Introduction

Although pediatric obstructive sleep apnea (POSA) was first described as soon as 1976, only in 2005 was this disorder listed separately from the adult form in the second international classification of sleep disorders (ICSD-2)1,2. POSA may be affecting as much as 1–5% of the pediatric population, with obesity, adenotonsillar hypertrophy, allergic rhinitis, inflammatory factors, and abnormal craniofacial anatomy being the most significant contributing factors1,2. Considering the differences in symptoms of obstructive sleep apnea (OSA) in children and adults, it usually takes a well-trained and attentive physician to register the increased risk of the presence of this disorder.

In developed countries, the awareness of OSA in adults is pretty good; the situation is, however, a bit more complicated when it comes to POSA. Multiple sleep scales for POSA have been developed, such as (but not limited to) the OSA Score, Sleep clinical record, OSA-18 questionnaire, and Pediatric Sleep Questionnaire (PSQ) with Sleep-Related Breathing Disorders Scale (SRBD scale)2,3,4,5,6,7. Based on a meta-analysis published in 20208 comparing the PSQ, OSA-18, and pulse oximetry, the PSQ was found to be the most sensitive for screening mild POSA. In patients with moderate and severe POSA, however, sensitivities of PSQ and pulse oximetry were comparable. Therefore, a combination of PSQ and pulse oximetry was recommended for early detection of POSA8. A review study identified the PSQ to be the only diagnostically sufficiently accurate questionnaire for screening of impaired breathing in children, and, at the same time, emphasized the importance and benefit of the involvement of dentists in primary screening of POSA9.

In the Czech Republic, the Epworth Sleepiness Scale10 is the most commonly used questionnaire for OSA screening; it, however, should be used only for adult patients. Even though there are several questionnaires used by specialists for screening disturbed sleep in children worldwide, no official pediatric sleep questionnaire has been introduced in the Czech Republic so far (not even an official translation of international questionnaires into Czech language).

The main aim of this study was to create a questionnaire in Czech language that could detect the increased risk of POSA and at the same time would: (1) be complex and reflect the multifactorial etiology of POSA, (2) be short and quick to fill in, (3) be easy to understand and complete by doctors with different specializations as well as by parents/caregivers of the child, (4) be used without the need to include clinical parameters or examination results, and (5) have sensitivity and specificity comparable with reputable pediatric sleep questionnaires.

Methods

The process of developing a Czech questionnaire for identifying children at risk of obstructive SleEp apNea (SEN CZ) is described in detail in the Supplementary material and illustrated also in the workflow scheme (see Fig. 1). The process also included a pilot study and testing of the SEN CZ (version 1 and version 2, respectively). SEN CZ was designed to primarily address the symptoms of POSA (both night and day symptoms) and their frequency. Our intention was to avoid the necessity of incorporating clinical parameters or examination results in the questionnaire items. The questionnaire specificity and sensitivity assessment were evaluated based on the result of the home sleep apnea test (HSAT); using the respiratory polygraphy, Alice NightOne (Respironics, Inc., PA, USA) and measured apnea–hypopnea index (AHI). The severity scale recommended by the American Academy of Sleep Medicine for use in children under the age of 12 is: AHI < 1 (physiological finding); AHI ≥ 1 to < 5 (mild POSA); AHI ≥ 5 to < 10 (moderate POSA); AHI ≥ 10 (severe POSA).

Figure 1
figure 1

Workflow of the Czech questionnaire for identifying children at risk of obstructive sleep apnea (SEN CZ) development; note that although collection of only 70 SEN CZ version 2 questionnaires were planned, 71 were included in the final analysis. AHI apnea–hypopnea index, CZ Czech language, EN English language; *The items were excluded if: (1) they did not use the Likert scale, (2) they had uniform distribution and, thus, low discriminatory power, (3) they were difficult to answer for the responders, (4) they had a weak correlation with AHI (r < 0.2), (5) they had a weak correlation with the total score of the questionnaire, (6) another item related to the same symptom had a higher discriminatory power, and/or (7) their contribution to creating the shortest yet functional screening was considered insufficient.

Full size image

The Supplementary material also include the detailed description of methods, participants enrolled in this study, and the clinical examinations they underwent. The questionnaires were completed by parents or legal guardians (hereinafter referred to as parents).

Ethics declarations

The study was performed with the approval of the Ethics Committee of the St. Anne’s Faculty Hospital (No. 33V/2020; date: 10th June, 2020). Written informed consent was obtained from all parents or caregivers of participating children in line with the Declaration of Helsinki before inclusion in the study.

Results

Characteristics of the participants and different versions of the SEN CZ

The demographic description of the enrolled children is presented in Table 1, the characteristics of the different versions of the SEN CZ in Table 2 and in detail described in Supplementary material. The results of the partial steps of the SEN CZ genesis are shown in Supplementary Table S1. The SEN CZ questionnaire with 17 items and a 5-item Likert scale is presented in Fig. 2; the non-tested English version is in Supplementary Fig. S1.

