Sleep quality is associated with emotion experience and adaptive regulation of positive emotion: An experience sampling study

1 INTRODUCTION

Sleep loss and poor sleep quality disrupt how the brain processes emotions (Walker & van der Helm, 2009). Much of the evidence on the impact of sleep loss on emotion, be it the processing of emotions, the recognition of emotional stimuli, or the capacity to regulate emotion, is derived from studies of full or partial sleep restriction (for review, see Beattie et al., 2015). Experimentally reducing sleep to a maximum of 2 h total has been associated with decreased positive affect, measured using the Positive and Negative Affect Schedule (PANAS), in adolescents and adults (Talbot et al., 2010). Even partial sleep restriction, such as reducing sleep by 2 h per night for 3 nights, has been associated with linear reductions in positive affect across study days (Saksvik-Lehouillier et al., 2020). Recent meta-analytic evidence from seven unique studies suggests that experimentally induced sleep loss has a modest, but significant, negative effect on emotion ratings (Hedge's g = −0.11). Combining studies in which participants were presented with positive and negative emotion stimuli (e.g., videoclips, or IAPS images), the authors reported no moderating effects of stimulus emotion (Tomaso et al., 2021).

While experimentally limiting sleep has disruptive effects on self-reported mood and emotion ratings (e.g., Haack & Mullington, 2005), less is known about the affective experiences associated with natural fluctuations in sleep patterns, particularly in young people. In adult samples (18–61 years), poorer self-reported sleep, both duration and quality, has been associated with reduced positive and increased negative self-reported emotion (de Wild-Hartmann et al., 2013b), and there is some evidence for similar effects in 13- to 16-year-olds (van Zundert et al., 2015). A 14-day diary study of 30 adults aged between 20 and 59 found that self-reported sleep quality was a small, but significant predictor of more positive next-day mood. Sleep quality was, in fact, the best predictor of mood from a range of additional sleep variables (e.g., awakenings, timing; Totterdell et al., 1994). However, as noted in a recent systematic review assessing the association between positive affect and sleep, the majority of studies have been cross-sectional, and have methodological challenges such as inadequate measurement of negative affect, or small heterogenous sample sizes (Ong et al., 2017).

1.1 What about emotion regulation?

Another issue in the burgeoning field of emotion-based sleep research is that most studies have focused on the experience of affect that occurs after sleep loss, rather than on the regulatory processes that may alter emotional experiences (Palmer & Alfano, 2017). Emotion regulation impairments are central to clinical models of anxiety and depression pathogenesis (Hofmann et al., 2012), and the role of disrupted negative emotion regulation is especially emphasised. As for studies examining emotion experience, the majority have used experimental reductions in sleep and then assessed negative emotion regulation at a single time point (e.g., Reddy et al., 2017; Tamm et al., 2019; Zhang et al., 2019).

Emotion regulation strategies can be broadly categorised as “adaptive”, when associated with long-term beneficial outcomes for mental wellbeing, or “maladaptive”, when associated with long-term negative outcomes (Schäfer et al., 2017). The habitual use of negative emotion regulation strategies has been associated with the likelihood of developing psychopathology in several meta-analyses. Among commonly assessed strategies for negative emotions, reappraisal can be considered an “adaptive” strategy (associated with reduced likelihood of psychopathology), whereas suppression is considered a “maladaptive” strategy (associated with increased likelihood of psychopathology; Aldao et al., 2010; Schäfer et al., 2017). Although investigated less, discussions of positive emotion regulation differentiate between strategies that either “enhance” or “dampen” positive emotions (Gilbert, 2012; Young et al., 2019), which are considered adaptive and maladaptive, respectively (Feldman et al., 2008). One recent study restricting sleep to 3 h in young adults found reduced self-reported emotion regulation success, specifically cognitive reappraisal, to negative stimuli (Tamm et al., 2019). In a study of partial sleep reduction to 6.5 h over 5 nights, Baum et al. (2014) reported emotion reaction difficulties in adolescents, as indicated by self-reported “easily upset” and unprovoked or disproportionate emotional reactions (interpreted as emotion regulation difficulties), relative to after 5 nights of typical sleep. Again, with a focus on negative emotion, an analysis of university students reported that more self-reported sleep difficulties at baseline were associated with reduced self-reported regulation effectiveness one year later (Tavernier & Willoughby, 2015). However, emotion regulation is not only dependent on the ability to manage responses to negative emotions, or negative feelings. Being able to notice, savour and reflect on positive emotions, is also important for affective functioning and mood, and may plausibly be disrupted with poor sleep, as for positive mood.

