Understanding Inpatient Sleep Studies for Sleep-Disordered Breathing

Abstract

Background

The utility of inpatient vs outpatient polysomnography (PSG) for individuals with sleep-disordered breathing is unclear.

Research Question

How do patient characteristics and sleep medicine care patterns differ between individuals undergoing inpatient vs outpatient PSG?

Study Design and Methods

We conducted a retrospective population-based health administrative database study on all adult Ontarians (Canada) hospitalized and/or who underwent PSG between 2012 and 2018. We compared individuals who underwent PSG: (1) during hospitalization (inpatient PSG), (2) within the first month after discharge (delayed PSG), and (3) were not hospitalized in the last year (outpatient PSG). Outcomes included the following: baseline characteristics at the time of PSG, outpatient follow-up rates, and positive airway pressure claims in the year after PSG.

Results

We identified 748 individuals in the inpatient group, 9,310 in the delayed group, and 730,967 in the outpatient PSG group. Compared with delayed or outpatient PSG groups, in unadjusted analyses, individuals in the inpatient PSG group were more likely to be older, previously assessed for sleep-disordered breathing, reside in a low-income neighborhood, and have greater comorbidity burden (standardized differences > 0.10). In adjusted analysis, individuals in the inpatient PSG group were less likely to be seen in the sleep clinic within the first year after PSG than the delayed or outpatient PSG group (hazard ratio [HR], 0.79; 95% CI, 0.71-0.87), with no difference between the delayed and outpatient PSG groups (HR, 1.00; 95% CI, 0.98-1.03). Compared with the delayed or outpatient PSG group, those in the inpatient PSG group were 21% (HR vs delayed group, 0.79; 95% CI, 0.67-0.94) to 53% (HR vs outpatient group, 0.47; 95% CI 0.27-0.82) less likely to initiate CPAP or auto-titrating positive airway pressure, and 2 to 10 times more likely to initiate bilevel positive airway pressure (HR vs outpatient group, 10.03; 95% CI, 7.30-13.77).

Interpretation

Our results indicate that individuals undergoing inpatient PSG represent a unique smaller subgroup with greater comorbidity and social disadvantage, whereas the delayed PSG group may represent an optimal model of care, informing directions for future prospective studies.

Take-Home Points.

Study Question: Do individuals undergoing inpatient vs outpatient polysomnography (PSG) differ in terms of baseline characteristics and sleep medicine health care patterns (outpatient follow-up rates and positive airway pressure therapy initiation)?

Results: Compared with those with outpatient PSG, individuals undergoing inpatient PSG had greater multimorbidity and marginalization, were less likely to receive sleep clinic follow-up, and were more likely to initiate bilevel positive airway pressure over CPAP/auto-titrating positive airway pressure, with no follow-up differences observed between delayed and outpatient groups.

Interpretation: The results of this study indicate that individuals undergoing inpatient PSG represent a unique, smaller subgroup with greater comorbidity and social disadvantage, whereas the delayed PSG group may represent an optimal model of care, informing directions for future prospective studies.

Sleep-disordered breathing (SDB) is the most prevalent sleep disorder,¹ characterized by abnormal respiratory patterns during sleep that cause sleep fragmentation, sympathetic activation, and hypoxemia. Untreated SDB is associated with adverse health outcomes and an increased burden on the health care system.^2,3 Positive airway pressure (PAP) therapy is the recommended treatment for SDB.^4,5 Diagnosing and treating SDB early could reduce morbidity, hospital readmissions, and mortality.6, 7, 8 Level 1 polysomnography (PSG) or an attended sleep study in a sleep laboratory setting has been used as a reference standard for the diagnosis of SDB⁹ and is the preferred test for diagnosing SDB in patients with significant comorbidities.¹⁰ Traditionally, PSG is offered to adults as an outpatient procedure, mostly due to feasibility and insurance coverage limitations.

