The effects of digital CBT intervention on attentional bias and sleep quality of poor sleepers with insomnia symptoms Authors: Yimei Wu, Jingwen Li, Hongying Liu, Junlong Luo, Wen He, Haijiang Li Published: 2024 ABSTRACT AbstractBackgroundAttentional bias is a salient manifestation of insomnia. Digital cognitive therapy for insomnia (dCBT-I) has been validated as effective in alleviating this cognitive dysfunction. However, the effect of dCBT-I on attentional bias among Chinese individuals with insomnia remains undiscussed. This research sought to investigate this effect via a pictorial dot-probe task.MethodsIn Study 1, the pattern of attentional bias among poor sleepers (N = 52) and normal sleepers (N = 56) was assessed by the dot-probe task. In study 2, dCBT-I and conventional education were received by the experimental group (N = 42) and control group (N = 25), respectively. The dot-probe tasks and sleep quality assessments were completed at baseline and post-test.ResultsThe results of Study 1 indicated that poor sleepers exhibited significant attentional bias, characterized by increased attentional vigilance. Compared to normal sleepers, they showed heightened attentional vigilance toward sleep-related cues. The results of Study 2 showed that both dCBT-I and conventional education led to improvements in PSQI scores. However, only dCBT-I training alleviated attentional vigilance toward sleep-related cues. Additionally, dCBT-I was uniquely effective in reducing feelings of fatigue.ConclusionsPoor sleepers had a significant attentional bias, marked by heightened vigilance toward sleep-related cues. Digital CBT-I effectively reduced attentional vigilance and fatigue, suggesting that dCBT-I targets the cognitive distortions associated with insomnia.Trial registrationChiCTR2100053172 (registered 13/11/2021).Keywords: Insomnia, Attentional bias, Attentional vigilance, Cognitive behavioral therapy for insomnia, Dot-probe task Introduction Insomnia is a well-known physical and psychological problem [1]. According to data collected by the American Academy of Sleep Medicine in 2020, it is estimated that 30% of adults experience symptoms of insomnia, with approximately 10% of adults experiencing chronic insomnia that lasts for at least three months [2]. Insomnia is defined as a predominant complaint of dissatisfaction with the quantity or quality of sleep and is associated with symptoms including difficulties in initiating and maintaining sleep, frequent awakenings, problems returning to sleep after awakenings, and other extensive symptoms [3]. The experience of non-refreshing or poor sleep is associated with impairments in daytime functioning, including reduced alertness, fatigue, exhaustion, dysphoria, and other dysfunctions [4]. In light of the far-ranging effects exerted by insomnia, it can be comorbid with other psychiatric disorders, thereby increasing the risk of psychiatric relapse [5]. Comprehending the etiology of insomnia through commonly discussed models is a viable way to unveil its nature. (a) The attention-intention-effort pathway [6, 7] elaborates on psychobiological inhibition. “Normal” sleep is a relatively automatic, subconscious process that willpower cannot induce. It is vulnerable to disruption caused by patients’ heightened arousal toward internal and external stimuli consistent or inconsistent with sleep-related signals and by direct attempts to control the process [8]. Subjective selective attention to the sleep-engaging state impedes this dearousal process and promotes the development of chronic insomnia. (b) Cognitive model [9]. The model posits that sleep-related worry, selective attention and monitoring, and detecting sleep-related threats continuously interfere with sleep initiation or maintenance. These factors prompt individuals to escalate their anxiety, culminating in a real deficit in sleep and daytime functioning [9]. Attentional bias and insomnia Among the predominant cognitive dysfunctions of insomnia, attentional bias has been repeatedly demonstrated by empirical studies [10–12]. Attentional bias refers to the preferential tendency to allocate excessive attention to specific stimuli related to their experience of the diseases [13]. Theoretical underpinnings for this study are derived from the models proposed by Espie [6] and Harvey [9], which suggest that individuals with insomnia tend to have an excessive preoccupation with sleep-related threatening cues during the pre-sleep period. The persistent and overwhelming attention to the sleep-related cues, in turn, reinforces sleep-specific physiological arousal during the daytime [12]. Sleep-related hyperarousal not only disrupts the automatic process of "normal" sleep [8] but also prompts maladaptive attention to sleep-related cues throughout the day [14]. The majority of studies adopted computerized reaction time tasks related to sleep-specific stimuli (e.g., dot-probe task [15], Stroop task [16], flicker task [10]). With different paradigms and samples, the two commonly discussed components of attentional bias (attentional vigilance and attentional disengagement [17] were identified in different studies. Researchers assessed sleep-related attentional bias with the emotional Stroop paradigm and event-related potentials and found that insomnia participants exhibited hypervigilance toward sleep-negative stimuli compared