Obstructive Sleep Apnea and Cognitive Impairment: Implications for Surgical Outcomes

ABSTRACT

The influence of coexisting obstructive sleep apnea (OSA) and cognitive impairment (CI) on perioperative outcomes is unclear, despite its potential clinical significance. We aimed to compare clinical outcomes between surgical patients with OSA or high risk OSA (OSA/HR‐OSA) who have CI and those without CI. We assessed 655 surgical patients aged ≥ 65 years. All participants completed the STOP or STOP‐Bang questionnaires to assess for risk of OSA and were classified as OSA/HR‐OSA or No‐OSA/LR‐OSA. To assess for probable CI, participants completed the in‐person or telephone Montreal Cognitive Assessment (MoCA) and were classified as CI or No‐CI. Among 655 participants, 377 (57.6%) had OSA/HR‐OSA and 216 (33.0%) had CI. Twenty‐one percent (137/655) presented with coexisting OSA/HR‐OSA and CI. More participants in the OSA/HR‐OSA group had CI compared to the No‐OSA/LR‐OSA group (36.3% vs. 28.4%, p = 0.041). After adjustment, each additional year of education was associated with a 12% decrease in the odds of CI among participants with OSA/HR‐OSA (p < 0.001). Participants with coexisting OSA/HR‐OSA and CI had about 2‐fold higher odds for all‐cause complications (OR: 2.01, 95% CI: 1.29, 3.13, p = 0.002) and 80% higher adverse composite outcomes (OR: 1.79, 95% CI: 1.16, 2.79, p = 0.009) at 30 days postoperatively than those with OSA/HR‐OSA and No‐CI, after adjustment for covariates. In conclusion, coexisting OSA/HR‐OSA and CI was common at 21% and associated with increased 30‐day all‐cause complications and composite outcomes. These findings emphasise the need for enhanced preoperative screening and multidisciplinary strategies to identify and manage such vulnerable patients.

1. Introduction

Obstructive sleep apnea (OSA) is characterised by recurrent upper airway collapse during sleep, resulting in periodic hypoxemia. OSA affects nearly 936 million adults worldwide, with prevalence rising sharply among older adults and surgical populations (Iannella et al. 2025; Benjafield et al. 2019; Heinzer et al. 2015). Older patients may be particularly vulnerable because aging, multimorbidity, and polypharmacy can weaken upper airway stability (Osorio et al. 2022). In the perioperative setting, OSA can exacerbate cardiovascular and respiratory instability, especially under the physiologic stress of surgery and anaesthesia. Unrecognised OSA has been associated with an increased risk of hypoxemia, unexpected intensive care unit (ICU) admission, and prolonged hospitalisation (Chaudhry et al. 2024; Pivetta et al. 2022; Chan et al. 2019; Bolden et al. 2020).

Beyond intraoperative and airway‐related concerns, OSA independently predicts a range of postoperative complications (Pivetta et al. 2022; Chan et al. 2019; Altree et al. 2021). Large cohort studies have reported higher rates of cardiovascular events, respiratory failure, and readmission among patients with polysomnography‐confirmed OSA compared to those without OSA (Chaudhry et al. 2024; Pivetta et al. 2022; Chan et al. 2019; Bolden et al. 2020; Altree et al. 2021). Patient‐reported outcome measures (PROMs) are likewise compromised. OSA has been associated with greater postoperative pain and higher opioid requirements, possibly due to altered nociceptive processing after chronic sleep disruption (Strutz et al. 2019; Lam et al. 2016). Coexisting fatigue and catabolic stress may further contribute to frailty, reducing functional reserve and delaying recovery in older adults (Makary et al. 2010; Lu et al. 2025).

Emerging evidence suggests that OSA may be associated with cognitive impairment (CI), an especially pertinent concern in older surgical patients, who are already at heightened risk for postoperative delirium and perioperative neurocognitive disorders (Marchi et al. 2024; Chen et al. 2022). Specifically, OSA has been identified as a potential risk factor for mild CI (MCI) and Alzheimer's disease (Wang et al. 2025). Studies have shown that moderate‐to‐severe OSA correlates with deficits in attention, executive function, and memory when assessed using validated screening tools such as the Montreal Cognitive Assessment (MoCA) in the general population (Gosselin et al. 2019; Legault et al. 2021; Olaithe and Bucks 2013).

Despite these associations, the impact of coexisting CI among older surgical patients with OSA or high risk of OSA remains incompletely defined. By integrating risk‐stratification approaches to OSA assessment, this study aimed to clarify how CI influences postoperative outcomes in older surgical patients with OSA. This will inform targeted perioperative management for this vulnerable population. This report presents a post hoc analysis of two multicenter, longitudinal prospective cohort studies (Yan et al. 2024, 2025). The primary objectives were to (1) compare postoperative clinical outcomes between older surgical participants with and without CI among those with OSA or high risk OSA (OSA/HR‐OSA) and (2) identify preoperative risk factors associated with coexisting CI and OSA/HR‐OSA.

