Article Text

Original research
How is the COVID-19 lockdown impacting the mental health of parents of school-age children in the UK? A cross-sectional online survey
  1. Austen El-Osta1,
  2. Aos Alaa1,
  3. Iman Webber1,
  4. Eva Riboli Sasco1,
  5. Emmanouil Bagkeris2,
  6. Helen Millar3,
  7. Charlotte Vidal-Hall4,
  8. Azeem Majeed5
  1. 1Self-Care Academic Research Unit (SCARU), Department of Primary Care and Public Health, Imperial College London Faculty of Medicine, London, UK
  2. 2National Heart and Lung Institute, Imperial College London, London, UK
  3. 3NHS Fife Mental Health Directorate, Fife Health & Social Care Partnership, Fife, Scotland, UK
  4. 4Brackenbury Primary School, London, UK
  5. 5Department of Primary Care and Public Health, Imperial College London Faculty of Medicine, London, UK
  1. Correspondence to Dr Austen El-Osta; a.el-osta{at}


Objective Investigate the impact of the COVID-19 lockdown on feelings of loneliness and social isolation in parents of school-age children.

Design Cross-sectional online survey of parents of primary and secondary school-age children.

Setting Community setting.

Participants 1214 parents of school-age children in the UK.

Methods An online survey explored the impact of lockdown on the mental health of parents with school-age children, and in particular about feelings of social isolation and loneliness. Associations between the UCLA Three-Item Loneliness Scale (UCLATILS), the Direct Measure of Loneliness (DMOL) and the characteristics of the study participants were assessed using ordinal logistic regression models.

Main outcome measures Self-reported measures of social isolation and loneliness using UCLATILS and DMOL.

Results Half of respondents felt they lacked companionship, 45% had feelings of being left out, 58% felt isolated and 46% felt lonely during the first 100 days of lockdown. The factors that were associated with higher levels of loneliness on UCLATILS were female gender, parenting a child with special needs, lack of a dedicated space for distance learning, disruption of sleep patterns and low levels of physical activity during the lockdown. Factors associated with a higher DMOL were female gender, single parenting, parenting a child with special needs, unemployment, low physical activity, lack of a dedicated study space and disruption of sleep patterns during the lockdown.

Conclusions The COVID-19 lockdown has increased feelings of social isolation and loneliness among parents of school-age children. The sustained adoption of two modifiable health-seeking lifestyle behaviours (increased levels of physical activity and the maintenance of good sleep hygiene practices) wmay help reduce feelings of social isolation and loneliness during lockdown.

  • public health
  • mental health
  • epidemiology

Data availability statement

No further data are available.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Strengths and limitations of this study

  • We surveyed 1214 parents of school-age children to assess the impact of lockdown measures on feelings of social isolation and loneliness.

  • We assessed direct and indirect measures of loneliness using the Direct Measure of Loneliness recommended by the Office for National Statistics and the validated UCLA Three-Item Loneliness Scale, and used Cohen’s kappa to determine whether both measures of loneliness are correlated.

  • We collected data on mental health, physical activity levels and other lifestyle factors in the first 100 days of the lockdown.

  • A key limitation of the study was lack of follow-up which restricted the assessment of the trajectory of feelings of social isolation and loneliness over time.

  • School closures have a significant impact on the mental health of parents of school-age children including feelings of social isoaltion and loneliness, and this should be taken into account when considering future COVID-19 risk mitigation strategies.


The COVID-19 pandemic has affected educational systems worldwide, leading to the near-total closures of educational institutions in the UK. As of 6 May 2020, schools were suspended in 177 countries affecting over 1.3 billion learners worldwide,1 and in many cases closures have resulted in the universal cancellation of examinations.2 3 UNICEF estimated that almost 4 months of education will be lost as a result of the first lockdown.4 School closures have far-reaching economic and societal consequences, including the disruption of everyday behaviours and routines. In the UK, over 2 million workers have already lost their jobs,5 6 and although the long-term impact of the pandemic on education is not yet clear, the pre-existing attainment gap between the poorest and richest children7 may widen significantly as a result of COVID-19.4 8 9 Children and young people make up 21% of the population of England,10 and by the time they returned to school after the summer break, some would have been out of education for nearly 6 months.

