Reliability and Validity of the Adaptive Cognition About Sleep-2, an Abbreviated Version of the Adaptive Cognition and Behaviors About Sleep-6
Article information
Abstract
Background and Objective
The Adaptive Cognition and Behaviors about Sleep-6 (ACBS-6) comprises two factors, adaptive sleep-related cognitions (Items 1–4) and behaviors (Items 5 and 6). We aimed to develop a shortened version.
Methods
Four anonymous online surveys of Korean adults were analyzed. In a general-population derivation sample (n=600), items most relevant to adaptive sleep cognition and most strongly correlated with the Insomnia Severity Index (ISI) were selected. Their reliability and validity were checked among three independent validation samples such as the general population (n=600), respondents reporting insomnia (n=600), and respondents reporting insomnia and taking hypnotics (n=500). Internal consistency was estimated with the Spearman–Brown-corrected split-half coefficient. Mediation analyses explored whether adaptive cognition mediates the association between insomnia severity and the two-item questionnaire version of the discrepancy between desired time in bed and desired total sleep time (DBSTq-2).
Results
ACBS-6 Items 2 and 4 were selected as the Adaptive Cognition about Sleep-2 (ACS-2). The split-half coefficient ranged from 0.59 to 0.75. The ACS-2 score was inversely correlated with the ISI in all four samples (r=−0.14 to −0.23, all p<0.01), and its correlation with the Dysfunctional Beliefs and Attitudes about Sleep-6 items ranged from −0.11 to 0.08. The ACS-2 did not mediate the association between the DBSTq-2 and the ISI.
Conclusions
The ACS-2 is a feasible brief rating scale measuring adaptive cognition about sleep. Adaptive cognition appears to be distinct from dysfunctional beliefs about sleep, and not the opposite pole of the same dimension.
INTRODUCTION
Cognitive-Behavioral Therapy for insomnia (CBT-I) is recommended as the first-line treatment of insomnia [1]. The 3P model [2] explains that insomnia is precipitated by a stressor, but it persists even after the stressor has resolved because of perpetuating factors such as maladaptive sleep habits and dysfunctional thinking about sleep. In the cognitive model of insomnia [3], patients with insomnia worry excessively about their sleep and about the daytime consequences of not sleeping enough, and this aggravates sleep disturbance. The anxious state then triggers selective attention to sleep-related threat cues and monitoring of them, which leads to a distorted perception of their sleep and daytime deficit. Dysfunctional beliefs about sleep and counterproductive safety behaviors exacerbate this process [3]. The attention–intention–effort model [4] extends this account by proposing that the intention to sleep and the effort to achieve it inhibit the automaticity of normal sleep onset. Thus, sleep-related cognition itself can act as a perpetuating factor of insomnia [5]. Cognitive therapy is one component of CBT-I, and the efficacy of CBT-I is well established [6]. It targets these beliefs to reduce insomnia severity.
Dysfunctional beliefs about sleep refer to unrealistic thoughts and attitudes about sleep which contribute to the onset and persistence of insomnia [7]. They are the most widely studied of the sleep-related cognitions. They include rigid expectations about how much sleep is necessary, exaggerated worries about the consequences of not sleeping enough, and a sense of helplessness about one’s own sleep. The Dysfunctional Beliefs and Attitudes about Sleep-16 items (DBAS-16) [8] is the rating scale most commonly used to measure them, and a 6-item version, the DBAS-6, was later derived from it using a machine learning approach [9]. Individuals with insomnia show higher scores on these beliefs than good sleepers, with several items discriminating the two groups well [10]. The scores also decrease over the course of CBT-I, and the degree of change is associated with sleep improvement [11,12]. However, reducing dysfunctional beliefs during treatment does not always improve insomnia [13], and the pattern of beliefs may be more informative than the total score [14].