2. Methods

This study is a post hoc analysis of two multicenter, longitudinal prospective cohort studies in Toronto, Ontario, Canada: The Postoperative Functional Disability in Unrecognised Cognitive Impairment (POF) (Yan et al. 2024) study and the Detection of Cognitive Impairment (Detect‐CI) study (Yan et al. 2025). Research ethics board approval was obtained from both participating institutions (Toronto Western Hospital and Mount Sinai Hospital) for both studies. The clinical trial numbers for the POF and Detect‐CI studies were NCT04850833 and NCT06030765 with registration on April 7, 2021, and August 11, 2023, Respectively. detailed methods have been previously described (Yan et al. 2024, 2025).

2.1. Participants

Participants were recruited consecutively and consented from July 23, 2021, to May 18, 2023, for the POF study and from August 30, 2023, to July 30, 2024, for the Detect‐CI study. Eligible participants were those aged ≥ 65 years, undergoing elective inpatient non‐cardiac surgery and could complete online and/or telephone assessments. Exclusion criteria included known major neurocognitive disorder diagnoses, sensory impairment, substance use disorder, or ambulatory, vascular, or neurosurgical procedures.

2.2. Procedures

This study was conducted as a post hoc pooled analysis of the POF and Detect‐CI cohorts. Since the POF study was conducted during the COVID‐19 pandemic, all assessments were administered remotely, either online through a survey administered on the Research Electronic Data Capture tool (REDCap) (Harris et al. 2009, 2019) or by telephone. Before surgery, participants from the POF study completed the STOP questionnaire, while Detect‐CI participants completed the STOP‐Bang questionnaire to assess OSA risk (Chung et al. 2016, 2008). To evaluate cognitive function, POF participants completed the Telephone Montreal Cognitive Assessment (T‐MoCA), whereas Detect‐CI participants completed the MoCA in person (Nasreddine et al. 2005; Pendlebury et al. 2013).

As this was a secondary analysis of two independent studies conducted before and after the COVID‐19 pandemic, the screening measures were not standardised across cohorts. For the pooled analysis, they were harmonised as conceptually related measures of OSA risk and global cognitive screening. All participants completed the following patient‐reported outcome measures (PROMs) on REDCap: the World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0) for functional disability (Üstün et al. 2010), the 5‐item FRAIL questionnaire for frailty (Van Kan et al. 2008), a visual analog scale (VAS) for pain (Scott and Huskisson 1976), and the single‐item Sleep Quality Scale (SQS) for sleep quality (Snyder et al. 2018).

Demographic and medical information (including age, sex, body mass index [BMI], American Society of Anesthesiologists [ASA] score, total years of education, medical history [including OSA treatment, if any, and compliance], and type of surgery) were obtained from electronic medical records or directly from participants. Clinical outcomes were collected within 30 and 90 days after surgery using chart review by research personnel blinded to initial assessment results. Outcomes included postoperative delirium, all‐cause complications (cardiac, respiratory, gastrointestinal, renal, and others such as wound infections or postoperative nausea/vomiting), hospital length of stay (LOS), non‐home discharge, readmission, mortality, and emergency department (ER) visits. Composite adverse outcomes, which included any of the aforementioned outcomes, were also reported. To supplement chart review, participants or their informants were contacted by telephone or online survey at each follow‐up interval to report any adverse events. Postoperative delirium was evaluated one to three days after surgery (or until discharge) using the Chart‐based Delirium Identification Instrument by trained study staff and the Confusion Assessment Method by nursing teams on each shift (Inouye et al. 2005, 1990).

2.3. Assessment of Obstructive Sleep Apnea and Cognitive Impairment

Participants were classified based on their preoperative OSA and CI status. For OSA, participants were classified as being either diagnosed or high‐risk OSA (OSA/HR‐OSA) or having no OSA diagnosis or at low risk (No‐OSA/LR‐OSA). Specifically, OSA/HR‐OSA was defined as patients with a documented diagnosis of OSA in their electronic medical records or those who screened as high risk based on validated tools—the STOP (for POF participants) or STOP‐Bang (for Detect‐CI participants) questionnaires. This grouping was chosen to capture both diagnosed OSA and patients at high risk of undiagnosed OSA, recognising that OSA is frequently underdiagnosed in clinical populations (Iannella et al. 2025). The STOP questionnaire assesses snoring, tiredness, observed apnea, and high blood pressure, while the STOP‐Bang has additional items on BMI, age, neck circumference, and sex (Chung et al. 2016, 2008, 2012). A STOP score ≥ 2 (out of 4) or a STOP‐Bang score ≥ 3 (out of 8) was used to define high risk (Chung et al. 2008, 2012). Thus, harmonisation for OSA was performed at the level of the instrument‐specific risk classification rather than by assuming direct equivalence between raw STOP and STOP‐Bang scores.

