Keywords:
Insomnia - anxiety - worry - factor analysis - Anxiety and Preoccupation about Sleep
Questionnaire - reliability
INTRODUCTION
Insomnia is a condition characterized by difficulty falling asleep or maintaining
sleep or early-morning awakening with an inability to return to sleep. When such symptoms
occur at least three times a week for more than three months, they meet the diagnostic
criteria of insomnia disorder, which has an estimated prevalence of 10% according
to large multicenter studies[1],[2]. Insomnia is associated with many health problems, including impaired daytime functioning,
mood disturbance, reduced cognitive functions, altered immune function, and fatigue,
and has many direct and indirect negative financial consequences. Insomnia disorder
is therefore considered one of the most serious public health problems[3],[4].
Many theories aim to explain the development and maintenance of insomnia. Among these,
the cognitive model - which purports that worry, selective attention, and the monitoring
of these cognitive processes are critical in the persistence of insomnia - has been
attracting attention in recent years[5]. According to this model, worry is the starting point of a vicious circle, as it
activates the sympathetic nervous system, causing increased arousal and emotional
distress. Then, selective attention on insomnia increases, which (along with arousal)
leads to heightened awareness of factors that disrupt sleep, such as bodily sensations
or sounds, causing more worry. Selective attention can also create bias or misperceptions
about insomnia or its consequences, increasing worry. The cognitive model of insomnia
thus argues that the combination of worry, arousal, and emotional stress increases
anxiety, causing difficulty achieving or maintaining sleep[6]-[8]. In addition, this model suggests that anxiety about insomnia encourages sleep-related
safety behaviors, which are adopted to prevent worry about insomnia but are not demonstrably
beneficial to sleep. For example, drinking alcohol is a classic sleep-related safety
behavior, but though it may initially facilitate sleep, alcohol disrupts sleep structure,
causing more nighttime awakenings. Such safety behaviors do not improve sleep and
can lead to additional mental health problems such as substance abuse[9],[10].
Given the importance of the relationship between insomnia and worry, there is a need
to develop and analyze psychometric tools that assess insomnia worry for research
and clinical purposes. Pre-sleep arousal scale (PSAS), thought control questionnaire
- insomnia revised (TCQI-R), insomnia daytime worry scale (IDWS) are scales developed
and validated for insomnia-related worry. The PSAS, TCQI-R, and IDWS have some limitations.
The PSAS has a cognitive subscale, but it focuses on pre-sleep worry and does not
measure daytime worry. The TCQI-R’s worry subscale evaluates worry management but
not the presence and intensity of sleep- related worry, while the IDWS focuses only
on daytime worry, not on nighttime worry[11]-[13].
A fourth insomnia worry scale, the anxiety and preoccupation about sleep questionnaire
(APSQ), was created by Tang and Harvey (2004)[14] to assess worry about sleep. It measures two subdimensions - worries about the consequences
of poor sleep and worries about the uncontrollability of sleep. The scale’s validity
and reliability have been evaluated in a large sample, and it has been shown to correlate
with anxiety and sleep quality[15].
The APSQ is promising compared to other insomnia-related worry scales, as it is short
and designed to measure both the intensity of anxiety and preoccupation with insomnia.
Most importantly, it addresses both daytime and nighttime sleep-related anxiety[16]. Additionally, it is sensitive to the effects of treatment. These features make
the APSQ potentially useful for research and clinical purposes[15],[16]. Also, the APSQ-brief version (2-item form) displayed acceptable psychometric properties[17].
The items of any sleep scales can mean different in various languages and the scale
suitable for one society may not be suitable for the other one, and therefore some
items need to be modified. Also determining psychometric properties is very important
to understand how much the scale can measure the subject to be assed, whether it is
reliable or not, whether it can be applied again after a certain period. From this
point of view, we aimed to first perform the Turkish translation process of the original
APSQ and then investigate the psychometric properties of the 10-item Turkish version
of the APSQ in clinical (among patients with the major depressive disorder) and non-clinical
samples (among university students).
MATERIAL AND METHODS
Participants and study protocol
The study sample included 183 clinical and non-clinical participants. Of these, 42
were patients diagnosed with major depression based on the fifth revision of the diagnostic
and statistical manual of mental disorders (DSM-5) and admitted to the psychiatric
clinic of Kahramanmaraş Necip Fazil City Hospital. The remaining 141 were non-clinical
volunteers recruited from various faculties of Kahramanmaraş Sütcü Imam University.