Lockdown measures significantly limit social interactions, opportunities for social intercourse or the ability to receive the social support needed to promote mental well-being.11 The temporary closure of schools also means that children miss out on vital social skills and physical activity which may cause further detriment to their mental health and the quality of their social interaction with their parents and other members of the household.12 Loss of routine social contact could also lead to different patterns of social response13 14 while increasing feelings of social isolation and loneliness.15–17 There is growing concern over the impact of school closures on the mental health and well-being of parents and school-age children,18–20 and in particular about increasing feelings of social isolation and loneliness.21

The impact of loneliness on public mental health is well characterised,22 and includes depression,23 24 anxiety25 and suicide,26 27 and is also linked with cardiovascular conditions28 29 and cancer.30 Prolonged periods of loneliness and social isolation are also associated with future mental health problems up to 9 years later,31 including a strong association with depression32 and stress.33 Although acknowledged to be different concepts, social isolation and loneliness may affect people of all ages,34 and the terms are used interchangeably such that they are often considered together.35 There have been numerous attempts in the literature to identify predictors of loneliness,31 36 37 but this subjective phenomenon remains difficult to measure, and its prevalence is thought to be significantly under-represented. Known predictors of loneliness include living alone, living in rented accommodation, household size, education level, self-reported health measures and, paradoxically, living in population-dense areas.

The measurement of social isolation and loneliness is challenging as it is largely subjective and qualitative in nature.38 The UK Office for National Statistics (ONS) recommends the use of the validated UCLA Three-Item Loneliness Scale (UCLATILS) as an indirect measure for loneliness, and an additional Direct Measure of Loneliness (DMOL) question.39 ONS recommends attempting to harmonise these indicators across the UK Government Statistical Service, but the recency of the recommendations may be a reason behind the lack of standardised and retrospective data on loneliness in the UK. Although both scores measure loneliness, they are fundamentally different. The composite score of UCLATILS measures general and indirect loneliness and feelings of social isolation, whereas the DMOL is a separate (single item) measure that assesses the current/temporal feeling of loneliness by the respondent and is recommended for use by ONS

Successful interventions aimed at tackling social isolation and loneliness include leveraging existing community assets such as parks and green spaces, befriending schemes, skill development strategies and psychological therapies.40–43 The UK government published its first Loneliness Strategy in October 2018, signalling the first important step in tackling this rising problem of society. Preventative measures that can be implemented to reduce the risk of social isolation and loneliness and bridge social distancing during lockdown include the use of digital technologies. China and Singapore established various initiatives to minimise outbreak-related stress and poor mental well-being including the deployment of enhanced social support networks and psychological services that could be delivered online.44–46 Teachers can also play an important role in alleviating a child’s sense of isolation at school,47 48 but the extent to which this can be accomplished with live or online lessons through remote learning remains unclear. Reports have already documented loneliness in the elderly as a result of the COVID-19 lockdown,49 but research regarding this aspect of mental health on parents with school-age children during the pandemic is scarce in the first 100 days after the lockdown and this population remains largely understudied.

Study objectives

The aim of this study was to explore how the lockdown is affecting the mental health of parents of school-age children, and in particular to assess the impact of an extended period of school closures on feelings of social isolation and loneliness.


Study design

We conducted a cross-sectional online survey of adult parents and legal guardians of children who were attending primary or secondary education in the UK.

The link to the electronic survey was published and available on the Imperial College Qualtrics platform between 29 May and 11 July 2020 (6 weeks). The survey was open and could be accessed by anyone with a link. Potentially eligible participants received an invitation email from the study team, and the head teacher of Brackenbury Primary School also disseminated the email and link to his counterparts in other schools. Study information was disseminated including the Participant Information Sheet (PIS) and link to the survey. The researchers’ personal and professional networks were also mobilised to respond and further disseminate the eSurvey among eligible participants. The PIS included information regarding the study’s aims, the protection of participants’ personal data, their right to withdraw from the study at any time, which data were stored, where and for how long, who the investigator was, the purpose of the study and survey length. Participants were informed that this was a voluntary survey without any monetary incentives but offering the possibility to access the findings at a later stage while underlying the potential collective benefits of taking part in terms of helping advance knowledge in this area and the formulation of future policies to tackle the COVID-19 pandemic. The data collected were stored on the Imperial College London secure database and only the team researchers could access the eSurvey results.

