The COVID-19 pandemic was anticipated to lead to a global ‘tsunami’ of mental health disorders,Reference Torjesen1 resulting from months of fear about the virus, the impact on relationships, loss of friends and family, loss of income or employment, post-viral effects on health and the psychological effects of stay-at-home orders.Reference Greenberg, Docherty, Gnanapragasam and Wessely2 Approximately 30% of adults appear to have experienced clinically significant symptoms (scoring above established thresholds on validated measures) of depression or anxiety, referred to as common mental disorders (CMDs) in the UK and USA, but this decreased after the initial lockdown restrictions eased.Reference Riehm, Holingue, Smail, Kapteyn, Bennett and Thrul3–Reference Pierce, McManus, Hope, Hotopf, Ford and Hatch5 Similarly, there was an initial increase in symptoms in mental health service attendees on average in the UK, but the effect was not maintained during the period of ‘lockdown’,Reference Saunders, Buckman, Leibowitz, Cape and Pilling6 and although antidepressant prescribing increased, this appears to be in line with trends from recent years rather than an indication of a step-change associated with the pandemic.Reference Rabeea, Merchant, Khan, Kow and Hasan7 As such, the anticipated ‘tsunami’ of mental health disorders is not apparent at the general population level. Yet, it is likely that mental health in some groups of people was disproportionately negatively affected by the pandemic, and it is important to identify these groups to better plan for future health emergencies. Studies conducted in different countries have consistently found heterogeneous trajectories of mental health symptoms during the pandemic, demonstrating that distinct subpopulations with elevated CMDs did not recover after the initial months of the pandemic.Reference Saunders, Buckman, Fonagy and Fancourt4,Reference Pierce, McManus, Hope, Hotopf, Ford and Hatch5,Reference Batterham, Calear, McCallum, Morse, Banfield and Farrer8–Reference Kimhi, Eshel, Marciano, Adini and Bonanno10 Factors associated with such negative symptom trajectories include younger age, female gender, lower income and a previous mental health diagnosis.Reference Saunders, Buckman, Fonagy and Fancourt4,Reference Shevlin, McBride, Murphy, Miller, Hartman and Levita11,Reference Luo, Guo, Yu and Wang12 However, few studies have investigated differential CMD symptom trajectories beyond the first 6 months of the pandemic,Reference Shevlin, Butter, McBride, Murphy, Gibson-Miller and Hartman13–Reference Parsons, Purves, Skelton, Peel, Davies and Rijsdijk15 and most have analysed changes in CMD symptom measure scores regardless of clinical significance. To assist mental health services in planning for a potential future surge event in a health emergency like the COVID-19 pandemic, it is important to model the trajectories of clinical need in the population rather than change in symptom scores. To better understand the potential need for clinical interventions for CMDs and how these have changed over the course of the COVID-19 pandemic, a focus on modelling levels of CMD symptoms that are likely to be clinically significant and data spanning a wider time interval is needed. The present study aimed to identify statistically distinct trajectories of clinically significant CMD symptoms and the associated characteristics of participants during the first 14 months of the pandemic (March 2020 to May 2021), across three national lockdowns in England.
Method
Participants
The analytic sample was drawn from the COVID-19 Social Study, a large panel study of the psychological and social experiences of over 70 000 adults (age ≥18 years) in the UK during the COVID-19 pandemic.Reference Fancourt, Bu, Paul and Steptoe16 The study has been described in detail elsewhereReference Bu, Steptoe and Fancourt17,Reference Fancourt, Steptoe and Bu18 and further information is provided in Supplementary Appendix A, available at https://doi.org/10.1192/bjo.2024.2. Participants were recruited using three main approaches: (a) convenience sampling, such as advertisement through social media and mailing lists; (b) targeted recruitment of people who were from a low-income background, had no or few educational qualifications or were unemployed; and (c) promotion to vulnerable groups (e.g. those with pre-existing mental health conditions, older adults, carers and people experiencing domestic violence or abuse).
