Obstructive sleep apnea (OSA), occurring in 26% of people aged 30-70 years, manifests as repetitive apnea and hypopnea episodes related to sleep-time upper-airway obstruction [1]. OSA causes sleep fragmentation, is associated with excessive daytime sleepiness and decreased concentration, and has been linked to metabolic disorders, cardiovascular and neurovascular diseases, and neurocognitive impairment [2-4].

The Epworth Sleepiness Scale (ESS), Berlin, STOP, and STOP-Bang questionnaires have been developed as OSA-screening-tool alternatives to the expensive and time-consuming polysomnography (PSG) test. There have already been several reports published on the sensitivity and specificity of these screening tests [5-8].

For most sleep questionnaires, respondent age can be problematic. Hypertension, which longitudinal and cross-sectional studies have definitively shown to be strongly dose-response-related to OSA severity, is included in most of the current questionnaires [9,10]. The problem that arises is that hypertension is significantly correlated not only with OSA severity but also with older age. Therefore many relatively young OSA patients, even those with severe OSA, do not have hypertension, which can lead to false negatives (i.e., OSA sufferers deemed “no OSA”). Also, age usually is related to both social activity and physical strength, both of which can affect the answers to questions on sleep-de-privation-caused daytime fatigue. However, age is not adjusted for in the STOP or Berlin questionnaires or indeed in several others. On this basis, we hypothesized that the reliability of sleep questionnaires could differ according to patient age. In our study, we evaluated the ESS, Berlin, STOP, and STOP-Bang questionnaire validities as OSA-screening tests for various adult age groups.

We retrospectively reviewed the medical records of patients who referred to a sleep clinic with symptoms such as insomnia, snoring, sleep apnea or excessive daytime sleepiness between April 2012 and May 2016. A total of 396 patients aged 20 years or over were enrolled in this study. After categorizing the patients into three groups according to age (20-39, 40-59, or ≥60 years), patient data, including age, sex, height, weight, history of hypertension and overall medical history, were collected. The ESS, Berlin, STOP, and STOP-Bang questionnaires, which had been translated into Korean and certified, and the diagnostic overnight PSG results were also collected. Those for whom the total sleep time was under 180 min in the PSG were excluded. The relevant Institutional Review Board approved this study.

The patients were each assigned to one of three groups based on age: 20-39 (young adult), 40-59 (middle-aged), and 60 and over (elderly). Table 1 summarizes the patients’ characteristics. The 396-patient cohort comprised 332 males and 64 females of 47 years mean age. Gender significantly differed among the three age groups (p=0.002). Hypertension, identified in 100 patients (25.3%) also showed significant differences by age group (p=0.000). The OSA group (AHI >5) included 313 patients (79.0%), and the normal (control) group, 83 patients (21.0%). There were significant OSA-distribution differences among the age groups (p= 0.022). Within the OSA group, the moderate-to-severe OSA sub-group (AHI >15) included 137 patients (34.6%), and the severe OSA sub-group (AHI >30), 64 patients (16.2%). There were also significant differences in these two groups’ respective distributions among the age groups (p=0.000) (Table 1).

For the OSA group (AHI >5), sensitivity and accuracy were highest for the STOP-Bang questionnaire, specificity for the ESS, and AUC for the Berlin questionnaire. The specificity and specificity of the Berlin and the STOP questionnaire significantly differed by age. As for the Berlin questionnaire’s AUC, the confidence interval did not include 0.5 for the young-adult or middle-aged group. However, it did not include 0.5 only for the young-adult in the STOP and STOP-Bang questionnaire’s AUC (Table 2).

Meanwhile, about the moderate-to-severe OSA sub-group (AHI >15), the STOP-Bang questionnaire showed the highest sensitivity, the ESS the highest specificity and accuracy, and the Berlin questionnaire the highest AUC. By age group, the Berlin, STOP, and STOP-Bang questionnaire’s specificity significantly differed. As for the Berlin and STOP-Bang questionnaires’ AUC, the confidence interval did not include 0.5 for the young-adult or middle-aged group (Table 3).

