Does pre-COVID impulsive behaviour predict adherence to hygiene and social distancing measures in youths following the COVID-19 pandemic onset? Evidence from a South African longitudinal study.

Background: Engagement in protective behaviours relating to the COVID-19 pandemic has been proposed to be key to infection control. This is particularly the case for youths as key drivers of infections. A range of factors influencing adherence have been identified, including impulsivity and risk taking. We assessed the association between pre-COVID impulsivity levels and engagement in preventative measures during the COVID-19 pandemic in a longitudinal South African sample, in order to inform future pandemic planning. Methods: Data were collected from N = 214 youths (mean age at baseline: M = 17.81 (SD =.71), 55.6% female) living in a South African peri-urban settlement characterised by high poverty and deprivation. Baseline assessments were taken in 2018/19 and the COVID follow-up was conducted in June–October 2020 via remote data collection. Impulsivity was assessed using the Balloon Analogue Task (BART), while hygiene and social distancing behaviours were captured through self-report. Stepwise hierarchical regression analyses were performed to estimate effects of impulsivity on measure adherence. Results: Self-rated engagement in hygiene behaviours was high (67.1–86.1% “most of the time”, except for “coughing/sneezing into one’s elbow” at 33.3%), while engagement in social distancing behaviours varied (22.4–57.8% “most of the time”). Higher impulsivity predicted lower levels of hygiene (β =.14, p =.041) but not social distancing behaviours (β = −.02, p =.82). This association was retained when controlling for a range of demographic and COVID-related factors (β =.14, p =.047) and was slightly reduced when including the effects of a life-skills interventions on hygiene behaviour (β = −.13, p =.073). Conclusions: Our data indicate that impulsivity may predict adolescent engagement in hygiene behaviours post COVID-19 pandemic onset in a high risk, sub-Saharan African setting, albeit with a small effect size. For future pandemics, it is important to understand predictors of engagement, particularly in the context of adversity, where adherence may be challenging. Limitations include a small sample size and potential measure shortcomings. The sample was drawn from a longitudinal intervention study conducted in peri-urban Khayelitsha, South Africa, which followed children and their families from before birth until current age (19–21 years; see Fig. 1 for a CONSORT flowchart of assessments). From the antenatal period until 6 months after birth, expectant mothers received either a parenting intervention (‘Thula Sana’; n = 220) aimed at improving parenting skills and attachment, or maternal services as usual (control group, n = 229). All mothers in the community who were eligible for study participation were invited, and group assignment was randomized. Families were followed up several times over the first 18 months of the child’s life [26] and again at 13 years of child age [27]. No effects of the early intervention on adolescent outcomes were identified [27] and we subsequently do not control for receipt of this intervention in the current study. The youths then underwent a second intervention (‘Zifune’; for details, see below), aimed at teaching life skills to improve pro-sociality and reduce violence behaviours, at ages 16–19 years (n = 319; re-randomized based on early intervention group allocation) (Skeen S, Du Toit S, Marlow M, Stewart J, Rabie S, Melendez-Torrez GJ, et al: Zifune: Does a second wave intervention delivered to former recipients of an early mother-infant attachment intervention reduce interpersonal violence during adolescence? A re-randomized controlled trial, in preparation). Data collection took place in 2018/2019, at three time-points: pre-intervention, directly post-interventions (n = 314) and at a 3-month follow-up (n = 307). At the post-intervention assessment, participants completed several behavioural tasks to investigate whether the intervention had led to any changes in risk taking and moral behaviours (n = 280). CONSORT Flow Diagram for Cohort Studies The Zifune life skills intervention was developed for use with youths living in high adversity contexts in low- and middle-income countries (LMIC) (Skeen S, Du Toit S, Marlow M, Stewart J, Rabie S, Melendez-Torrez GJ, et al: Zifune: Does a second wave intervention delivered to former recipients of an early mother-infant attachment intervention reduce interpersonal violence during adolescence? A re-randomized controlled trial, in preparation). An adolescent advisory board provided feedback to ensure applicability and acceptability of its contents. The intervention utilises a collaborative approach, incorporates principles of cognitive behaviour therapy, and employs creative and fun methods to allow youths to reflect on their relationships and behaviours and to devise future plans. Eight group-based sessions for groups of approximately 20 youths each were provided by trained facilitators from the local community. Sessions covered six main themes: vision for the future, time management, financial planning, mindfulness, risk-taking behaviour and interpersonal violence, with sessions about long-term planning and risk-taking behaviours in particular potentially affecting impulsivity levels. An intervention facilitator remained in regular contact with and provided support to the youths throughout the course of the study via phone calls. Following the outbreak of the COVID-19 pandemic, South Africa went into a strict lockdown in March 2020. Brief telephonic interviews were conducted with participants (n = 237) in June to October 2020 through remote-working data collectors. During this time, South Africa’s first large case wave took place (July–August 2020), followed by a strong decline in cases. Participants were assessed on a range of COVID-related variables, including social distancing and hygiene behaviours, household food security, mental health, and schooling outcomes. We utilize data from those who took part in the behavioural tasks at the post-intervention assessment of the Zifune study and completed the COVID-related questionnaire after the pandemic outbreak (n = 214). All participants provided written consent at each wave of the data collection. Assessments were conducted in the participants’ language of choice, predominantly isiXhosa. All data were