The present study included participants residing in England, recruited between 21 March 2020 and 6 September 2020. As recruitment to the study was not random, the sample was weighted by the proportions of gender, age, ethnicity and education in England, obtained from the Office for National Statistics.19 From a total of 58 485 participants with available data, 21 051 did not provide data for at least three time points and were therefore excluded. Of the remaining individuals, 3731 people with missing data on predictors including variables for weighting were additionally excluded, resulting in a study sample of 33 703 participants (flow diagram presented in Supplementary Fig. 1). All participants gave written consent when they signed up to the study. The study was approved by the University College London Research Ethics Committee (approval number 12467/005).
Measures
At recruitment, participants were asked to self-report their sociodemographic characteristics and previous mental health diagnoses via an online survey. The following factors were self-reported: age, gender, ethnicity, income, educational attainment, living arrangements with others, overcrowding (i.e. fewer than one room per person in household), history of mental health condition (yes or no) or chronic physical health diagnoses (yes or no), pre-pandemic amount of social contact, keyworker status (e.g. health and social care worker) and personality (Big Five Inventory).Reference Soto and John20
Participants were asked to complete measures assessing symptoms of depression (Patient Health Questionnaire-9 (PHQ-9))Reference Kroenke, Spitzer and Williams21 and anxiety (Generalised Anxiety Disorder-7 (GAD-7))Reference Spitzer, Kroenke, Williams and Löwe22 at recruitment and at each follow-up via the same online survey application. Follow-up data collection was initially conducted each week, with participants receiving an automatic email invitation to the next wave of data collection 7 days after completing each survey. Reminders were sent 24 and 48 h following the first automatic invitation, and if participants did not complete the survey after these reminders, they stopped receiving further invitations for follow-up. After 22 weeks, follow-up moved to monthly intervals in August 2021.Reference Fancourt, Bu, Paul and Steptoe16 At this point, individuals who had been lost to follow-up but did not formally unsubscribe from the study were invited to participate again. From then, automatic invitations for follow-up data collection were sent 28 days after completion of each survey. Retention rate at follow-up surveys was high throughout the pandemic (see Supplementary Appendix B).
The thresholds to indicate clinically significant levels of depression or anxiety symptoms used for the current study were ≥10 on the PHQ-9 and ≥8 on the GAD-7, in line with those used by English national psychological treatment services23 and the scale developers.Reference Kroenke, Spitzer and Williams21 Clinical caseness refers to symptom levels likely to meet the diagnostic criteria for the measured disorder. Throughout this study, we used the term ‘clinically significant’ to describe self-reported symptoms of depression or anxiety above the threshold of clinical caseness on the PHQ-9 or GAD-7. However, we acknowledge there is evidence that self-report measures have been found to overestimate prevalence of CMDs compared with diagnostic interviews.Reference Scott, Stevelink, Gafoor, Lamb, Carr and Bakolis24 For further details on measures, participant characteristics and retention rate see Supplementary Appendix B and Supplementary Table 1.
Analysis
Growth mixture modelling (GMM) was used to identify different trajectories of the incidence of reporting symptoms of depression and anxiety that were likely to be clinically significant, according to validated thresholds. Depression and anxiety scores were dichotomised using the thresholds listed above to create a binary variable indicating the presence or absence of symptoms likely to be in the clinically significant range. This information is likely to be more relevant than changes in symptom scores for future health service organisation and care planning, as it indicates the number of people with degrees of symptom severity more likely to require clinical care. GMM was used to identify statistically distinct subgroups of individuals based on their patterns of change in the likelihood of reporting clinically significant symptoms of depression or anxiety.Reference Rubel, Lutz, Kopta, Köck, Minami and Zimmermann25,Reference Muthén, Brown, Masyn, Jo, Khoo and Yang26 Although GMM can be specified with pre-assumed forms of change, we made no prior assumption for the current analysis, instead leaving these to be determined by the data (by setting time scores as free parameters, achieved by specifying the first two time scores to 0 and 1 to allow model identification, and all others set to free parameters [*]).Reference Bu, Steptoe and Fancourt17 The following model fit statistics were considered for the GMM: the Vuong-Lo-Mendell-Rubin likelihood ratio test,Reference Lo, Mendell and Rubin27 the Akaike Information Criterion, the Bayesian Information Criterion and entropy values. Further information on the GMMs and model selection decisions is presented in Supplementary Appendix C. Full information maximum likelihood through the expectation–maximisation algorithmReference Dempster, Laird and Rubin28 was used to handle missing data in the GMMs, and survey weights were trimmed at the top 90% to minimise the impact of extreme weights.Reference Saunders, Buckman, Fonagy and Fancourt4,Reference Asparouhov and Muthen29 GMM analyses were conducted in Mplus for Windows, version 8 (Muthén & Muthén, Los Angeles, USA; see www.statmodel.com).Reference Muthén and Muthén30
Following the identification of depression and the anxiety trajectories, multinomial logistic regression models were constructed to explore associations between baseline participant characteristics and trajectory class. We used characteristics that have been associated with increased risk of clinically significant symptoms during the COVID-19 pandemic in previous studies, including age, gender, income, education, living situation, previous mental or physical health conditions, keyworker status, personality characteristics and the reported amount of social contact before the pandemic.