Screening positive for probable CI was defined as a T‐MoCA score of $\le$ 18/22 in POF participants and a MoCA score of $\le$ 25/30 in Detect‐CI participants (Legault et al. 2021; Olaithe and Bucks 2013). An additional point was added to the total T‐MoCA/MoCA score of participants with less than 12 years of education. For simplicity, we will refer to a positive screen as CI. The MoCA is an in‐person cognitive screening tool that assesses multiple domains including visuospatial/executive function, naming, memory, attention, language, abstraction and orientation (Nasreddine et al. 2005; Danquah et al. 2024). It has been validated in the surgical population with good sensitivity and specificity (87% and 72%, respectively) (Danquah et al. 2024). The T‐MoCA is an abbreviated version designed for telephone administration that excludes the visuospatial and naming tasks (Pendlebury et al. 2013). Prior work has shown that the T‐MoCA has acceptable validity as a remote cognitive screening tool and comparable screening performance for detecting CI across diverse populations (Katz et al. 2021). It is a reasonable alternative when in‐person assessment is not feasible. Accordingly, harmonisation for CI was performed using instrument‐specific validated cutoffs to define a common binary screening category, rather than by treating T‐MoCA and MoCA scores as directly interchangeable.

2.4. Statistical Analysis

Statistical analyses were conducted in R (version 4.4.2). Descriptive statistics for demographic characteristics were analysed for the study population, comparing participants with OSA/HR‐OSA to those with No‐OSA/LR‐OSA. Categorical variables are reported as counts and percentages; continuous variables are presented as means and standard deviations or medians and interquartile ranges, as appropriate. Additionally, demographic characteristics were analysed comparing participants with and without CI, among those with OSA/HR‐OSA. Between group comparisons were made using Chi‐squared tests, Wilcoxon‐rank and Fisher's exact tests, as appropriate. There were no missing data for the main classification variables (OSA and CI status). Missingness for clinical outcomes and patient‐reported measures was low (< 5%); therefore, no imputation was performed and analyses were conducted using available‐case data for each outcome.

To identify preoperative predictors of coexisting CI and OSA/HR‐OSA, univariable logistic regression for each clinically relevant candidate factor was performed. A priori and other clinically relevant variables with a p‐value $\le$ 0.2 in univariable analyses were included in the multivariable logistic regression model. Clinical outcomes within 30 and 90 days after surgery were compared between the OSA/HR‐OSA and No‐OSA/LR‐OSA participants using generalised linear models. Outcomes included delirium, hospital LOS, non‐home discharge, all‐cause complications, readmission, mortality, ER visits and composite outcomes (any adverse outcome). All analyses were adjusted for age, sex, ASA score, BMI, hypertension and coronary artery disease (CAD) except for outcomes of readmission and mortality due to low event counts.

An identical analysis was performed to compare clinical outcomes between participants with CI + OSA/HR‐OSA and No‐CI + OSA/HR‐OSA, with similar adjustment for covariates except for outcomes of delirium, readmission and mortality due to low event counts. We also performed a subgroup analysis including only participants with a pre‐existing diagnosis of OSA. Clinical outcomes were compared between those who were treatment compliant (i.e., continuous positive airway pressure [CPAP], mandibular advancement device or those who were assessed as not requiring treatment) versus those who were non‐compliant with treatment. All tests were two‐tailed, and statistical significance was defined with an alpha value of 0.05.

3. Results

3.1. Study Population

A total of 655 participants were included in this analysis (POF study: n = 262; Detect‐CI study: n = 393) (Figure 1). Of all participants, 377 (57.6%) had OSA/HR‐OSA. Among them, 135 (35.8%) had diagnosed OSA and 242 (64.2%) were classified as HR‐OSA. The remaining 278 participants (42.4%) had No‐OSA/LR‐OSA. The demographics and characteristics for the study population are shown in Table 1. The median age was 72.0 [68.0–77.0] years with 42% males. Most participants had an ASA physical status III/IV (85.8%) and underwent orthopaedic surgery (61.7%). The proportion of males was higher in the OSA/HR‐OSA group than the No‐OSA/LR‐OSA group (Table 1). In addition, more participants in the OSA/HR‐OSA group had hypertension, CAD, and diabetes mellitus.

FIGURE 1.

Study flowchart.

TABLE 1.

Demographics and baseline characteristics of study population.

