Modified Berlin Score for predicting sleep apnea in patients with acute ischemic stroke

Abstract

Background

Sleep-disorder breathing is common among ischemic stroke patients. It associated with increasing risk of stroke and poor functional outcome after stroke. Polysomnography (PSG) is the diagnostic standard but the application for acute stroke patients is limited. The Berlin Questionnaire (BQ) is a tool used for screen obstructive sleep apnea in general population, however, the diagnostic accuracy in acute ischemic stroke patients is inconsistent.

Objects

The study aims to explore the accuracy of Berlin Questionnaire (BQ) for detecting the sleep apnea for acute ischemic stroke patients, using home-based sleep apnea testing as a reference. We compare the accuracy between BQ and a modified BQ.

Methods

We conducted a prospective observational study between March 2023 and February 2025 at a teaching hospital in central Taiwan. Patients aged 20–85 years with MRI-confirmed acute ischemic stroke admitted within 48 h of onset were included. Each participant completed the Berlin Questionnaire and underwent home-based sleep apnea testing (Alice NightOne, Philips Respironics). Sleep apnea severity was defined by the respiratory event index (REI). The accuracy of BQ and mBQ in predicting moderate-to-severe sleep apnea (REI ≥ 15) was assessed using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results

Forty-six patients completed both assessments. Twenty-one patients (45.7%) had no or mild sleep apnea (REI < 15), and 25 (54.3%) had moderate-to-severe sleep apnea (REI ≥ 15). Alcohol use and higher body mass index (BMI) were significantly associated with moderate-to-severe apnea (p = 0.01 for both). The BQ demonstrated a sensitivity of 45.8%, specificity of 72.7%, PPV of 64.7%, and NPV of 55.2% for predicting moderate-to-severe sleep apnea. The mBQ improved performance to a sensitivity of 54.1%, specificity of 72.7%, PPV of 68.4%, and NPV of 59.3%. Patients with moderate-to-severe apnea exhibited significantly lower minimum oxygen saturation compared with those with no or mild apnea (p = 0.03).

Conclusions

The accuracy of the Berlin Questionnaire for detecting sleep apnea in acute ischemic stroke is moderate but improves with modification of BMI weighting. Moderate-to-severe apnea is associated with lower oxygen saturation, highlighting the importance of early screening and targeted management in stroke units. Larger, multicenter studies are warranted to validate these findings and explore the prognostic implications of nocturnal hypoxemia in stroke recovery.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12883-026-04873-x.

Introduction

Sleep-disordered breathing, particularly obstructive sleep apnea (OSA), is one of the most prevalent sleep-related disorders and has been recognized as both a risk factor and a modifiable comorbidity in ischemic stroke [1]. OSA is characterized by recurrent episodes of upper airway obstruction during sleep, leading to intermittent hypoxemia, hypercapnia, and sleep fragmentation. The consequences include increased sympathetic activity, endothelial dysfunction, and systemic inflammation, which contribute to elevated risks of cardiovascular and cerebrovascular disease [2, 3].

Epidemiological studies have demonstrated that patients with OSA have a two- to fourfold increased risk of stroke and higher all-cause mortality [4]. Moreover, in individuals with acute ischemic stroke, OSA has been linked to early neurological deterioration, poor functional outcomes, and increased risk of recurrent stroke [5–7]. Despite its clinical relevance, sleep apnea remains underdiagnosed and undertreated in acute stroke populations.

Polysomnography is performed overnight and recording the biophysiological sleep-related changes including sleep stages, electroencephalography, cardiovascular parameters, oximetry, respiratory effort and airflow. Polysomnography (PSG) remains the gold standard for diagnosing sleep apnea. However, its use in acute stroke is limited by several factors, including patient immobility, communication difficulties, and the need for specialized sleep laboratory facilities and trained technicians. Consequently, portable or home-based diagnostic systems have gained attention as practical alternatives for sleep assessment in hospitalized or subacute patients [8].

The Berlin Questionnaire (BQ), developed in 1998, is one of the most widely used screening tools for identifying individuals at high risk for OSA [9, 10]. It consists of three categories; the first category is snoring behavior, the second category is daytime sleepiness, and the third category is obesity or hypertension. Category 3 was defined as positive when the subject had hypertension or a BMI greater than 30 kg/m². In the general population, the BQ has demonstrated reasonable accuracy, with reported sensitivity and specificity of 58.8% and 77.6%, respectively, for detecting moderate-to-severe OSA (apnea-hypopnea index ≥ 15).

