Background: Given nurses often face sleep problems, this study compares two internet-delivered Acceptance and Commitment Therapy (ACT) interventions to improve sleep quality (SQ), psychological flexibility (PF), reduce rumination, and explore neural mechanisms. Methods: 477 nurses were randomly assigned to ACT linear psychotherapy model (LINEAR), ACT loop psychotherapy model (LOOP) and wait-list group. SQ, rumination, and PF were assessed with questionnaires. Prefrontal cortical activation changes were measured using functional near-infrared spectroscopy. Results: The linear mixed-effects model demonstrated significant improvements in SQ, PF, and reduced rumination compared to pre-intervention for both models through enhanced psychological flexibility. LOOP showed a significantly superior effect compared to LINEAR. DLPFC activation increased following both interventions, with LOOP additionally stimulating the frontopolar area. Changes in the DLPFC mediated the relationship between intervention and outcome improvements. Frontopolar changes mediated SQ improvements but not rumination or PF. No significant changes in functional connectivity were observed during the verbal fluency task. Conclusions: Both interventions improved outcome variables, with LOOP being notably more effective, offering a novel approach. Mediation analyses highlight the role of DLPFC activation in understanding ACT's mechanisms and targeting insomnia treatment, while the mechanisms of LOOP's superior effect warrant further research. Trial Registration: Chinese Clinical Trial Registry (ChiCTR2200063533). https://www.chictr.org.cn/
Nurses, the largest healthcare group, are crucial in supporting doctors during treatment and recovery (Althobaity & Alshammari, 2020). However, factors like rising infection rates, heavy workloads, inadequate protective gear, demanding shifts, and bed shortages increase the risk of sleep problems among nurses (Firew et al., 2020; Zhang et al., 2021), with 42.9 % reporting poor sleep quality (Han et al., 2016). Prolonged sleep deprivation can lead to cognitive impairments and other serious issues (Mieda & Sakurai, 2013; Rosado et al., 2015). Therefore, improving nurses’ sleep is essential.
Rumination is central to widely accepted insomnia models (Harvey, 2002; Morin, 1993; Spielman et al., 1987) and the transdiagnostic model of insomnia (Harvey et al., 2005). According to the response styles theory (Nolen-Hoeksema, 1991), dwelling on stressful events and their consequences (i.e., rumination) prolongs and intensifies distress, leading to hyperarousal and difficulty sleeping (Brosschot et al., 2006; Guastella & Moulds, 2007). Rumination not only predicts insomnia, impaired subjective and objective sleep quality (Frøjd et al., 2022; Pillai et al., 2014; Winzeler et al., 2014), but also serves as a significant factor in perpetuating insomnia (Ballesio et al., 2021). Therefore, addressing rumination holds promise for improving sleep among nurses (Kurebayashi, 2020).
While cognitive and behavioral therapy for insomnia (CBT-I) is empirically supported, it faces adherence challenges, including withdrawal due to difficulties with stimulus control and sleep restriction techniques (Bouchard et al., 2003). Some patients do not experience significant improvements or continue to have sleep disturbances post CBT-I (Geiger-Brown et al., 2015; Ong & Smith, 2017). Acceptance and Commitment Therapy (ACT) emerges as a promising alternative to CBT-I (Martin et al., 2023; Shin et al., 2023), potentially addressing adherence issues and premature discontinuation (Martin et al., 2023; Saldaña et al., 2023), and improving various aspects of sleep (Paivi et al., 2019). According to the sleep effort model, excessive control and effort to sleep can hinder the natural sleep process (Broomfield & Espie, 2005). Unlike CBT-I, ACT prioritizes acceptance over modifying thoughts (Salari et al., 2020), promoting acceptance of insomnia and reducing excessive control (Shin et al., 2023). This approach can alleviate symptoms, increase sleepiness (Dalrymple et al., 2010), and demonstrate sustained intervention effectiveness (Vendela et al., 2018). Additionally, enhancing psychological flexibility (PF), the primary goal of ACT (Hayes et al., 2006, 2012), is significantly associated with improved sleep outcomes post-ACT (Daly-Eichenhardt et al., 2016; Zakiei et al., 2021) and better sleep quality (Wang et al., 2023). Given that nurses reportedly have lower PF (Duarte & Pinto-Gouveia, 2017; Wang et al., 2023), ACT's emphasis on enhancing this trait is particularly relevant. Therefore, ACT may offer a useful addition to CBT-I. Moreover, this study also targets reducing rumination. ACT has demonstrated promise in reducing both rumination (Kocovski et al., 2009) and repetitive thinking (Ruiz et al., 2018), emphasizing heightened awareness and present-focused mindfulness to foster acceptance, which effectively reduces rumination (Reyes et al., 2020; Shahkaram et al., 2024). Together, ACT is chosen as the foundation of our intervention protocol.