Characteristic	OSA/HR‐OSA (n = 377)	No‐OSA/LR‐OSA (n = 278)	p‐value
Age, years	72.0 [68.0–77.0]	73.0 [68.3–77.0]	0.74
Sex, male	198 (52.5)	80 (28.8)	< 0.001
Education a , years	15.5 [13.0–17.0]	16.0 [13.0–17.0]	0.76
BMI b , kg/m2	31.0 ± 6.6	27.6 ± 6.0	0.42
ASA class
I/II	34 (9.1)	59 (21.2)	< 0.001
III/IV	343 (91.0)	219 (78.8)	< 0.001
Medical history
Hypertension	256 (67.9)	112 (40.3)	< 0.001
CAD	46 (12.2)	11 (4.0)	< 0.001
Asthma/COPD	64 (17.0)	45 (16.2)	0.87
Diabetes mellitus	87 (23.1)	20 (7.2)	< 0.001
Stroke	17 (4.5)	12 (4.3)	> 0.99
Surgery type
Orthopaedic	232 (61.5)	172 (61.9)	> 0.99
General	33 (8.8)	38 (13.7)	0.66
Urological	39 (10.3)	17 (6.1)	0.020
Spinal	15 (4.0)	15 (5.4)	0.96
Otolaryngological	25 (6.6)	23 (8.3)	0.52
Gynaecological	9 (2.4)	21 (7.6)	> 0.99
Other	6 (1.6)	10 (3.6)	0.17
Smoking	9 (2.4)	14 (5.0)	0.11
Alcohol use	247 (65.5)	162 (58.3)	0.07
CI (MoCA)	137 (36.3)	79 (28.4)	0.041
WHODAS c	8.85 ± 7.87	6.93 ± 7.05	< 0.001
Disability exceeding patient‐acceptable level ($\mathbf{\ge }$ 16%)	143 (38.9)	61 (22.3)	< 0.001
Clinically significant disability ($\mathbf{\ge }$ 35%)	45 (12.2)	16 (5.8)	0.009
FRAIL c	1.32 ± 1.19	0.99 ± 1.01	< 0.001
Prefrail	187 (50.8)	140 (51.1)	> 0.99
Frail	63 (17.1)	21 (7.7)	< 0.001
Pain d	4.00 ± 3.04	3.86 ± 3.02	0.59
SQS e	6.31 ± 2.38	6.77 ± 2.26	0.014

Note: Values expressed as mean ± SD, median [interquartile range (IQR)], or n (%) where appropiate. Bold values indicate p < 0.05 is significant.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; FRAIL, fatigue, resistance, ambulation, illnesses, and loss of weight; HR, high risk; LR, low risk; OSA, obstructive sleep apnea; SQS, sleep quality scale; WHODAS, World Health Organization Disability Assessment Schedule.

^a Data missing in one participant.

^b Data missing in three participants.

^c Data missing in 13 participants.

^d Data missing in 16 participants.

^e Data missing in 24 participants.

For PROMs, participants with OSA/HR‐OSA had worse preoperative functional disability, frailty, and sleep quality scores than those in the No‐OSA/LR‐OSA group (Table 1). Additionally, more participants with OSA/HR‐OSA had functional disability, clinically significant functional disability, and frailty than those with No‐OSA/LR‐OSA (Table 1).

The total prevalence of CI was 33.0% (Table 1). Almost 21% (137/655) of participants had coexisting CI and OSA/HR‐OSA. More participants in the OSA/HR‐OSA group had CI compared to the No‐OSA/LR‐OSA group (36.3% vs. 28.4%, p = 0.041). Among the OSA/HR‐OSA participants, those with CI had lower median education than those with No‐CI. All other characteristics and PROMs were comparable between those with and without CI (Table 2).

TABLE 2.

Demographics and characteristics comparing those with and without CI.

Characteristic	OSA/HR‐OSA (n = 377)
	CI (n = 137)	No‐CI (n = 240)	p‐value
Age, years	72.0 [69.0–77.0]	72.0 [68.0–76.3]	0.32
Sex, male	77 (56.2)	121 (50.4)	0.33
Education a , years	15.0 [13.0–16.0]	16.0 [13.0–18.0]	< 0.001
BMI b , kg/m2	29.6 [26.3–34.2]	29.7 [26.6–34.6]	0.66
ASA class
I/II	8 (5.8)	26 (10.8)	0.10
III/IV	129 (94.2)	214 (89.2)	0.10
Medical history
Hypertension	97 (70.8)	159 (66.3)	0.43
CAD	20 (14.6)	26 (10.8)	0.36
Asthma/COPD	20 (14.6)	44 (18.3)	0.43
Diabetes mellitus	37 (27.0)	50 (20.8)	0.21
Stroke	9 (6.6)	8 (3.3)	0.23
Surgery type
Orthopaedic	82 (59.9)	150 (62.5)	0.69
General	16 (11.7)	26 (10.8)	0.35
Urological	15 (10.9)	25 (10.4)	0.91
Spinal	4 (2.9)	13 (5.4)	0.60
Otolaryngological	13 (9.5)	12 (5.0)	0.14
Gynaecological	4 (2.9)	11 (4.6)	0.50
Other	3 (2.2)	3 (1.3)	0.67
Smoking	2 (1.5)	7 (2.9)	0.50
Alcohol use	89 (64.9)	158 (65.8)	0.95
WHODAS c	9.87 ± 8.56	8.29 ± 7.43	0.09
Disability exceeding patient‐acceptable level ($\mathbf{\ge }$ 16%)	55 (42.3)	88 (37.0)	0.37
Clinically significant disability ($\mathbf{\ge }$ 35%)	18 (13.8)	27 (11.3)	0.59
FRAIL c	1.48 ± 1.20	1.24 ± 1.18	0.06
Prefrail (1–2)	67 (51.5)	120 (50.4)	0.92
Frail ($\mathbf{\ge }$ 3)	28 (21.5)	35 (14.7)	0.13
Pain d	4.39 ± 3.36	3.79 ± 2.84	0.09
SQS e	6.09 ± 2.51	6.43 ± 2.30	0.18