However, the applicability of the BQ in stroke populations is uncertain. Studies suggest that the correlation between questionnaire-based risk and PSG-confirmed OSA is moderate, with sensitivity and specificity values ranging between 55 and 67% [11, 12]. Stroke-related cognitive deficits, aphasia, and lack of self-awareness may further compromise reliability [13].

Sleep apnea significantly elevates stroke risk and impairs recovery outcomes for those who have suffered one. Early screening and intervention are essential to improve patient prognosis and minimize the risk of recurrence [14].

Given the high prevalence of sleep apnea in stroke and the limitations of PSG, refining a screening tool for early detection is clinically important [15]. Different countries define obesity or the standard body size range of BMI differently. If people from different countries use the original Berlin questionnaire standard to assess patients, it may affect the accuracy of the Berlin questionnaire. In 2022, the officially recommended normal (healthy) BMI range for adults in Taiwan, defined by the Ministry of Health and Welfare, is 18.5 ≦ BMI < 24 Kg/m². This standard categorizes under 18.5 as underweight, 24–27 as overweight, and 27 or higher as obese [16]. We hypothesize that modifying the BMI component of the BQ may improve the predictive accuracy for populations in Asia, particularly the Far East. We Therefore, this study aimed to (1) assess the diagnostic accuracy of the Berlin Questionnaire in patients with acute ischemic stroke, (2) assessing whether separating hypertension and BMI into two categories could increase the accuracy of the Berlin questionnaire (mBQ) in predicting sleep apnea. and (3) explore the relationship between sleep apnea severity and oxygenation parameters in this population. According to Taiwan’s definition of obesity, when BMI > 27 kg/m2, mBQ category 4 is defined as positive. When a person has high blood pressure, mBQ category 3 is defined as positive.

Methods

Study design and participants

This prospective observational study was conducted at a teaching hospital in central Taiwan between March 1, 2023, and February 28, 2025. The study enrolled a total of 459 participants. The study protocol was approved by the Institutional Review Board of Chia-Yi Christian Hospital (IRB No. 2022110). All participants provided written informed consent before enrollment.

Eligible participants were adults aged 20–85 years with MRI-confirmed acute ischemic stroke who were admitted within 48 h after symptom onset. Stroke severity was assessed using the National Institutes of Health Stroke Scale (NIHSS) [17], and stroke subtype was classified according to the Trial of Org 10,172 in Acute Stroke Treatment (TOAST) criteria [18].

Exclusion criteria

Patients were excluded if they had any of the following:

• Impaired consciousness or aphasia preventing questionnaire completion.

• Severe dementia (Mini-Mental State Examination score < 10).

• Congestive heart failure, chronic obstructive pulmonary disease, or asthma.

• Major depressive disorder.

• Inability to provide informed consent.

Sleep disorder risk assessment

Within the first week of admission, patients completed a structured interview including the Berlin Questionnaire and the Pittsburgh Sleep Quality Index. For this study, a Modified Berlin Questionnaire (mBQ) was also applied, in which the body mass index (BMI) category was weighted separately as an independent risk domain.

There are three categories of BQ: snoring behavior, fatigue during the day, and obesity or hypertension. When two or more categories are positive, the patient is considered at high risk for sleep apnea. A patient was considered to have low risk of sleep apnea if the score was positive in only one category or not at all.

mBQ differs from BQ in category 3. The BQ defines category 3 as positive if the patient answers yes to item 10 or if his or her BMI exceeds 30 kg/m2. mBQ category 3 is considered positive if item 10 is answered positively. In category 4, patients with a BMI greater than 27 kg/m2 are considered to be positive. Whenthree or more categories are positive, the patient is considered at high risk of sleep apnea. Those patients who score positive in only two or fewer categories were considered to have a low risk of sleep apnea.

Home-based sleep apnea testing

Sleep studies were conducted using the Alice NightOne device (Philips Respironics, Murrysville, PA, USA), a portable system recording nasal airflow, thoracoabdominal movement, and pulse oximetry [19]. Automated analysis generated a report including total sleep time, respiratory event index (REI), mean oxygen saturation, lowest oxygen saturation, and the number of hypoxic episodes.

Sleep apnea severity was classified by REI as:

• None: <5 events/hour.

• Mild: 5–14 events/hour.