Results
Descriptive statistics
A total of 33 703 participants met inclusion criteria, with descriptive statistics presented in Table 1. There was an overrepresentation of women (76%) and individuals with university-level degrees (70%). There was an underrepresentation of individuals from minority ethnic groups (5%) compared with the general population. We applied weights that better represented the general population. The weighted sample consisted of 20 490 (61%) women, 2777 (8%) individuals from minority ethnic groups and 1650 (49%) individuals with university-level degrees. A previous mental health diagnosis at the point of study entry was reported by 6880 (20%) individuals.
Descriptive statistics
Trajectories of depression and anxiety symptom change
GMM was performed on both the GAD-7 and PHQ-9 data independently. A total of 321 333 PHQ-9 scores (indicating clinically significant symptoms n = 60 872[19%]) and 321 333 GAD-7 scores (indicating clinically significant symptoms N = 59 085[18%]) were analysed. For both depression and anxiety symptoms, the five-class solution appeared to be the best fitting (see Supplementary Table 2 for fit statistics). The trajectories (in Fig. 1) were similar between measures. The classes are described as follows: Class 1, low likelihood of clinically significant symptoms of depression or anxiety throughout the study period; Class 2, high likelihood of clinically significant symptoms throughout the study period; Class 3, clinically significant symptoms early in the pandemic, which reduced within the early months; Class 4, clinically significant symptoms that emerged after 5 months of the pandemic; and Class 5, moderate likelihood of incident depression or anxiety throughout the study period.
Identified trajectories for depression (top panel) and anxiety (bottom panel) symptom incidence.
Class 1 was the largest trajectory class, indicating people who were extremely unlikely to report levels of depression (62% of the total sample) or anxiety (63% of the total sample) that were likely to be clinically significant throughout the pandemic period studied. Nearly 40% of people were members of the four other trajectory classes, which reported clinically significant levels of depression or anxiety at least once during the first 14 months of the COVID-19 pandemic. In particular, 13% of people belonged to Class 2, meaning they consistently reported levels of depression or anxiety likely to be clinically significant throughout the observation period. Other groups reported a moderate likelihood of clinically significant symptoms of depression or anxiety for several months during the pandemic. Class 4 was the only group that appeared to show an increase in the likelihood of clinically significant mental health symptoms in the lead up to the November 2020 lockdown in the UK, representing 6% of the total sample. Supplementary Appendix D presents correspondence between depression and anxiety classes (i.e. the proportion of individuals across to two sets of classes), and it was observed that the levels of similarity when individuals were in the same depression and anxiety class was higher for classes 1 and 2, with less correspondence for the other classes (Supplementary Table 4).
Associations with trajectory classes
Descriptive statistics of each identified trajectory class are presented in Supplementary Table 5. Sociodemographic and personality characteristics independently associated with the likelihood of belonging to depression trajectory classes and anxiety trajectory classes are presented in Tables 2 and 3, respectively. Class 1 (low incidence) is used as the reference class in these multinomial regression models. Further logistic regression analyses were performed comparing Class 2 and Class 3 for both the anxiety and depression trajectories, given the similar proportion of individuals in the clinical range at the start of the study period for these two classes (presented in Supplementary Table 6). Classes 4 and 5 were then compared because of the similar final proportion of individuals in each class observed to be scoring in the clinical range in the last available data in the study period (presented in Supplementary Table 7).