Note: Values expressed as mean ± SD, median [interquartile range (IQR)], or n (%) where appropriate. Bold value indicate p < 0.05 is significant.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; CAD, coronary artery disease; CI, cognitive impairment; COPD, chronic obstructive pulmonary disease; FRAIL, fatigue, resistance, ambulation, illnesses, and loss of weight; HR, high risk; LR, low risk; OSA, obstructive sleep apnea; SQS, sleep quality scale; WHODAS, World Health Organization Disability Assessment Schedule.

^a Data missing in one participant.

^b Data missing in three participants.

^c Data missing in 13 participants.

^d Data missing in 16 participants.

^e Data missing in 24 participants.

3.2. Preoperative Factors Associated With Cognitive Impairment and OSA/HR‐OSA

A multivariable logistic regression model was performed to explore clinically relevant preoperative factors associated with CI in participants with OSA/HR‐OSA (Table 3). For each additional year of education, the odds of CI in participants with OSA/HR‐OSA decreased by 12% after adjusting for diabetes, stroke, frailty, pain, and sleep quality.

TABLE 3.

Univariable and multivariable analyses of preoperative factors associated with CI and OSA/HR‐OSA.

Covariate	Comparison	Univariable analysis		Multivariable analysis a
		OR (95% CI)	p‐value	OR (95% CI)	p‐value
Demographic characteristic
Surgery	Orthopaedic vs. other	0.89 (0.58, 1.38)	0.61	—	—
Age	+5 years	1.12 (0.93, 1.35)	0.25	—	—
Education	+1 year	0.85 (0.79, 0.92)	< 0.001	0.88 (0.80, 0.94)	< 0.001
Sex	Male vs. female	1.26 (0.83, 1.93)	0.28	—	—
BMI	+1 kg/m2	0.99 (0.96, 1.02)	0.59	—	—
ASA	+1	1.11 (0.75, 1.65)	0.60	—	—
Hypertension	Yes vs. no	1.24 (0.79, 1.96)	0.36	—	—
Diabetes	Yes vs. no	1.41 (0.86, 2.29)	0.172	1.42 (0.83, 2.43)	0.19
Asthma	Yes vs. no	0.76 (0.42, 1.34)	0.35	—	—
CAD	Yes vs. no	1.41 (0.75, 2.62)	0.28	—	—
Stroke	Yes vs. no	2.04 (0.76, 5.56)	0.153	2.34 (0.83, 6.65)	0.10
Alcohol consumption	Yes vs. no	0.96 (0.62, 1.50)	0.86	—	—
Smoking	Yes vs. no	0.49 (0.07, 2.07)	0.38	—	—
Preoperative assessments
Functional disability	Yes vs. no	1.25 (0.81, 1.93)	0.32	—	—
FRAIL	+1	1.18 (0.98, 1.41)	0.074	0.97 (0.77, 1.20)	0.76
Pain	+1	1.07 (0.99, 1.15)	0.074	1.06 (0.98, 1.15)	0.15
SQS	+1	0.94 (0.86, 1.03)	0.185	0.95 (0.87, 1.05)	0.34

Note: Bold values indicate p < 0.05 is significant.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; CAD, coronary artery disease; CI, confidence interval; FRAIL, fatigue, resistance, ambulation, illnesses and loss of weight; OR, odds ratio; SQS, sleep quality scale.

^a The multivariable model was adjusted for diabetes, stroke, frailty, pain and sleep quality.