to good sleepers [18]. Fabbri and colleagues [19] found that individuals with more severe insomnia symptoms had a stronger attentional disengagement to sleep-related words than those with milder symptoms. Using a tired-face pictorial dot-probe task, individuals with insomnia displayed difficulty in both orienting to and disengaging attention from tired faces compared to normal sleepers [15]. However, some findings have dampened the consistency of the presence of sleep-related attentional bias in experiments. Lundh [20] and Barclay [16] failed to detect evidence of attentional bias via the Stroop task. Coincidentally. Spiegelhalder and colleagues [21] used a mixed modality (visual-auditory) task and found no significant differences in attentional performance between insomniac and normal sleepers. While the different effects of commonly used paradigms cannot be ignored. It is to date controversial whether the flicker paradigm, the dot-probe task, Stroop task, and the Posner task can measure the attentional processes and states uniformly [14]. Among them, dot-probe task was tested less sensitive to group effects [12] and was also shown a more robust validity across the experiments [14]. Therefore, it is possible to use the dot-probe task in the current study to examine a particular pattern of attentional bias among poor sleepers with insomnia symptoms. Digital cognitive behavioral therapy and insomnia At present, the treatment methods for insomnia include psycho-behavioral intervention and drug therapy, and psycho-behavioral intervention mainly refers to cognitive behavioral therapy for insomnia (CBT-I). CBT-I is a multicomponent intervention [22] that includes sleep hygiene education, stimulus control, sleep restriction, relaxation training, and cognitive therapy. CBT-I is one of the most effective treatments for insomnia [23]. However, nonnegligible factors, including increasing labor costs and objective limitations (e.g., prolonged face-to-face interactions, tight supply of traditional CBT to demands), of traditional CBT-I call for an improved pattern. Fortunately, digital cognitive behavioral therapy (dCBT-I) has been greatly promoted and developed [24]. Digital cognitive behavioral therapy for insomnia can improve the symptoms of insomnia and health resilience in the context of the COVID-19 pandemic [25]. This newly viable therapy was also effective in reducing nocturnal and diurnal functional impairments associated with insomnia [24, 26]. A meta-analysis of 11 randomized controlled trials showed a statistically significant effect of dCBT-I on insomnia compared to other interventions, such as sleep hygiene [27]. However, the efficacy of dCBT-I in the treatment of the Chinese population with poor sleep still needs the support of further research. In view of the above description, the current study aims to investigate whether there is an attentional bias toward sleep-related cue pictures among poor sleepers who have relatively mild symptoms of insomnia compared to normal sleepers and whether the dCBT-I intervention could improve biased attention and sleep quality. We hypothesized that (a) individuals with poor sleep quality would show attentional bias toward sleep-related stimuli compared to normal sleepers and (b) the digital CBT-I intervention would improve the biased attention toward sleep-related stimuli and sleep quality of people with poor sleep. Materials and methods Study 1: biased attention of individuals with insomnia Participants The current study (Study 1 and Study 2) completed prospective clinical trial registration and was carried out according to relevant guidelines and regulations, ensuring that participant recruitment did not begin until the trial protocol and registration date were publicly available. This study protocol was reviewed and approved by Shanghai Normal University Ethics Association. Informed consent was obtained from all participants prior to group allocation. The sample was drawn from 1698 participants recruited via the Questionnaire Star platform (a professional online survey platform in China: https://www.wjx.cn/app/survey.aspx) at Shanghai Normal University and universities in several cities. Full-time college students who were at least 18 years old were invited to participate in this study. To prevent possible contamination of statistical results from atypical samples, standard exclusion criteria were established. Standard exclusion criteria were as follows: (a) Epworth Sleepiness Scale (ESS) score greater than or equal to 10 [28]; (b) Patient Health Questionnaire-9 (PHQ-9) questionnaire score greater than or equal to 20 [29]; (c) Generalized Anxiety Disorder (GAD-7) score greater than or equal to 19 [30]; (d) received psychological and physical treatment within one month; I Insomnia due to academic, life or physical illness (e.g., cardiovascular, neurovascular disease, etc.); and (f) caffeine abuse and alcohol abuse [23]. The exclusion criteria had been tested in cited articles as creditable efficacy to filter out individuals with comorbidities that are accompanied by difficulties with sleep onset, sleep maintenance, or early morning awakening [31, 32]. These atypical samples had a certain possibility of contaminating the statistical results. The confirmation of psychological and physical treatment within one month, the confirmation of insomnia due to