Despite the existence of a substantial evidence base pointing to the positive sequelae of pandemics (e.g. increased resilience and optimism, better social support and bonding, etc.; Chen & Bonanno, 2020; Drury & Tekin Guven, 2020; Solnit, 2010), widespread concern has been expressed about the protracted nature of the COVID-19 pandemic, and its potentially significant negative socio-economic and health-related impact on the lives of citizens over the medium to long term (Gayer˗Anderson et al., 2020; Ornell et al., 2020; Shah et al., 2020). By June 2020, over a quarter of a million people in the UK had contracted COVID-19, and approximately 40,000 COVID-19 related deaths had been registered (Office for National Statistics, 2020a). Approximately 8.9 million people were in receipt of income support via the government’s Coronavirus Job Retention Scheme (HM Revenue and Customs, 2020), and the UK debt level, which was estimated to be £1.95trn, was larger than the economy for the first time in over 50 years (Office for National Statistics, 2020b). Recent commentaries argue that the socio-economic consequences of the pandemic are exposing and exacerbating existing societal inequalities, with the pandemic having a disproportionately negative impact on the lives of more vulnerable members of society (Morgan & Rose, 2020). Amidst these growing concerns, there is a pressing need to develop a robust evidence base, derived from analyses of high-quality, population-level data, to determine how the public are adapting to life and the many public-health restrictions imposed throughout the course of the pandemic (Davis, 2020).
In March 2020, the longitudinal COVID-19 Psychological Research Consortium (C19PRC) Study was designed and launched with the aim of collecting high-quality data (via self-report questionnaires, qualitative interviews, and quasi-experimental studies) to test a range of theoretically-informed research questions to obtain a greater understanding of the adult population’s psychological and social adjustments to the pandemic. Two core aspects of the C19PRC Study design will help ensure that this aim is achieved, and that the study’s data is well placed to contribute significantly to the knowledge base surrounding the mental health impacts of the COVID-19 pandemic. First, a broad array of standardised measures were used to capture the prevalence of common mental disorders including major depressive disorder (MDD), generalized anxiety disorder (GAD), and posttraumatic stress disorder (PTSD), as well as other important experiences such as somatisation and paranoia (McBride et al., 2020). These core measures facilitate the assessment of a variety of mental disorders and experiences commonly investigated in previous infectious respiratory disease outbreaks (Cheng, 2004; Gardner & Moallef, 2015). They also offer a more detailed interrogation of these diagnostic constructs compared to other leading national longitudinal mental-health studies currently being conducted during the pandemic, which have, in many cases, relied on established but general measures of psychological distress (Pierce, Hope, et al., 2020) or very short screening tools for MDD and GAD (Henderson et al., 2020).
Second, the inclusion of a battery of psychometric measures to assess individual-level psychological factors (e.g., personality, memory, cognitive reasoning ability, locus of control, death anxiety, happiness, and resilience), political attitudes and behaviours (e.g., voting behaviour, political predispositions, nationalism, and patriotism), COVID-19 health-related knowledge and behaviours, as well as the collection of geo-spatial data to facilitate linkage of individual-level survey data to important macro-level data (e.g., country-level COVID-19 related statistics including geographically-framed infection rates, mortality rates, and lockdown status), ensures that the C19PRC Study possesses explanatory potential beyond that of most other studies and surveys established during the pandemic.
As detailed elsewhere, the C19PRC study commenced in the UK, but has since expanded to include international partners in the Republic of Ireland (RoI), Spain, Italy, Saudi Arabia, and the United Arab Emirates (UAE). The UK strand of the Study, to which we refer as C19PRC-UK, is the ‘parent’ survey of the Consortium and is funded by the Economic and Social Research Council in the UK. Where possible/appropriate, international partners model their fieldwork procedures and survey content for each wave on the C19PRC-UK design, although there are important differences between the countries in terms of the timing of fieldwork and survey content. For example, in the RoI and Spain, the first two waves were conducted during March/April and May 2020 (Hyland et al., 2020; Valiente et al., 2020), which was consistent with the UK, whereas in Italy, the UAE and Saudi Arabia, baseline and follow-up waves were conducted between April and August 2020 (Bruno et al., 2021). Whereas the UK survey has a strong focus on collecting socio-political survey content (McBride et al., 2020), a key priority for the Spanish team was to measure and assess positive psychosocial responses to the pandemic (e.g., posttraumatic growth, hedonic and eudaimonic well-being, openness to the future, primal positive beliefs, etc.; Valiente et al., 2020, 2021). The Consortium is committed to data harmonisation (where possible) to facilitate multi-country research studies, and this complex programme of work is on-going. Between April and September 2020, the Consortium produced 14 academic papers analysing the rich survey data, and several of these involved multi-country data analysis (Hartman et al., 2020; Hyland et al., 2021; Murphy et al., 2021). All outputs are accessible via the dedicated OSF, COVID-19 Psychological Research Consortium (C19PRC) Panel Study (2020) hosted with the Open Science Framework.