In the present study, we aimed to address the gaps in knowledge around the role of sleep in positive emotion regulation, while simultaneously recording negative affect and negative emotion regulation. We focused on young adults (aged 18–24 years) because late adolescence to early adulthood is a developmental period where there are changes in emotion regulation capacities (Young et al., 2019), and striking shifts in sleep patterns (Roenneberg et al., 2004). Young adults are estimated to have a greater sleep need compared with older adults (Short et al., 2018), and from the onset of puberty, show a preference for a delayed sleep onset timing, leading to the characteristic “owl”-like behaviour of adolescents that persists into the early 20s (Skeldon et al., 2016). Furthermore, younger adults face different daily affective challenges compared with adults, such as the pursuit of autonomy (Weinstein & Mermelstein, 2007), and the neural architecture to support emotion regulation is still maturing in this period (Blakemore & Choudhury, 2006; Guyer et al., 2016). Finally, adolescence and young adulthood is the developmental window where mood disorders most typically emerge, underscoring the importance of this period for understanding sleep and emotion regulation (Paus et al., 2008).

We used experience sampling to obtain more frequent measurements of young adults' emotional experiences and regulation strategies than afforded by traditional questionnaire measures. Experience sampling also allowed us to measure naturally occurring fluctuations in participants' sleep in their normal environments. We expected that higher ratings of sleep quality and longer sleep duration would be associated with increased levels of positive affect, and decreased negative affect, consistent with previous studies (de Wild-Hartmann et al., 2013b; Reddy et al., 2017). We also expected that higher ratings of sleep quality and longer sleep would be associated with increased use of adaptive regulation of positive and negative emotions and decreased use of maladaptive strategies, in line with previous studies of negative emotion regulation (Mauss et al., 2013; Palmer et al., 2018).

2 METHODS 2.1 Participants

Participants were recruited using a research volunteer system at the Institute of Psychiatry, Psychology and Neuroscience, King's College London, which includes both university students and volunteers signing up to the database who are not university students. We also used additional convenience sampling methods, via internal departmental webpages, and social media. Inclusion criteria for participation were: aged 18–24 years, access to an internet-enabled smartphone, self-reported as English speaking, and UK residence. Compensation for participation was monetary vouchers (£8), with an additional £2 reward for completion of over 70% of EMA surveys (67% of participants completed >70% of surveys). Ethical approval was obtained via the Research Ethics Minimal Risk Self-Registration protocol (Daily Emotional Experiences Study, MRA-19/20-19272, Psychiatry, Nursing and Midwifery Research Ethics Subcommittee, King's College London). Participants gave informed consent online via Qualtrics (Qualtrics International Inc.) prior to completing the baseline assessment.

We aimed to collect data from at least 100 participants, based on the sample size reported in a previous daily diary study (N = 98), taking measures of emotion intensity and duration over the course of 1 week (Verduyn & Brans, 2012). One hundred and forty individuals completed the baseline questionnaire, of whom 15 were excluded for not meeting the eligibility criteria (wrong age, n = 10; incorrect email, n = 2; not resident in the UK, n = 3). Of the remaining 125 participants at baseline, 115 individuals registered for the EMA component of the study. Eight of these participants were later excluded because their residence was subsequently reported as outside the UK, and six participants provided insufficient data, resulting in a final sample of 101 individuals.