Despite the importance of SDB, most patients (approximately 80%-90%) with SDB remain undiagnosed.¹¹ Among admitted patients, this diagnostic gap is even more pronounced due to clinical findings overlapping with exacerbations of other medical conditions.12, 13, 14, 15 There can also be a risk of overdiagnosing SDB during hospitalizations due to hypoxemia or fluid overload from other conditions (eg, COPD, chronic heart failure [CHF]). Therefore, it is common that SDB may be suspected during hospitalization, leading to a request for PSG. Inpatient PSG is intended to diagnose and treat SDB early, preventing SDB-related complications and reducing readmissions.¹⁶ However, PSG is complicated to administer, resource intensive, and expensive for inpatients given their health issues and care needs. In response to these challenges, some centers do not perform any inpatient services, whereas others determine on a case-by-case basis (private correspondence). Despite potential benefits, evidence on the utility of inpatient SDB diagnosis and management is limited by sample size, short follow-up time, single-center study design, and selection bias.^6,12, 13, 14^,17, 18, 19, 20 As a result, the use of inpatient PSG is not established as the standard of care, leading to practice variations across sleep centers and jurisdictions.

To address this knowledge gap, we conducted a population-based study to characterize patients (objective 1) and sleep medicine health care patterns (objective 2) for those undergoing inpatient vs outpatient PSG. We hypothesized that individuals who underwent an inpatient vs outpatient PSG would be more likely to have greater morbidity and be marginalized. Because these characteristics can affect access to care and transitions between inpatient and outpatient care, we also hypothesized that individuals who underwent an inpatient PSG would have decreased follow-up rates in the sleep clinic and initiate PAP treatment less frequently than those who underwent PSG as an outpatient.

Study Design and Methods

Study Design and Populations of Interest

We conducted a retrospective population-based study using health administrative databases on all Ontarians (Canada) ≥ 18 years of age who were hospitalized for any cause and/or underwent any sleep study (level 1 PSG) between January 1, 2012, and December 31, 2018. Three distinct groups were created: (1) individuals who underwent a PSG during hospitalization, including the discharge date (inpatient PSG group); (2) individuals who underwent a PSG within the first month postdischarge (delayed PSG group); and (3) individuals who underwent a PSG, but were not hospitalized in the last year (outpatient PSG group). The index date was defined as the PSG date. All individuals were followed through health administrative data until loss of follow-up, death, or December 31, 2019 (the end of the follow-up period), whichever occurred first. The time frame for the study (2012-2019) was chosen (1) to allow 12 months of follow-up to identify outcomes, (2) given that the Ministry of Health changed rules in the PSG reimbursement in 2011, and (3) because the COVID-19 pandemic since March 2020 affected health care delivery in sleep medicine.

Ethical Publication Statement

ICES is an independent, nonprofit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement. The use of anonymized data in this project was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a research ethics board.

Ontario Model of Sleep Medicine Care

In Ontario, an individual with suspected SDB is referred by a physician to a sleep physician for a consultation and PSG, which takes place in a sleep clinic.²¹ Only level 1 PSGs are covered by the government. All insured residents who undergo a level 1 PSG and are diagnosed with SDB by a sleep physician registered with the Assistive Devices Program (ADP) receive funding for PAP systems documented in the ADP database.^22,23 Funding eligibility does not dependent on SDB severity, daytime sleepiness, or adherence to PAP therapy. The choice of PAP therapy (CPAP, auto-titrating positive airway pressure [APAP], or bilevel positive airway pressure [BPAP]) is at the discretion of the sleep physician. The ADP's Respiratory Equipment Pool covers 75% of the cost for PAP systems (CPAP, APAP, or BPAP without a backup rate [BUR]). For those on social assistance, 100% of the cost is covered. The Ventilator Equipment Pool (VEP), fully funded by the government, provides ventilators, including BPAP with BUR, free to eligible individuals. To qualify an individual for BPAP with BUR via the VEP, a level 1 PSG is only required to confirm the diagnosis of SDB, whereas for other conditions, BPAP with BUR could be prescribed without PSG (see details in the E-Text 1).