Furthermore, nurses’ irregular schedules, including long night shifts, make traditional face-to-face interventions challenging to implement (Brunet et al., 2020; Feng et al., 2021). Internet- delivered interventions offer flexibility and have proven effective in improving sleep and well-being among shift workers (Newby et al., 2021; Peter et al., 2019). Given the high workload, limited rest time, and low salary of many nurses in China (Ye et al., 2020), these interventions are cost-effective, eliminating travel time and costs (Hedman et al., 2013). Therefore, nurses could benefit significantly from Internet-delivered interventions.
Our study developed two internet-delivered ACT (iACT) protocols: iACT loop (LOOP) and iACT linear (LINEAR) models. Both incorporate core ACT processes, including the six facets of psychological flexibility (PF) and inflexibility (PI) (Hayes et al., 2006) and daily mindfulness practices (Philip & Cherian, 2022). Each model consists of 6 lessons, supported by research on the feasibility of brief six-session ACT (Jäger et al., 2021). Despite these similarities, LINEAR and LOOP differ notably. LINEAR adopts a sequential structure, addressing one facet of PF and PI per lesson, providing structured support. LOOP, culturally adapted and inspired by ancient Chinese Daoism, uses the four steps of “启-承-转-合(means starting- proceeding-transforming-integrating)”to explain the entire process of PF and PI per lesson. This allows participants to prioritize their most inflexible facet first, exploring internal solutions and cultivating flexibility through a recurring process. While LINEAR follows a fixed sequence, LOOP offers a more adaptive therapeutic journey.
Moreover, insomnia is associated with cerebral cortical dysfunction, with reduced prefrontal activation during tasks and decreased dorsolateral prefrontal cortex (DLPFC) recruitment during cognitive control (Sun et al., 2017; Wu et al., 2020). EEG, fNIRS, and fMRI studies highlight the prefrontal cortex's (PFC) pivotal role in sleep maintenance and regulation (Altena et al., 2008; Csipo et al., 2021; Leerssen et al., 2020; Massimini et al., 2004; Wu et al., 2020). Meanwhile, few studies have explored ACT's neural impact on sleep, primarily focusing on chronic pain Lee et al. (2023). While some found increased PFC activation post-ACT intervention among fibromyalgia patients (Jensen et al., 2012) and obsessive-compulsive disorders (Lee et al., 2023), others reported decreased activation in various brain regions (Aytur et al., 2021). Despite these findings, there is a lack of well-controlled studies investigating ACT's neural mechanisms, particularly in individuals with sleep problems. This study aims to fill this gap by examining the neural effects of ACT in nurses with poor sleep quality.
The verbal fluency task (VFT) is widely used to assess PFC activation and cognitive performance (Liu et al., 2020; Wewege et al., 2017). fNIRS, a safe and non-invasive neuroimaging technique, measures changes in brain hemodynamic responses (Sanders et al., 2019), providing insights into brain energy demand (Herold et al., 2018). With superior spatial resolution to EEG and enhanced temporal resolution compared to fMRI (Pinti et al., 2020). In this study, we use the fNIRS-VFT paradigm to induce cortical brain activation.
We hypothesize that both LOOP and LINEAR will effectively reduce rumination and improve sleep quality (SQ) and PF. Furthermore, we anticipate differences in intervention efficacy between LOOP and LINEAR. Additionally, we expect both models to induce changes in prefrontal cortex activation.
MethodDesignThis was a single center double-blinded randomized controlled trial. It received approval from the Institutional Review Board of the Medical Ethics Committee of the first author's affiliation (Approval No of Ethics Committee: S2021–568–01) and was pre-registered at the Chinese Clinical Trials Registry (ChiCTR2200063533) following RCT guidelines. The study adhered to ethical standards outlined in the Helsinki Declaration of 1975, as revised in 2008.
ParticipantsParticipants from Beijing Class 3A Comprehensive Hospital were recruited through online WeChat ads promoting a novel psychotherapy model aimed at improving sleep, psychological flexibility, and reducing rumination. Interested nurses joined via a WeChat Mini Program (see supplementary Fig. S1) and completed an online screening questionnaire. Recruitment began in September 2022.
Inclusion criteria: (1) ages 19 to 65; (2) access to the internet, understanding instructions in Chinese; (3) Pittsburgh Sleep Quality Index score above 5, indicating poor sleep quality (Buysse et al., 1989; Patel et al., 2018); (4) right-handed. Exclusion criteria: (1) organic mental disorder, schizophrenia, or mood disorder; (2) visual or hearing impairment; (3) drug or alcohol dependence; and (4) prior ACT intervention experience. The psychiatrist and clinical psychology team from the hospital used the Structured Clinical Interview for DSM-5 Disorders (American Psychiatric Association, 2013) to assess exclusion criteria. Participants meeting these criteria were informed of their results and instructed to contact counseling services. All personal information and screening results were kept strictly confidential, and participants provided informed consent.
