Mutua, Newton and Abubakar17 epilepsy,Reference Ibinda, Wagner, Bertram, Ngugi, Bauni and Vos18 suicideReference Bitta, Bakolis, Kariuki, Nyutu, Mochama and Thornicroft19 and emotional and behavioural problems is high in KilifiReference Kariuki, Abubakar, Kombe, Kazungu, Odhiambo and Stein20 and people with such disorders experience stigmaReference Mbuba, Abubakar, Odermatt, Newton and Carter21 which is linked to cultural explanations of the causes of these disorders, such as witchcraft and curses.Reference Ongeri, Nyawira, Kariuki, Theuri, Bitta and Penninx22 The mental health infrastructure and human resources are limited,Reference Bitta, Kariuki, Chengo and Newton23 with only three psychiatric out-patient units, all based in large general hospitals in large town centres.
The intervention
The intervention was a combination of a participatory educational material (participatory videos) and a face-to-face contact-based approach. The participatory videos were created and disseminated by people with lived experience of mental illness, who are referred to here as content creators. The target audience were influential stakeholders who were familiar with the community's social structure and who were likely to initiate change in the community. They included people with lived experience of mental illness either as patients or caregivers, village elders, religious leaders, youth representatives and healthcare workers.
The study team included a project coordinator with postgraduate training in health research and two field assistants with tertiary level training in community development and mass media. Content creators were identified through the psychiatric out-patient units. Before recruitment, members of the study team gave a general talk about the study's objectives to all patients attending the psychiatric out-patient units. Those who were interested in the study were then given additional information and were invited to attend study-specific sensitisation meetings. A final list of participants who would be available for the entire duration of the study was derived and these participants were trained in the participatory video techniques.
The target audience was identified either through the County's administrative units or through snowballing. Training in the process of participatory video was conducted by InsightShare (insightshare.org/services) and all content creators plus all study team members were trained. Training on the use of filming equipment was provided by the Documentary Institute of Eastern Africa (diea.info). In total, six participatory videos focusing on schizophrenia, suicide attempts, bipolar mood disorder, substance use disorder, epilepsy and depression were created, and 420 participants (target audience) were included in the study over 55 group sessions.
Content creators chose various filming venues within Kilifi and all dissemination sessions were conducted at the Kenya Medical Research Institute's Wellcome Trust premises. This dissemination venue was selected because it was centrally located relative to the target audience and it offered amenities such as electricity, running water, security of filming equipment and strict control of COVID-19 prevention measures. The COVID-19 measures included maintaining physical distance of at least 1.5 m between individuals, use of masks and having up-to-date vaccination records. During each dissemination session, at least one content creator was present and they moderated the sessions with a member of the study team. The content creator led the film viewing and question-and-answer sessions and the study team members administered the stigma questionnaires and set up the equipment before the viewing sessions started.
Measures
Baseline data on the following sociodemographic variables were collected: age, gender, occupation, religious affiliation, level of education, marital status, house tenure (rental or owner), experience with mental illness and exposure to other project activities before participating in the participatory video exercises.
To measure stigma, we evaluated mental health-related knowledge using the Mental Health Knowledge Schedule (MAKS),Reference Evans-Lacko, Little, Meltzer, Rose, Rhydderch and Henderson24 attitudes using the Community Attitudes to Mental Illness (CAMI) scaleReference Taylor and Dear25 and behaviour towards people with mental illness using the Reported and Intended Behaviours Scale (RIBS).Reference Evans-Lacko, Rose, Little, Flach, Rhydderch and Henderson26 All tools were translated to Kiswahili, back-translated into English and were validated for Kilifi populations and found to possess good psychometric properties.Reference Bitta, Baariu, Fondo, Kariuki, Lennox and Newton15 The MAKS contains 12 questions: 6 measure general mental health knowledge and 6 measure knowledge about specific mental illnesses. The adapted and validated version of the CAMI used in this analysis contains 23 questions which measure 3 domains: authoritarianism, tolerance and sympathy. Tolerance and sympathy are positive traits, whereas authoritarianism is a negative trait which reinforces isolation of people with mental illness. Therefore, for ease of interpretation, we reverse scored the total scores in the authoritarianism domain and reported the domain as ‘liberal attitudes’.