Table 1 Characteristics of the 100 children enrolled in the pilot study and testing of the SEN CZ questionnaire.
Full size table
Table 2 Sensitivity and specificity of the SEN CZ questionnaire.
Full size table
Figure 2
figure 2

Czech questionnaire for identifying children at risk of obstructive sleep apnea (SEN CZ).

Full size image

Identification of key clinical and behavioral factors and their correlations with AHI

After the pilot study, we continued with explorative factor analysis and performed the analysis of the final SEN CZ on 71 child–parent pairs.

While the Kaiser criterion suggests considering three factors, the sum of the squares of the factor loadings suggests inferring a maximum of five factors. Therefore, we decided to continue with the exploratory factor analysis up to five factors. As we did not find any outliers, we used the ML estimator. Simultaneously, we chose the oblimin rotation due to the assumption that the factors may intercorrelate. To evaluate the model fit, we monitored the change of the following fit indices: χ2, TLI and RMSEA11, see Supplementary Table S2. The fit indices provide a sufficient fit with two interrelated factors (breathing and neurobehavioral problems) and satisfactory fit with three interrelated factors (breathing problems, inattention, and hyperactivity). This shows that there may be three symptomatic factors which together create a higher-order factor of POSA risk.

The validity of these three factors was further supported by the correlations of their total scores with AHI in Table 3. We also provide the correlation of the SEN CZ with AHI for comparison. Moreover, each of the three factors also achieves high reliability, see Table 3. More details are described and shown in Supplementary material.

Table 3 Correlation of the individual factors with the total score of SEN CZ questionnaire and AHI.
Full size table

Prediction of the POSA risk based on the SEN CZ total score

To further evaluate the ability of the SEN CZ to predict the increased POSA risk, we performed a logistic regression—analysis of the relationship between one or more existing independent variables (such as the SEN CZ total score, POSA risk according to AHI). After verifying that our data meet the assumptions of logistic regression, we proceeded to the analysis itself. In the first step, we entered the total score from the questionnaire into the regression, and in the second, we entered the control variables age and maxillary constriction, see Table 4.

Table 4 Logistic regression of the total score of the SEN CZ questionnaire and the risk of POSA according to AHI.
Full size table

Model 1, considering the SEN CZ score as a predictor, was statistically significant compared to the null model (χ2(1) = 29.79, p < 0.001) and explained 48% of the variance (Nagelkerke R2) of POSA risk. Model 2, considering the SEN CZ score and potential intervening variables, did not significantly improve the fit of the original model (Nagelkerke R2 = 0.49, χ2(3) = 29.98, p < 0.001). Therefore, we continued to work with Model 1. Model 1 yielded a sensitivity of 86%, specificity of 77%, and classification accuracy of 83% at a cut-off value corresponding to a 50% probability of finding POSA. The increasing score in the SEN CZ significantly predicted the diagnosis of the disease, thus meeting the criterion validity of the questionnaire. The regression that illustrates this relationship corresponds to the expression logit(POSA) = − 4.92 + 0.18*SEN_SCORE where the parameter SEN_SCORE is the SEN CZ questionnaire total score. We may convert this logit using an exponential function to get the odds of POSA risk. Therefore, if the child reaches a score of 40 in the SEN CZ according to the parent's report, the chance (odds of POSA = exp(logit(POSA))) that this child would be diagnosed with POSA in polygraphy is 9.78–1. Therefore, we have a roughly 91% probability of finding POSA in this child. A 50% probability of finding POSA risk approximately corresponds to the score of 23 in the SEN CZ.

Using the cut-off value for mild, moderate, and severe POSA according to AHI, the approximate cut-off values in the SEN CZ itself can be determined based on the linear regression of AHI on the SEN CZ score. This relationship can be expressed by the equation SEN_SCORE = 23.98 + 6.11*AHI. AHI values above 1 (mild POSA), therefore, correspond to SEN CZ score above 30.09.

The calculated SEN CZ total score cut-off values for POSA risk assessment are shown in Table 5. However, the given cut-off values should be considered with a grain of salt due to the sample size, exploratory factor analysis, possible measurement error and the pilot nature of the research. If we also adjust the logistic regression according to the cut-off value of at least mild POSA (31; 66% probability of finding), the model correctly determines 85% of cases, with the sensitivity of 84% and specificity of 86%. Although there is an improvement in specificity, this result also indicates the need for a larger sample with a sufficient representation of the healthy population.

Table 5 Cut-off scores in the SEN CZ questionnaire for POSA risk prediction according to AHI.
Full size table

Discussion

In this pilot study, our multidisciplinary team created a Czech questionnaire for identifying children at risk of POSA. In view of the limited availability of PSG and polygraphy, such a questionnaire could facilitate the identification of such children. Considering the existence of several pediatric sleep questionnaires in English2,3,4,5,6,7, we initially intended to simply translate a highly evaluated questionnaire and adapt it from English to Czech. However, in view of the linguistic and cultural differences, such a simple translation of the PSQ-SRBD scale4, which was deemed to be the most suitable according to our original literature review, was in the end (based on our collaboration of professionals and experts from different medical fields associated with pediatric sleep medicine in the Czech Republic) found to be insufficient. Hence, we decided to take further steps and design a questionnaire for identifying POSA risk for Czech users, which we named SEN CZ.