Procedures and randomizationAmong the 734 nurses who responded to online ads, 257 were excluded based on criteria. Subsequently, 477 participants were randomly assigned to the LINEAR psychotherapy intervention group (n = 159), LOOP psychotherapy intervention group (n = 159), and control group (n = 159) at a 1:1:1 ratio (see Fig. 1). Randomization, conducted by an independent person using computer-generated pseudo-random numbers, ensured allocation concealment. All participants completed pre-intervention measurements (T1). The two intervention groups received distinct online self-help interventions, while the control group waited for one month. Post-intervention questionnaires were administered (T2), and 13 participants in the LINEAR group, 10 in the LOOP group, and 29 in the control group did not complete the intervention and post-intervention. Thus, the final post-intervention participant count was 425.
At trial inception, all the participants were offered the opportunity of undergoing fNIRS. After the enrollment of the first 80 participants, we estimated the proportion of individuals in each group willing to undergo fNIRS. A nonrandom subgroup of participants was then selected to undergo fNIRS to achieve the target sample size of 240 (approximately half the enrolled participants) (Barnes et al., 2023). fNIRS scans were conducted during both pre-intervention and post-intervention phases. These scans were administered by trained researchers who were unaware of the trial-group assignments.
InterventionThe one-month self-help online interventions (Table 1; supplementary Table S1 for detailed protocol) were delivered via the internet. The intervention for LINEAR group was based on the ACT LINEAR psychotherapy model, consisting of six online lessons delivered every five days in a fixed order, following the six facets of ACT. The content, provided through text, video, audio, and illustrations, required participants to practice Mindfulness daily for 20–30 min over 30 days to enhance present-moment awareness. Participants accessed one lesson at a time, progressing to the next lesson upon completion.
Overview of iACT intervention content.
| Linear psychotherapy protocol | Loop psychotherapy model | Daily Mindfulness Practice (Homework) |
|---|---|---|
| lesson 1: Experiential Avoidance and Acceptance.Lecture and explain what is Experiential Avoidance and Acceptance, and how to apply Acceptance to life.lesson 2: Cognitive Fusion and Defusion.Lecture and explain what is Cognitive Fusion and Defusion, and how to apply Defusion to life.lesson 3: Dominance of the Conceptualized Past and Feared Future, and Contact with the Present Moment.Lecture and explain what the Dominance of the Conceptualized Past and Feared Future is, Contact with the Present Moment, and how to Contact with the Present Moment.lesson 4: Attachment to the Conceptualized Self and Self as Context.Lecture and explain what is Attachment to the Conceptualized Self and Self as Context, and how to achieve Self as Context.lesson 5: Lack of Values and ValuesLecture and explain what is Lack of Values and Values is and how to clear values.lesson 6: Inaction and Committed Action.Lecture and explain what is Inaction and Committed Action, and how to commit to action. | lesson 1: Psychological flexibility and psychological inflexibility.Firstly, let the participants set a goal to achieve after this lesson.Secondly, guide them to think about their past coping styles, and the consequences.Thirdly, Lecture and explain the whole six facets of psychological flexibility and psychological inflexibility, and how to apply them to life.Fourthly, let them choose and use the one facet of PF that is most suitable for them in the following 5 days.lesson 2–6: Psychological flexibility and psychological inflexibility reinforcement.Following the four steps, different examples, text, video, audio, and illustrations were used to illustrate the six facets, which provides the participants with a solution from the first lesson. Each lesson is a further explanation of the six facets, step by step, which gives participants more autonomy to start with the most inflexible facet, then guides them to explore their internal resources to proceed with the whole process. | Day 1–5: Body Scan in Mindfulness.Day 6–10: Body Stretch in Mindfulness.Day 11–15: Mindful Breathing.Day 15–20: Mindful Walking.Day 21–25: Metta Meditation in Mindfulness.Day 25–30: Breathing Space: the Door to Actions in Mindfulness. |
The intervention for LOOP group was based on the ACT LOOP psychotherapy model (Table 1; supplementary Table S2 for detailed protocol). Notably, the LOOP model is culturally adapted. Cultural adaptation involves the systematic modification of an evidence-based treatment to consider language, culture, and context, ensuring compatibility with the client's cultural patterns, meanings, and values (Bernal et al., 2009). This includes integrating cultural factors such as interdependence and spirituality (Nagayama Hall et al., 2011). Our LOOP model, inspired by ancient Chinese Daoism principles that align with the spirit of constantly changing thinking patterns, use the four steps of “启-承-转-合(starting-proceeding-transforming-integrating)” to present content through text, video, audio, and illustrations, encouraging participants to select the suitable dimension based on different situations (Fig. 2). The six facets of psychological flexibility and inflexibility were introduced in the first lesson. Subsequent lessons illustrated these facets with examples, progressing from straightforward to more sophisticated. Participants identified their issues related to the six facets of psychological inflexibility and chose the appropriate facet of psychological flexibility to practice over the next five days. Like the LINEAR group, the LOOP group practiced daily Mindfulness for 20–30 min over 30 days. Participants accessed one lesson at a time and proceeded to the next lesson upon completion.