The survey comprised a total of 51 questions displayed on one page and was accessible using a personal computer or smartphone. Questions regarding demographic characteristics of the users included information on gender, age, ethnicity, educational level, number of people living in the household, first part of postal code and employment status. Participants could review their answers before submitting them. All data collected through the survey were anonymised and not personally identifiable. The online survey technical functionality was tested before being published. The first question asked participants to confirm their consent to participate in the eSurvey.

Experiences and perceptions related to the impact of the lockdown on the mental health of parents and other members of their household were evaluated through a number of questions concerning self-reported or perceived levels of depression, stress, feeling of loneliness, social isolation and boredom. Indirect measures of loneliness were measured using the validated UCLATILS with responses never/hardly ever (score of 1), some of the time (score of 2) and often (score of 3).50 The questions were each scored 1–3, then totalled to a score ranging from 3 to 9. Indirect measure of loneliness using UCLATILS was subsequently categorised as follows: no loneliness (score=3), moderate loneliness (score=4–6) and severe loneliness (score=7–9). An additional one-item DMOL was also used as recommended by the ONS.51 Questions concerning users’ experiences were scored on a 1–5 Likert scale. Respondents were able to refrain from providing an answer by selecting ‘no opinion’. Such answers were treated as missing data in all the analyses (listwise exclusion) but due to the small number of missingness (<1.5%) the data were not imputed.52 53 The association of the two scores was tested using the Cohen’s kappa test of agreement.

The survey included 11 additional questions to explore perceptions of feelings of social isolation before and after school closures. Perceptions on remote learning were explored through questions related to whether or not their child received regular homework, live or online lessons, had access to technology (personal computer, tablet or phone), time spent studying and whether the child had access to a dedicated space to study. Perceptions on the impact of school closures on the lifestyle behaviours of respondents and their schoolchildren were recorded by asking questions relating to prelockdown and postlockdown self-reported measures of physical activity levels of both parents and children, the children’s sleeping patterns and how children spent their leisure time. The quality of the survey was assessed by completing the Checklist for Reporting Results of Internet E-Surveys.

Statistical analysis

Analyses were conducted separately for the UCLATILS and DMOL as recommended by the ONS.51 Parent and child characteristics were described using frequencies and percentages. Pearson’s χ2 test was used to identify differences of statistical significance. Associations between the UCLATILS, DMOL and the characteristics of the study participants were assessed using ordinal logistic regression models. The factors that were significant in the univariable models (p value <0.05) were considered in the multivariable analyses. All analyses were performed using Stata V.15 statistical software (StataCorp).

Patient and public involvement

No patient was involved. The study protocol and online survey were developed in collaboration with the Governing Board of Brackenbury Primary School in the London Borough of Hammersmith & Fulham where the lead author is also a co-opted school governor.


Demographic profile of respondents

The electronic survey captured responses from 1214 respondents from across England (table 1). More than half (53.1%) were aged 40–49 years, whereas 2.5%, 29.2%, 14.4% and 0.9% were in the second, third, fifth or sixth decade of age, respectively. Eighty-seven per cent of respondents were female, and 80.5% identified as white ethnic background. Sixty-six per cent were educated to university degree, 70.9% were in full-time or part-time employment and 87.1% had a partner that was employed. A fifth (20.8%) had one child, 53.5% had two children and 25.8% had three or more children. Only 3.8% were a single-parent family, whereas 75.3% of respondents were living in households consisting of four or more people.

Table 1

Respondent characteristics

School and children characteristics

Nine out of 10 (89.5%) children attended a state-funded school. More than half (54.1%) of respondents had a child receiving primary education, 22.3% in secondary school and 23.6% had more than one child, one attending either primary or secondary schools. Eleven per cent of respondents had a child a special educational need or disability. Sixty-eight per cent indicated that their child had access to a dedicated space where they can learn or study at home. The vast majority (97.9%) of children had access to a personal computer, laptop, tablet or smartphone, of whom 54.0% had their own devices and 43.9% did not have their own but could access devices belonging to other members of their household and 2% did not have access to any technology. Remote learning was accessed by 90.7% of children, but only 47.7% of respondents reported their child was receiving live or online lessons. Only 9.5% of children received private tuition. The time spent on remote learning ranged from 0 to 8 hours/day, with 36.8% studying for less than 2 hours, 30.7% studying between 2 and 4 hours and 32.5% studying more than 4 hours.

