Insomnia is defined as a frequent and persistent difficulty initiating or maintaining sleep, resulting in general sleep dissatisfaction, distress with regard to poor sleep, and broad impairments in daily function. About 10% of the general population experiences chronic insomnia, making insufficient sleep one of the most prevalent global health concerns [1]. Insomnia can occur in isolation or arise concurrently with a mental disorder, medical condition, or substance use, all of which are frequently exacerbated by an insomnia comorbidity [2]. A widely accepted cognitive model posits that five aberrant cognitive processes underlie and sustain insomnia: worry (also known as cognitive arousal), selective attention and monitoring, misperceptions of sleep and daytime deficits, unhelpful beliefs about sleep, and counterproductive safety behaviors [3]. In particular, it has been proposed that sleep-related worry can trigger autonomic arousal and emotional distress, culminating in an anxious state in which sleep is likely to be disturbed and daytime functioning can become impaired [4]. Further to this, experimental manipulations that increase or decrease sleep-related worry have been reported to respectively prolong or reduce sleep-onset latency. Worry that is related to a lack of sleep can be a target for interventions such as cognitive behavioral therapy for insomnia (CBT-I).

The Anxiety and Preoccupation about Sleep Questionnaire (APSQ) is a brief (10-item) self-report survey specifically designed to assess the severity of sleep-related worry among patients with insomnia [5] and also evaluate the effects of insomnia treatment. The 10 items in APSQ were designed to assess the frequency of sleep-related worry throughout nighttime and daytime as well as a broad spectrum of insomnia symptoms such as poor mood (anxiety and depression), arousal, abnormal sleep parameters, and sleep disorders. A factorial validity study has previously reported that a two-factor model accounts for 70.7% of the score variance, and that both factors demonstrate high internal consistency (Cronbach’s alpha values of 0.91 and 0.86). Appropriate discriminant and convergent validity have also been demonstrated based on strong correlations with sleep parameters and daytime impairments [6].

Sleep-related worry and maladaptive cognition regarding sleep can also be measured using several established assessment tools, including the Dysfunctional Beliefs and Attitudes about Sleep (DBAS)-16 items [7], the Metacognitions Questionnaire-Insomnia (MCQ-I), which measures an individual’s metacognitive processes concerning sleep [8], and the Glasgow Sleep Effort Scale (GSES), which measures an individual’s persistent preoccupation with sleep or sleep effort [9]. All these measures have been translated into Korean, with confirmed validity and reliability for these Korean versions reported by prior domestic studies [10-12]. Notably, however, the APSQ has yet to be translated into Korean. We speculate that a Korean APSQ could provide complementary data on sleep-related worry for a more comprehensive description of maladaptive sleep-related cognition and for assessing the effects of sleep-related worry on insomnia severity.

The aims of this current study were to formulate a Korean version of the APSQ scale and examine its reliability and validity for the general population by comparing its response patterns with those of the aforementioned questionnaires on sleep-related cognition. We further examined whether the APSQ scale is a more useful tool for measuring sleep-related worry compared to these other measurements.

The first 300 completed surveys from the general population had an exact 1:1 ratio of males and females and 36.7% of these respondents were single. The mean age of the subjects was 49.0±16.3 years, 18.7% had experienced past psychiatric symptoms, 13.0% had current psychological distress, and 29.0% were currently experiencing insomnia. The demographic and clinical characteristics of these participants are summarized in Table 1.

This study assessed the reliability and validity of a new Korean-language version of the APSQ scale by comparing its responses to multiple preexisting scales and exploring its psychometric properties using CFA and IRT. The APSQ scale showed satisfactory reliability and validity, and psychometric properties that were highly similar to the original English-language version. We also demonstrated that the APSQ is an informative rating scale. Notably, the APSQ scores were strongly correlated with DBAS-16, confirming their utility in assessing sleep-related cognition. These strong correlations indicated that sleep-related worry might be associated with an individual’s dysfunctional beliefs about sleep. Hence, the proposed APSQ tool provides information for the research and clinical evaluation of sleep-related worry that complements existing measures of various other sleep problems, thereby supporting the comprehensive assessment of insomnia symptoms and underlying maladaptive beliefs regarding sleep.