There were no specific inclusion or exclusion criteria for non-clinical participants.
Inclusion criteria for the clinical group were an age between 18 and 65 and a diagnosis
of major depressive disorder (MDD) according to the DSM-51. Clinical and non-clinical participants were informed about the study, their written
consent was obtained and then they completed a psychological test battery including,
the socio-demographic form, Turkish APSQ, insomnia severity index (ISI), Pittsburg
sleep quality index (PSQI). Students filled the psychological test battery in a quiet
classroom environment, and the patients filled it in the outpatient clinic. To examine
test-retest reliability, the Turkish APSQ was readministered to 20 randomly selected
MDD patients and 25 randomly selected healthy volunteers fifteen days should be written
instead of three weeks. after the initial survey.
The study was conducted between June 2019 to December 2019, and the study protocol
was approved by the local ethics committee of Kahramanmaras Sutcu Imam University
(approval number: 29.05.2019/10; decision number: 05).
Translation process and content validity
Permission to adapt the APSQ to Turkish was obtained via email from Allison G. Harvey
(2002)[5], who developed the original scale. The scale was independently translated from English
into Turkish by three Turkish native speakers. These translations were examined by
the research team, who formed the Turkish scale by adopting the Turkish expressions
determined to best represent each English item. The Turkish form of the scale was
then translated back into English by a native English speaker, and the original scale
was compared with this translation for consistency. Finally, the scale items were
discussed by the research team and final versions were chosen.
After this translation process, the Davis technique was used to assess the scale’s
content validity. Through this technique, which has experts evaluate a scale’s suitability
and understandability, the scale’s 10 items were examined by 10 specialist psychiatrists
who determined whether the content of items was appropriate, highly appropriate, slightly
appropriate, in need of serious review, or non-compliant[18]. Based on these ratings, a content validity index was calculated. Next, a pilot
study that included face-to-face interviews with 10 university students and 10 MDD
patients assessed how subjects understood each item.
Measurement tools
Anxiety and preoccupation about sleep questionnaire
The 10-item APSQ was developed by Tang and Harvey (2004)[14] to measure sleep-related anxiety. Participants rate how true each statement is for
them during the last month on a 10-point Likert-type scale (from 1 “strongly disagree”
to 10 “strongly agree”). The scale has two subdimensions. Items 1-5 and 7 measure
the first subdimension (worries about the consequences of poor sleep), while items
6 and 8-10 measure the second (worries about the uncontrollability of sleep). Summing
the ratings of all items yields a total scale score. Higher APSQ scores indicate higher
insomnia-related anxiety and preoccupation. Sample items are “I worry about the amount
of sleep I am going to get every night” and “I worry about my loss of control over
sleep”. The internal consistency of the scale is 0.92, and its psychometric properties
have been examined in a large sample. The APSQ has been found to correlate with cognitive
arousal, sleep quality, sleep-related beliefs, anxiety, and depression[15].
Insomnia severity index
The ISI is a measurement tool with highly valid and reliable that evaluates the severity
of insomnia. The index consists of 7 questions and each is scored between 0 and 4
points. The total score ranges from 0 to 28 points. The higher the total score, the
greater the severity of insomnia. Scores of 15 and above are considered as clinical
insomnia[19]. Boysan et al. (2010)[20] conducted the Turkish version of the index.
Pittsburg sleep quality index
The PSQI is a self-administered questionnaire that evaluates sleep quality and disturbances
over a one-month time interval[21]. The scale consists of twenty-four items, nineteen of which are self-assessment
questions also five of which are questions answered by a roommate or a subject’s spouse.
Self-rated nineteen questions include the following seven subscales: (i) positive
quality of sleep; (ii) sleep latency; (iii) sleep duration; (iv) sleep efficiency;
(v) sleep disturbances; (vi) medication use for sleep; and (vii) daytime dysfunction.
The seven subscales are scored between 0 and 3 points. The total score ranges from
0 to 21 points with high scores indicate poor sleep quality. The validity and reliability
of the PSQI was shown acceptable in Turkish population[22].