2.2 Study design

Figure 1 presents the study design, which consisted of a baseline assessment and a longitudinal EMA component, carried out using the MetricWire (MetricWire Inc.) smartphone application.

image

Study design. At baseline, participants completed the Emotion Regulation Questionnaire (ERQ), the Responses to Positive Affect questionnaire (RPA), and the Insomnia Severity Index (ISI). The experience sampling component of the study consisted of daily measures of emotion intensity, emotion duration, and emotion regulation strategy use (middle panel). For the emotion regulation strategies (middle panel), adaptive strategies are in blue, maladaptive in red, and neutral in purple. The first survey of each day contained both the sleep assessment and emotion regulation questions, and the other three surveys (final panel) measured emotion levels and regulation only (created using Biorender)

2.3 Baseline measures

At baseline, participants completed sleep and emotion-related measures via Qualtrics (see Table 1 for psychometric properties of each scale). Demographic data were also obtained at baseline (gender, age, and education level).

TABLE 1. Descriptive statistics for baseline measures of sleep and trait-level emotion regulation (N = 101) Mean (SD) Scale range Sample range Cronbach's α ISI 9.40 (5.46) 0–28 0–24 0.86 RPA: emotion-focused 12.65 (2.85) 5–20 5–20 0.73 RPA: self-focused 8.22 (2.69) 4–16 4–16 0.80 RPA: dampening 18.61 (4.93) 8–32 9–29 0.82 ERQ: cognitive reappraisal 27.40 (6.31) 6–42 11–38 0.84 ERQ: expressive suppression 14.75 (5.62) 4–28 4–25 0.84 Abbreviation: ERQ, emotion regulation questionnaire; ISI, insomnia severity index; RPA, responses to positive affect scale. 2.3.1 Sleep

The Insomnia Severity Index (ISI; Morin et al., 2011) is a 7-item instrument, widely used to measure the severity of sleep difficulties and to assess the impact these have on an individual's everyday functioning. Participants responded to seven questions using a 5-point Likert scale (0 = no issues, 4 = very severe issues).

2.3.2 Emotion regulation

Positive emotion regulation was assessed using the Responses to Positive Affect questionnaire (RPA; Feldman et al., 2008), which comprises 17 items evaluating the degree to which participants habitually engage in three different types of positive emotion regulation: (i) adaptive emotion-focused regulation (5 items); (ii) adaptive self-focused regulation (4 items); (iii) and maladaptive dampening (8 items). Each of these regulation types are reported on a 4-point Likert scale (1 = almost never, 4 = almost always) and as separate subscales within the RPA.

Negative emotion regulation was assessed using the 10-item Emotion Regulation Questionnaire (ERQ, Gross & John, 2003) which evaluates individual differences in the habitual use of two emotion regulation strategies: (i) adaptive cognitive reappraisal and (ii) maladaptive expressive suppression. Each of these regulation types are reported on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree).

2.4 EMA measures

Questionnaire items administered during the daily EMA surveys were: (i) one item on sleep quality from the Consensus Sleep Diary (Carney et al., 2012), a widely used self-report instrument designed for daily sleep recording, and two items asking about sleep schedule (sleep time, awakening time); (ii) five items assessing positive emotion regulation adapted from the RPA and one novel item (“I have been thinking that I deserve these feelings”); and (iii) six items assessing negative emotion regulation, including four items adapted from the ERQ, assessing cognitive reappraisal (2 items) and suppression (2 items) and two additional items to measure distraction (“I have been trying to feel less negative by doing something unrelated”; “I have been trying to feel less negative by thinking about something unrelated”), where the first item measured activity-related distraction, derived from Stone et al. (2019), and the second item measured thought-related distraction (see Figure 2, Table S1). The RPA and ERQ items were modified so that the present perfect tense was used instead of the present tense or the imperative. Participants could therefore answer the instruction, “think about your experiences in the last few hours or since the last survey” (see Table S1 for all the items). However, this modification, along with the other minor language simplifications detailed in Table S1 should be considered as non-validated changes.