Data Source

Residents of Ontario have universal public health insurance under the Ontario Health Insurance Plan (OHIP), the single-payer for all medically necessary services. Outpatient and inpatient service details are retained in health administrative databases²⁴ housed at the ICES (https://datadictionary.ices.on.ca). For this project, we used individual-level details on outpatient and inpatient visits, procedures, and claims for PAP therapy retained in health administrative databases, such as the Registered Persons Database (demographics and vital statistics), the Discharge Abstract Database (hospitalizations), the National Ambulatory Care Reporting System Database (emergency department visits), the OHIP database (physician billing and technical fees for procedures, including PSG), the Ontario Cancer Registry, the Narcotics Monitoring System (dispensed prescriptions for opioids and other monitored drugs), the Canadian Census (socioeconomics), and the ADP database (PAP claims and home oxygen therapy). These databases were linked using unique encoded identifiers and analyzed at ICES.

Baseline Characteristics (Objective 1)

Baseline characteristics included the following: (1) sociodemographics (sex, age, place of residence [rural vs urban], neighborhood income quintile,^25,26 Ontario Marginalization Index,^27,28 and Ontario geographic regions based on Home and Community Care Support Services), (2) prior comorbidities likely associated with increased risk of SDB (obesity, hypertension, diabetes, cardiovascular disease [CVD]-relevant hospitalizations, CHF, COPD, asthma, dementia, cancer,²⁹ neuromuscular disease, mood and anxiety disorders, end-stage renal disease/hemodialysis, and alcohol use disorder), (3) opioid use in the last year,³⁰ (4) the Charlson Comorbidity Index (CCI) based on hospitalizations within 2 years before the index date,³¹ (5) prior health care utilization measured by the number of outpatient and inpatient visits in the last year, (6) prior supplemental oxygen use, (7) probability of severe SDB at the index PSG,³² and (8) being diagnosed or treated for SDB in the last 5 years before the index date, defined as the presence of prior PSGs (any type), prior PAP treatment claims, outpatient or inpatient visits with an SDB diagnosis, or surgical interventions for SDB. For hospitalized individuals (inpatient and delayed PSG groups), we also derived information on length of stay (LOS) at the index hospitalization.

Where applicable, we used validated algorithms to ascertain comorbid conditions.33, 34, 35, 36, 37, 38, 39 More details on definitions are provided in the e-Table 1.

Sleep Medicine-Related Services Within the First Year After the Index PSG (Objective 2)

Follow-Up Appointments

In a sleep clinic, the number of outpatient sleep clinic-related visits and the time between the index date and the first clinic visit were reported. An outpatient clinic visit with an SDB diagnosis and a physician registered with the ADP were considered sleep clinic-related. Given that there is no need to be registered with the ADP to prescribe BPAP with a BUR from the VEP, for the sensitivity analyses, we considered an outpatient clinic visit with an SDB diagnosis with any physician (ie, any specialty and not necessarily registered with the ADP).

PAP Therapy Initiation (Claim)

We identified patients who initiated PAP therapy as those with any PAP claim in the ADP databases after the index date considered combined and separately: CPAP, APAP, or BPAP with or without BUR.

Analysis

To address objective 1, descriptive statistics were used to characterize and compare the study populations. A threshold > 0.10 in standardized difference indicated a meaningful difference between groups.⁴⁰ Among hospitalized individuals, we then used a multivariable logistic regression model to identify baseline characteristics associated with being in inpatient vs delayed PSG groups.

To address objective 2, a cumulative incidence function and Gray’s test were used to compare outcomes between groups in unadjusted analyses. To compare follow-up rates in the sleep clinic in the first year, adjusting for the confounders previously listed, we assessed both the time to event and the frequency of events per person-time, using 2 analytical approaches: (1) cause-specific Cox regression (outcome: time to event) and (2) Poisson regression (outcome: count of visits and offset the natural log of person-time). To compare PAP therapy initiation in the first year between groups, we used multivariable cause-specific Cox regression to adjust for the baseline characteristics previously listed. Because in Ontario, only 1 device is covered by the provincial insurance plan within 5 years by the ADP, for individuals who claimed > 1 device in 1 year (< 1%), likely due to clerical error (because the dates for claims were mostly close or the same), for the time to event analyses, we selected a single device based on a prespecified hierarchy order BPAP > CPAP > APAP (because a BPAP device is the most expensive). Only 1 device would be received per individual via the VEP.