Participant characteristics associated with depression trajectory classes
Numbers in bold are statistically significant at P < 0.05. RRR, relative risk ratio.
Participant characteristics associated with anxiety trajectory classes
Numbers in bold are statistically significant at P < 0.05. RRR, relative risk ratio.
Depression trajectories
Class 1 (low incidence) were older than the other classes, and had a higher ratio of men compared with women (Table 2). Being from a minority ethnic group was associated with following a Class 2 (high incidence) or Class 5 (moderate incidence) trajectory compared with Class 1. Previous mental health diagnoses were very prevalent in Class 2, but all classes were more likely to report previous mental health conditions, as well as physical health conditions, compared with Class 1. Being on a low income and being a carer were also independently associated with increased likelihood of following all trajectories, compared with Class 1. Differences in personality characteristics were also observed: higher neuroticism and openness, as well as lower conscientiousness, scores were associated with a greater likelihood of following trajectories for classes 2–5 compared with Class 1. Being a keyworker was associated with an increased likelihood of following trajectories for Class 3 (early clinically significant symptoms) and Class 5. Socialising every day before the pandemic was associated with increased likelihood of following all trajectories (compared with Class 1), whereas socialising less than once a month was also associated with following Class 2 and Class 5 trajectories. Living alone was associated with following all trajectory classes except Class 4 (emergent clinically significant symptoms).
Further analyses (presented in Supplementary Table 6) exploring the characteristics associated with Class 3 (early clinically significant symptoms) versus Class 2 (high incidence) trajectories, which had similar rates of clinically significant symptoms initially, identified that being a woman, a keyworker and having higher extraversion scores were associated with a greater likelihood of being in Class 3. Being from a minority ethnic group, lower levels of education, lower income, having a previous mental health diagnosis, being a carer, having higher neuroticism scores and having a long-term physical health condition were all associated with a decreased likelihood of being in Class 3. Supplementary Table 7 presents the comparison of factors associated with being a member of Class 5 (moderate clinically significant symptoms) versus Class 4 (emergent clinically significant symptoms). It was found that being younger, having a previous mental health diagnosis, having a long-term physical health condition, socialising less than once a month, having higher neuroticism and openness scores, and having lower conscientiousness scores were associated with a greater likelihood of being in Class 5.
Anxiety trajectories
There were many similarities in the characteristics associated with anxiety trajectory classes 2–5 compared with Class 1 (Table 3). Younger age, previous mental and physical health conditions, lower income, overcrowded living conditions and higher openness and neuroticism scores were all associated with increased odds of following trajectory classes 2–5, indicating potentially clinically significant symptoms of anxiety at some point in the pandemic. Exceptions were that female gender was not associated with higher odds of being in Class 2 (high incidence), minoritized ethnic background was not associated with Class 3 and socialising every day or being a carer were not associated with Class 4 – these characteristics were each linked with higher likelihood of membership to all other classes (versus Class 1).
In further analyses comparing the characteristics of participants in classes 3 and 2, being a woman, having a higher income, not having a previous mental health diagnosis or long-term physical health condition, and socialising once or twice a week (versus less than once a month) was associated with greater odds of being in Class 3 (early clinically significant symptoms). Higher extraversion, lower neuroticism and being a keyworker (versus not) were also associated with an increased likelihood of being in Class 3 compared with Class 2. Finally, Class 4 (emergent clinically significant symptoms) and Class 5 (moderate clinically significant symptoms) were compared (see Supplementary Table 7). Individuals in Class 5 were more likely to be younger, have lower education, have a previous mental health diagnosis, be a carer, socialise more frequently and have higher neuroticism and openness scores compared with those in Class 4.