The control group did not receive any intervention and waited one month after completing the pre-intervention measurement.
Outcome measuresSleep qualityThe Chinese version of Pittsburgh Sleep Quality Index (PSQI), translated by Liu (1996) from the original English version (Buysse et al., 1989), assesses participants’ sleep quality. It shows high reliability and validity (Yan et al., 2021), with 18 items scored from 0 to 3. Higher scores indicate poorer sleep quality (T1Cronbach's α = 0.86, T2Cronbach's α = 0.83).
RuminationThe Chinese version of Ruminative Response Scale (RRS), translated by Yang (2009) from the original English version (Nolen-Hoeksema & Morrow, 1991), measures participants’ level of rumination. It exhibits strong reliability and validity (Han & Yang, 2009), with 22 items graded from 1 (never) to 4 (always). Higher scores indicate higher levels of rumination (T1Cronbach's α = 0.89, T2Cronbach's α = 0.83).
Psychological flexibilityThe Chinese version of the Multidimensional Psychological Flexibility Inventory (MPFI), translated by Lin (2020) from the original English version (Rolffs et al., 2018), includes two subscales: psychological flexibility and psychological inflexibility. Its internal consistency in Mandarin was excellent (Lin et al., 2020). With 60 items graded from 1 to 6, each subscale covers six dimensions, reflecting psychological flexibility or inflexibility. This study focused on the psychological flexibility subscale, where higher scores indicate greater psychological flexibility (T1Cronbach's α = 0.78, T2Cronbach's α = 0.86).
Verbal fluency taskA Chinese version of the Verbal Fluency Task (VFT) was conducted during daylight hours, consisting of a 30-second pre-task baseline, a 60-second task period, and a 60-second post-task baseline. Participants counted from 1 to 5 during both baseline periods and generated phrases using three Chinese characters in the task period, with characters changed every 20 s to minimize silent periods.
fNIRS measurementParticipants sat in a quiet room while a multi-channel continuous-wave near-infrared spectrometer (NirSmart-6000A, Danyang Huichuang Medical Equipment Co., Ltd., China) recorded changes in oxygenated hemoglobin (HbO) and deoxyhemoglobin (HbR) concentrations during the VFT to infer brain activity. The system used LED light sources and avalanche photodiodes (APD) at 730 nm and 850 nm wavelengths, with a sampling rate of 11 Hz. 14 source-detector (SD) probes with a fixed 3-cm inter-probe distance were arranged around the FPz channel (10/20 international system), covering bilateral prefrontal and temporal cortices, the lowest probes along the Fp1-Fp2 line (see supplementary Fig. S2). This established a total of 19 NIRS channels, with the channels and corresponding brain regions presented in supplementary Table S3. Oxy-Hb signals, known for their higher signal-to-noise ratio (Strangman et al., 2002) and retest reliability (Plichta et al., 2006), were employed as markers of regional cortical activity in this study.
Preprocessing included motion artifacts correction using moving SD and cubic spline interpolation, and a bandpass filter (0.01–0.20 Hz) to remove physiological noise. The modified Beer-Lambert law converted optical densities into changes in Oxy-Hb and Deoxy-Hb concentrations. VFT block waveforms were calculated with a block range set of 0–125 s, a pre-baseline range set of 0–10 s, and a post-baseline range set of 70–125s. We used a 60 s task period of constructing phrases as the time window to analyze mean Oxy-Hb changes. Linear fitting was applied to the data between these two baselines.
Statistical analysesTo account for variations within individuals in a longitudinally repeated study, we built a Linear Mixed-effects Model using the lme4 and lmerTest package in R. Fixed effects consisted of time effect, group effect, and their interaction, with random effects accounting for individual differences (Voldsbekk et al., 2021). This approach allowed us to assess the impact of timeframes (pre- and post-intervention) and different groups on changes in outcome variables. The value β in the mixed-effects model represented the estimated effect. Pairwise comparisons of estimated marginal means among treatments were made for each group using the emmeans package (Lenth, 2024). Statistical significance set at p < 0.05.
Mediation models were tested using Process 3.5 (Hayes & Preacher, 2014). We first examined the mediating effect of PF changes on outcome changes for LOOP versus control and LINEAR versus control. Next, we assessed LOOP's superior efficacy by testing the mediating effect of PF changes for LOOP versus LINEAR. We then explored the neural mechanism of ACT by analyzing the mediating effect of brain activation changes on outcome changes for both LOOP versus control and LINEAR versus control, as well as comparing LOOP with LINEAR to further clarify LOOP's superior efficacy. Both direct and indirect effects were assessed, using bootstrapping with 5000 resamples and 95 % confidence intervals, fully adjusted for covariates. The indirect effect was deemed significant if 0 was not included in the confidence interval.