3.3. Association Between OSA/HR‐OSA, CI and Postoperative Clinical Outcomes

A total of 17 patients (2.7%) experienced postoperative delirium, 13 (2.0%) were readmitted within 30 days postoperatively and 75 (11.8%) were discharged to non‐home facilities (Table S1). Regardless of CI status, participants with OSA/HR‐OSA had about 2‐fold higher odds of all‐cause complications (OR: 1.90, 95% CI: 1.30, 2.78, p < 0.001) at 30 days postoperatively compared to those with No‐OSA/LR‐OSA after adjusting for age, sex, ASA score, BMI, hypertension and CAD. All other clinical outcomes had no significant differences between the two groups at either 30 or 90 days postoperatively.

Among participants with OSA/HR‐OSA, those with CI had about 2‐fold higher odds of all‐cause complications such as cardiac, pulmonary, gastrointestinal, renal or other (e.g., wound infection) complications (OR: 2.01, 95% CI: 1.29, 3.13, p = 0.002) than those with No‐CI at 30 days postoperatively, after adjusting for age, sex, ASA score, BMI, hypertension and CAD (Table 4). Additionally, those with CI had about 80% greater composite adverse outcomes at 30 days (delirium, all‐cause complications, non‐home discharge, ER visits, 30‐day readmission and/or mortality) (OR: 1.79, 95% CI: 1.16, 2.79, p = 0.009) than those with No‐CI after adjusting for age, sex, ASA score, BMI, hypertension and CAD. All other clinical outcomes had no significant differences between the two groups at either 30 or 90 days postoperatively.

TABLE 4.

Clinical outcomes in OSA/HR‐OSA participants with and without CI.

Clinical outcome	CI + OSA/HR‐OSA (n = 137)	No‐CI + OSA/HR‐OSA (n = 240)	OR/Estimate (95% C.I.)	p‐value
30 days postoperatively
Delirium a	4 (3.0)	2 (0.9)	—	0.15
LOS b , days	2 [1–4]	2 [1–4]	0.66 (−0.46, 1.78)	0.25
Non‐home discharge a	13 (9.8)	27 (11.6)	0.79 (0.38, 1.58)	0.51
All‐cause complications c , d	70 (52.2)	81 (34.8)	2.01 (1.29, 3.13)	0.002
30‐day readmission b	1 (0.7)	5 (2.2)	—	0.32
Mortality c	0 (0.0)	2 (0.9)	—	—
ER visits b	14 (10.4)	23 (9.9)	0.99 (0.48, 2.01)	0.99
Composite outcomes c , f	75 (56.0)	95 (40.8)	1.79 (1.16, 2.79)	0.009
90 days postoperatively
All‐cause complications d , e	9 (6.8)	19 (8.2)	0.81 (0.33, 1.82)	0.62
Mortality e	1 (0.8)	1 (0.4)	—	—
ER visits e	6 (4.5)	9 (3.9)	1.10 (0.36, 3.17)	0.86
Composite outcomes e , g	11 (8.3)	19 (8.2)	0.99 (0.44, 2.15)	0.98

Note: Values expressed as median [interquartile range (IQR)], or n (%) where appropriate. Bold values indicate p < 0.05 is significant.

Abbreviations: C.I., confidence interval; CI, cognitive impairment; ER, emergency room; HR, high risk; LOS, length of hospital stay; OSA, obstructive sleep apnea.

^a Data missing in 12 participants.

^b Data missing in 11 participants.

^c Data missing in 10 participants.

^d All‐cause complications include cardiac, pulmonary, gastrointestinal, renal, or other (e.g., wound infection) complications.

^e Data missing in 13 participants.

^f Composite outcomes at 30 days postoperatively included delirium, all‐cause complications, non‐home discharge, ER visits, 30‐day readmission and/or mortality.

^g Composite outcomes at 90 days postoperatively included all‐cause complications, ER visits, and/or mortality.

For the subgroup analysis of OSA participants based on treatment compliance, we found that 90 participants were treatment compliant whereas 42 were non‐compliant (Table S2). Among those who were treatment compliant, 50 (55.6%) participants had CI whereas 40 (44.4%) had No‐CI. Alternatively, among those who were treatment non‐compliant, 26 (61.9%) participants had CI whereas 16 (38.1%) had No‐CI. Participants who were treatment compliant with CI had lower hospital LOS (median [IQR]: 1 [1–3] vs. 3 [1–4.75], p = 0.033) but higher composite outcomes (64.0% vs. 42.1%, p = 0.043) than those with No‐CI at 30 days postoperatively (Table S2). Analysis at 90 days could not be performed due to low counts (Table S2).

4. Discussion

In this study of 655 older surgical patients, 58% of participants had OSA/HR‐OSA and 21% had coexisting CI and OSA/HR‐OSA. Multivariable logistic regression demonstrated that lower education was associated with increased adjusted odds of CI and OSA/HR‐OSA. Importantly, patients with CI and OSA/HR‐OSA had approximately two‐fold higher odds for all‐cause complications and 80% higher odds of composite adverse outcomes at 30 days after surgery compared with those without CI. These findings suggest that CI could be a potential independent risk factor for OSA patients who are at an elevated perioperative risk.