The RIBS contains eight questions: four measure reported behaviours and four measure intended behaviours. Reported behaviours are calculated only as frequencies and are not used to calculate overall scores, hence only intended behaviours are included in the final scores. Responses for all tools are on a Likert scale and higher scores endorse positive traits, for instance a high score on the RIBS indicates greater likelihood of positive intended behaviour towards people with mental illness, whereas a low score indicates greater likelihood of stigmatising behaviour. Overall scores are calculated by summing total scores for all questions. Domain-specific scores are calculated by summing the scores from questions related to that domain.
Statistical analysis
Descriptive participant sociodemographic statistics were presented for baseline and for time points 2 and 3, respectively. Assuming heterogeneous variance,Reference Zhao, Wang, Totton, Cullen and O'Connor27 mixed linear models were used to compare standardised scores for knowledge, attitudes and behaviour at time points 2 and 3. In all the models, time was included as a fixed effect using a categorical dummy variable. All measured sociodemographic variables were included as random covariates in the models to control for differences in distribution in the study sample. Results were reported as standard deviation units. Using a two-sided test, and assuming a standard deviation of the difference between means to be 2, the mean difference to be 0 under the null hypothesis, a correlation of 0.9 between the paired observations and a 5% significance level, a sample size of 420 was required to detect a difference in means with 100% power, after accounting for 20% attrition.
To determine whether the associations between baseline characteristics and stigma outcomes were affecting the effectiveness of the intervention, post hoc analyses were conducted to explore interactions between the intervention effectiveness and baseline characteristics. A negative coefficient indicated reduced stigma scores compared with baseline. For the time variable, the assumption was that there was no departure from a linear trend. All analyses were conducted in the statistical software STATA version 17 for Microsoft Windows 11.
Results
Sociodemographic characteristics of study participants
In total, 420 participants were recruited. All 420 completed follow-up at time point 2 and 399 (95%) completed follow-up at time point 3. There was no difference in the gender distribution of participants (P = 0.567) at baseline and time point 3. Participants at baseline had a median age of 38 years (interquartile range IQR = 27–51 years). There was no difference in the mean ages at baseline and at time point 3 (P = 0.554). The majority (92.8%) of the participants had completed at least primary level education, which is equivalent to 8 years of formal education. At baseline 69% of participants were either married or had been previously married; the rest (30.5%) were single. There were no significant differences between baseline and time point 3 distributions of levels of education, marital status, religious affiliation or occupation (Table 1).
Table 1 Sociodemographic characteristics of study participants
a. These are people who may have joined the community awareness campaigns before being recruited to the study.
Changes in levels of knowledge and its sociodemographic correlates
In the univariable analysis, general mental health-related knowledge significantly improved at both follow-up time points compared with baseline, as indicated by a positive effect size of β = 0.30 (95% CI 0.26– 0.34; P < 0.01) at time point 2 and β = 0.31 (95% CI 0.26–0.35; P < 0.01) at time point 3. Knowledge about specific mental illnesses also improved at both follow-up time points compared with baseline (Table 2). Being Christian (β = 0.17, 95% CI 0.01–0.34; P = 0.04), having a tertiary level of education (β = 0.15, 95% CI 0.03–0.27; P = 0.02) and having additional contact with the project (i.e. engaging in activities other than the participatory videos) (β = 0.08, 95% CI 0.02–0.14; P = 0.01) were significantly associated with higher general mental health-related knowledge scores. There were no significant associations between knowledge about specific mental illnesses and any sociodemographic variables.
Table 2 Baseline and post-intervention knowledge scores using the Mental Health Knowledge Schedule
ref, reference. Bold denotes P<0.05.
a. Zero denotes that estimates and both upper and lower confidence intervals were less than +/− 0.01.
In the multivariable analysis, general mental health-related knowledge scores improved immediately after watching the participatory videos (β = 0.30, 95% CI 0.26–0.34; P < 0.01) and remained higher than baseline scores 4 months later (β = 0.31, 95% CI 0.27–0.35; P < 0.01). Knowledge about specific mental illnesses also significantly improved at both follow-up time points (Table 2). Additional contact with the project (β = 0.08, 95% CI 0.02–0.14; P = 0.02) and tertiary level of education (β = 0.18, 95% CI 0.05–0.31; P = 0.01) remained significantly associated with higher general mental health-related knowledge. Additionally, people of Christian and Islamic faiths exhibited higher general mental health-related knowledge (Table 2). There were no significant sociodemographic correlates of scores on knowledge about specific mental illnesses.