academic tasks, life affairs, or physical illness, and the assessment of caffeine abuse and alcohol abuse were conducted through general self-report questions. Eligible participants completed the group allocation—poor sleepers (PS) or normal sleepers (NS). Figure 1 provides a flowchart and the number of participants excluded in the whole recruiting process. Thus, a sample of 52 participants with potential insomnia (mean age (SD) = 21.02 (1.88) years; 78% female) and 56 normal sleepers (mean age (SD) = 20.84 (1.68) years; 83% female) were included. Please see Table 1 for demographic information. Flowchart of the recruiting process Table 1Descriptive statistic results of Study 1PS(n = 52)NS(n = 56)F/t/ χ2 dfpM(SD)M(SD)Age (years)21(1.88)21(1.68)2.98a(106)0.84Gender (male: female)11: 419: 470.46b10.50Income (level)2(1.05)2(1.22)0.08a(106)0.78PSQI score9(2.32)3(1.27)18.19a(106)0.001FSS Score44(10.51)34(12.76)4.22a(106)0.001Overall attentional bias27(45.78)5(52.55)2.36d(106)0.02Selective attention toward sleep-related pictures27(45.78)n/a4.30c(51)< 0.001Attentional vigilance19.00(58.86)n/a2.33c(51)0.02Attentional disengagement8.28(53.65)n/a1.11c(51)0.27Abbreviation: NS Normal Sleepers group, PS Poor Sleepers group, PSQI Pittsburgh Sleep Quality Index, FSS Fatigue Severity ScaleaF TestbChi-square TestcOne-sample t-testdTwo-sample t-test Descriptive statistic results of Study 1 Abbreviation: NS Normal Sleepers group, PS Poor Sleepers group, PSQI Pittsburgh Sleep Quality Index, FSS Fatigue Severity Scale aF Test bChi-square Test cOne-sample t-test dTwo-sample t-test Measurements Dot-probe task The present study employed a modified dot-probe task, a computerized speeded reaction-time task in which two cues are simultaneously presented at different locations on a computer screen [33]. This task was used to examine the attentional bias of individuals with insomnia and was administered via a testable online experimental platform. (https://www.testable.org). Modifications to the task were made regarding threatening stimuli depicting daytime fatigue/malaise [14]. The pictorial stimuli embedded in the current dot-probe task have been validated and used by a previous study examining daytime attentional bias among primary insomnia patients [14]. We retained only 12 sleep-related images, depicting to-be-yawning scenarios, and matched them with 12 neutral images to rule out the influence of non-image-related factors. The experiment of each participant started after a screen correction. In this study, both practice and experimental trials followed a specific sequence: a fixation cross ( +) appeared at the center of the screen for a random duration between 400–600 ms, a balanced picture pair was displayed for 500 ms, followed by a pair of detection points and dots presented randomly on the right or left for 2000 ms, and finally, a blank screen appeared for 500 ms. During the experiment, participants were asked to respond as fast as possible to a certain feature of the probe, namely, whether the dots were shown vertically or horizontally after the appearance of the cues (picture pairs in the current study). The break was given in the middle of the experiment. Forty-eight trials were random pairs of neutral vs. sleepy or sleepy vs. neutral images, and 24 trials were random pairs of neutral vs. neutral images. The size of all images was 360 × 270 pixels. The experimental flow of the dot-probe task is shown in Fig. 2. The experimental image stimuli were 12 sleep-related and 12 neutral images, which had been used and evaluated in a previous study, and the sleep-related pictures were mostly of people yawning or wanting to get to sleep [14]. All the participants were required to complete 72 trials of full-random stimulus pairs. Here, to ensure that the results were not concealed by order effects, the picture pairs were presented in automatically random order. Flow of dot-probe task Questionnaires The Pittsburgh Sleep Quality Index (PSQI) was utilized to measure the quality of sleep [34]. The Chinese version of the Pittsburgh Sleep Quality Index (C-PSQI) was demonstrated as a psychometrically sound measure of disturbance for patients with primary insomnia [35]. An array of studies simply utilized the PSQI as a creditable sleep measure in studies of insomnia [16, 36]. The C-PSQI score may not function as a benchmark of diagnostic criteria as the DSM-5 does because of its low specificity, but it can be a sensitive and valid outcome assessment tool for use in community-based studies of primary insomnia [35]. It is also a reliable tool for distinguishing between good and poor sleepers [37]. The PSQI is a 0–3 Likert scale with a total scale ranging from 0 to 21. In the current study, the scale had good internal consistency (Cronbach’s α = 0.75). In the group allocation of the current study, PSQI scores greater than or equal to 7 were regarded as poor sleepers, while scores less than 7 were regarded as normal sleepers. The level of fatigue of all participants was also measured via the Fatigue Severity Scale [38], as we set fatigue as an ancillary parameter of sleep quality. Based on previous findings, fatigue was not only a symptom of insomnia but also regarded as a result of accumulated sleep deprivation among affected patients [4, 23], and subjective sleep quality was tested as a predictor of fatigue [39]. The FSS consists of 9 