Sensitivity Analyses

As a secondary definition for the delayed PSG group, we identified individuals who underwent a PSG within the first 3 months after discharge. We have also compared baseline characteristics of individuals who underwent a PSG within 1, 3, and 6 months after discharge. As a secondary definition for an outpatient sleep clinic-related visit, we considered any outpatient clinic with an SDB diagnosis, regardless of ADP registration.

Results

Between 2012 and 2018, 1,212,771 adults underwent a PSG, and 820,182 were hospitalized for any cause in Ontario (Canada) (Fig 1). Among 1,212,771 individuals who underwent a PSG, we identified 748 adults (0.1%) in the inpatient PSG group (median age, 65 years; 55.9% male; and 27.9% in the lowest income quintile), 9,310 individuals (0.8%) in the delayed PSG group (median age, 60 years; 63.6% male; and 22.2% in the lowest income quintile), and 730,967 adults (60.3%) in the outpatient PSG group (median age, 51 years; 58.1% male; and 17.6% in the lowest income quintile).

Figure 1.

Study flow diagram. We conducted a retrospective population-based study using health administrative databases on all insured Ontarians (Canada) ≥ 18 y of age who were hospitalized for any cause and/or underwent any sleep study (level 1 PSG) between January 1, 2012, and December 31, 2018. Three distinct groups were created: (1) individuals who underwent a PSG during hospitalization, including the discharge date (inpatient PSG group); (2) individuals who underwent a PSG within the first month after discharge (delayed PSG group, primary definition); and (3) individuals who underwent a PSG, but were not hospitalized in the last year (outpatient PSG group). Secondary definition for the delayed PSG group: a PSG within 3 mo postdischarge. OHIP = Ontario Health Insurance Plan; PSG = polysomnography.

Differences in Baseline Characteristics Between Groups (Objective 1)

Compared with individuals with a delayed or outpatient PSG, in unadjusted analyses, individuals in the inpatient PSG group were more likely to be older; reside in a low-income neighborhood and the lowest households and dwellings quintile; have higher CCI; have prior diabetes, hypertension, COPD, asthma, CHF, neuromuscular disease, end-stage renal disease, or dementia, be on supplemental oxygen, and be hospitalized for any CVD; and be assessed for SDB diagnosis in the last 5 years (e-Table 2, Table 1). Compared with the outpatient PSG, those in the inpatient or delayed PSG groups were more likely to be marginalized, have a prior diagnosis of cancer or alcohol use disorder, be seen by a primary care physician or be on opioids in the last year, and undergo a PSG in the last 5 years. Compared with others, individuals in the inpatient PSG group were more likely to reside in a rural area. At index PSG, a higher probability of severe SDB was noted in the inpatient or delayed PSG groups compared with the outpatient group; those hospitalized were also more likely to have in-laboratory titration PSG. Finally, among hospitalized individuals, compared with the delayed PSG group, those in the inpatient PSG group were more likely to stay longer in a hospital (median LOS, 14 days; interquartile range [IQR], 8-26 vs 4 days; IQR, 2-7).

Table 1.

Baseline Characteristics^a by PSG Groups (Objective 1)