• Moderate: 15–29 events/hour.

• Severe: ≥30 events/hour.

Apnea type was determined by event characteristics:

• Central sleep apnea (CSA): ≥5 central events/hour.

• Obstructive sleep apnea (OSA): ≥5 obstructive events/hour.

• Hypoxia was defined as ≥ 5 desaturation episodes per hour.

Statistical analysis

Continuous variables were presented as mean ± standard deviation, and categorical variables as frequencies and percentages. Between-group differences were analyzed using Student’s t-test for continuous data and chi-square or Fisher’s exact test for categorical data. Percentages are used to represent the proportion of categorical variables. Diagnostic performance of BQ and mBQ for predicting moderate-to-severe sleep apnea (REI ≥ 15) was evaluated by calculating sensitivity, specificity, PPV, and NPV. The area under the curve (AUC) was used to evaluate the accuracy of the Berlin Questionnaire and the modified Berlin Questionnaire in predicting moderate to severe sleep apnea. Statistical analyses were performed using SPSS version 27.0 (IBM Corp., Armonk, NY, USA). A two-tailed p-value < 0.05 was considered statistically significant.

Results

Patient characteristics

Of the 459 patients, 309 were men and 150 were women. 103 men and 64 women were classified as low risk, 206 men and 86 women were classified as high risk of sleep apnea by BQ.

A total of 46 patients completed both questionnaire assessments and home-based sleep testing. Of these, 21 patients (45.7%) had no or mild sleep apnea (REI < 15), and 25 (54.3%) had moderate-to-severe sleep apnea (REI ≥ 15).

As shown in Table 1, alcohol consumption (p = 0.01) and higher BMI (p = 0.01) were significantly associated with moderate-to-severe sleep apnea. Age, sex, smoking status, diabetes, hypertension, cardiovascular disease, prior stroke, and hyperlipidemia did not differ significantly between groups.

Table 1.

Baseline characteristics of the populations (n = 46)

	REI < 15 (n = 21)	REI ≧ 15 (n = 25)	P value
Age	62.9 ± 10.4	63.4 ± 10.6	0.76
Sex
F	9	7	0.29
M	12	18
Smoker	7	10	0.64
Alcohol	2	11	0.01*
Diabetes mellitus	6	8	0.80
Hypertension	13	15	0.89
Cardiovascular disease	1	1	0.9
Stroke	1	5	0.12
Hyperlipidemia	2	4	0.51
BMI	25.3 (20.6 27.4)	28.2 (20 30.5)	0.01*
NIHSS (0–20)	3 (1–6)	2.5 (0–6)	0.63
Education
No formal education	1	2
Less than 12 years	7	9	0.87
12 years or more	13	14
Berlin questionnaire
High risk	6	11	0.29
Low risk	15	14
Berlin questionnaire 2
High risk	6	13	0.10
Low risk	15	12

REI, respiratory–event index, BMI, body mass index, *: P < 0.05

Relationship between sleep apnea risk and respiratory patterns

Patients categorized as high-risk by the original Berlin Questionnaire were not significantly different from low-risk patients regarding mean oxygen saturation, lowest saturation, or frequency of hypoxic events (Table 2). Similarly, there were no significant differences in obstructive or central apnea frequencies between BQ risk groups (p > 0.05 for all).

Table 2.

Relationship between sleep apnea risk (Berlin questionnaire) and respiratory pattern in ischemic stroke patients

	CSAI$	OSAI$	Hypoxia#	High O2#	Low O2#	Mean O2#
Low risk(n = 29)	5	16	13	97.97 ± 1.01	82.90 ± 6.29	93.61 ± 2.15
High risk(n = 17)	0	11	9	98.65 ± 0.93	85.41 ± 5.13	93.29 ± 3.03
P	NA	0.72	0.15	0.51	0.40	0.15

CSAI, central sleep apnea index, OSAI, obstructive sleep apnea index, #: t-test, $: Chi-square or fisher exact test

Using the Modified Berlin Questionnaire, diagnostic trends improved modestly, with a clearer distinction between low- and high-risk groups, although the differences did not reach statistical significance (Table 3).

Table 3.