Mental health and physical well-being

The vast majority of respondents felt their children were experiencing medium to high levels of boredom (93.8%) and medium or high levels of stress (82.3%) during the lockdown compared with before school closures. Almost half of the participants (48.1%) have reported a shift in the sleeping pattern of children by staying up until much later in the evening during the lockdown. Only 37.2% of respondents reported that the sleeping patterns of their children did not change during the lockdown. Forty-five per cent reported that their levels of physical activity were low during the lockdown. Seventy per cent of respondents felt that school closures also reduced the physical activity of their child.

Loneliness and social isolation

The Cohen’s kappa test between the direct and indirect measures of loneliness (UCLATILS and DMOL) suggested lack of agreement (kappa=−0.34) and therefore it was deemed important to explore the two scores separately. On the UCLATILS, which is the indirect measure of loneliness, 46% (46.3%) of respondents felt they lacked companionship, whereas 52.4% reported having feelings of being left out, and 58% reported feeling isolated from others (table 1; online supplemental table 1). More than half (58.9%) reported they felt lonely often or most of the time on the direct measure (DMOL). Parents reported that 58.5%, 71.0% and 72.2% of children felt they lacked companionship, had feelings of being left out or feeling isolated from others in that same order, whereas 46.9% showed signs of feeling lonely often or most of the time on DMOL. Overall, 43.3% of respondents confirmed that their children were experiencing feelings of social isolation. More than two-thirds (68.8%) felt that video calls where their child could see their teacher could help reduce feelings of social isolation, whereas 60.6% felt this could reduce feelings of loneliness. Overall, 43.9% and 33.0% felt that the lockdown and school closures, respectively, had caused them and their child to feel significantly more depressed (online supplemental table 1).

UCLA Three-Item Loneliness Scale

The multivariable ordinal logistic model suggested that the main factors associated with significantly higher odds of having a higher level of UCLATILS (the indirect measure of loneliness) were female gender of the respondent, having a child with special needs, lack of a dedicated space, a change in the child’s sleeping patterns and having low or medium physical activity during the lockdown (table 2). The univariably significant association of age, level of schooling (primary or secondary education) and access to technology with UCLATILS were attenuated and became non-significant in the multivariable model. Compared with male respondents, females were 82% more likely to have a higher score on UCLATILS. Parents of children who had special needs, and those who lacked a dedicated space to study had 44.0% and 33% higher odds of scoring higher on UCLATILS, respectively. Parents with a low or medium level of physical activity had 53% and 45% higher odds of reporting a higher UCLATILS, respectively, compared with respondents who had high levels of physical activity during lockdown (table 2). Households who reported a disruption in the sleeping pattern of their children were 90% more likely to report a higher UCLATILS.

Table 2

Univariable and multivariable association of UCLATILS with characteristics of study participants

Direct Measure of Loneliness

The factors associated with higher DMOL (the direct measure) were gender, employment status, physical activity level, household size, having children with special needs, having dedicated space to study and changes in sleeping patterns during the lockdown (table 3). In particular, female respondents and those who were unemployed were 52.0% and 70.0% more likely to report a higher DMOL in that same order. Respondents with low or medium levels of physical activity during the lockdown had a 53% increase in the odds of scoring a higher DMOL. Having a child with special needs increased the odds of scoring higher on DMOL by 45%, whereas single-parent families and those whose children changed their sleeping patterns had 2.1-fold higher odds of scoring a higher DMOL.

Table 3

Univariable and multivariable association of ONS Direct Measure of Loneliness (DMOL) score with characteristics of study participants

Households who reported a lack of a dedicated space to study scored 59.0% higher on DMOL (table 3). The associations of other parent and child characteristics that were significantly associated in the univariate analysis with a DMOL (age, education, level of schooling, access to technology and distance learning) were attenuated and became non-significant in the multivariable model.