Statistical analysis
The Statistical Package for the Social Sciences (SPSS) version 23 and SPSS Amos 24
were used to analyze the data from the present study. The sample variances were first
evaluated for normality and homogeneity, and all were in acceptable ranges. The presence
of outliers was also controlled. To meet the normal distribution assumption, kurtosis
and skewness parameters were set within ±2[23].
We started by evaluating participants’ sociodemographic characteristics using descriptive
statistics (number, percentage, average, and standard deviation). We used the Kaiser-Meyer-Olkin
(KMO) test to assess the adequacy of the sample size[24]. Suitability of the scale for factor analysis was evaluated via Bartlett’s test,
and explanatory factor analysis (EFA) was first applied to determine the factor structure[25]. Principal component analysis (PCA) and direct oblimin rotation methods were used
in the EFA. As confirmatory factor analysis (CFA) is recommended to further support
models, CFA was performed via SPSS Amos 24 to test the factor structure obtained via
EFA Many fit indices are used in CFA to reveal the adequacy and compatibility of the
tested model[26],[27].
Our analyses included the chi-square fit test, the comparative fit index (CFI), the
general fit index (GFI), the root mean square error of approximation (RMSEA), and
the incremental fit index (IFI). Cronbach’s alpha and item-total score correlations
were used to determine the scale’s reliability, and its temporal stability was assessed
via the test-retest method. Fifteen days after the baseline assessment, the Turkish
APSQ was readministered to 20 randomly selected MDD patients and 25 randomly selected
controls.
To determine the Turkish APSQ’s criterion validity, we examined its relationship to
scores on the PSQI, and the ISI using Pearson correlations. An independent samples
t-test was computed to compare the APSQ scores of MDD patients and controls to assess
the scale’s predictive validity. We divided all participants into good-sleepers group
and clinical insomnia group according to ISI scores that 15 point and above indicates
clinical insomnia. Then we compared mean scores of sleep measures between these groups
via independent samples t-test.
RESULTS
Description statistics and sleep parameters
All participants’ ages ranged from 18 to 53 with a mean of 24.10 years (SD ± 8.30).
Most were female (n=140, 76.5%) and university students (n=141, 82%). The mean age of the control group was 20.5, while the mean age in the
depression group was 35.9. Most of the control group and depression group were women.
Sleep parameters of the depression group were worse than the control group. Also,
the clinical insomnia rate was higher in the depression group according to the ISI.
[Table 1] shows the sociodemographic data and sleep parameters of both groups.
Table 1
Sociodemographic data and sleep parameters.
|
Control group
(n=141)
|
Depression group
(n=42)
|
|
Age (M (SD))
|
20.5 (1.84)
|
35.9 (10.3)
|
|
Gender (female)
|
%78
|
%71.4
|
|
Clinical insomnia
|
%17.8
|
%75
|
|
PSQI M(SD)
|
5.90 (4.58)
|
13.69 (3.43)
|
|
ISI M(SD)
|
9.52 (5.80)
|
17.78 (4.96)
|
|
APSQTOTAL M(SD)
|
33.35 (24.92)
|
67.54 (24.61)
|
|
APSQF1 M(SD)
|
21.47 (16.14)
|
37.78 (16.62)
|
|
APSQF2 M(SD)
|
11.41 (10.04)
|
25.83 (11.10)
|
Notes: APSQ = Anxiety and preoccupation about sleep questionnaire; F1 = Factor 1: worries
about the consequences of poor sleep; F2 = Factor 2: worries about uncontrollability
of sleep; PSQI = Pittsburg sleep quality index; ISI = Insomnia severity index; n =
Number of individuals; M = Mean; SD = Standard deviation.
Content validity
Based on expert scores, each item on the Turkish APSQ yielded a content validity index
above 0.8, exceeding accepted standards[18]. Thus, no item was removed. From a pilot administration with 20 individuals, it
was determined that all items were understandable and none required changes, and the
scale was finalized.