Intention-to-treat (ITT) analysis was employed, and missing values were imputed using the last observation carried forward method. Descriptive statistics were used for between-group differences at baseline. Categorical variables, such as gender, employment, and other demographics, were analyzed using the chi-square test (χ²), while continuous variables, such as age and outcome measures, were analyzed using one-way ANOVA for comparisons among three groups and independent-sample T-test for comparisons between two groups.
ResultsDemographics and baseline dataAs presented in Table 2, participants in the three groups did not exhibit significant differences in demographics and outcome variables prior to the intervention.
Demographics and baseline data.
| Linear intervention Group (n = 159) | Loop intervention Group (n = 159) | Control Group (n = 159) | χ² or F (p) | |
|---|---|---|---|---|
| Age — yr | 0.29 (0.75) | |||
| M±SD | 27.36 ± 6.45 | 27.14 ± 5.87 | 26.84 ± 6.64 | |
| Range | 18 - 46 | 18 - 39 | 18 - 49 | |
| Sex, n (%) | 1.68 (0.43) | |||
| Male | 9 (5.70 %) | 9 (5.70 %) | 14 (8.80 %) | |
| Female | 150 (94.30 %) | 150 (94.30 %) | 145 (91.20 %) | |
| Employment, n (%) | 0.42 (0.81) | |||
| Employed | 139 (87.40 %) | 135 (84.90 %) | 137 (86.20 %) | |
| Further education or internship | 20 (12.60 %) | 24 (15.10 %) | 22 (13.80 %) | |
| Only child, n (%) | 1.45 (0.48) | |||
| Yes | 28 (17.60 %) | 36 (22.60 %) | 35 (22 %) | |
| No | 131 (82.40 %) | 123 (77.40 %) | 124 (78 %) | |
| Having children, n (%) | 3.54 (0.17) | |||
| Yes | 66 (41.50 %) | 64 (40.30 %) | 51 (32.10 %) | |
| No | 93 (58.50 %) | 95 (59.70 %) | 108 (67.90 %) | |
| Marital status, n (%) | 1.59 (0.45) | |||
| Married | 68 (42.80 %) | 75 (47.20 %) | 64 (40.30 %) | |
| Unmarried | 91 (57.20 %) | 84 (52.80 %) | 95 (59.70 %) | |
| Years of working, n (%) | 3.95 (0.41) | |||
| 1–10 years | 100 (62.90 %) | 99 (62.30 %) | 113 (71.10 %) | |
| 11–15 years | 40 (25.20 %) | 37 (23.30 %) | 29 (18.20 %) | |
| over 15 years | 19 (11.90 %) | 23(14.50 %) | 17 (10.70 %) | |
| PF score, (M±SD) | 74.67±13.73 | 73.75±12.48 | 72.93±13.40 | 0.69 (0.50) |
| RRS score, (M±SD) | 48.42±10.64 | 46.84±11.15 | 46.72±10.61 | 1.22 (0.30) |
| SQ score, (M±SD) | 10.03±3.04 | 9.43±3.04 | 9.59±3.10 | 1.39 (0.25) |
Notes. PF = Psychological Flexibility; RRS = Rumination; SQ = Sleep quality.
The linear mixed effect model examined the effects of linear psychotherapy model (LINEAR) and loop psychotherapy model (LOOP) on PF, rumination and SQ.
As shown in Table 3. For PF, time effect showed no statistically significant differences between pre- and post-intervention means across the control group [β (95 %CI): 1.830 (−0.548, 4.208), p = 0.133]. Group effect was not significantly different between intervention and control means at baseline [LINEAR-CON: β (95 %CI): 1.736 (−1.516, 4.987), p = 0.297; LOOP-CON: 0.824 (−2.428, 4.075), p = 0.620]. Group * time estimated the difference in mean change from pre-to post-intervention for LINEAR and LOOP compared to control group. Our results demonstrated that LINEAR and LOOP could significantly improve PF [LINEAR: β (95 %CI): 21.296 (17.932, 24.659), p < 0.001; LOOP: −4.340 (−5.133, −3.546), p < 0.001]. As shown in Supplementary Table S4, LINEAR and LOOP differed significantly post-intervention (p < 0.001). The LOOP [Emmeans = 110.3 (108.0–112.6)] had a better effect on increasing PF compared to the LINEAR [Emmeans = 97.8 (95.5–100.1)].