In this paper, we report the protocol for the third wave of C19PRC Study in the UK (C19PRC-UKW3), which was conducted during July and August 2020. As described elsewhere (McBride et al., 2020), at baseline (C19PRC-UKW1), 2025 adults aged ≥18 years, who were representative of the UK adult population with respect to gender, age, and household income, were recruited via an internet-based panel survey in March 2020. Towards the end of April 2020, 1406 of these adults were recontacted for the first follow-up survey (C19PRC-UKW2), representing a 69.4% retention rate. The first two waves of the C19PRC Study were conducted at the beginning and peak of the first wave of COVID-19 in the UK, respectively, whereas fieldwork for C19PRC-UKW3 commenced at the tail end of the first wave (see Figure 1).
Graphical presentation of the number of daily COVID-19 cases and deaths in the UK, sourced from Our World in Data, 2020, aligned to the C19PRC Study survey waves. New daily deaths and cases depicted as 7-day rolling average
Despite the decline in daily COVID-19 transmission and death rates, important social, economic, and political events rapidly unfolded during the period between the end of C19PRC-UKW2 and C19PRC-UKW3. These included, but were not limited to: (1) the relaxation of the first national lockdown; (2) commencement of human trials for a COVID-19 vaccination in the UK; (3) social and political unrest during the pandemic; (4) the gradual return to school for children before the 2020 summer holidays; (5) announcement of a timeline to end the Coronavirus Job Retention Scheme, and (6) the introduction of travel-related quarantine restrictions and bans (see Table S1 for a detailed timeline). As with previous waves, the content of the C19PRC-UKW3 was considered carefully to capture the impact of these events on the lives of survey participants.
A key methodological concern of longitudinal panel studies is sample attrition (Lynn, 2009), and studies initiated during the COVID-19 pandemic are not immune to this challenge. Attrition in a panel survey tends to increase as the number of follow-up periods increases, and it has considerable potential to negatively impact on the generalisability of findings if participants who stay in the study differ from those who drop out in relation to core study outcomes (Gustavson et al., 2012). Whilst the C19PRC Study team works closely with our fieldwork partner, Qualtrics, to maximise the retention of adults across waves to protect and sustain the longitudinal credentials of the survey, refreshment or ‘top-up’ sampling was conducted at C19PRC-UKW3. Refreshment sampling recruits new respondents into the panel to match specific characteristics of adults who were lost to follow-up. This process, which is common in established panel surveys such as the American National Election Study, ensures that the C19PRC panel sample will remain sufficiently large to conduct meaningful longitudinal analyses for the core study outcomes of common mental disorders, as well as being as representative as possible of the baseline target population (adults aged 18 years and older living in the UK). This paper describes the C19PRC team’s work to (i) examine the level of attrition in the C19PRC by the third wave and whether this could be predicted by important baseline mental-health attributes, psychological characteristics, as well as socio-demographic factors; (ii) conduct weighting procedures to formally manage attrition in the longitudinal panel; and (iii) determine the success of sample refreshment procedures conducted at C19PRC-UKW3.
2 METHOD 2.1 C19PRC-UKW3: Fieldwork procedures 2.1.1 Fieldwork organisation overviewFieldwork for the C19PRC Study was conducted by the survey company Qualtrics. Qualtrics partners with over 20 online sample providers to supply a network of diverse, quality respondents to their worldwide client base and, to date, has completed more than 15,000 projects across 2,500 universities worldwide.
2.1.2 ProcedureC19PRC-UKW3 survey data collection commenced on 9 July 2020, approximately 10 weeks after the completion of C19PRC-UKW2. In Phase 1, Qualtrics re-contacted all adults who participated in previous waves (N = 2025) via email, SMS, or in-app notifications and invited them to participate. The survey was released to a sub-sample of participants initially for a ‘soft launch’ (see Quality Control Section) prior to the full launch of the survey wave later that day.
Qualtrics’ partners released invitations in batches and, after the initial invitation was received, respondents who had not completed the survey were sent two reminders to encourage them to participate. The first reminder was sent approximately 36–48 h after the initial survey invite, with the second reminder sent another 36–48 h after this first reminder. Phase 1 fieldwork lasted two weeks (9–23 July 2020).