Construct validity
We assessed the suitability of our sample size for factor analysis with the KMO test
and Bartlett’s significance test. The KMO coefficient should be above 0.60, and Bartlett’s
sphericity test calculated in the chi-square test should be statistically significant[28]. In our study, the KMO coefficient was 0.94, and Bartlett’s significance test was
statistically significant (χ2=1652.80, df=45, p<.001), indicating that our data were suitable for factor analysis. The 10 items were
thus analyzed via EFA, using PCA with direct oblimin rotation. Inconsistent with the
original scale, PCA revealed only one factor with an eigenvalue above 1. However,
as the scree plot revealed two factors, we decided to rerun our PCA analysis and finally
identified two factors, as found in the original scale[15]. The scree plot is presented in [Figure 1] and the scale’s factor structure is outlined in [Table 2].
Table 2
Factor structure of Turkish APSQ with principle components analysis.
|
Items
|
Factor 1
|
Factor 2
|
|
1. I worry about the amount of sleep I am going to get every night
|
.629
|
|
|
2. I worry about how the amount of sleep I had last night is going to affect my day
time performance
|
.983
|
|
|
3. I worry about how the amount of sleep I get is going to afflict my health
|
.888
|
|
|
4. I worry about how much the amount of sleep I get will weaken my social ability
|
.907
|
|
|
5. I worry about how much the amount of sleep I get will shake my moo
|
.661
|
|
|
6. I worry about my loss of control over sleep
|
|
.649
|
|
7. I worry about my ability to stay awake and alert during the day
|
.838
|
|
|
8. I put great effort into trying to rectify my sleep problems
|
|
.566
|
|
9. My failure to rectify my sleep problems troubles me a lot
|
|
.832
|
|
10. My worry about sleep is persistent
|
|
.825
|
Notes: Only loadings above 0.3 are presented in the table. Factor 1: Worries about the consequences
of poor sleep; Factor 2: Worries about uncontrollability of sleep.
Figure 1 Scree-Plot Graph of the Factor Analysis.
In addition to EFA, we performed CFA to further test the two-factor model. The standardized
factor loads of scale items were found to vary from 0.72 and 0.89. CFI, GFI, IFI>0.900,
χ2/df<5, and RMSEA<0.0854 values are recommended as criteria for indicating a good fit
indices except if RMSEA were suitable[26],[28]. CFA results suggested a modification to improve the goodness of fit indices. After
combining the error variances of items 1 and 2 and of items 9 and 10, all fit indices
reached acceptable levels. Confirmatory factor analysis path diagram is shown in [Figure 2] and the fit indices of the scale are shown in [Table 3].
Table 3
Model-fit results of confirmatory factor analysis for Turkish version APSQ, primary
and after modification.
|
Model
|
RMSEA
|
CFI
|
GFI
|
IFI
|
χ2/df
|
p
|
|
Two-factor model with 10 items before the first modification
Two-factor model with 10 items 1st modification
|
.099
.089
|
.963
.971
|
.908
.918
|
.963
.971
|
2.796
2.434
|
<0.000
<0.000
|
|
Two-factor model with 10 items 2nd modification
|
.078
|
.979
|
.930
|
.979
|
2.100
|
<0.000
|
Notes: RMSEA = Root mean square error of approximation; CFI = Comparative fit index; GFI
= Goodness of fit index; IFI = Incremental fit index; χ2/df = Normalized chi-square.
Figure 2 Confirmatory factor analysis Path diagram for two factor model of Turkish version
APSQ and standardized factor loadings - F1: Worries about the consequences of poor
sleep and F2: worries about uncontrollability of sleep.
Internal consistency and test-retest reliability
Cronbach’s alpha correlation analysis was used to determine the scale’s internal consistency,
and item analysis was used to determine the quality of items. Cronbach’s alpha values
of total scale, factor 1 and factor 2 were 0.95, 0.935, and 0.9, respectively, indicating
excellent internal consistency. [Table 4] lists the item-total score correlations and Cronbach’s alpha values that would result
from deleting each item. The test-retest correlation for MDD patients coefficient
was calculated as r=0.661 (n=20; p<0.001) and for non-clinical population as r=0.828 (n=25; p<0.001).