items and is rated on a 7-point scale (1 = “strongly disagree”, 7 = “strongly agree”). The internal consistency of FSS was satisfactory in the present study (Cronbach’s α = 0.88). In this study, higher FSS scores indicated greater fatigue, namely participants were more tired. The FSS was used to explore whether scores would decrease as sleep quality improved following dCBT-I interventions. Procedure The whole procedure was completely remote. The self-reported C-PSQI was administered to all participants (N = 1698) to assess their sleep quality in the last 1 month, along with FSS and a general information survey. All the questionnaires were administered during a regular weekday [37], followed by an online dot-probe task on the same day. A randomized controlled single-blind manipulation was used. All participants were given a monetary reward at the end of the experiment. Statistical analysis All statistical analyses in Study 1 and Study 2 were performed via SPSS 25.0. To avoid the influence of extreme values on the data analysis, trials with too low overall accuracy (< 75%), latencies less than 200 ms or greater than 1300 ms, or being out of ± 3 standard deviations were discarded as outliners before computing the score of biased attention. Subsequently, the attentional bias and related components were subjected to standardized quantification based on reaction responses. Based on previous studies [15, 40] and the assumptions of the current study, PS will manifest one or more components of biased attention: attentional vigilance and attentional disengagement. To quantify these components, we measured participants' response times to sleep-related and neutral pictures. The overall attentional bias score was calculated by subtracting the average latencies for consistent probes (i.e., probe at the location of the sleep-related picture) from inconsistent probes (i.e., probe at the location of the neutral picture). The attentional vigilance score was calculated by subtracting the average latencies for consistent probes from the average latencies for neutral probes. The attentional disengagement score was calculated by subtracting the average latencies for neutral probes from the average latencies for inconsistent probes. A positive score indicates that sleep-related pictures received more attention than neutral pictures. All components were measured in milliseconds. According to the mean value of overall attentional bias and the result of the one-sample t-test, a positive value indicates that individuals would show selective attentional bias toward sleep-related stimuli; otherwise, they would show attentional avoidance of the stimulus. A two-sample t-test was used to explore the differences in age, income, and attentional bias between PS and NS, and a chi-square test was used to confirm that there was no significant difference in gender between PS and NS. Furthermore, one-sample t-tests were performed on overall attentional bias, attentional vigilance score, and attentional disengagement score in PS. Study 2: the effects of dCBT-I on individuals with insomnia Participants To increase the number of participants in the digital CBT-I, the inclusion threshold according to the PSQI score was expanded to be no less than 5 [34, 37], and other selection criteria were the same as in Study 1, which left us with a total of 67 participants with poor sleep who met the experimental criteria and signed informed consent. Among this group of samples, 41 participants were recruited from Study 1. To minimize practice effects, we replaced 12 yawning pictorial stimuli in the dot-probe task for Study 2. All 67 participants showed an attentional bias toward sleepy-related stimuli at baseline. These participants were randomly divided into the dCBT-I group (DI, N = 42) and the conventional education group (CE, N = 25) as the control group. Table 2 shows the demographic information of the participants. All participants were given a monetary reward at the end of the experiment. Table 2Descriptive statistic results of Study 2DI(n = 42)CE(n = 25)F/ χ2 dfpM(SD)M(SD)Age (years)21(1.60)21(2.26)2.47(65)0.46Gender (male: female)10: 324: 210.5810.54aIncome (level)2(1.13)3(1.00)0.02(65)0.52Abbreviation: DI dCBT-I intervention group, CE conventional education groupaChi-square Test Descriptive statistic results of Study 2 Abbreviation: DI dCBT-I intervention group, CE conventional education group aChi-square Test Procedure The whole procedure was completely remote. All participants in the two groups were asked to complete a dot-probe task and questionnaires same as in Study 1. Subsequently, the DI and CE groups completed three weeks of dCBT-I intervention and conventional education, respectively. The dCBT-I intervention was conducted in an online applet of WeChat (a Chinese Social Network Application) five times per week, 10–15 min each time. They needed to finish the task 30 min before bed, between 8 p.m. and 12 p.m. The topic and tasks of intervention each week were different and were set before the study. The intervention incorporated the use of cognitive and behavioral techniques, including sleep restriction, stimulus control, bedtime routine, etc. In the first week, topics were focused on sleep hygiene education, including how to assess sleep conditions; in the second week, topics were focused on sleep restriction, stimulus controlling, exercise