CFA indicated that the Korean APSQ fits a two-factor model similar to the original language version [5]. The two-factor model of the APSQ showed good model fit among the whole sample population with good loading values (factor I: 0.705 to 0.850 and factor II: 0.707 to 0.859). The reliability was also good as indicated by Cronbach’s alpha values of 0.965 for the full scale APSQ, 0.954 for factor I, and 0.928 for factor II. Convergent validity was high as evidenced by the strong correlations with multiple previously established rating scales. Further to this, we re-confirmed the validity and reliability of the ISI, DBAS-16, MCQI-14, and GSES prior to these comparisons (Table 4). While multiple studies have confirmed the reliability and validity of the ISI and DBAS-16 for the Korean population [10,16], the MCQI-14 is a newly developed shortened version of the MCQ-I scale [11], and neither has been widely used in Korea until recently. Moreover, only one prior study has investigated the psychometric properties and validated the reliability of the GSES for the Korean general population [12]. Hence, this present study provides needed support for the validity and reliability of the MCQI-14 and GSES.

We also here observed excellent fit indices (CFI=1.000, TLI=1.002, RMSEA=0.000, SRMR=0.038) and needed to check for the possibility of overfitting. In the CFA, we used DWLS estimation, which is particularly relevant for understanding these excellent fit statistics. DWLS tends to produce better fit indices compared to Maximum Likelihood estimation. DWLS estimation also provides more robust and accurate model fits of the CFA model using categorical data. We therefore considered that the risk of overfitting was low. Furthermore, with a sample size of 300 and 33 degrees of freedom, our model had sufficient statistical power while maintaining parsimony.

The GRM results showed that all APSQ items had very high slope parameters, indicating strong sensitivity for gauging the severity of worry related to sleep. Further, scale information curves showed that APSQ provides significantly more information on the latent trait than the DBAS-16, MCQI-14, or GSES, particularly in the θ range of -2.5 to 2.5 (covering more than 95% of the general population). This indicated that the APSQ offers the most precise measurement for most individuals. We also found that DBAS-16 and MCQI-14 are less informative, while GSES is the least informative in this regard. Overall, therefore, the APSQ is the most effective measure of this sleep-related trait for the majority of the general population.

This study had several limitations of note. First and foremost was the potential sampling bias introduced by the online surveys. While these surveys did not target a specific population in order to assess the APSQ efficacy for identifying sleep-related worry among the wider community, and used sex and age quotas to adequately represent the characteristics of the general population (a method commonly used in public opinion research to reduce survey time and cost), open voluntary participation may have selected for respondents with a heightened interest in sleep-related issues [19]. Indeed, the average age of our present study cohort was above the median and a greater proportion of our subjects reported current insomnia than is usually detected in the general population. Despite these issues, however, quota sampling is considered useful when there is a clear model for analysis [20]. Nonetheless, the potential for response bias in our current data cannot be completely ruled out, and caution should be exercised in interpreting and generalizing these results. Second, while high CFI and TLI values combined with a low RMSEA are typically desirable, a perfect or near-perfect fit can indicate overfitting, which limits the applicability of the findings to other samples. We suggest, however, that the risk of overfitting in this case was low because the hypothesized model has a solid theoretical basis (i.e., the product of APSQ analyses), is not overly complex, and does not have an excessive number of parameters. As more data become available, cross-validation should be used to test the robustness and generalizability of the APSQ across different samples [21]. Third, we explored the reliability and validity of the Korean version of the APSQ among the general population and not a clinical sample of insomnia. We explored the conduct validity of the scale in the general population to examine its distribution, reliability, and factor structure across a broad range of insomnia severity, including individuals without insomnia and those with varying degrees of symptoms. In a future study, we will explore the reliability and validity of this Korean version of the APSQ scale among clinical samples of insomnia.

In conclusion, we have produced and validated a Korean version of the APSQ self-reported rating scale for measuring sleep-related worry in a general population. This APSQ version showed both good discriminative and informative capacity, highlighting its potential as an effective tool for assessing sleep-related worry in heterogeneous populations.