image

Smartphone EMA measurement: (a) Assessing emotion intensity using a sliding visual analogue scale (VAS; prompts asked “how positive” or “how negative”). (b) Assessing emotion duration using a sliding VAS. (c–d) Assessing strategies used for regulating emotions using a tick box format, for both positive (c) and negative (d) emotion regulation strategies

2.4.1 Sleep parameters

Measures of sleep quality and duration were obtained each morning (10 am–12 pm) with the items (i) “How would you rate the quality of your sleep?” (5-point Likert scale, 1 = very poor, 5 = very good) and (ii) “What time did you go to sleep?” and “What time did you wake up?”, with duration calculated from the participant-estimated sleep and wake times.

2.4.2 Emotion and emotion regulation

Positive and negative emotion intensity were assessed using the two ratings: “How positive have you been feeling?” and “How negative have you been feeling?” on a sliding scale between 0–100. Positive and negative emotion duration were assessed using the item “How long has this feeling been going on?”, on a sliding scale ranging from “Seconds” to “Hours”, for ease of participant interpretability. We then converted duration-based responding (seconds-hours) to a score between 0–100, noting that this mapping is conceptually imperfect. The order of the positive and negative items was the same across days and participants. Regardless of the answer provided to the emotion intensity or duration question, the use of regulation strategies was then assessed using the question, "What types of thoughts have you been having about these feelings?" for each emotion type. Participants were able to select as many strategies as applicable (see Figure 2).

2.5 Procedure

After completion of the initial baseline measures, the participants were contacted by a researcher by email and familiarised with the EMA procedures. After registration within the MetricWire application, participants received four questionnaires each day for 7 consecutive days (28 questionnaires in total, see Figure 1). A notification for the first survey of each day was sent to participants at a random timepoint between 10 am and 12 pm, while notifications for the following three surveys were sent at pseudo-random times (at least 2 h apart) between 1 pm and 10 pm. Links to surveys expired after 20 min and were recorded as missed if not returned within this period. A threshold of 20% of completed surveys was required in order for data to be incorporated into subsequent analyses, as per recommendations for EMA procedures to ensure sufficient power (Edwards et al., 2016). Finally, to encourage engagement over the duration of the experiment, participants whose responses dropped below 40% by the third day of surveys were contacted by email and reminded of the participation bonus.

2.6 Statistical analysis

Data processing and analysis was carried out in RStudio version 4.0.2 (The R Development Core Team, 2020). Data collected within the EMA portion of the study were nested in three levels: observations (i.e., surveys), within days (note sleep parameters were measured just once per day), within participants. Multilevel modelling was used because it can accommodate the hierarchical structure of EMA data. All R code is available from: https://osf.io/urjsf/.

For each of the outcomes (EMA measures of emotion intensity, emotion duration, and emotion regulation strategies), three-level, random-intercept models were created, with each of the day-level sleep parameters (sleep duration, sleep quality) added as predictors (10 models in total, 4 for emotion intensity and duration, 6 for regulation strategies). We used baseline participant-level ISI scores, RPA scores, and ERQ scores as covariates. “Participant” and “Days within participant” were added as random intercepts to account for between-person and between-day differences, respectively, throughout these analyses, as per previous three-level EMA procedures (Geyer et al., 2018). For multilevel models, we used the lme4 package within R, and an optimisation by quadratic approximation (BOBYQA) with a set maximum of 20,000 iterations. Missing data were handled using listwise deletion for individual assessments (i.e., a missing item on a single ESM questionnaire resulted in removal of that assessment point for that individual, see Table S2 for details on rates of missing data).

We first tested whether the daily EMA measures of sleep (duration and quality) were associated with emotion intensity (positive or negative; 0–100 scales) and emotion duration (positive or negative; 0–100 scales). Next, we tested whether sleep (duration, quality) was associated with emotion regulation (use of strategies coded as binary variables: 0 = not used, 1 = used either/both).