routine, relaxation training, and food intake control; in the third week, topics were focused on gratitude journaling, cognitive restructuring, correct thinking about one's sleep medication and mind relaxation. To ensure that the DI group gained relevant knowledge, the participants were assigned corresponding videos and weekly intervention tasks. After the participants logged in to the applet, their daily log-in histories were recorded on the website, and they were asked to complete a daily video and punch in their online daily sleep hygiene diary. Conversely, the CE group was provided with comprehensive articles on how to adopt a regular lifestyle, the importance of exercise, basic lifestyle, and behavioral requirements to improve sleeping quality without other online tools. Participants in the CE group were asked to read 1 article and complete an online sleep hygiene diary daily with the identical form of sleep diary of the two groups. There were 3 weeks between the baseline and post-test, which is the duration of dCBT-I intervention or conventional education. After all the participants completed the above experimental procedure, they were asked to complete the posttest, including questionnaires and the dot-probe task, as in Study 1, within 5 days. Moreover, the research team recorded the time that participants logged in and supervised their task completion rate (> 95%) to ensure the validity of all interventions. Statistical analysis A 2 (group: DI, CE) × 2 (time: baseline, posttest) MANOVA was conducted to investigate the influence of dCBT-I, with PSQI scores, FSS scores, attentional bias, attentional vigilance, and attentional disengagement as dependent variables. Group allocation was treated as a categorical predictor. Then, a simple effect analysis was performed to explore the specific influence of dCBT-I on participants’ status with a 2-tailed threshold of p < 0.05. Results Study 1: biased attention of individuals with insomnia Participant characteristics Demographic and behavioral characteristics in the poor sleepers (PS) and normal sleepers (NS) are presented in Table 1. PS and NS did not differ in age (p = 0.84), gender (p = 0.50), or household income (p = 0.78), but they differed in PSQI scores (F(1, 106) = 18.19, p < 0.001) and FSS scores (F(1, 106) = 4.22, p < 0.001), confirming an accurate group allocation. Attention bias in participants The results of two-sample t-test showed that there was a significant difference in the value of overall attentional bias between PS and NS (t(1, 106) = 2.36, p = 0.02), validating the existence of attentional bias among individuals with potential insomnia. The results of the one-sample t-test indicated that the average response time on the probe for the sleepy pictures (M(SD) = 27.00(45.78)) was significantly lower than that on the probe for the neutral pictures (t(1, 51) = 4.30, p < 0.001), indicating explicit sleep-related selective attention. In addition, the PS group showed significant attentional vigilance on sleepy pictures (M(SD) = 19.00(58.86), t(1, 51) = 2.33, p = 0.02) but no attentional disengagement (p = 0.27). Study 2: the effects of dCBT-I on individuals with insomnia Participant characteristics Demographic and behavioral characteristics of participants in the dCBT-I intervention group (DI) and conventional education group (CE) are presented in Table 2. There were no significant differences in age (p = 0.46), gender (p = 0.54), or household income (p = 0.52) between the two groups. The effects of dCBT-I on participants The results of MANOVA showed that (a) for PSQI scores, the main effect of time reached significance (F(1, 65) = 54.06, p < 0.001, ηP2 = 0.01), but the interaction of group × time was not significant (p = 0.52); (b) for FSS scores, there were no significant results for the main effect of time (p = 0.07) or the interaction of group × time (p = 0.60); (c) for attentional bias, the interaction of group × time was significant (F(1, 65) = 5.33, p = 0.02, ηP2 = 0.08); (d) for attentional vigilance, the interaction of group × time was significant (F(1, 65) = 5.04, p = 0.03, ηp2 = 0.07); and (e) for attentional disengagement, there were no significant results for the main effect of time (p = 0.52) or the interaction of group × time (p = 0.76). The subsequent simple effect analysis indicated significant differences in both the DI and CE groups' PSQI scores at baseline and post-test (FDI(1, 65) = 31.23, p < 0.001, ηp2 = 0.32; FCE(1, 65) = 24.76, p < 0.001, ηp2 = 0.28); specifically, both groups showed significantly lower PSQI scores at post-test than at baseline (MDI(SD) = 5.10 (2.00); MCE(SD) = 4.92(2.10)). Although the interaction of group × time on FSS was not significant, the DI group did show a significantly reduced FSS score (F(1,65) = 4.26, p = 0.04, ηp2 = 0.06; M(SD) = 10.65 (60.11)). According to the intervention of attentional bias, the DI group had significantly decreased attentional bias scores in the post-test (F(1,65) = 6.58, p = 0.01, ηp2 = 0.01; M(SD) = -1.00 (65.02)), and their attentional vigilance scores were significantly reduced after dCBT-I in the DI group (F(1,65) = 4.00, p = 0.05, ηp2 = 0.06; M(SD) = 10.65 (60.11)). Detailed information on the MANOVA results is presented in Table 3. Table 3Repeated measure MANOVA results of Study 2GroupBaselinePost-testFpTimeGroup × TimeM(SD)M(SD)FpFpPSQIDI7.83(2.58)5.10(2.00)31.230.00154.060.0010.280.60CE8.08(2.71)4.92(2.10)24.760.001FSSDI41.52(8.77)37.40(10.08)4.260.043.510.070.420.52CE43.56(12.62)41.56(9.01)0.600.44Attentional biasDI39.02(69.37)-1.00(65.02)6.580.010.680.415.330.02CE14.18(54.32)33.08(101.88)0.880.35Attentional vigilanceDI46.66(114.13)10.65(60.11)4.000.050.040.845.040.03CE4.43(54.48)34.54(68.81)1.670.20Attentional disengage-mentDI-7.64(71.37)-11.64(41.66)n/an/a0.410.520.090.76CE9.75(60.20)-1.46(78.61)n/an/aAbbreviation: DI dCBT-I intervention group, CE conventional education group, PSQI Pittsburgh Sleep Quality Index, FSS Fatigue Severity Scale Repeated measure MANOVA results of Study 2 Abbreviation: DI dCBT-I intervention group, CE conventional education group, PSQI Pittsburgh Sleep Quality Index, FSS Fatigue Severity Scale Discussion Study 1: biased attention of individuals with insomnia In line with our hypothesis, individuals with potential insomnia showed selective attention toward sleepy pictures, which was characterized as attentional vigilance toward sleep-related cues. The current study demonstrated that one of the clinical manifestations of insomnia is dysfunctional sleep-related cognition. This finding is consistent with well-validated empirical research [7, 12] and enriches the theory of the evolution of psychophysiologic insomnia following the A-I-E pathway raised by Espie and his colleagues [7, 8]. The results of Study 1 reinforced that sleep quality was influenced by selective attention, reflecting the reasonableness of the "A" in the A-I-E pathway. More importantly, we further elaborated that the individuals who only experienced transient sleep disturbances in the A-I-E pathway not only had selective attention to stressors such as illness, mental, and emotional responses but also showed selective attention shifting toward sleep-related cues, similar to those with persistent psychophysiologic insomnia. This suggests that selective attention to sleep-related information, which is the first-factor affecting sleep–wake automaticity in the A-I-E pathway, may serve as a critical bridge linking transient poor sleepers and primary insomnia patients. This provides new insights into refining the A-I-E pathway model. While the A-I-E pathway is not a linear functioning but three intersecting stages [41], more studies in the future can prioritize exploring the interaction between attentional bias, distorted explicit intention, and counterproductive engagement for sleep [7]. The association between poor sleepers and attentional bias, as examined in Study 1, confirms that both the A-I-E pathway and Harvey's model demonstrate the critical role of distorted cognitive components in insomnia symptoms [7, 9]. Despite differences in sleep habits between Eastern and Western individuals, both models exhibit cross-regional applicability [42], making them effective in explaining sleep-related attentional bias in Chinese populations. The result that the PS group had a higher FSS score was in line with the finding that there was a significant correlation between symptoms of insomnia (i.e., poor sleep quality) and daytime fatigue, and individuals with mild symptoms of insomnia tend to emphasize their feelings of fatigue in the daytime [43]. However, the causal relationship between fatigue and symptoms of insomnia is unclear. Hence, the causality of these two factors should be explored in future studies (e.g., utilizing a cross-lagged panel model to probe the mutual influence of insomnia and fatigue). The PS group only presented attentional vigilance in dot-probe tasks rather than attentional disengagement, which is inconsistent with previous studies [14, 19]. The inconsistency may derive from different preconditions in general insomniac symptoms [19]. Espie and colleagues [7] conceptualized selective attention related to insomnia by describing different preconditions for sleep-related attentional bias. For instance, among patients having a stronger threatening driver, increasing distress related to sleeplessness, their inability to resolve sleeplessness is perceived as a threat, and the resulting sleep deprivation leads to heightened anxiety [7]. Conversely, for patients holding a stronger craving driver or preoccupying longing for sleep, sleep is their primary reinforcer [7]. Building on the notion of threatening drivers, it can be referred that in the study reflecting attentional disengagement among individuals with insomnia [44]. they were more likely motivated by their anxiety associated with sleeplessness. The disappearance of a significant effect of attentional disengagement after controlling for depression and anxiety scores in Jansson-Fröjmark's research [14] supports this inference. Following the same reasoning, any participants with severer than mild state anxiety were excluded from analysis and the results in Study 1 showing attentional vigilance but not disengagement could be explained in much the same light. It can be further deduced that participants in the current study who had heightened attentional vigilance toward sleep-related cues were probably driven more by a craving for sleep, so their goal-directed behavior is centered on obtaining sleep. In line with this reasoning, it is natural to disclose a potential limitation of Harvey’s cognitive model of insomnia [9], which overemphasized state anxiety in precipitating the selective