Effects of iACT on psychological flexibility, rumination and sleep quality.
| SQ | PF | RRS | |||||||
|---|---|---|---|---|---|---|---|---|---|
| β | Lower 95 CI | Upper 95 CI | β | Lower 95 CI | Upper 95 CI | β | Lower 95 CI | Upper 95 CI | |
| Time points | −0.465 | −1.027 | 0.096 | 1.830 | −0.548 | 4.208 | −1.579 | −3.547 | 0.390 |
| LINEAR-CON | 0.346 | −0.290 | 0.982 | 1.736 | −1.516 | 4.987 | 1.591 | −0.489 | 3.671 |
| LOOP-CON | −0.201 | −0.837 | 0.434 | 0.824 | −2.428 | 4.075 | 0.780 | −1.301 | 2.860 |
| LINEAR-CON: Time points | −3.730 | −4.523 | −2.936 | 21.296 | 17.932 | 24.659 | −13.170 | −15.954 | −10.386 |
| LOOP-CON: Time points | −4.340 | −5.133 | −3.546 | 34.723 | 31.360 | 38.087 | −13.849 | −16.633 | −11.065 |
Notes. 95 % CI = 95 % confidence intervals; CON = control group; LINEAR = linear intervention group; LOOP = loop intervention group; PF, Psychological Flexibility score; RRS = Rumination score; SQ = Sleep quality score.
For rumination, group * time estimated mean change differences between LINEAR and LOOP compared to control. The effect of LINEAR was significant [β (95 %CI): −13.170 (−15.954, −10.386), p < 0.001], as well as LOOP [β (95 %CI): −13.849 (−16.633, −11.065), p < 0.001]. LINEAR and LOOP differed significantly post-intervention (p < 0.001). The LOOP [Emmeans = 29.8 (28.4–31.2)] had a better effect on reducing rumination compared to the LINEAR [Emmeans = 34.7 (33.3–36.1)].
For SQ, group * time estimated mean change differences between LINEAR and LOOP compared to control. The effect of LINEAR and LOOP psychotherapy was significant [LINEAR: β (95 %CI): −3.730 (−4.523, −2.936), p < 0.001; LOOP: −4.340 (−5.133, −3.546), p < 0.001]. LINEAR and LOOP differed significantly post-intervention (p < 0.001). The LOOP [Emmeans = 4.69 (4.23–5.14)] had a better effect on enhancing SQ compared to the LINEAR [Emmeans = 5.84 (5.39–6.29)].
Mediation analysis of ACT's efficacyThe mediating effect was significant when intervention membership (LINEAR verses control) was the predictor, PF changes were the mediator, and SQ changes were the outcome, with a mediating effect of 0.618 (SE = 0.131; 95 % CI: 0.375, 0.886), accounting for 45 % of the total effect of 1.377. For rumination changes as the outcome, the mediating effect was 1.833 (SE = 0.618; 95 % CI: 0.616, 3.026), accounting for 33 % of the total effect of 5.598. When intervention membership (LOOP verses control) was the predictor, PF changes were the mediator, and SQ changes were the outcome, with a mediating effect of 1.086 (SE = 0.271; 95 % CI: 0.528, 1.612), accounting for 43 % of the total effect of 2.497. For rumination changes as the outcome, the mediating effect was 4.123 (SE = 1.197; 95 % CI: 1.762, 6.431), accounting for 41 % of the total effect of 9.962. Further analysis compared the superiority of LOOP to LINEAR interventions, with intervention membership (LINEAR verses LOOP) as the predictor, PF changes as the mediator, and SQ changes were the outcome, with a mediating effect of 0.421 (SE = 0.119; 95 % CI: 0.203, 0.669), accounting for 38 % of the total effect of 1.12. For rumination changes as the outcome, the mediating effect was 1.352 (SE = 0.535; 95 % CI: 0.340, 2.443), accounting for 31 % of the total effect of 4.365 (see Supplementary Table S5).
iACT interventions on cortical activationOf the 477 participants, 210 agreed to baseline brain imaging, and 152 completed follow-up imaging post-intervention. Two participants were excluded due to poor-quality baseline fNIRS data, leaving 150 participants for analysis. The three fNIRS groups showed no significant differences in demographics, outcome variables, or cortical activation across the 19 channels prior to the intervention (see Supplementary Table S6 and Fig. 3). They also did not differ significantly from the trial's baseline sample (Supplementary Table S7).
As shown in Figs. 4, 3B and supplementary Table S8, statistically significant differences of oxyHb activations in Channel 1 (mainly located in the DLPFC) were observed in both LINEAR and LOOP post-intervention [LINEAR: β (95 %CI): 0.053 (0.021, 0.085), FDR-adjusted p = 0.007; LOOP: β (95 %CI): 0.051 (0.024, 0.079), FDR-adjusted p = 0.007]. Additionally, significant oxyHb activation differences were observed in LOOP at Channel 5 (mainly located in the frontopolar area) [β (95 %CI): 0.052 (0.019, 0.083), FDR-adjusted p = 0.002]. However, no significant changes were identified in control group.