Unlike models with continuous outcomes, which represent expected effects at both the participant and sample level, the results of multilevel models with binary outcome variables are participant-specific only (Hox et al., 2010). Therefore, sample-level trends were calculated in the form of predicted probabilities as per recommendations for logistic models (Persoskie & Ferrer, 2017). All analyses were conducted using maximum likelihood estimates (McCulloch, 2003). All predictors were centred, with time-varying predictors centred using the individual's mean and time-invariant predictors using the grand mean (Snijders & Bosker, 2012). Sleep duration was assessed as a linear predictor in all models, consistent with previous EMA designs (e.g., Littlewood et al., 2019).

Outliers were calculated by generating weighted averages (using the number of surveys returned by each participant) for each variable. Outliers were defined as where the participant's average exceeded IQR +1.5 or fell below IQR −1.5. We repeated analyses with and without outliers and found similar patterns of significant effects (see Tables S3 and S4). We also conducted sensitivity analyses removing participants with ISI scores in the moderate to severe range (>14), to test whether the effects were driven by these higher-symptom individuals (n = 19). As for analyses removing outliers, we found largely comparable patterns of results when excluding these 19 individuals (see Tables S5 and S6).

3 RESULTS 3.1 Participant characteristics

One hundred and one individuals (85 female, 16 male) took part in this study, with a mean age of 21.69 years (SD = 1.91; range 18–24 years). Approximately half had completed secondary school level education (45.54%), and half had obtained higher education qualifications (bachelor's degree: 37.62%, master's degree: 14.85%). Descriptive statistics for baseline assessments of sleep and trait-level emotion regulation tendencies are shown in Table 1.

The mean experience sampling survey completion rate was 72.84% (SD = 17.54). The mean sleep quality score was 3.46 (SD = 0.95; median = 4, range = 1–5, IQR = 3–4) and the mean sleep duration was 8.01 h (SD = 0.99; median = 8.06 h, range = 4.00–10.50 h, IQR = 7.61–8.57 h). The mean ISI scores reported (Table 1) were comparable to those reported in a recent large student sample (mean ISI score = 8.99; Walsh et al., 2020). Most participants' ISI scores fell below the clinical cut-off for moderately severe, or severe insomnia symptoms (absence of insomnia, <8, n = 44; sub-threshold insomnia, 8–14, n = 38; moderate insomnia, 15–21, n = 16; severe insomnia, 22–28, n = 3). Mean emotion and regulation ratings are displayed in Figure 3.

image

Mean ratings of reported emotion intensity and duration (a) and regulation strategy use (b and c). Error bars indicate mean ± standard deviation, “+” in panel (a) denotes outlier values. Note that individuals were able to select more than one regulation strategy at each EMA prompt (therefore total across strategy usage >100%)

For positive regulation strategies, the most frequently used was emotion-focused strategies (57.69% of all reports). Self-focused emotion regulation (19.94%) and dampening (11.66%) were less frequently used. For negative emotion regulation, the frequency of reported use was similar for all three regulation strategies (reappraisal: 29.32%, suppression: 23.66%, distraction: 39.08%). Further descriptive statistics for the experience sampling emotion regulation ratings are presented in Table 2. Within and between-person standard deviations indicate that strategy use varied at least as much within individuals as between individuals. Intra-class correlations show that 12%–47% of the variance within each individual strategy was accounted for by between-person variation.