Relationship between sleep apnea risk (modified Berlin questionnaire) and respiratory pattern in ischemic stroke patients

	CSAI$	OSAIS	Hypoxia$	High O2#	Low O2#	Mean O2#
Low risk(n = 27)	3	16	12	98.11 ± 0.84	83.70 ± 6.44	93.77 ± 2.260
High risk(n = 19)	1	13	10	98.37 ± 1.25	84.00 ± 5.37	93.09 ± 2.85
P	0.48	0.52	0.58	0.08	0.38	0.19

CSAI: central sleep apnea index, OSAI: obstructive sleep apnea index, #: t-test, $: Chi-square or fisher exact test

Sleep apnea severity and oxygenation

Patients with moderate-to-severe sleep apnea had a significantly lower minimum oxygen saturation compared with those with no or mild apnea (p = 0.03), as shown in Table 4; Fig. 1. These patients also exhibited higher frequencies of obstructive events and hypoxic episodes (p < 0.01).

Table 4.

Relationship between AHI and oxygen saturation

	CSAI$	OSAI$	Hypoxia$	High O2#	Low O2#	Mean O2#
REI 1–14 (n = 21)	2	7	2	98.25 ± 1.16	85.50 ± 6.21	93.80 ± 2.66
REI ≧ 15 (n = 25)	3	22	20	98.27 ± 0.98	82.23 ± 5.21	93.58 ± 1.63
P	0.71	0.003*	< 0.01*	0.47	0.03*	0.37

 REI respiratory-event index

#: t-test, $: Chi-square or fisher exact test, CSAI: central sleep apnea index, OSAI: obstructive sleep apnea index, *: P < 0.05

Fig. 1.

Mean of the lowest oxygenation saturation. REI: respiratory-event Index The oxygenation saturation shown represents the mean of the minimum oxygen saturation values

Diagnostic performance

The diagnostic accuracy metrics for both questionnaires are summarized in Table 5.

Table 5.

Compare between Berlin questionnaires and modified Berlin questionnaires for the predict sleep apnea

	Sensitivity	Specificity	PPV	NPV
BQ	45.8%	72.7%	64.7%	55.2%
mBQ	54.1%	72.7%	68.4%	59.3%

BQ, Berlin questionnaire, mBQ, modified Berlin questionnaire, PPV, positive predict value, NPV, negative predict value

• Berlin Questionnaire: Sensitivity 45.8%, specificity 72.7%, PPV 64.7%, NPV 55.2%.

• Modified Berlin Questionnaire: Sensitivity 54.1%, specificity 72.7%, PPV 68.4%, NPV 59.3%.

These results indicate that isolating BMI as an independent risk category modestly improves sensitivity and overall predictive performance for identifying moderate-to-severe sleep apnea.

Discussion

This study evaluated the predictive accuracy of the Berlin Questionnaire and a Modified Berlin Questionnaire in identifying sleep apnea among patients with acute ischemic stroke. The main findings are: (1) The prevalence of sleep apnea (REI ≥ 5) in this cohort was high (80.4%), consistent with prior studies. (2) The standard Berlin Questionnaire demonstrated moderate accuracy in predicting moderate-to-severe sleep apnea. (3) Modifying the BMI category as an independent risk factor slightly improved sensitivity and overall diagnostic performance. (4) Moderate-to-severe sleep apnea was associated with lower minimum oxygen saturation, highlighting the role of nocturnal hypoxemia in stroke pathophysiology.

Comparison with previous studies

Our results align with earlier research reporting modest diagnostic accuracy of the Berlin Questionnaire in clinical populations. Ng et al. reported a sensitivity of 81% and specificity of 48% when compared with home polygraphy [20], while Sforza et al. found 77% and 39%, respectively, among elderly subjects [21]. In contrast, studies in atrial fibrillation cohorts have demonstrated higher sensitivity (76–78%) but lower specificity, suggesting that comorbid conditions may influence questionnaire performance [22]. Srijithesh et al. examined the BQ’s validity in stroke patients four weeks after onset and found sensitivity and specificity values of 66.7% and 55.6%, respectively, which are comparable to our findings [11]. However, the present study focused on the acute stage of stroke, when sleep-disordered breathing may be more severe due to transient neurological dysfunction.

Clinical implications

Hypoxia is associated with an increased risk of mortality in patients with acute ischemic stroke [23]. A study by Roffee et al. found that 23% of patients with acute ischemic stroke spent more than 30 min with oxygen saturation below 90% [24]. Our study found that patients with moderate to severe sleep apnea in the acute stroke phase had significantly lower oxygen saturation levels.