(A) Mean Oxy-Hb changes of linear group in pre-intervention. (B) Mean Oxy-Hb changes of linear group in post-intervention. (C) p-values of each channel for pre- and post- comparison in linear group. (D) Mean Oxy-Hb changes of loop group in pre-intervention. (E) Mean Oxy-Hb changes of loop group in post-intervention. (F) p-values of each channel for pre- and post-comparison in loop group. (G) Mean Oxy-Hb changes of control group in pre-intervention. (H) Mean Oxy-Hb changes of control group in post-intervention. (I) p-values of each channel for pre- and post-comparison in control group.
After functional connectivity calculation, two 19 × 19 correlation matrices were generated (supplementary Fig. S3). Supplementary Table S8 shows no significant channel-to-channel connectivity changes in either the LINEAR or LOOP groups post-intervention [LINEAR: β (95 %CI): −0.032 (−0.129, 0.647), FDR-adjusted p = 0.621; LOOP: β (95 %CI): −0.067 (−0.164, 0.030), FDR-adjusted p = 0.268].
Mediation analysis of neural mechanism in act efficacyThe mediating effect was significant when intervention membership (LINEAR verses control) was the predictor, Channel 1 changes were the mediator, and SQ changes were the outcome, with a mediating effect of 1.144 (SE = 0.290; 95 % CI: 0.609, 1.743), accounting for 33 % of the total effect of 3.48. For PF changes as the outcome, the mediating effect was 4.918 (SE = 1.564; 95 % CI: 2.037, 8.137), accounting for 27 % of the total effect of 18.14. For rumination changes as the outcome, the mediating effect was 2.427 (SE = 1.004; 95 % CI: 0.517, 4.415), accounting for 26 % of the total effect of 9.2. Similarly, when intervention membership (LOOP verses control) was the predictor, Channel 1 changes were the mediator, and SQ changes were the outcome, the mediating effect was 1.361 (SE = 0.345; 95 % CI: 0.718, 2.101), accounting for 70 % of the total effect of 1.94. For PF changes as the outcome, the mediating effect was 9.371 (SE = 1.450; 95 % CI: 6.680, 12.369), accounting for 31 % of the total effect of 29.78. For rumination changes as the outcome, the mediating effect was 7.886 (SE = 1.085; 95 % CI: 5.782, 10.064), accounting for 50 % of the total effect of 15.84 (see Supplementary Table S9).
Further exploration of the superior effect of LOOP with Channel 5 changes as the mediator showed a mediating effect of 0.876 (SE = 0.329; 95 % CI: 0.286, 1.580), accounting for 26 % of the total effect of 3.36, partially mediating the relationship between intervention and SQ changes (see Supplementary Table S10). This effect was not observed for rumination or PF.
DiscussionThe study explored the efficacy of self-developed iACT loop (LOOP) and linear (LINEAR) psychotherapy models for nurses with poor sleep quality compared to a wait-list control. Both models significantly improved PF, SQ, and reduced rumination, with the culturally adapted LOOP model showing greater efficacy through enhanced psychological flexibility. Both models enhanced activation in the DLPFC, LOOP also stimulated the frontopolar area. Changes in the DLPFC mediated the relationship between intervention and outcome improvements, highlighting its role in ACT's neural mechanism. Frontopolar changes mediated only SQ improvements, not rumination or PF, suggesting the need for further research on LOOP's neural mechanism of superior effects. Moreover, there were no significant changes in functional connectivity during VFT following the intervention, indicating an unclear link between ACT and task-state functional connectivity.
The study demonstrated that both models enhanced PF, with LOOP showing superior efficacy, likely due to its cultural adaptation. Systematic reviews and meta-analyses have shown that culturally adapted treatments are more effective, acceptable, and satisfying for patients (Chowdhary et al., 2014; Hall et al., 2016; Soto et al., 2018). Participants find culturally adapted ACT enjoyable and acceptable (Chowdhary et al., 2014; Perry et al., 2019; 2024). The LOOP, follows the four step of “启-承-转-合 (Starting-Proceeding-Transforming-Integrating)” to deliver contents, aligns well with Chinese cultural context, making it more easily accepted by participants (Ding et al., 2020) and helping them manage distress and feel more comfortable seeking social support (Xu et al., 2020). Our findings suggest that this model could be applicable to other Chinese cohorts.
Both models improved sleep quality, with LOOP proving more effective through enhanced psychological flexibility. Its unique design introduced all six facets in the first session, with subsequent sessions reinforcing core ACT concepts. This approach encouraged participants to focus on their most inflexible facet early on, potentially increasing psychological flexibility from the outset and supporting a comprehensive therapeutic journey (Janeslätt et al., 2019; Mendoza et al., 2020). Given the close association between psychological flexibility and sleep, enhancing overall psychological flexibility contributes to improved sleep quality (Kato, 2016). LOOP potentially empowers participants to actively accept sleep difficulties from the first session, reshaping their beliefs and attitudes, which is a well-established treatment strategy (Chapoutot et al., 2021).