Prior to Phase 2, Qualtrics compared the characteristics of the Phase 1 sample to the pre-determined sampling quotas set at baseline. As previously described (McBride et al., 2020), the target population for the C19PRC-UKW1 survey was the UK adult population aged ≥18 years, and quota sampling methods were employed to achieve a representative sample in terms of age and gender (using 2016 population estimates from Eurostat, 2020) and household income (using 2017 income bands from the Office for National Statistics, 2017). Phase 2 fieldwork was therefore organised to recruit new respondents according to gaps in the sampling quotas following the completion of Phase 1. New respondents for Phase 2 were alerted to the C19PRC-UKW3 by Qualtrics in one of two ways: (1) they opted to enter studies they were eligible for by signing up to a panel platform; or (2) they received automatic notification through a partner router which alerted/directed them to studies for which they were eligible. To avoid self-selection bias, survey invitations to eligible participants only provide general information and do not include specific details about the contents of the survey. Participants were required to be adults, able to read and write in English, and resident in the UK. No other exclusion criteria were applied. Panel members routinely receive an incentive for survey participation (e.g., gift cards), based on the length of the survey, their specific panellist profile, and target acquisition difficulty.
Phase 2 fieldwork commenced on 23 July 2020 with a ‘soft launch’ (see Quality Control Section) and the full survey was launched on 24 July 2020. Qualtrics proceeded as follows during the Phase 2 fieldwork: (1) adults in ‘hard to reach’ quota groups (e.g., young people in the highest income bands) were targeted first; (2) the focus then shifted to allow the quotas to ‘fill up’ naturally; before (3) switching back to targeting respondents to fill incomplete quotas. Adults who chose to participate followed a link to a secure website and completed all surveys online. The invite link only remained active for a participant until a quota they would have qualified for was reached.
2.1.3 Informed consent processParticipants were informed about the purpose of the C19PRC Study, that their data would be treated in confidence, that geolocating would be used to determine the area in which they lived (in conjunction with their residential postcode stem), and of their right to terminate participation at any time. Participants were also informed that some topics may be sensitive or distressing. Information about how their data would be stored and analysed by the research team was also provided. Participants were also informed that they would be re-contacted at a later date to invite them to participate in subsequent survey waves. Participants provided informed electronic consent prior to completing the survey and were directed to contact the NHS 111 helpline upon completion if they had any concerns about COVID-19.
2.1.4 Compliance with general data protection regulation (GDPR)C19PRC data will be stored confidentially in line with GDPR. When the study data is deposited with the UK Data Service, location data will be removed and replaced with relevant socioeconomic summary data (e.g. area-level deprivation and population density data). All other personal data will also be removed.
2.1.5 Quality controlQualtrics conducted validation checks on the C19PRC-UKW3 data, though this varied slightly across the Phases. In Phase 1, the ‘soft launch’ was conducted with 100 respondents and this data was screened for technical errors and omissions in the survey measures and/or filtering processes prior to the full launch. Adults who participated in the ‘soft launch’ were retained in the Phase 1 sample.
Qualtrics routinely analyses survey completion times to ensure that respondents spend sufficient time providing high-quality responses. For longitudinal surveys, this process is completed once only, at baseline. Once a participant satisfies the minimum survey completion time, which is set at half the median time of the soft launch for that wave (11 min 11 s for C19PRC-UKW1; McBride et al., 2020), the data they provide in subsequent waves is not subject to a minimum completion time restriction. Thus, the respondent’s completion time at baseline serves as an indicator of their status as a legitimate survey respondent which they carry with them across subsequent waves.
For Phase 2, Qualtrics screened the ‘soft launch’ data (n = 47) for technical errors and/omissions before the full launch and a survey completion time was again set based on half the median time for the soft launch (9 min, 42 s). Phase 2 ‘soft launch’ respondents were included in the main Phase 2 sample. Following the completion of Phase 2 fieldwork, Qualtrics removed any participants who (1) completed the survey in less than the minimum completion time or (2) were potentially duplicate respondents.
2.2 MeasuresTable 1 provides an overview of the C19PRC-UKW3 survey content by Phase (see Supplementary Materials for specific details of all measures administered).