Meanwhile, the adaptive side of sleep-related cognition has rarely been explored. Adaptive cognition about sleep refers to accepting sleep as it occurs and refraining from the effort to control it. The DBAS-16 and its shortened versions measure only the dysfunctional pole of sleep-related thought. A low DBAS score means that a person does not have dysfunctional beliefs about sleep, but it does not mean that the person has adaptive cognition about sleep. In the coping literature, adaptive and maladaptive coping styles were reported to be distinct constructs and not simply opposites of each other [15]. A similar asymmetry has been proposed for sleep appraisal, and a rating scale measuring both the positive and the negative appraisal of sleep was developed for this reason [16], with a Korean version subsequently validated [17]. Acceptance-based models of insomnia also emphasize that letting go of the effort to control sleep can be a therapeutic target [18]. Psychological inflexibility has been reported to be associated with insomnia severity, and this association was mediated by dysfunctional beliefs about sleep and by sleep-related effort or preoccupation [19,20]. Thus, dysfunctional beliefs may be located between other sleep-related cognitions and insomnia severity. Insomnia is not driven by cognition alone. The proportion of slow wave sleep decreases with aging, and sleep–wake regulation is also altered among older adults [21]. In such cases, an adaptive attitude toward sleep may be more helpful than the correction of distorted beliefs.
The Adaptive Cognition and Behaviors about Sleep-6 (ACBS-6) was developed to measure adaptive cognition and behavior about sleep [22]. It comprises two factors, adaptive sleep-related cognitions and adaptive sleep-related behaviors. The cognition factor targets core CBT-I goals such as accepting the natural variability of sleep, reducing sleep-related worry, and reframing daytime fatigue. The behavior factor reflects the homeostatic principle that daytime activity promotes sleep. The two factors capture conceptually different aspects, and summing them may obscure the contribution of each. In the development study, the ACBS-6 total score was not significantly correlated with insomnia severity, although it contributed inversely to insomnia severity in a regression model that included other sleep-related cognitions [22]. Similar to the DBAS [14], the cognition factor of the ACBS-6 may be more related to insomnia severity than the total score.
When CBT-I is delivered digitally [23], sleep-related cognition can be assessed at every session to follow how it changes, and a scale short enough to bear that frequency is required. Brief and stepped forms of behavioral treatment have been proposed to improve the accessibility of CBT-I [24,25], and shortened versions of existing rating scales have been developed with the same purpose [26]. We therefore aimed to develop a shortened version of the ACBS-6 and to examine its reliability and construct validity. It was derived in one general-population sample and examined in three independent validation samples, which comprised the general population, respondents reporting insomnia, and respondents reporting insomnia and taking hypnotics. We also explored whether adaptive cognition mediates the association between insomnia severity and the discrepancy between desired time in bed and desired total sleep time, measured with a 2-item questionnaire (DBSTq-2). Adaptive cognition was expected to mediate this association, as dysfunctional beliefs about sleep have been reported to do [19,20].
METHODS
Participants
This study used data from four independent anonymous online surveys of the Korean general population conducted through a professional survey company (EMBRAIN) in 2025. Across all surveys, eligible respondents were aged 18 to 80 years, reported a habitual bedtime between 9 pm and 1 am (to exclude participants with circadian rhythm sleep–wake disorders), and were not shift workers. The derivation sample (Sample I) and one validation sample (Sample II) were general-population samples; a third sample (Sample III) comprised respondents reporting insomnia (difficulty initiating or maintaining sleep over the past three months); and a fourth (Sample IV) comprised respondents reporting insomnia and taking hypnotics. Target sample sizes were set by quota allocation across demographic cells following the central limit theorem [27]. Samples I, II, and III each targeted 600 participants, allocated as 50 per 12 demographic cells (sex×six age groups), and Sample IV targeted 500 participants, allocated as 100 per five age strata (18–29, 30–39, 40–49, 50–59, and 60 years or older). Sample I was surveyed from June 16 to 24, 2025; a total of 6,087 enrollment emails were sent to about 1.8 million registered panel members, 1,308 accessed the survey, and 639 completed it, yielding 600 analyzed responses. Sample II was surveyed from October 23 to November 4, 2025; 6,472 emails were sent, 1,062 accessed the survey, and 633 completed it, yielding 600 responses. Sample III was surveyed from June 18 to 24, 2025; 7,558 emails were sent, 1,882 accessed the survey, and 640 completed it, yielding 600 responses. Sample IV was surveyed from July 30 to August 14, 2025; 81,571 emails were sent, 19,072 accessed the survey, and 532 completed it, yielding 500 responses.