Both models reduced rumination, with LOOP outperforming LINEAR by enhancing psychological flexibility. This advantage likely stems from its repetitive presentation of intervention materials, which allows participants to master core ACT content through six repetitions, potentially making it more effective in increasing psychological flexibility. This repetitive intervention mode has been effectively applied in various psychological domains. For instance, exposure therapy involves repeatedly exposing patients to anxiety-provoking stimuli to reduce sensitivity (McLean et al., 2022; van Loenen et al., 2022). Similarly, CBT employs repeated cognitive restructuring and behavioral exercises to alter maladaptive cognitive patterns and behaviors (Cuijpers et al., 2020; Pardos-Gascón et al., 2021). In educational psychology, repeated reading enhances reading fluency and comprehension by having students read the same text multiple times (Alqahtani, 2020; O'Connor et al., 2007). Additionally, repetitive practice strengthens neural pathways, facilitating skill automation and proficiency (Dayan & Cohen, 2011; Macnamara et al., 2016). Our research applied this repetitive mode in LOOP, demonstrating promising results, suggesting its potential replication in future ACT interventions to enhance efficacy.
This study is the first to investigate the neural mechanism of ACT on nurses with poor sleep quality. Our findings indicated increased activation in the DLPFC post-ACT intervention, suggesting enhanced cognitive resource recruitment during VFT following ACT. The significant mediating effects imply that ACT interventions may enhance cognitive resources and emotional regulation through DLPFC activation, leading to improvements in sleep quality and psychological flexibility, and reductions in rumination. These neural changes align with previous research highlighting the role of the DLPFC in cognitive control and emotion regulation (Lévesque et al., 2003; Shackman et al., 2009). Additionally, prior research has shown a weakened right prefrontal lobe network in the development of chronic insomnia disorder (Gong et al., 2022), with the DLPFC being a target in sleep intervention studies (Gong et al., 2020). fMRI studies have consistently demonstrated increased activity in the prefrontal cortex, including the DLPFC, and decreased subcortical activity resembling the Defusion facet of ACT (Creswell et al., 2007; Koenigsberg et al., 2010; Lee et al., 2023; Wang et al., 2022). Overall, our results suggest that the DLPFC could be a potential neural mechanism of ACT. Future research should continue to validate these findings and further explore how this neural pathway can be leveraged to optimize the efficacy of ACT.
Our study also reveals that the LOOP stimulates frontopolar activation, aligning with previous research linking reduced local computational deep sleep in the frontopolar area with obstructive sleep apnea (Jussila et al., 2016) and its association with slow-wave sleep (Ferrara et al., 2002). However, our mediation analysis indicates that changes in the frontopolar area only partially mediate the relationship between the intervention and SQ improvements, but not rumination or PF. This suggests that while frontopolar activation contributes to the enhanced sleep quality observed with the LOOP intervention, it does not significantly impact rumination or PF. These findings highlight the specificity of mechanisms underlying LOOP's superior effect and underscore the need for further research to clarify them.
Lastly, our study found no significant changes in task-related functional connectivity. Prior evidence on ACT's impact in this context is limited. For instance, in patients with obsessive-compulsive disorder, post-ACT fMRI revealed strengthened connectivity in certain brain regions (Lee et al., 2023), suggesting an unclear link between ACT and task-related functional connectivity. Our study is the first to report VFT-related brain connectivity following ACT in individuals with poor sleep, contributing to the growing body of evidence on ACT's impact on task-independent connectivity.
The study has several notable limitations. Firstly, the nurses were exclusively from one Hospital, suggesting the need for future research involving nurses from diverse geographic areas to enhance the generalizability of findings. Secondly, this study solely analyzed pre- and post-intervention data, without incorporating follow-up measurements. Future research should include follow-up assessments to further investigate the efficacy of iACT. Lastly, the cultural adaptation of the LOOP model was based on traditional Chinese culture, potentially limiting its generalizability to populations outside of China. Future research should explore its applicability in other cultural contexts to validate its effectiveness.
ConclusionBoth models significantly improved sleep quality, psychological flexibility, and reduced rumination through enhanced psychological flexibility, with LOOP exhibiting superior efficacy. While both models increased activation in the prefrontal cortex, LOOP additionally stimulated the frontopolar area. Mediation analyses underscore the role of DLPFC activation in understanding ACT's neural mechanisms for insomnia treatment, highlighting the need for further research into the mechanisms of LOOP's superior effect.
FundingThis work was supported by the first batch of Scientific Research Projects [grant numbers 0 × 12,205].
Data statementThe raw data of the present study are available from the corresponding author on reasonable request.
CRediT authorship contribution statementDifan Wang: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing – original draft, Writing – review & editing. Bingyan Lin: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing – original draft, Writing – review & editing. Jiaxue Du: Formal analysis, Writing – review & editing. Wenyu Liu: Formal analysis, Writing – review & editing. Tong Sun: Data curation. Qingyi Li: Data curation. Lijun Xiao: Funding acquisition, Supervision, Validation, Writing – review & editing.