A comprehensive approach including physical examination, thorough history-taking, and overnight PSG is required for accurate diagnosis of OSA. Complaints of daytime sleepiness, fatigue and snoring should be considered suspicious of OSA. Additionally, the presence of any underlying disease, not to mention apnea-related information from a family member of a person living with the patient, also should be considered [11]. However, history-taking and physical examination are insufficient for distinguishing primary snoring from sleep apnea. Thus, supplementary overnight PSG, the current standard, is considered necessary for definitive OSA [12].

Because nearly 80% of men and approximately 93% of women with moderate-to-severe sleep apnea go undiagnosed, necessary screening tests must be further developed and improved to increase diagnosis rates [13]. Indeed, several studies focusing on the various questionnaires and the clinical OSA-associated factors have done just this [5-8].

The Berlin questionnaire has been widely employed since its introduction at the 1996 Conference on Sleep in Primary Care in Berlin, Germany. It is used to investigate excessive daytime sleepiness, the presence of fatigue and hypertension, snoring, and sleep apnea. One OSA-diagnostics study reported sensitivities ranging between 54 and 86%, specificities between 43 and 87%, and a PPV of 89% [6]. The OSA-screening STOP questionnaire was designed to reduce some of the risk associated with general anesthesia. Significantly, this tool’s sensitivity has been found to improve with higher AHI: with an AHI ≥5, the sensitivity was 65.6%, with an AHI of 15, 74.3%, and with an AHI of 30, 79.5% [7]. The STOP-Bang questionnaire has been demonstrated to offer higher sensitivity and NPV, especially for moderate-to-severe OSA patients [7]. In light of these three (Berlin, STOP, STOP-Bang) questionnaires’ relative commonness, less complicated criteria, and relative ease of application in the clinical setting, we evaluated them in the present study. We additionally evaluated the ESS that was designed originally for assessment of excessive daytime sleepiness, as it has been widely utilized by primary care providers to identify potential sleep disorders.

Two studies that compared STOP and STOP-Bang with Berlin and/or ESS, determining STOP-Bang to have the highest sensitivity and AUC, recommended it as the better OSA-screening tool [14,15]. Another, a meta-analysis of OSA-screening tests found Berlin to be the most accurate questionnaire for prediction of OSA diagnosis. The least precise questionnaire was determined to be ESS, due possibly to the fact that excessive daytime sleepiness is a common syndrome among obese persons without OSA, as driven by other-than-nighttime-sleep-deprivation mechanisms. The simplest questionnaire, STOP, was deemed a relatively poor predictor of OSA, while STOP-Bang was judged to be average [16].

As for the effect of age on OSA, increasing age has been correlated with both increasing prevalence and decreasing severity [17]. Male dominance for OSA prevalence and severity disappears above 55 years of age [18,19]. The present data showed similar results: the female ratio was found to have been significantly increased in the elderly group. Overall OSA prevalence increased significantly with age.

Also in the current study, hypertension prevalence increased significantly with age; in fact, it was markedly higher in the elderly group than in either of the other age groups. OSA prevalence, meanwhile, was much higher in both the middle-aged and elderly groups than in the young-adult group. These results strongly suggest the possibility of age-related prevalence differences between hypertension and OSA. Notably, this might have influenced the Berlin and STOP questionnaire’s sensitivity and specificity differences among the different age groups because hypertension plays a relatively significant role in the Berlin and STOP questionnaires [9,10].

The ideal diagnostic test for an otherwise healthy population would offer the following: relatively high sensitivity, sufficient specificity, minimal intrusiveness, relative cheapness, and a utility for identification of patients early in the disease course [20]. We found that the ESS showed the lowest sensitivity; the STOP-Bang questionnaire, while affording the highest sensitivity, as found in previous studies, offered only inadequate specificity; the Berlin questionnaire’s AUC results showed it to be recommendable for evaluation of patients with suspected OSA in young-adult and middle-aged groups, notwithstanding its relatively lengthy question list and complicated scoring system [14,15].

There are some research limitations. For example, both the OSA patients and the normal (control) group showed sleep-related symptoms. Also, the number of patients in the elderly group was small. Furthermore, we were unable to examine patient occupation and lifestyle, both of which factors might have affected the results.

In conclusion, the Berlin and the STOP questionnaire showed significant sensitivity and specificity differences among the age groups. Also, for the moderate-to-severe OSA patients, there were the Berlin, STOP, and STOP-Bang questionnaire differences in specificity among the age groups. The ESS questionnaire, however, did not show any age-related differences.