attention to sleep-related cues and overlooked the mixed influences of different thinking styles. Based on the previous discussion, it can be concluded that selective attention is a composite expression of various bias patterns [45], which are associated with different preconditions. Distressed well-being is not inherently linked to the concept of selective attention but rather serves as a precondition for a specific bias pattern (i.e., attentional disengagement). This cognitive model overlooks the fact that sleep loss, disruption, and perceived sleep inadequacy, which give rise to a longing for sleep [7], form a precondition opposite to one like worry or distress. Combined with the results of Study 1, which controlled anxiety levels, it can be inferred that a certain pattern of attentional bias (i.e., attentional vigilance) still evolves independent of state anxiety or distress. Under such conditions, Harvey’s model may not fully account for this aspect. Study 2: the effects of dCBT-I on individuals with insomnia Both the dCBT-I intervention group (DI) and conventional education group (CE) experienced a significant reduction in PSQI scores after the intervention. DI had a significantly lower attentional bias and vigilance after the intervention, indicating the efficacy of dCBT-I in addressing biased attention among participants with insomnia. Additionally, DI also showed a decrease in the level of fatigue, duplicating another study on the spillover effect of dCBT-I [26]. The parallel reduction in PSQI score is similar to Lancee’s finding that attentional bias modification training and placebo training both led to a general reduction in insomnia symptoms [33]. In this study, since participants were required to maintain a detailed sleep diary in a user-filled pattern, the expectancy effect may have influenced them to improve the quality of their diary entries and rate certain questionnaire items higher than they otherwise would have. Our study found that the dCBT-I intervention was more effective than conventional education in reducing attentional bias, which was consistent with our hypothesis. This finding is in line with previous studies demonstrating that CBT-I can effectively improve sleep efficiency and reduce insomnia symptoms in individuals with comorbid psychiatric disorders [46]. Moreover, this finding bridges the gap between dCBT-I and insomnia treatment. The dCBT-I intervention required adherence to a standard sleep schedule and restrictions on bedtime behaviors while also reducing attention to room light, time-checking, and other obsessive behaviors. Moreover, the cognitive reconstruction components of the intervention helped to reduce excessive self-monitoring of the internal environment and then be less alert to sleep-related cues. Additionally, the dCBT-I intervention's focus on cognitive change and behavioral control may not directly affect feelings of fatigue and cannot regulate participants' physical activities during the day in a remote setting, where uncontrollable factors can interfere. Limitations The current study has limitations worth noting. First, in this study, we opted not to use DSM-5 criteria due to ethical and logistical considerations. The considerations about the potential psychological impact and stigma of formal diagnostic labels guided the decision to avoid using these criteria. Additionally, our team did not possess the clinical qualifications necessary to apply DSM-5 accurately, which could have compromised data validity and posed a risk of false diagnoses. We chose alternative criteria that better aligned with the study's ethical standards and research context. Instead, we used the PSQI as a moderate yet less specific assessment tool to identify poor sleepers with insomnia symptoms. In future studies, researchers can use both DSM-5 diagnostic criteria and the PSQI to select participants, creating a spectrum from normal sleepers, and poor sleepers to primary insomnia to disclose the developmental trajectory of insomnia. This would enable the research findings to be generalized to individuals with diagnosed insomnia. Second, the remote nature of the dot-probe task and dCBT-I, conducted with college students during the COVID-19 pandemic, both indeed present potential limitations that deserve further consideration. Conducting this task in an uncontrolled home environment, rather than a standardized laboratory setting, could have introduced variability in the results due to factors such as distractions, inconsistent equipment (e.g., varying screen sizes or internet speeds), network stability, and differing levels of participant engagement (e.g., emotional salience). Such factors could affect reaction times and focus, potentially skewing the assessment of attentional biases toward sleep-related cues. On the other side, college samples have more flexible schedules than community samples. According to previous findings [47, 48], individuals with more control over their schedules tend to exhibit fewer symptoms of depression and experience better sleep. In the current study, the effects of dCBT-I may be diminished due to the relatively mild nature of sleep disturbances performed in the sample. Third, although dCBT-I was shown to be a viable substitute for in-person therapy [49], further research is needed to compare its efficacy to in-person CBT-I, which is still