Discussion
In this study, we observed that nearly four in ten people were members of trajectory classes that reported symptoms of depression or anxiety likely to be in the clinically significant range at least once during the first 14 months of the COVID-19 pandemic in England. However, there were changing patterns of incidence, with <20% of total individual PHQ-9 and GAD-7 scores during the study period likely to be above clinical levels. The majority of participants were members of Class 1 and extremely unlikely to report clinical levels of depression or anxiety (62 and 63% of the total sample, respectively); conversely, 13% of people (Class 2) were likely to report clinical levels of depression or anxiety throughout the observation period. Other groups reported a moderate likelihood of incident depression or anxiety for several months during the pandemic. Class 4 were the only group who appeared to show an increase in the likelihood of clinically significant mental health symptoms in the lead up to the November 2020 lockdown in the UK, but it represented just 6% of the sample. Sociodemographic factors, such as younger age, having previously been diagnosed with a mental or physical health condition, and having a low income, were independently associated with being in trajectory classes with greater likelihood of reporting depression or anxiety symptoms in the clinical range. In addition, higher neuroticism and openness scores, and socialising on a daily basis before the pandemic, were associated with an increased likelihood of following trajectories at greater risk of clinically significant depression and anxiety scores throughout the study period.
These findings confirm heterogeneous mental health trajectories identified earlier in the pandemic.Reference Saunders, Buckman, Fonagy and Fancourt4,Reference Pierce, McManus, Hope, Hotopf, Ford and Hatch5 Four of the five trajectories identified here (all except Class 4) are similar to ones observed in studies conducted with data from the first 6 months of the pandemic,Reference Saunders, Buckman, Fonagy and Fancourt4,Reference McPherson, McAloney-Kocaman, McGlinchey, Faeth and Armour31 as well as two studies that used data spanning the first 12 months of the pandemic.Reference Ellwardt and Präg14,Reference Parsons, Purves, Skelton, Peel, Davies and Rijsdijk15 Class 4, representing a group of adults whose clinically significant symptoms emerged during the course of the pandemic, appears to be a particularly at-risk group regarding mental health. This trajectory class is somewhat similar to a group labelled as those with ‘deteriorating’ mental health, identified in a smaller study conducted over the first 12 months of the pandemic that used less frequent measurement of symptoms with the PHQ-ADS (Patient Health Questionnaire-Anxiety Depression Scale).Reference Shevlin, Butter, McBride, Murphy, Gibson-Miller and Hartman13 The study found that having a history of mental health treatment, higher levels of loneliness and death anxiety, and lower levels of resilience were associated with a ‘deteriorating’ trajectory class compared with the ‘resilient’ class (corresponding to Class 1 in our study). Here, we found that individuals from Class 4 were, on average, younger than participants in Class 1, with higher levels of education but lower incomes, suggesting that perhaps working-age adults who were more likely to have insecure employment or to be on zero-hours contracts comprised this group, including young adults under 30 years of age (albeit only 13% of Class 4) who may have been particularly affected by interruptions to educational courses and graduate employment schemes.Reference Ellwardt and Präg14 They may therefore have been more vulnerable to changes in employment and financial insecurity brought about by the pandemic, and thus to the subsequent effects on their mental health. It may also be that that the prospect of long-lasting cyclical lockdowns may have posed a particular challenge to this group, with the psychological challenges more about the impact of lockdowns and associated burnout, rather than fear of the virus. Although future health emergencies will undoubtedly follow different patterns, the identification of a group at high risk of deteriorating mental health over time is key to isolating who could benefit from early mental health service intervention to reduce incidence and clinical costs later on.
Another key contribution of this study is identifying factors associated with those who had high levels of distress at the start of the pandemic but soon recovered, compared with those who had high likelihood of symptoms in the clinical range throughout. Those who recovered were on average more likely to be keyworkers and socialise more often before the pandemic. It is possible that these individuals were initially distressed because of strict lockdown measures, the uncertainty of the pandemic, the experience of stressful life eventsReference Kendler and Gardner32 and, potentially, the added burden or difficulty faced in their work.Reference Kinman, Teoh and Harriss33 However, they appear to have quickly recovered as restrictions eased. This may have been because they were then able to utilise their social support networks or that they were better able to adjust to the ‘new normal’ because of their own resilience or other resources.Reference Pierce, McManus, Hope, Hotopf, Ford and Hatch5 Furthermore, some of the initial symptoms captured were likely a result of adjustment problems triggered by the first lockdown, which by definition should abate within 6 months of the stressor being removed (i.e. end of lockdown restrictions). The identification of these positive trajectories after the initial shock is important in showing how some individuals may not need mental health service intervention.