collected by trained and supervised data collectors, with at least a high school diploma and with prior experience in working with vulnerable populations. For the current phase of the study, ethical approval was obtained from the Health Research Ethics Committee (HREC) from Stellenbosch University (Ref: N17/10/094). Information on the gender, age, level of education (utilised in the form of a “correct grade for age” variable), housing (formal vs informal housing); number of household members the individual was living with during the COVID-19 pandemic, HIV status, and household receipt of any form of government-provided cash grants was collected. The BART [28] is a naturalistic computer task measuring impulsive and risk-taking behaviours. Participants are presented with a balloon, which they can enlarge in a step-wise fashion by pressing a button. Each pump increases the reward pay-off that the participant receives, but also the chance of the balloon popping, which leads to no rewards for the trial. Participants have the choice to step away after each button press, and collect the already accrued rewards for the trial, or to keep pumping. In the current study, all participants were asked to complete 30 trials. They were told that one trial would be chosen at random in the end, for which they would receive the earned monetary reward. To even out expectations, all participants observed 12 balloons being inflated to their bursting point before commencing the task. The bursting point was set to be identical in each trial between participants. The overall number of pumps was used as a predictor of interest, with a higher number of pumps reflecting higher risk taking. The extent to which participants engaged in each of four hygiene behaviours (hand washing, hand sanitising, coughing/sneezing into one’s elbow, and wearing a face mask) during the past week was measured on a scale from 0 “never” to 3 “most of the time” for each item (see Additional file 1: Appendix 1 for full item list and rating scale). A total score (0–12) was calculated. An exploratory factor analysis revealed that the items did not load well onto a single underlying factor, potentially due to participants picking and choosing certain behaviours or adhering less stringently to measures as the pandemic situation in South Africa relaxed towards September/October 2020. As a result, we decided to investigate the total score, reflecting the overall extent of hygiene behaviours each participant engaged in, but also analysed the four behaviours separately to see whether any effects found were driven by high scores on particular items. The extent to which participants engaged in five social distancing practices during the past week was assessed: keeping a 1–2 m distance, and avoiding public transport, going to the shops/pharmacy, public spaces and going for a walk in the neighbourhood. Items were rated from 0 “never” to 3 “most of the time” and summed up into a total score (0–15). Exploratory factor analysis suggested that the latter three items loaded onto a potential “avoidance of public outings” factor, though individual item loadings were small. Therefore, we chose to investigate the total score, indexing the extent of overall social distancing behaviours, and to additionally explore single-item effects. We added age and sex to the analyses, since risk behaviours in the BART have been shown to be influenced by both factors. We furthermore controlled for education (being in the correct grade for age) and timing of the assessment, since the COVID situation changed substantially in South Africa throughout our data collection, from the first case wave in June/July 2020 to level 1 restrictions in September 2020. In terms of COVID-related factors that could have influenced participants’ abilities to engage in hygiene and social distancing behaviours, we adjusted our analyses for household food security as a measure of deprivation (Household Food Insecurity Access Scale (HFIAS, [29]), and the number of individuals living in the participant’s household, which could have desensitized participants to being around large groups of people, or heightened worries and subsequent measure engagement, especially in multi-generational households. Finally, we controlled for receipt of the life skill intervention at ages 16–19 years, as it was found to influence risk taking in males particularly (Mikus N, Skeen, S, Stewart J, Marlow M, DuToit S, Rabie S, Mendelez Torres GJ, et al: Psychosocial intervention improved self-control in adolescents, in preparation). Analyses were conducted using StataSE 16 and R 4.1.1 In a first step, we investigated descriptive characteristics of the sample and compared it to participants who had completed the BART impulsivity measure and were not included in the COVID follow-up on relevant demographic factors, using t-tests and χ2 tests as appropriate. We then performed Pearson’s correlation analyses between the key variables. Finally, based on findings from the correlation analyses, a hierarchical linear regression analysis was performed, with hygiene behaviours as the key outcome. In the first step, impulsivity was added as a predictor, with higher pumps on the BART indexing higher impulsivity/ risk taking. Secondly, the demographic factors of age, sex and correct class for age were included. In a third step, COVID-related factors (food security, number of people living in the household, time to level 1 restrictions) were added to the model. In a last step, receipt of the life skills intervention was added, to see whether any effects found may be explained by exposure to its contents. Finally, since the intervention was found to affect BART-measured impulsivity in a previous study (Mikus N, Skeen, S, Stewart J, Marlow M, DuToit S, Rabie S, Mendelez Torres GJ, et al: Psychosocial intervention improved self-control in adolescents, in preparation) and showed close to significant predictions of hygiene behaviours (β = .09, p = .190) in the current study, we conducted secondary exploratory causal mediation analyses, using the “mediation” package in R 4.1.1 [30]. The aim was to investigate whether the life skills intervention may be able to buffer potential associations between higher impulsivity and lower protective behaviour engagement. For this, we explored whether any indirect effects of exposure to the life skills intervention on hygiene behaviours through impulsivity would be found. However, we acknowledge limited power due to a small sample size.