According to the guidelines for the early management of patients with acute ischemic stroke, supplemental oxygen is recommended for those with hypoxia to maintain oxygen saturation above 94% [25]. The results of our study emphasize the importance of screening for sleep apnea in patients with acute stroke. Modified Berlin Questionnaire could be a useful tool for screening patients at risk of sleep apnea and for helping doctors improve outcomes for patients with ischemic stroke.

OSA is highly prevalent in stroke patients, with meta-analyses reporting rates between 67% and 75% [26]. Consistent with these data, 80.4% of our participants demonstrated some degree of sleep apnea(REI ≧ 5). Early identification of sleep-disordered breathing is critical, as timely intervention with continuous positive airway pressure (CPAP) therapy may improve neurological recovery, reduce cardiovascular risk, and enhance functional outcomes.

The Modified Berlin Questionnaire represents a simple, cost-free, and easily administered screening approach suitable for bedside use in stroke units. Although its sensitivity remains moderate, even modest improvements in early detection could increase the number of stroke patients referred for definitive testing and management.

Mechanistic insights

The finding that patients with moderate-to-severe sleep apnea had significantly lower minimum oxygen saturation supports the concept that nocturnal hypoxemia contributes to neuronal injury and delayed recovery. These results align with the findings of Wali et al., who demonstrated that sleep apnea severity is significantly associated with the desaturation index, time spent below 90% oxygen saturation, and oxygen saturation nadir [27].Intermittent hypoxia triggers oxidative stress and inflammatory cascades that may exacerbate ischemic brain injury [28]. These mechanisms warrant further exploration in longitudinal studies linking oxygen desaturation patterns with neurological outcomes.

Study limitations

Several limitations should be acknowledged.

First, this was a single-center study with a relatively small sample size, which may limit generalizability. Second, both the questionnaire and sleep study were conducted during the acute stroke phase; longitudinal follow-up could provide insights into recovery-related changes. Third, patients with severe neurological deficits or impaired consciousness were excluded, potentially underestimating the true prevalence of sleep apnea in more severe strokes. Finally, we did not correlate sleep apnea severity or oxygenation metrics with long-term functional outcomes, which should be addressed in future research.

Future directions

Larger multicenter trials incorporating both PSG and home-based testing are needed to validate the Modified Berlin Score. Combining clinical questionnaires with physiological markers (e.g., oxygen desaturation index, nocturnal heart rate variability) could yield more robust prediction models. Moreover, integrating such screening tools into acute stroke workflows may facilitate early initiation of CPAP therapy and secondary prevention strategies.

Conclusion

The Berlin Questionnaire demonstrates moderate accuracy in screening for sleep apnea among patients with acute ischemic stroke. Modification of the BMI component improves its diagnostic performance modestly. Moderate-to-severe sleep apnea is associated with lower nocturnal oxygen saturation, suggesting potential mechanistic links to stroke progression and recovery. Implementing simple screening tools such as the Modified Berlin Questionnaire in acute stroke settings may enhance detection and management of sleep-disordered breathing, ultimately improving patient outcomes.

Authors’ contributions

Augustus CK, Wong: Writing – original draft, Writing – review & editing, Statistical analysis, Data collection.Yi-Sin Wong: Writing – original draft, Writing – review & editing, Statistical analysis, Data collection.Sheng-Feng Sung: Writing – review & editing, Supervision, Funding acquisition.Chi-Shun Wu: Writing – review & editing, Project administration, Data collection.Yu-Hsiang Su: Writing – review & editing, Project administration, Data collection.Mao-Hsun Lin: Writing – review & editing, Project administration, Data collection.Chien-Yu Su: Writing – review & editing, Project administration, Data ollection.Cheung-Ter Ong: Writing – original draft, Supervision, Methodology, Conceptualization.

Funding

This study was supported by a research grant from Ditmanson Medical Foundation Chiayi Christian Hospital (R112-03).

Data availability

The data that support the finding of the study are available from the corresponding author at **ctong98@yahoo.com.tw** under reasonable request.

Declarations

Ethics approval and consent to participate

This study followed the ethical standards of the Research Ethics Committee of Chia Yi Christian Hospital and complied with the 1964 Declaration of Helsinki and its subsequent revisions or equivalent ethical standards. The research plan was approved by the Institutional Review Committee of Chia Yi Christian Hospital (IRB No.: 2022110). All participants signed written informed consent forms before enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.