For each survey, the fastest 5% of respondents within each quota and responses whose average time spent between questions exceeded three times the overall average were excluded to enhance data quality.
Rating Scales
Insomnia Severity Index
The Insomnia Severity Index (ISI) is a rating scale which assesses an individual’s insomnia severity over the previous two weeks [28]. It includes seven items which can be scored on a 5-point Likert scale (0–4), and a higher total score reflects a more severe degree of insomnia. In this study, we applied the Korean version of the scale, whose psychometric properties have been reported previously [26]. Cronbach’s alpha of the ISI ranged from 0.82 to 0.89 across the four samples.
Adaptive Cognition and Behaviors about Sleep-6
The ACBS-6 is a 6-item self-report rating scale which measures the adaptive cognition and behavior about sleep [22]. It has two factors, adaptive sleep-related cognitions (Items 1–4) and adaptive sleep-related behaviors (Items 5 and 6). Each item can be scored from 0 (strongly disagree) to 10 (strongly agree), and a higher averaged score reflects a more adaptive cognition or behavior about sleep.
Dysfunctional Beliefs and Attitudes about Sleep-6 items
The DBAS-6 is a data-driven shortened version [9] of the DBAS- 16 [8], which is the rating scale most commonly used to measure an individual’s dysfunctional beliefs about sleep. It includes six items which can be scored from 0 (strongly disagree) to 10 (strongly agree), and a higher averaged score reflects a more dysfunctional belief about sleep. Cronbach’s alpha of the DBAS-6 ranged from 0.75 to 0.82 across the four samples.
Discrepancy between desired time in bed and desired total sleep time questionnaire-2
This index is calculated as the difference between the desired time in bed (dTIB) and the desired total sleep time (dTST) [29]. The DBSTq-2 is a two-item questionnaire version of this index, which consists of two items: “I desperately need at least five hours of sleep” and “I want to fall asleep early in the evening and sleep well into the late morning.” Each item is scored from 0 (strongly disagree) to 10 (strongly agree), and the total score is the average of the two items, with a higher score reflecting a larger discrepancy.
Statistical Analysis
Item distributions were checked for approximate normality using skewness and kurtosis within ±2. In the derivation sample, we first screened the items for content relevance, and among them selected the two items most strongly correlated with the ISI. Because the shortened version includes only two items, internal consistency was estimated with the Spearman–Brown-corrected split-half coefficient. Pearson correlations with the ISI and the DBAS-6 were computed to examine convergent and discriminant validity in all samples, and the correlation of each ACBS-6 item with the DBSTq-2 was also examined. Mediation analyses were performed in each sample to explore whether adaptive sleep cognition, measured either with the ACBS-6 or with the shortened version, mediates the association between the DBSTq-2 and insomnia severity. The same mediation models were additionally applied to all participants of the four samples combined (n=2,300) to summarize the overall pattern. All variables were standardized within each sample before the mediation analyses, and the 95% confidence interval of the indirect effect was estimated with 5,000 bootstrap resamples. Statistical analyses were performed using JASP version 0.98 (JASP Team; https://jasp-stats.org), and a two-tailed p-value <0.05 was considered statistically significant.
RESULTS
The demographic and clinical characteristics of the four samples are presented in Table 1. Samples I to III were comparable in age and sex, whereas Sample IV was younger and included more women. Past psychiatric history and reported insomnia became more frequent from Sample I to Sample IV, and the mean ISI and DBAS-6 scores were higher in Samples III and IV than in the two general-population samples. The mean ACBS-6 score ranged from 3.8 to 4.3 across the four samples.