4.1. Association Between Obstructive Sleep Apnea and Cognitive Impairment

The association between OSA and CI, as measured by the MoCA, aligns with prior work demonstrating OSA‐related deficits in attention, executive function and memory. A recent study of 1142 participants referred to a Canadian sleep clinic for suspected OSA found that MoCA scores decreased significantly with increasing OSA severity, with the severe OSA group having the lowest scores and highest prevalence of CI (Beaudin et al. 2024). Another study found that patients with severe OSA had greater CI, with a lower median MoCA score of 23.5 (IQR: 20.0, 25.0) compared to those without OSA (median: 28.5, IQR: 27.8, 29.2) or with mild‐to‐moderate OSA (median: 26.5, IQR: 21.0, 28.0) (Davidescu et al. 2024).

Intermittent hypoxemia and sleep fragmentation may contribute to neuroinflammation, hippocampal vulnerability, and subtle cognitive decline, potentially increasing susceptibility to postoperative delirium and perioperative neurocognitive disorders (Gosselin et al. 2019). Emerging interventional data suggest that effective OSA treatment may mitigate cognitive deficits or slow cognitive decline. In a systematic review of older patients with OSA and MCI or dementia who received continuous positive airway pressure treatment, significant improvements in neurocognitive performance were reported, particularly in verbal, visuospatial, and working memory domains (Montalvão Neves et al. 2025). Although longer‐term, adequately powered trials remain limited, these findings raise the possibility that perioperative identification and management of OSA can improve surgical outcomes and preserve cognitive reserve in older patients.

4.2. Impact of Obstructive Sleep Apnea and Cognitive Impairment on Postoperative Outcomes

Among participants with OSA/HR‐OSA, those with CI experienced significantly higher rates of all‐cause complications and composite adverse outcomes at 30 days after surgery compared with those without CI. The increased complication burden among participants with both CI and OSA/HR‐OSA suggests a potential synergistic vulnerability and bidirectional relationship. CI may impede adherence to postoperative instructions or optimal use of positive airway pressure therapy, while OSA‐induced hypoxemia may further exacerbate neurocognitive stress, creating a cycle of poor recovery.

Another important consideration is the relatively low incidence of postoperative delirium (2.7%) and 30‐day readmission (2.0%) in our cohort compared with the literature (Yan et al. 2023). These low event rates may partly reflect the perioperative practices at our hospitals, including enhanced recovery pathways, multimodal analgesia, and early mobilisation and discharge (Aarts et al. 2018; Wong et al. 2022; Ljungqvist et al. 2017). However, ascertainment bias remains possible, particularly for delirium, as events occurring after discharge may not have been captured despite structured in‐hospital assessment. Readmission ascertainment is less vulnerable to this issue because it was based on both chart review and patient follow‐up, although incomplete capture remains possible. Importantly, the small number of events limits statistical power and precision, increasing the possibility of unstable estimates and type II error. Given the post hoc and observational nature of this pooled observational analysis, residual confounding cannot be excluded despite multivariable adjustment. Hence, associations involving delirium, readmission and other infrequent outcomes should be interpreted cautiously.

4.3. Association Between Obstructive Sleep Apnea, Cognitive Impairment and Education

Our multivariable analysis identified lower education as a predictor for CI among participants with OSA/HR‐OSA, after adjustment for diabetes, stroke, frailty, pain and sleep quality. This finding aligns with literature examining risk factors for CI and OSA. A recent meta‐analysis found that patients who progressed from subjective cognitive decline to MCI or dementia had two fewer years of schooling than patients who did not progress (Arora et al. 2024). Education is thought to contribute to cognitive reserve, which may delay the clinical expression of neurodegenerative changes. in addition, another study demonstrated that among middle‐aged and older adults with untreated OSA, those with higher education performed significantly better on semantic fluency and visuospatial tasks than patients with lower education (Hlaing et al. 2021). This suggests that education may buffer the neurocognitive impacts associated with OSA (Hlaing et al. 2021). Overall, our findings support the hypothesis that lower educational attainment may increase vulnerability to CI among older surgical patients with OSA/HR‐OSA, potentially through reduced cognitive reserve.