TABLE 1. Overview of content of C19PRC Study Wave 3 (Phases 1 & 2), United Kingdom (UK), July–August 2020 Theme Content C19PRC wave 3 Phase 1 Phase 2 Demographics Age, gender, country of residence, marital status, economic activity, key/essential worker status, born in the UK†, grow up in the UK†, urbanicity†, level of education†, religion† X X†only Housing characteristics Living alone X X Number of adults living in household X X Number of children living in household X X Ages of children living in household X X Housing tenure - X Residential details (type of property; number of bedrooms; length at property) X X Household finances Estimated annual gross household income - X Change in monthly household income during pandemic X X Use of savings/increasing debt during pandemic X X Made saving due to pandemic X X Perceived future financial security X X Working hours Changes in working hours (self) X X Health conditions Existence of any major underlying health conditions–self - X Existence of any major underlying health conditions–immediate family member - X Currently pregnant–self (partner) X X Number of weeks pregnant, if applicable X X Currently pregnant–immediate family member X X Children in household Childcare for children in household during lockdown X X Use of childcare facilities/services X X COVID-19 Sourcing of information (newspapers, TV, radio, social media, Internet, etc.) - X Level of trust in information source - X Engaging in behaviour to reduce risk of contracting COVID-19 (e.g., wearing face mask) X X Engagement with lockdown restrictions X - Anxiety-level relating to COVID-19 X X Perceived individual risk contracting COVID-19 over next 6 months X X Experiences of self-isolation X X Experience of being infected with COVID-19 (including testing) - self X X Experience of having COVID-19 (feeling unwell, admitted to hospital) X X Knowing someone close (family member/friend) who has tested positive for COVID-19 X X Knowing someone close (family member/friend) who has tested died due to COVID-19 X X COVID-19 vaccine acceptability (self) X X COVID-19 vaccine acceptability (child) X X Preference for schools reopening X X Comfort engaging in activities (e.g. socialising, shopping, going to the gym etc.) X - Preference for pace of easing lockdown restriction X - Predicted course of the pandemic X X Living in a local lockdown area X - Concern about second coronavirus wave X X Support/opposition for restrictions in case of second wave X - Support/opposition for air bridges and quarantine X - Contact tracing: Knowledge and willingness to engage X - Perceived compliance with social distancing: Neighbourhood, country, UK X X Perceived compliance with health and safety guidance: Neighbourhood, country, UK X - Going on holiday/travel abroad X X Mental health Depression: Patient health questionnaire-9 (Kroenke et al., 2001) X X Anxiety: Generalized anxiety disorder scale-7 (Spitzer et al., 2006) X X Traumatic stress international trauma questionnaire (Cloitre et al., 2018) X X Paranoia: Persecution and deservedness scale (Melo et al., 2009) - X Somatic symptoms: Patient health questionnaire-15 (Kroenke et al., 2002) X X Self-harm, suicidal thoughts and suicide attempts X X Social anxiety: Mini social phobia inventory (mini-SPIN) (Connor et al., 2001) X - Autistic traits: Autism spectrum quotient (AQ-10) (Allison et al., 2012) X X Psychological factors Personality: Big-fiveiinventory-10 (Rammstedt & John, 2007) - X Loneliness: Loneliness scale (Hughes et al., 2004) X X Death anxiety: Death anxiety inventory (Tomás-Sábado et al., 2005) - X Locus of control: Locus of control scale (Sapp & Harrod, 1993) - X Self-esteem: Single-item self-esteem scale (Robins et al., 2001) X X Resilience: Brief resilience scale (Smith et al., 2008) - X Attachment style: Relationships questionnaire (Bartholomew & Horowitz, 1991) X X Hopefulness: Brief-H-positive scale (Fraser et al., 2014) X X Happiness: Subjective happiness scale (Lyubomirsky & Lepper, 1999) X X Life satisfaction X X Aspects of life better/worse since pandemic X - Social support: Modified medical outcome social support survey (mMOS-SS) (Ganz et al., 2003) X X Health-related behaviours Alcohol use: AUDIT-C (Bush et al., 1998) X X Height and weight X X Socio-political views/related behaviours Voting behaviour last general election X X Political party identification X X Voting behaviour European referendum - X Measure of ‘left-wing’ or ‘right-wing’ on social and economic issues - X Satisfaction with how government/institutions handling pandemic X - Child rearing views - X Experiences of discrimination (pre & during pandemic): Everyday Discrimination Scale (short-form) (Sternthal et al., 2011) X - Future voting behaviour X X Trust Institutions X X a Refer to Supplementary Material for detailed information on all study measures. † Variables indicates with this symbol were only administered at Phase 2. 