All four cognition-factor items (Items 1–4) were judged relevant to adaptive sleep cognition on content review, and among them the two most strongly correlated with the ISI in the derivation sample were selected (n=600) (Table 2): Item 2 (“I do not worry much about not being able to sleep,” r=−0.26) and Item 4 (“I feel tired not because I lack sleep, but because I have too much to do,” r=−0.12). These became Items 1 and 2 of the new scale (hereafter Q1 and Q2, respectively). The two excluded cognition- factor items showed correlations of −0.09 for Item 1 and −0.11 for Item 3 with the ISI, and the behavior-factor items were not associated with it (r=0.02 for Item 5 and −0.05 for Item 6). Item 5 was also positively correlated with the DBAS-6 (r=0.10, p=0.012).
The two selected items, named the Adaptive Cognition about Sleep-2 (ACS-2) and scored as their average, were then examined in the derivation sample and three independent validation samples (Table 3). The two items satisfied the normality assumption (skewness 0.33 to 0.82, kurtosis −0.80 to −0.17) and yielded Spearman–Brown-corrected split-half coefficients of 0.60 in the derivation general-population sample (Sample I), 0.59 in the general population (Sample II), 0.65 in respondents reporting insomnia (Sample III), and 0.75 in respondents reporting insomnia and taking hypnotics (Sample IV). The ACS-2 score was inversely correlated with the ISI in each validation sample (r= −0.14, −0.18, and −0.21, respectively; all p<0.01), reproducing the association observed in the derivation sample (r=−0.23, p<0.01). The correlation with the ISI increased from the ACBS-6 total score to the four cognition-factor items and again to the two items of the ACS-2 (r=−0.05 to −0.15, −0.08 to −0.19, and −0.14 to −0.23, respectively). Cronbach’s alpha of the ACBS-6 was 0.751, 0.769, 0.733, and 0.808 in Samples I to IV, respectively, which was higher than the split-half coefficient of the ACS-2. The correlation of the ACS-2 with the DBAS-6 was −0.11 in the derivation sample and ranged from −0.01 to 0.08 in the validation samples.
Mediation analyses were performed to explore whether adaptive sleep cognition mediates the association between the DBSTq- 2 and insomnia severity (Table 4). The DBSTq-2 score was significantly correlated with the ISI in all four samples (r=0.42 to 0.50, all p<0.01), but it was not correlated with the ACS-2 (r= −0.01 to 0.06), and its correlation with the ACBS-6 ranged from 0.05 to 0.13. When the ACBS-6 was entered as a mediator (Table 4), the indirect effect was significant in Samples II and III (−0.014 in both) but not in Samples I and IV, and it accounted for less than 3% of the total effect in every sample. Because these indirect effects were in the inverse direction, the direct effect slightly exceeded the total effect in all four samples. When the ACS-2 was entered as a mediator (Table 4), the indirect effect was not significant in any of the four samples. In both models, the direct effect of the DBSTq-2 on the ISI remained significant, and the path from the mediator to the ISI was significant in all four samples (β=-0.18 to −0.11 for the ACBS-6 and −0.22 to −0.17 for the ACS-2). The same pattern was observed when all participants of the four samples were combined (n=2,300) (Fig. 1). In the combined analysis, the indirect effect through the ACBS-6 was significant but accounted for 2.8% of the total effect. At the item level, the positive association between the DBSTq-2 and the ACBS-6 came from Item 1 and the two behavior-factor items. In the derivation sample, these three items were positively correlated with the DBSTq-2 (r=0.09, 0.10, and 0.09, respectively; all p<0.05), and Item 1 showed the same pattern in the other three samples (r=0.09 to 0.21). In contrast, the two items forming the ACS-2 were not correlated with the DBSTq-2 in any of the four samples (r=−0.03 to 0.06).