TABLE 2. Descriptive statistics for the experience sampling Emotion Regulation (ER) strategies ER strategy Item Ma 95% CI lower 95% CI upper SD between-person SD within-person residual ICC Emotion-focus 1 0.43 0.38 0.48 0.25 0.29 0.42 2 0.33 0.28 0.38 0.22 0.28 0.38 Self-focus 1 0.14 0.10 0.17 0.15 0.21 0.34 2 0.11 0.08 0.14 0.13 0.18 0.32 Dampening 1 0.08 0.06 0.10 0.06 0.17 0.12 2 0.05 0.03 0.07 0.08 0.13 0.29 Reappraisal 1 0.16 0.12 0.20 0.18 0.22 0.40 2 0.21 0.16 0.26 0.22 0.23 0.47 Suppression 1 0.18 0.15 0.21 0.14 0.24 0.25 2 0.12 0.10 0.15 0.10 0.21 0.20 Distraction 1 0.20 0.16 0.24 0.19 0.23 0.41 2 0.32 0.27 0.37 0.25 0.27 0.47 Abbreviation: ICC: intra-class correlations at the subject level, denoting the amount of variance attributable to between-person effects. a Mean proportion of strategy use per day, averaged across participants. 3.2 How are baseline trait measures and EMA measures related?

Bivariate correlations between baseline and EMA measures are presented in Figure 4. Overall, trait and EMA measures of the same constructs showed significant positive correlations. Baseline insomnia severity was significantly associated with EMA-measured sleep quality (r = −0.34), but not sleep duration (r = −0.13). Baseline trait emotion regulation strategy use was significantly associated with EMA measured use of the same strategy for both positive emotions (emotion-focused: r = 0.32; self-focused: r = 0.33; dampening: r = 0.28) and negative emotions (reappraisal: r = 0.24; suppression: r = 0.27; see Supplementary Materials for full description of these findings).

image

Heatmaps of bivariate correlations between sleep (ISI and EMA-reported Sleep quality, Sleep duration), emotion (EMA-reported intensity, duration) and state (EMA) and trait emotion regulation measures (RPA, ERQ) presented separately for positive and negative emotion and regulation strategies (N = 101). [* denotes significance at p < 0.05; EMA: ecological momentary assessment, ER: emotion regulation, ERQ: emotion regulation questionnaire, ISI: insomnia severity index, RPA: responses to positive affect scale]

3.3 Effects of sleep on positive and negative emotion intensity and duration

Four multilevel models examined the effect of daily variation in self-reported sleep quality and duration on daily levels of: (i) positive emotion intensity, (ii) positive emotion duration, (iii) negative emotion intensity, and (iv) negative emotion duration.

3.3.1 Positive emotion: experience and duration

For positive emotion intensity, there was a small, but statistically significant effect of prior night's sleep quality. Higher quality sleep was associated with higher daily ratings of positive emotion intensity (Table 3; Table S2 for results of models including outliers). In this model, the ISI and two subscales of the RPA (emotion-focused and self-focused positive emotion regulation) were also statistically significant with small effect sizes: lower levels of insomnia symptoms and greater trait use of both emotion-focused and self-focused regulation were associated with higher positive emotion. There was no statistically significant effect of sleep duration or the RPA dampening subscale.

TABLE 3. Results from multilevel models examining positive emotions and positive emotion regulation (outliers removed) Model variables β SE Lower CI Upper CI p ES Positive emotion intensity Sleep quality 6.16 1.18 3.85 8.47 <0.001 0.03 Sleep duration 1.19 0.81 −0.39 2.78 0.14 0.00 ISI −0.64 0.25 −1.13 −0.15 0.01 0.02 RPA-EF 1.04 0.47 0.12 1.95 0.03 0.02 RPA-SF 1.10 0.49 0.14 2.07 0.03 0.02 RPA-DP 0.23 0.28 −0.32 0.78 0.41 0.00 Positive emotion duration Sleep quality 2.52 1.04 0.48 4.56 0.02 0.01 Sleep duration 0.14 0.71 −1.25 1.53 0.84 0.00 ISI −0.47 0.28 −1.03 0.08 0.10 0.01 RPA-EF 1.19 0.53 0.14 2.24 0.03 0.02 RPA-SF 0.52 0.56 −0.59 1.62 0.36 0.00 RPA-DP −0.13 0.32 −0.76 0.50 0.68 0.00 Emotion-focus Sleep quality 0.32 0.14

留言 (0)

沒有登入
gif