Changes in attitudes towards people with mental illnesses and sociodemographic correlates of attitude scores
All three domains of attitude improved at both time points after the intervention in both the univariable and multivariable analyses (Table 3). In the univariable analysis, compared with those with no formal education, having any form of education was significantly associated with a more liberal and sympathetic attitude but not with a more tolerant attitude. Compared with single people, people who were married (β = −0.21, 95% CI −0.36 to −0.07; P < 0.01) or who were divorced, widowed or separated (β = −0.27, 95% CI −0.48 to −0.06; P = 0.01) were significantly less liberal. Additional contact with the project through attending mass campaigns or other events was associated with a more tolerant attitude (β = 0.14, 95% CI 0.01–0.28; P = 0.04) but there were no associations with liberal or sympathetic attitudes. Older age was associated with a less liberal (β = −0.01, 95% CI −0.01 to 0.00; P < 0.01) but more sympathetic attitude (β = −0.01, 95% CI −0.01 to 0.00; P < 0.01). In the multivariable analysis, having any level of formal education was significantly associated with a more liberal and more sympathetic attitude. Age was associated with a more sympathetic attitude (β = 0.01, 95% CI 0.01–0.02; P < 0.01).
Table 3 Baseline and post-intervention attitude scores using the Community Attitudes to Mental Illness scale
ref, reference. Bold denotes P<0.05.
a. Zero denotes that the upper confidence interval was less than 0.01.
Changes in behaviour towards people with mental illnesses and sociodemographic correlates of behaviour scores
Compared with baseline, intended behaviour scores improved immediately after watching the participatory videos (β = −0.29, 95% CI 0.23–0.35; P < 0.01) and 4 months later (β = −0.37, 95% CI 0.31–0.43; P < 0.01), suggesting that the participatory videos may have been effective in lowering discriminatory behaviours among participants (Table 4). In both the univariable and multivariable analyses, secondary and tertiary levels of education were associated with higher intended behaviour scores, i.e. less discriminatory behaviour. In both models, compared with people with no religion, Christianity was associated with higher intended behaviour scores, i.e. Christians reported that they did not intend to engage in discriminatory behaviour.
Table 4 Baseline and post-intervention behaviour scores using the Reported and Intended Behaviours Scale
ref, reference. Bold denotes P<0.05.
Interaction between intervention outcome and baseline characteristics in the knowledge domain
In the general mental health-related knowledge domain, significant differences were found in three baseline characteristics as follows: compared with those who rented houses, homeowners had lower levels of knowledge post-intervention (β = 0.05, 95% CI 0.00–0.09; P = 0.04); compared with those who were employed, those who were unemployed had higher levels of knowledge post-intervention (β = −0.06, 95% CI −0.10 to −0.01; P = 0.02); and compared with those who were not exposed to any additional components of the intervention, the exposed groups had higher levels of knowledge post-intervention (β = −0.09, 95% CI = −0.14, −0.04; P < 0.01) (Table 5). In the knowledge about specific mental illnesses, significantly higher scores post-intervention were found only in those who were exposed to additional components of the intervention (β = −0.07, 95% CI −0.13 to −0.02; P = 0.01) (Table 5).
Table 5 Interaction analysis between sociodemographic variables and participatory video intervention in the knowledge domain
Bold denotes P<0.05.
a. Zero denotes that estimates and both upper and lower confidence intervals were less than +/− 0.01.