Mediation models showing the pathway from the DBSTq-2 to insomnia severity through the ACBS-6 (A) and the ACS-2 (B), among all participants of the four samples combined (n=2,300). **p<0.01. DBSTq-2, discrepancy between desired time in bed and desired total sleep time questionnaire-2; ACBS-6, Adaptive Cognition and Behaviors about Sleep-6; ACS-2, Adaptive Cognition about Sleep-2.
DISCUSSION
In this study, we developed the ACS-2, a 2-item rating scale drawn from the ACBS-6, and explored its reliability and validity. The ACS-2 score was inversely correlated with insomnia severity in all four samples. Its correlation with the DBAS-6 was not consistent in direction across the four samples, and it did not mediate the association between the DBSTq-2 and insomnia severity. However, the path from the ACS-2 to the ISI remained significant after the DBSTq-2 was controlled for.
The two items of the ACS-2 reflect the reduction of sleep-related worry and effort. Q1 (“I do not worry much about not being able to sleep”) addresses sleep anxiety and rumination, which is a core target of cognitive therapy in CBT-I and a central mechanism in the cognitive model of insomnia [3]. Q2 (“I feel tired not because I lack sleep, but because I have too much to do”) reframes daytime tiredness by shifting the focus away from sleep loss toward lifestyle or stress factors. The absence of sleep effort is also the core of the attention–intention–effort model [4]. Both items describe the cognitive state which cognitive therapy aims to promote, and we can assume that this is why they were inversely correlated with insomnia severity. The two behavior-factor items, however, state a homeostatic principle as a fact, and they were not correlated with the ISI in the derivation sample. This may explain why the ACBS-6 total score was not significantly correlated with the ISI in the development study, in which it contributed to insomnia severity only after other sleep-related cognitions were controlled for [22]. In the present study, however, the total score was significantly correlated with the ISI in three of the four samples.
The correlation between the ACS-2 and the DBAS-6 ranged from −0.11 to 0.08 across the four samples, and its direction was not consistent. This result can be expected if adaptive and dysfunctional sleep-related cognitions are distinct constructs, and not two ends of a single dimension. The same asymmetry has been reported for coping styles and for the appraisal of sleep [15,16]. An individual may hold few dysfunctional beliefs about sleep without holding a high level of adaptive cognition about sleep, and such an individual cannot be identified by measuring only the dysfunctional pole. Thus, cognitive change in CBT-I may require two distinct processes: the reduction of dysfunctional beliefs and the promotion of adaptive cognition. However, the two items did not behave in the same way. In the derivation sample, Q1 was significantly correlated with the DBAS-6 (r=−0.18, p< 0.01), whereas Q2 was not (r=−0.00). Whether the two items measure the same aspect of adaptive cognition needs to be explored in further studies.
The results of the mediation analyses support this interpretation. The DBSTq-2 score was significantly correlated with the ISI in all four samples. However, almost none of this association was mediated by adaptive cognition, because the path from the DBSTq-2 to the ACS-2 was not significant in any of the four samples. In contrast, the path from the ACS-2 to the ISI remained significant in all four samples even after the DBSTq-2 was controlled for. It can be assumed that adaptive cognition is related to insomnia severity independently of the discrepancy between the dTIB and the dTST. Dysfunctional beliefs about sleep have been reported to mediate the influence of psychological inflexibility on insomnia [19,20]. Thus, the absence of such mediation for adaptive cognition suggests that adaptive cognition occupies a different position in the cognitive model of insomnia.