This study is also unique in identifying factors associated with moderate clinically significant symptoms throughout the pandemic compared with emergent clinically significant symptoms. People with emergent clinically significant symptoms were on average less likely to have had a previous mental health diagnosis, and scored marginally lower on openness and neuroticism at the start of the pandemic. This supports the notion that a group of people who did not have clinically significant levels of psychological distress before the pandemic have been disproportionately negatively affected, resulting in the development of incident mental health conditions. The identification of personality-based factors associated with the risk of poorer CMD outcomes during the pandemic can be used to inform formulation-based practice, highlighting individuals based on their personality characteristics who might need further support, or who may be a risk of increasing symptoms.
This study replicated findings in a number of other studies that have suggested that females, younger adults, Reference Shevlin, McBride, Murphy, Miller, Hartman and Levita11,Reference Luo, Guo, Yu and Wang12,Reference Fancourt, Steptoe and Bu18 those with lower incomes and those with physical health concerns have been at increased risk of mental health disorders during the pandemic.Reference Saunders, Buckman, Fonagy and Fancourt4,Reference Shevlin, McBride, Murphy, Miller, Hartman and Levita11,Reference Ellwardt and Präg14,Reference Parsons, Purves, Skelton, Peel, Davies and Rijsdijk15 It is worth noting many of these factors had been associated with poorer mental health before the pandemic too.Reference Allen, Balfour, Bell and Marmot34–Reference Riecher-Rössler36 In particular, we found that individuals with previous mental health diagnoses were at higher risk of incident mental health disorders during the study period. This is especially concerning as the reduction in referrals to primary and secondary care mental health services during the pandemicReference Saunders, Buckman, Leibowitz, Cape and Pilling6,Reference Tromans, Chester, Harrison, Pankhania, Booth and Chakraborty37 suggests that this group may not have been able to access care as easily, or that they were reluctant to do so, whether because of fears of burdening healthcare systems or some other reason.
Limitations
The present study used a large data-set, bringing greater precision to previous estimates, and included more recent and frequent data than has been available in other studies. However, there are a number of limitations to note. The raw sample was not fully representative of the UK general adult population. We used survey weights and targeted sampling to improve the representativeness, but these methods cannot remove all sampling biases, and sample weighting could not account for all characteristics that were investigated in this study (e.g. the representativeness of people with a pre-existing mental health diagnosis). Moreover, the sample was limited to people living in England, affecting the wider generalisability of the findings. Additional biases may have been introduced by selecting a minimum of data completed at three time points during the study period as an inclusion criterion. This was selected as it was the minimum requirement for the models fitted, but is likely to have had only a small influence on the results, as the majority of participants provided data for at least 12 time points and sampling weights were generated after sample selection. We chose not to apply corrections for multiple testing in our multinomial regression models, given the exploratory nature of these analyses in identifying potential risk factors, and acknowledge that future studies should consider corrections, especially in more hypothesis-driven research. Residual confounding cannot be ruled out, and we could only explore a range of important covariates available in the data. In particular, depression and anxiety symptom scores before the pandemic were not available. Further detailed information about individualised COVID-19 risk and impact (such as levels of illness, direct impact on family, etc.) were not available, but could be informative about fluctuations in CMD symptom trajectories in future studies. It was decided to model the PHQ-9 and GAD-7 separately in this analysis, as previous analysis using weekly data identified distinct trajectories between the measures;Reference Saunders, Buckman, Fonagy and Fancourt4 however, trajectories observed in the current analysis were similar. Analyses modelling simultaneous change in depression and anxiety scores might add further nuance. A further limitation was that self-reported symptom scores were the only measures of CMD symptoms in this study. Other means of identifying symptoms (e.g. via structured clinical interviews) might have provided a somewhat different picture of participants’ CMD symptom experiences,Reference Hobbs, Lewis, Dowrick, Kounali, Peters and Lewis38 and studies have found that self-report measures may overestimate the prevalence of mental health issues compared with diagnostic interviews in some circumstances.Reference Scott, Stevelink, Gafoor, Lamb, Carr and Bakolis24 However, the PHQ-9 and GAD-7 are widely used in healthcare settings and in epidemiological research, and have been shown to be valid and reliable for identifying clinically meaningful levels of symptoms in various populations.Reference Spitzer, Kroenke, Williams and Löwe22,Reference Saunders, Moinian, Stott, Delamain, Naqvi and Singh39,Reference Shevlin, Butter, McBride, Murphy, Gibson-Miller and Hartman40 Finally, we focused on modelling symptoms that are likely to be clinically significant to inform the future organisation of mental health services, but acknowledge the loss of statistical power from using binary rather than continuous outcomes.