4.4. Limitations and Clinical Implications

Several limitations should be acknowledged. First, CI and OSA risk were assessed using different screening instruments across the two parent cohorts. Participants in the POF study completed the T‐MoCA and STOP questionnaire, whereas participants in the Detect‐CI study completed the MoCA and STOP‐Bang questionnaire. In this post hoc analysis, harmonisation was performed at the level of binary screening classification using instrument‐specific validated cutoffs, rather than through direct equivalence of raw scores. Although these instruments assess similar underlying constructs, they differ in content, administration format, and operating characteristics and are not perfectly interchangeable. The STOP‐Bang incorporates additional anthropometric and demographic variables beyond the STOP and may classify OSA risk differently, while the T‐MoCA excludes visuospatial and naming tasks included in the MoCA and may not identify impairment identically. As a result, this pooled analysis may be subject to measurement heterogeneity and misclassification, and the observed associations may partly reflect instrument‐related differences rather than true differences in OSA risk or cognitive status. Accordingly, our findings should be interpreted as exploratory and hypothesis‐generating.

Second, although our cohort included patients from multiple surgical specialties, including orthopaedic, general, urological, spinal, otolaryngological, and gynaecological, the sample was weighted toward orthopaedic patients. Thus, our findings may be reflective of a predominantly orthopaedic perioperative population, and caution is warranted when extrapolating these results to different surgical settings. Third, frailty may represent an important source of residual confounding, or a potential effect modifier, given its close relationship with both CI and adverse postoperative outcomes. Although frailty was assessed preoperatively and included in selected multivariable analyses, including the model exploring preoperative factors associated with CI among participants with OSA/HR‐OSA, it was not examined as an effect modifier and was not uniformly incorporated across all analyses as it was not the primary focus of the present analysis. Finally, due to the observational design of our studies and the post hoc nature of this pooled analysis, causality cannot be established between OSA/HR‐OSA, CI, and adverse postoperative outcomes.

Nevertheless, our results suggest that coexisting OSA/HR‐OSA and CI may identify older surgical patients at elevated perioperative risk. The significance of the observed associations should be interpreted with caution, particularly for outcomes with low event rates, where statistical estimates may be less precise. Future investigations should incorporate objective sleep assessments such as polysomnography and neuropsychological testing, postoperative cognitive follow‐up, and interventional designs to determine whether management of OSA or CI improves outcomes.

5. Conclusion

In this study of older surgical patients, coexisting OSA/HR‐OSA and CI was common at 21%. These patients had approximately two‐fold higher odds of all‐cause complications and 80% greater composite adverse outcomes at 30 days after surgery than those with OSA/HR‐OSA but no CI. These findings support enhanced preoperative screening and multidisciplinary strategies to better risk‐stratify and manage this vulnerable population. Further research with larger prospective cohorts is needed to confirm these findings.

Author Contributions

Yasmin Alhamdah: conceptualization, methodology, investigation, data curation, formal analysis, visualization, writing – original draft, writing – review and editing. Ellene Yan: conceptualization, methodology, data curation, investigation, writing – review and editing. Paras Kapoor: conceptualization, methodology, data curation, investigation, writing – review and editing. Aparna Saripella: investigation, data curation, writing – review and editing. Nina Butris: conceptualization, methodology, investigation, data curation, writing – review and editing. David He: conceptualization, methodology, investigation, writing – review and editing. Leif Erik Lovblom: formal analysis, writing – review and editing. Sazzadul Islam: investigation, data curation, writing – review and editing. Jean Wong: conceptualization, methodology, investigation, writing – review and editing. Frances Chung: conceptualization, methodology, funding acquisition, project administration, writing – review and editing, supervision.

Funding

This research was supported by Ontario Ministry of Health and Long‐Term Care Innovation Fund, University Health Network Foundation (G‐585005300860), ResMed Foundation (G‐575225462760), Canadian Institutes of Health Research (CIHR) (funding reference number: FBD‐195197), Ontario Graduate Scholarship, and Margaret & Howard Gamble Research Grant for the salary of Y.A., E.Y., P.K., N.B., A.S., and S.I. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of Interest

Y.A. reports research support from the Ontario Graduate Scholarship. E.Y. reports research support from the Canadian Institutes of Health Research (CIHR) Canada Graduate Scholarships Doctoral Award (funding reference number: FBD‐195197). P.K. reports research support from the Ontario Graduate Scholarship and Margaret & Howard Gamble Research Grant. J.W. reports research support from the Ontario Ministry of Health and Long‐Term Care Innovation Fund, Merck Inc. and is supported by a Merit Research Award from the Department of Anesthesiology and Pain Medicine, University of Toronto. F.C. reports research support from the Ontario Ministry of Health and Long‐Term Care Innovation Grant, University Health Network Foundation (G‐585005300860), ResMed Foundation (G‐575225462760), and STOP‐Bang Questionnaire proprietary to University Health Network. The remaining authors have no conflicts of interest to declare.

Supporting information

Table S1: Clinical outcomes in participants with OSA/HR‐OSA compared to No‐OSA/LR‐OSA.

Table S2: Clinical outcomes at 30 and 90 days postoperatively comparing those with CI and no CI in those who are treatment vs. non‐treatment compliant OSA participants.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.