An indirect effect was observed when the ACBS-6 was entered as a mediator in Samples II and III, but this is not evidence that the full scale performs better than the ACS-2. The path from the DBSTq-2 to the ACBS-6 was positive in all four samples, which is difficult to explain with the concept of the scale. A larger discrepancy between the dTIB and the dTST is not expected to accompany a more adaptive cognition about sleep. In the derivation sample, Item 1 (“I am satisfied even with a small amount of sleep”) and the two behavior-factor items were positively correlated with the DBSTq-2, whereas the two items of the ACS-2 were not. Item 1 and the first item of the DBSTq-2 (“I desperately need at least five hours of sleep”) both refer to a small amount of sleep, one as acceptance and the other as a minimum requirement. Thus, participants may have responded to these superficially similar statements in the same direction. In addition, Item 5 was positively correlated with the DBAS-6 (r=0.10, p=0.012) in the derivation sample, and the adaptive behavior factor of the ACBS-6 was also reported to be correlated with the DBAS-6 in the development study [22]. For a rating scale measuring adaptive cognition, an inverse correlation with the DBSTq-2 is expected and the absence of a correlation is acceptable, but a positive correlation is difficult to explain. Therefore, we consider this pattern to be a limitation of the total score rather than an advantage of the full scale.
The correlation coefficient with the ISI was numerically higher for the four cognition-factor items than for the ACBS-6 total score, and higher again for the two items of the ACS-2. This ordering can be expected in the derivation sample, in which the two items were selected on the basis of their correlation with the ISI, but it was reproduced in each of the three validation samples. The internal consistency of the ACS-2, however, was lower than that of the ACBS-6 in all four samples. The ACBS-6 gains internal consistency from its greater number of items, but it also includes content which is not correlated with insomnia severity. Adaptive cognition is an explicit target of cognitive therapy, and a 2-item scale may be useful when it needs to be assessed repeatedly, such as at every session of CBT-I [23] or in an electronic sleep diary.
This study has several limitations. First, the analyses rest on cross-sectional data, so the relationship between adaptive cognition and insomnia severity should be confirmed longitudinally, and the direction of the mediation models could not be established. Second, the items were selected partly on the basis of their ISI association, and the margin between the selected Item 4 and the excluded Item 3 was narrow (r=−0.12 versus −0.11), so the same selection procedure needs to be repeated in an independent sample. Third, the split-half coefficients ranged from 0.59 to 0.75, and a two-item scale cannot be evaluated by factor analysis on its own because such a model is not identified; the ACS-2 is therefore best regarded as a brief screener rather than a precise trait estimate. Its correlations with insomnia severity ranged from −0.14 to −0.23. Test–retest reliability and responsiveness to CBT-I were also not assessed. Fourth, all four samples were drawn from the same general-population online panel rather than from a clinical setting, and all data were self-reported without objective sleep measurement or clinical interview. The insomnia status of Samples III and IV therefore rests on self-report, and it was not confirmed whether their symptoms reached a clinically significant threshold; the hypnotic use of Sample IV was likewise self-reported, so neither the prescription nor the type of medication could be verified. The findings should be replicated in a clinically diagnosed insomnia population. Finally, all samples were Korean, and the cultural generalizability of the item content remains to be established.
In conclusion, the ACS-2 is a feasible brief rating scale measuring adaptive cognition about sleep, and it was inversely correlated with insomnia severity among the general population and among respondents reporting insomnia. However, the ACS-2 consists of only two items, and the scale is best regarded as a brief screener. The ACS-2 was independent of dysfunctional beliefs about sleep, and it did not mediate the association between the DBSTq-2 and insomnia severity. These findings suggest that adaptive cognition is a separate dimension of sleep-related thought, and not merely the absence of dysfunctional beliefs about sleep.
Availability of Data and Material
Data will be available from the author when requested.
Notes
Ethics Statement
The study was conducted in accordance with the Declaration of Helsinki, and the requirement for written informed consent was waived; participation was contingent on voluntary agreement to the consent questions. All protocols were approved by the Institutional Review Board of Asan Medical Center (2025-0696 for Sample I, 2025-1331 for Sample II, 2025-0607 for Sample III, and 2025-0937 for Sample IV).
Conflicts of Interest
Seockhoon Chung, an editor of the Sleep Medicine Research, was not involved in the editorial evaluation or decision to publish this article.
Funding Statement
None
Acknowledgements
None