Table 4
Item-total statistics for Turkish APSQ in study group.
|
Item
|
Mean
|
Mean
|
Corrected Item-Total
Correlation
|
Cronbach’s Alpha
if item deleted (α)
|
|
Item 1
|
4.18
|
3.51
|
.726
|
.952
|
|
Item 2
|
4.16
|
3.25
|
.825
|
.948
|
|
Item 3
|
4.28
|
3.30
|
.801
|
.949
|
|
Item 4
|
4.48
|
3.37
|
.842
|
.948
|
|
Item 5
|
4.51
|
3.52
|
.837
|
.948
|
|
Item 6
|
3.97
|
3.38
|
.833
|
.948
|
|
Item 7
|
4.36
|
3.47
|
.816
|
.949
|
|
Item 8
|
3.98
|
3.48
|
.717
|
.953
|
|
Item 9
|
3.71
|
3.43
|
.850
|
.947
|
|
Item 10
|
3.53
|
3.42
|
.767
|
.951
|
APSQ:Anxiety and Preoccupation about Sleep Questionnaire; SD:Standart deviation.
Criterion and predictive validities
To assess criterion validity, Pearson correlations were computed between Turkish APSQ
scores and scores on the PSQI, and the ISI. As it can be observed in the [Table 5], criterion validity was demonstrated because high correlations were found between
scores of the Turkish APSQ, PSQI, and the ISI.
Table 5
Correlations of Turkish APSQ with other measures.
|
Scale and sub-scale
|
PSQI
|
ISI
|
|
APSQ
|
.787*
|
.749*
|
|
Worries about the consequences of poor sleep
|
.688*
|
.655*
|
|
Worries about uncontrollability of sleep
|
.756*
|
.739*
|
Notes:
*p<.01; APSQ = Anxiety and preoccupation about sleep questionnaire; PSQI = Pittsburg
sleep quality index; ISI = Insomnia severity index.
In addition, the predictive validity of the scale was analyzed via independent-samples
t-test and we found significantly higher scores among the depression group (M=67.54,
SD=24.61) compared to university students (M=33.35 SD=24.92; t=7.826, p<0.0001). According
to ISI, while 75% (n=30) of the depression group had clinical insomnia, 17.8% of the
control group was detected clinical insomnia. Turkish APSQ scores as other sleep scales
were significantly higher in the clinical insomnia group compared to good-sleepers
group. [Table 6] shows comparisons of sleep scale scores between good-sleepers group and clinical
insomnia group.
Table 6
Comparisons of sleep scale mean scores between good-sleepers group and clinical insomnia
group.
|
Good-sleepers
|
Clinical insomnia
|
p
|
|
Control group n(%)
|
116(82.2)
|
25(17.8)
|
|
|
Depression group n(%)
|
12(25)
|
30(75)
|
|
|
APSQTOTAL
|
28.69
|
70.30
|
<0.001
|
|
APSQF1
|
18.71
|
40.36
|
<0.001
|
|
APSQF2
|
9.42
|
27.07
|
<0.001
|
|
PSQI
|
5.13
|
13.67
|
<0.001
|
|
ISI
|
7.88
|
19.65
|
<0.001
|
Notes: APSQ = Anxiety and preoccupation about sleep questionnaire; F1 = Factor 1: worries
about the consequences of poor sleep; F2 = Factor 2: worries about uncontrollability
of sleep; PSQI = Pittsburg sleep quality index; ISI = Insomnia severity index; n:
Number of individuals.
DISCUSSION
The aim of the current study was to adapt the APSQ to Turkish and examine its psychometric
properties in clinical and non-clinical samples. To the best of our knowledge, only
two studies have examined the psychometric properties of the APSQ, and they only evaluated
the English language form. No clinical samples were assessed in these studies, and
the temporal stability of the scale was not analyzed[14],[15]. Therefore, this is the first study to investigate the validity and reliability
of the scale in a different language, perform test-retest analysis, and assess the
scale with a clinically diagnosed group.
First, we examined the factor structure of the scale. The KMO value greater than 0.60
and a significant Bartlett’s test are required[24],[25]. Our analyses yielded a KMO coefficient of 0.94 and a statistically significant
Bartlett’s test (χ2=1652.80, df=45, p<.001), indicating that the scale was suitable for factor analysis.
After applying EFA and PCA with direct oblimin rotation, only one factor emerged with
an eigenvalue above 1. However, since the scree plot revealed a two-factor structure
and the original scale has a two-factor structure, we forced the scale into a two-factor
structure with PCA and direct oblimin, achieving alignment with the original scale.
This two-factor structure allows standardization across studies using this scale.