2.2.1 Study variablesThe following C19PRC-UKW1 variables were used for attrition analyses for C19PRC-UKW3: gender (females vs. males); age (18–24 years olds vs. 25–34 years, 35–44 years, 45–54 years, 55–64 years, and 65+ years groups); household income (≤£15,490 per annum vs. £15,491–£25,340, £25,341–£38,740, £38,741–£57,903, and ≥£57,931 bands); ethnicity (White vs. other); education (post-secondary education vs. other); economic activity (employed vs. other); urbancity (living in city vs. suburb, town or rural location); household composition (living alone vs. other; children <18 years living in household vs. other); living in UK (born or raised before aged 16 years in UK vs. other); physical health (self-reported chronic health condition vs. other); probable MDD diagnosis (score of ≥10 on the Patient Health Questionnaire-9 vs. other); probable GAD diagnosis (score of ≥10 on the Generalised Anxiety Disorder-7 vs. other); probable PTSD diagnosis (using the International Trauma Questionnaire’s diagnostic algorithm for PTSD caseness vs. other); mental health treatment (current or past treatment for mental health problems vs. other); loneliness (score of ≥6 on the Loneliness Scale); somatisation (total score on the Patient Health Questionnaire-15); neuroticism (total score on the neuroticism subscale of the Big-Five Inventory-10); resilience (total score on the Brief Resilience Scale); paranoia (total score on the Persecution and Deservedness Scale); death anxiety (total score on the Death Anxiety Inventory); intolerance of uncertainty (total score on the Intolerance of Uncertainty Scale); and COVID-19 anxiety (total score on single item indicator).
2.3 Ethical approvalEthical approval for the project was provided by the University of Sheffield (Reference number 033759).
2.4 Data analysis plan and weighting proceduresData analyses were conducted in a number of stages. First, the re-contact rate for Phase 1 was calculated, and responders and non-responders were compared on a range of baseline socio-demographic, mental health, and psychological characteristics, using chi-square tests and independent samples t-tests. Second, a binary logistic regression analysis was conducted to assess the association between baseline characteristics and attrition at C19PRC-UKW3. Regression coefficients (odds ratios and 95% confidence intervals) were plotted using the coefplot in Stata 15 (Jann, 2017; StataCorp., 2017).
Third, post-stratification survey weighting was conducted for the Phase 1 sample using a technique known as survey raking or sample-balancing, using the ‘anesrake’ package in R (Pasek & Pasek, 2018). Raking is a common method of adjusting survey data to ensure that the distribution of the characteristics of a sample closely mirror the known population distribution. In practice, this means the C19PRC-UKW1 sampling quotas for age, gender, and household income, as well as the baseline proportions achieved for ethnicity, urbanicity, household composition, and being born or raised in the UK, were imposed on the sample obtained at Phase 1. The raking algorithm assessed which of these selected sociodemographic variable distributions at C19PRC-UKW3 deviated from their target distribution at C19PRC-UKW1 by 5% or more, and subsequently iteratively adjusted to produce a weight value for each case in the sample until the sample distribution aligned with the population distribution for the chosen characteristics (DeBell & Krosnick, 2009; Pasek & Pasek, 2018). Raking is considered an ideal method for weighting survey data given that it is relatively easy to implement, but also since it only requires the marginal population proportion for each variable used in the weighting procedure (Mercer et al., 2018). Weighted frequencies were calculated for baseline characteristics for C19PRC-UKW3 Phase 1 sample to assess the success of the raking procedure.
And fourth, the representativeness of the combined C19PRC-UKW3 Phase 1 and Phase 2 samples was assessed by comparing the characteristics of the sample to the UK general population. Standardised difference scores were computed using the stddiffi command in Stata 15 (Bayoumi, 2016; StataCorp., 2017) to test for differences in relation to specific socio-demographic characteristics between the two data sources. Unlike other statistical tests (e.g. chi-square), the standardised difference score approach is not influence by sample size (Austin, 2009), and can be more informative than p-values for comparing across data sources that differ in relation to sample size (Harron et al., 2017). Standardised differences of 0.2, 0.5, and 0.8 represent small, medium, and large standardised differences respectively (Cohen, 1988); standardised difference scores of less than 0.1 suggests no meaningful differences between data sources in relation to the distribution of the variable under consideration (Normand et al., 2001).
3 RESULTS 3.1 Retention of respondents from previous wavesAs illustrated in Figure 2, at Phase 1, 1211 adults who participated in one or both of the previous waves were successfully recontacted (59.8% recontact rate) and 1166 adults provided full interviews at C19PRC-UKW3 (i.e., 5
留言 (0)