Interaction between intervention outcome and baseline characteristics in the attitude domain
In the liberal attitude domain, significant differences were found in five baseline characteristics as follows: with each unit increase in age, the intervention was less likely to improve participants’ attitude (β = 0.01, 95% CI 0.01–0.01; P < 0.01); compared with those who had no education, the intervention was more likely to improve scores among those with any level of formal education (Table 6); compared with those who were single, those who were presently married (β = 0.15, 95% CI 0.04–0.25; P = 0.01) or those who were divorced/widowed/separated (β = 0.33, 95% CI 0.17−0.48; P < 0.01) had significantly lower scores after the intervention; compared with those who rented houses, homeowners had lower levels of knowledge post-intervention and compared with those who were not exposed to any additional components of the intervention, the exposed groups were more liberal post-intervention (Table 6). In the tolerant attitude domain, significant differences were found in the following categories: marital status had a negative interaction with the domain, whereas additional contact had a positive interaction with the intervention outcome (Table 6). In the sympathetic attitude domain, experience with mental illness and additional contact with the project interacted with the intervention outcomes.
Table 6 Interaction analysis between sociodemographic variables and participatory videos intervention in the attitude domain
Bold denotes P<0.05.
Interaction between intervention outcome and baseline characteristics in the behaviour domain
In the intended behaviour domain, level of education was the only baseline characteristic for which significant differences were found post-intervention (Table 7).
Table 7 Interaction analysis between sociodemographic variables and participatory videos intervention in the behaviour domain
Bold denotes P<0.05.
Discussion
This study assessed the efficacy of a digital strategy utilising the method of participatory video to reduce public stigma against individuals with mental disorders in a resource-limited setting. Similar to a Canadian studyReference Whitley, Sitter, Adamson and Carmichael11 employing participatory video for mental illness stigma reduction, we observed significant enhancement in knowledge, attitudes and intended behaviour after participatory video exercises. The effectiveness of our participatory video approach could be attributed to multiple interconnected components. First, involving individuals with lived experience as content creators is likely to have contributed positively to the outcomes. Content generated by those with lived experience of mental disorders tends to evoke emotions and is perceived as ‘real and relatable,’ ‘attention-grabbing,’ and ‘change-inducing’.Reference Whitley, Sitter, Adamson and Carmichael11 Second, dissemination of the participatory videos in our study was conducted by individuals with lived experience, facilitating direct contact with the target audience. The efficacy of contact-based methods as stigma reduction strategies is well-documented.Reference Thornicroft, Sunkel, Aliev, Baker, Brohan and El Chammay28 Third, the intervention targeted community opinion leaders, a proven strategy for enhancing mental health awareness and facilitating early intervention for common mental disorders such as depression.Reference Ford-Paz, Kuebbeler, Contreras, Garduño and Sánchez29
The successful implementation of the intervention hinged on the interplay between the technology (participatory video) and individuals embedded within a complex ecosystem of interacting factors. For instance, we capitalised on the administrative and social structures at our study site to engage community leaders as a target audience. However, infrastructure-related elements such as a consistent power supply and individual-level factors such as participants’ willingness to integrate the lessons into daily routines influenced the intervention outcomes. This implies that participatory video should not be seen merely as a product, but rather as a technology-enabled service with results contingent on addressing the needs of the target audience and adapting to the implementation environment. Thus, for widespread and sustainable adoption, future studies should customise such services to both audience needs and the implementation context.
Our findings of significant correlations between stigma scores and certain sociodemographic variables align with those of a study that conducted a digital stigma reduction campaign in Kenya and Ghana.Reference Potts and Henderson4 However, these correlations were not uniform across all three stigma domains. This implies that when employing participatory video, it might be necessary to stratify the target population based on sociodemographic variables highly linked to the specific domain. For instance, our study revealed that practising any religion was associated with improved knowledge and behaviour outcomes, whereas a recent study in Kenya targeting youth aged 18–34 found better outcomes among those with no religion.Reference Potts and Henderson4 This discrepancy might be attributed to the age-related mediation of religious affiliation,Reference Joseph and Schuyler30 which is itself correlated with stigma.Reference Abuhammad and Al-Natour31 Consequently, for future studies aiming to enhance knowledge and intended behaviour in this context, age-based stratification of the target audience could be considered. Furthermore, the variability in sociodemographic correlates among the three stigma domains indicates that a singular intervention might not comprehensively address all aspects of stigma. This suggests that the application of the ‘what matters most’ principle in selecting the outcome variables of interest is important.Reference Yang, Chen, Sia, Lam, Lam and Ngo32–Reference Kohrt, Turner, Rai, Bhardwaj, Sikkema and Adelekun34
The participatory video intervention was embedded within a broader community-level awareness campaign. Engaging in other aspects of the extensive campaign correlated with enhanced knowledge and improved attitudes, but not with positive intended behaviour; these outcomes were consistently observed in the interaction analysis. This indicates that further approaches are required for improved attitudes to translate into positive behaviour. When targeting opinion leaders, a more effective strategy for enhancing their intended behaviour might involve amalgamating mass campaigns with contact-based participatory video approaches. A future study including a control group is necessary to delve deeper into this observation. In contrast to other surveys, there was no correlation between lived experience of mental illness and overall stigma scoresReference Potts and Henderson4,Reference Robinson and Henderson35 in our study. This may be due to lack of statistical power in our study to detect differences linked to lived experience.