Factor loading values are coefficients that describes the relationships between items
and factors. In general, a loading value at or above 0.60 is considered high, regardless
of its positive or negative values[27]. In our analysis, all items included in a factor had factor loadings greater than
0.60, indicating that the scale has a two-factor structure like the original. CFA
was used to reanalyze the factor structure obtained from EFA, resulting in suggested
modifications to reach acceptable fit indices. After two modifications (the error
variances between item 1 - item 2 and item 9 - item 10 were combined) the fit indices
according to CFA are as follows: χ2/df=2.100, CFI=979, GFI=.930, IFI=979, RMSEA=.078. The scale items’ standardized factor
loads ranged from 0.72 to 0.89. Therefore, through EFA and CFA methods, the Turkish
APSQ showed good fit indices and appropriate factor loads in our sample.
We also used Cronbach’s alpha to assess the scale’s internal consistency. Cronbach’s
alpha coefficients were .952, .935, and .906 for the total scale and two subscales,
respectively. These values are similar to those found in previous studies (.93, .91,
and .86)[14],[15]. As a Cronbach’s alpha coefficient above 0.80 indicates that the scale is highly
reliable[28], the Turkish APSQ has very good internal consistency.
In reliability analysis, item-total score correlation coefficients must be greater
than 0.30. Our item analysis results revealed corrected item-total score correlation
coefficients between 0.71 and 0.84, high correlations that suggest that the scale
items were understood correctly by participants[29]. One important parameter of scale reliability is yielding similar results from the
same participant at different times. It is desirable for a scale’s test-retest correlation
to be at least 0.70[30].
Our study is the first to examine the test-retest reliability of the APSQ by readministering
the scale to randomly selected samples from both groups (20 MDD patients and 25 controls)
two weeks after the initial administration and analyzing results for temporal stability.
The Pearson correlation coefficients were 0.661 for MDD patients and 0.828 for non-clinical
sample. Psychiatric interventions such as psychopharmacological treatments and/or
sleep hygiene training with the MDD patients may have affected the stability of their
scores over these two weeks. This suggests that the scale may be useful in reflecting
a course of treatment.
To determine criterion validity, relationships between Turkish APSQ scores and scores
from the PSQI, and the ISI were examined. As found in previous studies, high levels
of anxiety and preoccupation about sleep were significantly correlated with bad sleep
quality, late sleep onset, and insomnia severity[14],[15]. Mean APSQ scores were significantly higher in the depression group than in the
non-clinical group (67.54 versus 33.35), indicating that sleep-related anxiety was
higher in MDD patients. While sleep-related anxiety and preoccupation among MDD patients
have not been investigated in previous studies, this result suggests that sleep-related
anxiety in depression patients may mediate insomnia[14],[15]. Depression and insomnia pose a chicken or egg question. Insomnia can lead to depression
but can also be a symptom of depression, and treating insomnia improves the treatment
of depression[31]. The APSQ can detect sleep-related anxiety in patients with depression, prompting
appropriate psychiatric interventions that will not only treat insomnia but also depression.
Limitations
Our study has some limitations. First, it includes a relatively small clinical sample
and the non-clinical sample included only university students. Therefore, its results
cannot be generalized to the general population or all patients with depression. Second,
the age and gender distribution were not balanced in both groups. Third, the present
study was cross-sectional, precluding the determination of causal relationships. Fourth,
correlations between APSQ scores and anxiety and depression levels could not be examined,
since scales measuring depression and anxiety were not administered. Future studies
should examine the psychometric properties of the APSQ in larger patient groups.
CONCLUSION
Here, we showed that Turkish APSQ is valid and reliable scale for assessing anxiety
and preoccupation about sleep in both clinical and non-clinical populations. We found
that anxiety and preoccupation about sleep are associated with impaired sleep quality
and severity of insomnia in both MDD patients and non-clinical controls. Thus, measuring
anxiety and preoccupation about sleep can guide insomnia treatment or clinical research.
It is therefore important to analyze the validity and reliability of the APSQ across
a variety of general populations and clinical groups.
Ethic approval
The study was conducted between June 2019 to December 2019, and the study protocol
was approved by the local ethics committee of Kahramanmaras Sutcu Imam University
(approval number: 29.05.2019/10; decision number: 05). All authors agree on the content
of the study.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research,
authorship, and/or publication of this article.