In conclusion, nearly 40% of participants followed trajectories of self-reported depression and anxiety symptoms that were likely to be clinically significant for at least some period of the COVID-19 pandemic. Five distinct trajectories were observed, highlighting the importance of considering heterogenous mental health outcomes rather than population averages. Although three-quarters of individuals showed either consistently high or low incidence of probable mental health disorder, the others showed patterns that fluctuated, either showing early clinically significant symptoms, emergent clinically significant symptoms or fluctuating likelihood of reporting clinically significant symptoms. In future health emergencies, it is important that mental health services differentiate between early and emergent clinically significant symptoms, as although the former are likely to recover on their own within the first few months of the start of an emergency, the latter have a high risk of deterioration over time that could be reduced if appropriate support is provided. The specific predictors of group membership identified in this paper could help inform the development and delivery of public health initiatives in future health emergencies to improve mental well-being, with the potential to prevent the emergence of clinically significant symptoms in a significant portion of the population. It may also support future planning for mental health services, informing estimates of expected levels of need.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjo.2024.2
Data availability
The data-sets generated and/or analysed in this study are not publicly available due to stipulations made by the ethics committee. Requests to access the data should be made to the corresponding author, R.S.
Acknowledgements
This report is independent research supported by the National Institute for Health and Care Research ARC North Thames. The views expressed in this publication are those of the author(s) and not necessarily those of the National Institute for Health and Care Research or the Department of Health and Social Care. Additionally, the research team appreciate the support of a number of organisations with recruitment for this study. This includes the UKRI Mental Health Networks, Find Out Now, UCL BioResource, SEO Works, FieldworkHub and Optimal Workshop.
Author contributions
R.S. is the lead author and was responsible for data analysis. J.E.J.B. and J.W.S. contributed to the write-up, interpretation and provided methodological expertise. P.F. and S.P. contributed to the write-up and interpretation. F.B. contributed to the write-up, interpretation, data curation and provided methodological expertise. D.F. contributed to the write-up, interpretation, secured original funding, data curation and provided methodological expertise.
Funding
The COVID-19 Social Study was funded by the Nuffield Foundation (grant WEL/FR-000022583), but the views expressed are those of the authors and not necessarily the Foundation. The study was also supported by the MARCH Mental Health Network, funded by the Cross-Disciplinary Mental Health Network Plus initiative, supported by UK Research and Innovation (grant ES/S002588/1) and by the Wellcome Trust (grant 221400/Z/20/Z). J.E.J.B. was funded by the Wellcome Trust (grant 201292/Z/16/Z) and the Royal College of Psychiatrists. S.P. was funded by the National Institute for Health Research University College London Hospitals Biomedical Research Centre and the Royal College of Psychiatrists. D.F. was funded by the Wellcome Trust (grant 205407/Z/16/Z). The study was also supported by HealthWise Wales, the Health and Care Research Wales initiative, which is led by Cardiff University in collaboration with SAIL, Swansea University. The funders had no final role in the study design; the collection, analysis and interpretation of data; the writing of the report or the decision to submit the paper for publication. All researchers listed as authors are independent from the funders, and all final decisions about the research were taken by the investigators and were unrestricted.
Declaration of interest
None.
Open access
eLetters
No eLetters have been published for this article.