Study strengths and limitations
Our study contributes knowledge on the effectiveness of a participatory digital stigma reduction strategy in Africa, a region lacking extensive research on digital anti-stigma interventions.Reference Corrigan, Morris, Michaels, Rafacz and Rüsch5 We employed globally validated stigma assessment tools,Reference Potts and Henderson4,Reference Ribeiro, Gronholm, de Paula C, Hoffmann, Rojas Vistorte and Zugman36,Reference Pingani, Sampogna, Evans-Lacko, Gozzi, Giallonardo and Luciano37 which were also validated in our study setting.Reference Bitta, Baariu, Fondo, Kariuki, Lennox and Newton15
However, our study does have limitations. The intervention comprised two key components: interactive educational participatory videos and in-person contact between the target audience and individuals with lived experience of mental illness. Although both strategies have shown lasting reductions in stigma,Reference Clement, van Nieuwenhuizen, Kassam, Flach, Lazarus and De Castro38,Reference Dalky39 the exact mechanisms driving change in our study could not be specifically identified, warranting further qualitative exploration. Certain participants were exposed to other facets of the overall awareness campaign due to community residence, potentially influencing observed effects. Interaction effects between the participatory videos and mass community campaigns may have occurred, but owing to mass administration of the latter, such interactions could not be measured. Furthermore, absence of control groups not exposed to any interventions hindered the delineation of each component's relative contribution to the overall effect size. Despite sustained stigma reduction post-intervention for 4 months, the long-term impact remains unknown. Although the short follow-up periods minimised potential modification by other interventions, direct quantitative attribution of effects to the intervention was challenging. Lastly, social desirability bias due to face-to-face questionnaire administration, as seen in other studies,Reference Henderson, Evans-Lacko, Flach and Thornicroft40 may have affected results.
Implications
Although our study establishes evidence for the feasibility and efficacy of participatory approaches in reducing stigma, the diverse sociodemographic associations with stigma across the three domains suggest that future interventions may require targeted strategies for both components and desired outcomes. In this study, stigma levels were assessed using the biomedical model of illness understanding. To comprehensively address stigma, future research should explore it within alternative frameworks of mental health and illness, particularly in settings like ours, where culture significantly perpetuates stigma.Reference Ongeri, Nyawira, Kariuki, Theuri, Bitta and Penninx22 Understanding the mechanisms underlying participatory video's efficacy as a stigma reduction strategy demands a collaborative effort involving mental health researchers, technology experts and behavioural scientists.
Data availability
All data are currently being curated to remove participant identifying information and will be made available on a publicly accessible website. Details of published data will be made available in future papers.
Acknowledgements
We would like to thanks Professor Hellen Cross, Professor Arjune Sen and Professor Lay Sander for their review and feedback on the manuscript.
Author contributions
M.A.B.: study design, data analysis, draft of first manuscript. J.B.: data collection, manuscript review. S.M.K.: data analysis, manuscript review. S.G., B.L. and C.R.J.C.N.: study design, manuscript review.
Funding
This study was funded by the Wellcome Trust grant number 213763/Z/18/Z and by the National Institute for Health and Care Research (grant number NIHR200134) using Official Development Assistance (ODA) funding. The views expressed in this publication are those of the author(s) and not necessarily those of the Wellcome Trust, National Health Service, the National Institute for Health and Care Research or the Department of Health and Social Care.
Declaration of interest
None.
Open access
eLetters
No eLetters have been published for this article.