considered critical in correcting maldeveloped behaviors of chronic insomnia [50]. A shift in using a longitudinal design to examine the long-term effects of a repeated, periodic dCBT-I on the improvement of attentional bias. Moreover, the nature of the sample in the present study posed a substantial influence on the results. On the one hand, in Study 1, researchers who excluded individuals with elevated PHQ-9 and GAD-7 scores may have created an inconclusive sample, as comorbid depression and anxiety are quite common among individuals with insomnia [46]. Even though the vigorous exclusion criteria prevent other general psychological factors, several studies have supported that individuals with insomnia frequently have comorbid mental disorders, particularly depression or anxiety disorders [11, 51]. In this case, rigorous exclusion criteria may give rise to biased groups of participants and exert substantial influence on the results. On the other hand, a previous study found that the small sample size may explain the marginal significance of the result [14], which may also explain the lack of a significant difference in participants' fatigue scores after the intervention. From another perspective, according to the updated psychobiological inhibition model by Espie in 2023 [50], he identified various emotional states, including anxiety and distress, physiological arousal, behaviors (e.g., unstable patterns of sleep), and cognitive dysfunctions (e.g., attentional bias), as four distinct maladaptive conditioning factors for sleep. They emphasized the specific contributions of each of these factors to insomnia. This suggests that emotions may not necessarily contribute to insomnia through cognitive distortions alone but might have a new pathway. Building on this idea, we can think beyond the craving and threatening hypotheses that are bound by cognitive processes in the insomnia realm [19]. Since this study only considered individuals with neutral emotional states, it could not independently explore the potential emotion-insomnia pathway. Therefore, in future research, we might include certain levels of anxiety and depression as the characteristics of insomnia and combine DSM-5 and PSQI to establish composite inclusion criteria to reset the standard of participant recruiting to enlarge the sampling and assess their sleep-related attentional bias. The longitudinal study is another viable orientation for the causal link between cognitive and emotional components, attentional bias, and insomnia. Conclusions In conclusion, our study supports the presence of explicit attentional bias in individuals with symptoms of insomnia via a sleep-related vocabulary dot-probe task, and dCBT-I is effective in correcting it, improving sleep quality, and reducing daytime fatigue. These findings align with research on individuals diagnosed with insomnia and models of insomnia. While not conclusive, these findings provide implications for the clinical characteristics of individuals with mild/episodic insomnia and psychological treatments. Furthermore, insomnia is a significant predictor of other psychiatric disorders [52], so our study may lay the foundation for early diagnosis of related morbidities. Acknowledgements We express our sincere gratitude to all the authors and organizations for their efforts and support. Abbreviations Cognitive behavioral therapy Cognitive behavioral therapy for insomnia Digital cognitive behavioral therapy for insomnia Epworth Sleepiness Scale Patient Health Questionnaire-9 Generalized Anxiety Disorder 7-item Poor sleepers group Normal sleepers group Standard deviation Millisecond Pittsburgh Sleep Quality Index Chinese version of The Pittsburgh Sleep Quality Index Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition Fatigue Severity Scale Statistical Package for the Social Sciences version 25 Digital cognitive behavioral therapy for insomnia intervention group Conventional education group Multivariate analysis of variance Attention-Intention-Effort pathway among insomnia Authors’ contributions Writing, Y.W. and JW.L. ; Review & editing, Y.W. and JW.L. ; Conceptualization, Y.W. and JW.L. ; Data curation, Y.W. and JW.L. ; Formal analysis, Y.W. and JW.L. ; Investigation, Y.W., HJ.L., W.H. and JL.L. ; Methodology, Y.W. and HY.L. ; Supervision, HJ.L., W.H. and JL.L. ; Project administration, HJ.L., W.H. and JL.L. ; Software, HY.L. ; The roles of the authors in this paper are all equally important. All authors have read and agreed to the published version of the manuscript. Funding This research received no external funding. Data availability The data used in this study are available upon reasonable request from the authors. Declarations Ethics approval and consent to participate The present study was carried out according to relevant guidelines and regulations. This study protocol was reviewed and approved by Shanghai Normal University Ethics Association. Before inclusion in the study, we received informed consent from all participants. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yimei Wu and Jingwen Li contributed equally to this work. Contributor Information Hongying Liu, Email: hongyingliu@91jkys.com. Haijiang Li, Email: haijiangli@shnu.edu.cn. References Associated Data Data Availability Statement The data used in this study are available upon reasonable request from the authors.