General perceptions about lockdown, school closures, cancellation of examinations and student preparedness for next academic year

Two-thirds of respondents (66.2%) said they were indifferent that end-of-year examinations were being cancelled, compared with 10.8% who were happy, and 23.0% who said they were unhappy with this decision. Parents felt that only 30% of children preferred examinations to be online as opposed to face to face. Fifty-six per cent of parents of secondary education children felt that their child would not be adequately prepared for examinations if they were to be taken online. Twenty-one per cent reported they would be unhappy or very unhappy to send their child back to school should the lockdown be lifted and schools reopen again before the end of the academic year 2019/2020.


We collected data for 6 weeks during the first 100 days of lockdown in the UK and found that female gender, lower levels of physical activity, parenting a child with special needs, lower levels of education, unemployment, reduced access to technology, not having a dedicated space where the child can study and the disruption of the child’s sleep patterns during the lockdown are the main factors associated with a significantly higher odds of parents reporting feelings of loneliness.

Our findings are consistent with the results of other studies54 55 and reviews56 57 including those that tracked the mental health of adults, children and young people aged 4–16 years throughout the COVID-19 crisis and showed that parents reported an increase in their child’s emotional, behavioural and restless/attentional difficulties.21 58 It is also corroborates existing data which show that access to personal computers, smartphones and tablets varies widely in relation to income levels, with private schools being significantly more likely to provide children with adequate equipment including laptops and tablets.7 It is unsurprising that appropriate access to technology has direct implications on the efficiency of online schooling since remote learning relies on digital access and electronic devices that the child can use at home.59

Another major issue with online provision and remote learning is access to a dedicated space for the child at home that will facilitate such learning. Our study highlighted a significant association between the lack of a dedicated space and increased measures of loneliness in adult respondents using both the direct and indirect measures of loneliness. The lack of a dedicated space may be a proxy measure for lower income in families who are more likely to live in an overcrowded environment.60 The pre-existing attainment gap which loomed between the poorest and richest children showed that children from disadvantaged backgrounds were twice as likely to leave formal education without General Certificate of Secondary Education in English and Math compared with their peers who live in less deprived areas or whose parents have a higher total household income.61 The Education Endowment Foundation has also suggested that school closures could reverse the progress made in the last decade to narrow this gap62 as children from better-off families will have received as much as 35% more home learning than children from the poorest households.63 This raises particular concerns for parents of low income who are less likely to be in a position to assist their children’s studies with financial resources and this can play a significant role in a child’s learning.64 School closures have thus shed a light on the subsequent social and economic consequences of the pandemic including a rise in inequalities and those factors that could be considered as a proxy measure of income deprivation such as digital exclusion, reduced access to tablets and smartphones or a dedicated space where the child can study.33

A recent study established that disruption of good sleep hygiene practices could lead to a behavioural profile of social withdrawal and loneliness,65 whereas loneliness is a known independent risk factor for physical inactivity.66 This was reflected in the findings of our study which showed that both modifiable risk factors (lower physical activity levels and disruption of sleep patterns) were independently associated with higher loneliness. Pertinently, both of these personal risk factors are modifiable and could be addressed through self-care practices. For example, exercise has long been associated with better sleep, and evidence is accumulating on the efficacy of exercise as a non-pharmacological treatment option for disturbed sleep.67 Physical activity interventions in particular have also been shown to reduce loneliness and improve psychological well-being.68 69

Social interaction and physical activity are also known key factors in promoting a healthy state of physical and mental well-being,70–72 but the unprecedented social distancing and lockdown measures have forced the vast majority of the UK population to stay at home for long periods of time. This significantly limited routine opportunities for social interactions with peers, while the closure of schools, gyms and some parks and play areas significantly reduced physical activity levels, including those of parents of school-age children since this group remains largely understudied. Many households were also faced with various issues including concern over job security coupled to the increased need to supervise their children’s learning and homework when one or both parents are required to work from home. Our study showed that these factors are likely to adversely affect the mental health of individuals, and in particular by increasing the prevalence of social isolation and loneliness in households.

Our UK study illustrated an increasing trend in the prevalence of social isolation and loneliness in parents of school-age children during the lockdown as was evidenced among emergency workers and other the quarantined populations.57 73 However, this is the first study that investigated the level of loneliness in a population of parents with school-age children in the UK using both a direct and an indirect measure of loneliness.