After preliminary testing of SEN CZ ver. 1, we proceeded to further improve it to produce SEN CZ ver. 2, which has already achieved very good results (Cronbach’s α = 0.928); however, it was quite lengthy, and we have decided to shorten it as much as possible. The final version of the SEN CZ kept the good sensitivity and specificity, as well as reliability but contains less than half the items compared with the SEN CZ ver. 2, making it more user-friendly. In the context of sensitivity and specificity, the performance of the final version of the SEN CZ is very good, comparable with the best sleep questionnaires included in the meta-analysis by Wu et al.8.

In our questionnaire, we originally intended to keep dichotomous items for the simplicity of assessment; however, the use of Likert scale is often preferred to dichotomous questions (yes/no/unknown) when analyzing children’s behavior12. For screening purposes, it is better if we capture the answer on a scale rather than dichotomously as it is more accurate.

The majority of the items in the final version of the SEN CZ cover very similar areas of POSA symptoms as items in foreign pediatric sleep questionnaires2,3,4,5,6,7. This supports their importance as factors contributing to this disease. In our questionnaire, based on the results of the validation, we have excluded several items associated with the nocturnal symptoms due to their high difficulty for parents. Even though snoring and difficulty breathing during sleep are probably the most common complaints voiced by parents of children with POSA13, more recent studies reported that parents were able to report snoring in only 15% of children diagnosed with POSA14,15. These findings are suggestive of the fact that parents may often fail to notice the nocturnal symptoms, especially if the children have a separate room, or if not enough attention is paid to this problem.

In our questionnaire, many items focus on the field of daytime symptoms. For example, in the case of nasal obstruction and adenotonsillar hypertrophy, a transition to mouth breathing occurs, which subsequently affects changes in the orofacial region, such as the narrow hard palate, posteriorotation of the mandible and/or disproportions in the sagittal plane, as well as the elongation of the lower third of the face, which can be expressed in varying degrees and forms16,17. These features may present risk factors for the development of POSA; moreover, mouth breathing preference has been identified as an early contributor to SRBD that can be often found in primary snorers and patients with POSA18,19. At the same time, these contributing factors may be, compared to nocturnal symptoms, easily detectable also by parents, which makes them important for screening.

Interrupted sleep and nocturnal hypoxia affect, among other things, the prefrontal cortex and thus the cognitive executive functions of the individual20. Affected children have impairments in planning, plan execution, organization, decision-making, and information processing. These factors, together with impaired attentional retention, impaired concentration, impulsivity, hyperactivity, and emotional lability, often cause errors of inattention and impaired school performance21,22,23.

Children with POSA also tend to have increased irritability (or even aggression) and may show disinterest in daily activities and tendencies to depression. It has even been suggested that sleep disruption and difficulty in early childhood may predict behavioral and emotional problems during adolescence24. There is also evidence of a significant association between reduced or disturbed sleep and the severity of behavioral changes. Studies also suggest that children presenting with POSA display symptoms of attention deficit and hyperactivity disorder (ADHD) more often than those without it25,26. Most studies agree on a bidirectional association between POSA and ADHD27.

The results of the factor analysis and logistic regression indicate that the SEN CZ is slightly more likely to recommend a comprehensive POSA examination when it is not needed than to fail to recommend an examination when it should be performed. Given the screening nature of the questionnaire, this is a very good result. Moreover, the SEN CZ could be also a good predictor of ADHD. Therefore, it might be useful to continue the research with data obtained from children diagnosed with ADHD. We might then be able to analyze which items discriminate well between POSA and ADHD (i.e., identify questions in which parents of children with POSA score significantly differently than parents of children with ADHD).

Even though according to the latest AASM guidance, PSG is the gold standard for diagnosing pediatric sleep disorders28 it may not be readily available in all clinical settings. In our study, we encountered practical challenges in the accessibility of PSG to children due to its limited availability in the Czech Republic and long waiting times. In light of these limitations and considering that HSAT is recommended as a screening tool before the referral for PSG29,30, we opted for including HSAT as a part of our validation process. It is also important to keep in mind that SEN CZ should only be used as the first-tier screening tool intended to identify children who should be recommended for further examinations, not to serve as a replacement for any sleep monitoring or as a precise diagnostic tool.

Conclusion

We have designed and validated the SEN CZ questionnaire for screening of POSA risk reflecting the multifactorial etiology of this disease. Our questionnaire with 17 items and a 5-item Likert scale is the first tested questionnaire for identifying children at risk of POSA in the Czech language and could be filled in not only by doctors of different specializations but also by parents/caregivers of the children. It performs excellently, achieving high reliability and high correlation with AHI as an objective criterion for the POSA diagnosis, and shows comparable sensitivity and specificity to the already existing foreign pediatric sleep questionnaires.