The findings of this study may be used to direct interventions aimed at reducing feelings of social isolation and loneliness and to promote good mental health of parents with school-age children. COVID-19 lockdown can be deemed as a period of crisis that has dramatically affected the dynamics of households with school-age children. It is very important to look into the needs of this population during the lockdown as studies have shown that crises, quarantining and restrictions among school-age children have both short-term and long-term effects on their mental health which may affect the mental health of their parents.74 75 Future studies should investigate the effect of remote education on the mental health of children taking into account the findings of Martin who reported that more than 2 hours of daily screen exposure can negatively affect the mental health of young children.76

The prevailing assumption that a resurgence of COVID-19 cases is expected in the winter months shortly after schools reopen in September has led to the development of a range of preparedness and risk mitigation strategies.77 Recent modelling studies predict that school closures alone would only prevent 2%–4% of deaths, which is significantly less than other social distancing interventions.78 Thus, whereas school closures present an apparently logical method of reducing virus transmission as evidenced from previous influenza outbreaks, they pose a dilemma for policymakers seeking measures to protect populations.78 This is reflected in the findings of our study which showed that one in five respondents may be unwilling to send their child back to school should schools reopen again for this academic year. Because school closures have a significant impact on public mental health and well-being20 and may exacerbate inequalities,62 63 this should be taken into account when considering future risk mitigation strategies to minimise virus transmission in the community and educational settings.

The principal limitation of our study was the lack of follow-up, and not recording information about household income and demographic and lifestyle factors such as nutrition, smoking, use of alcohol and recreational drugs which may have enabled a fuller exploration of the factors that could influence the primary outcome measures examined. Further, the demographic profile of study participants largely consisted of white and employed female parents implying that this cross section may not be representative of the wider UK parent population. We also acknowledged that since this was an online survey, we may have excluded parents with little or no digital access. These limitations restrict the generalisability of our findings to the wider population of parents across the UK. In spite of these limitations, our findings echo the results of other studies which show that lockdown measures are negatively impacting the public mental health of individuals across all age groups and may be significantly increasing the prevalence of social isolation and loneliness.18–20

Parents of school-age children remain an understudied population, especially in that they are raising the ‘next generation’ of young adults. The mental health of parents during the lockdown is of major importance because it can significantly impact the psychosocial development and mental health of their children. The extraordinary measures introduced to control the COVID-19 pandemic have exacerbated pre-existing inequalities within society.79 When coupled with social distancing measures, the school closures have negatively impacted the mental health of schoolchildren and their parents and increased the prevalence of social isolation and loneliness in the community setting.


School closures and social distancing measures implemented during the first 100 days of the COVID-19 lockdown significantly impacted the daily routines of many people and influenced various aspects of government policy. Policy prescriptions and public health messaging should encourage the sustained adoption of good health-seeking self-care behaviours including increased levels of physical activity and the maintenance of good sleep hygiene practices to help prevent or reduce the risk of social isolation and loneliness, and this applies in particular where there is a single parent. Policymakers need to balance the impact of school closures on children and their families, and any future risk mitigation strategies should ideally not be a further disadvantage to the most vulnerable groups in society.

Data availability statement

No further data are available.

Ethics statements

Ethics approval

The study was given ethical approval by Imperial College Research Ethics Committee (ICREC number: 20IC5978). Participants consented to take part in the survey.


The authors thank David Collins (head teacher) and the Governing Board of Brackenbury Primary School for disseminating the survey.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @austenelosta, @ebagkeris, @charlottev_h, @Azeem_majeed

  • Collaborators Douglas Yohei Gordon (AND) David Collins.

  • Contributors All authors provided substantial contributions to the conception (AEO, CVH, HLM), design (AEO, EB, CVH), acquisition (IW, AA, ERS) and interpretation (EB, IW, HLM, AM) of study data and approved the final version of the paper. AEO took the lead in planning the study with support from coauthors. EB carried out the data analysis with support from AA. AEO is the guarantor of the paper.

  • Funding AEO and AM are supported by the National Institute for Health Research (NIHR) Applied Research Collaboration (ARC) North West London.

  • Disclaimer The views expressed are those of the authors and not necessarily those of the NHS or the NIHR or the Department of Health and Social Care.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.