Characteristic	Inpatient PSG (n = 748)	Delayed PSG (n = 9,310)	Outpatient PSG (n = 730,967)	Inpatient vs Delayed PSG	Inpatient vs Outpatient PSG	Delayed vs Outpatient PSG
				Standardized Difference
Sociodemographic characteristics
Age, y, median (IQR)	65 (55-75)	60 (51-69)	51 (41-61)	0.32	0.90	0.61
Sex, male	418 (55.9)	5,920 (63.6)	424,523 (58.1)	0.16	0.04	0.11
Neighborhood income quintiles
Q1	209 (27.9)	2,068 (22.2)	128,996 (17.6)	0.13	0.25	0.11
Q2	142 (19.0)	1,972 (21.2)	142,220 (19.5)	0.05	0.01	0.04
Q3	139 (18.6)	1,960 (21.1)	148,397 (20.3)	0.06	0.04	0.02
Q4	137 (18.3)	1,715 (18.4)	154,405 (21.1)	0.00	0.07	0.07
Q5	119 (15.9)	1,576 (16.9)	154,949 (21.2)	0.03	0.14	0.11
Rurality, yes	134 (17.9)	1,219 (13.1)	77,312 (10.6)	0.13	0.21	0.08
ON-Margb
Summary score	3.25 (2.50-3.75)	3.00 (2.50-3.75)	3.00 (2.50-3.50)	0.06	0.21	0.15
Age and labor force quintiles
Q1	146 (19.5)	1,856 (19.9)	186,180 (25.5)	0.01	0.14	0.13
Q2	142 (19.0)	1,734 (18.6)	146,102 (20.0)	0.01	0.03	0.03
Q3	105 (14.0)	1,710 (18.4)	130,931 (17.9)	0.12	0.11	0.01
Q4	127 (17.0)	1,731 (18.6)	126,510 (17.3)	0.04	0.01	0.03
Q5	211 (28.2)	2,204 (23.7)	136,666 (18.7)	0.10	0.23	0.12
Material resources quintiles
Q1	114 (15.2)	1,575 (16.9)	158,588 (21.7)	0.05	0.17	0.12
Q2	125 (16.7)	1,798 (19.3)	155,099 (21.2)	0.07	0.12	0.05
Q3	157 (21.0)	1,776 (19.1)	143,368 (19.6)	0.05	0.03	0.01
Q4	157 (21.0)	1,974 (21.2)	137,779 (18.8)	0.01	0.05	0.06
Q5	178 (23.8)	2,112 (22.7)	131,555 (18.0)	0.03	0.14	0.12
Racialized and newcomer populations quintiles
Q1	169 (22.6)	1,884 (20.2)	118,325 (16.2)	0.06	0.16	0.11
Q2	159 (21.3)	1,701 (18.3)	122,448 (16.8)	0.08	0.12	0.04
Q3	136 (18.2)	1,632 (17.5)	133,586 (18.3)	0.02	0.00	0.02
Q4	142 (19.0)	1,798 (19.3)	155,148 (21.2)	0.01	0.06	0.05
Q5	125 (16.7)	2,220 (23.8)	196,882 (26.9)	0.18	0.25	0.07
Comorbidities
Charlson Comorbidity Index, past 2 y
Low (1 comorbidity)	132 (17.6)	1,665 (17.9)	5,665 (0.8)	0.01	0.61	0.62
Moderate (2)	131 (17.5)	1,478 (15.9)	3,573 (0.5)	0.04	0.62	0.58
High (≥ 3)	302 (40.4)	1,089 (11.7)	2,064 (0.3)	0.69	1.15	0.50
Separate comorbidities
Prevalent
COPD	396 (52.9)	2,720 (29.2)	78,782 (10.8)	0.50	1.01	0.47
Cancer	117 (15.6)	1,220 (13.1)	41,027 (5.6)	0.07	0.33	0.26
Asthma	220 (29.4)	2,099 (22.5)	136,613 (18.7)	0.16	0.25	0.10
Hypertension	590 (78.9)	6,510 (69.9)	286,079 (39.1)	0.21	0.88	0.65
CHF	415 (55.5)	2,403 (25.8)	18,816 (2.6)	0.63	1.43	0.71
Diabetes	411 (54.9)	3,541 (38.0)	127,334 (17.4)	0.34	0.85	0.47
Dementia	65 (8.7)	278 (3.0)	7,607 (1.0)	0.25	0.36	0.14
In the past 5 y
Neuromuscular disease	102 (13.6)	724 (7.8)	33,345 (4.6)	0.19	0.32	0.13
End-stage renal disease/hemodialysis	106 (14.2)	404 (4.3)	2,532 (0.3)	0.34	0.55	0.27
Any CV hospitalizations	395 (52.8)	4,398 (47.2)	64,298 (8.8)	0.11	1.08	0.95
Alcohol dependence/intoxication	41 (5.5)	422 (4.5)	14,291 (2.0)	0.04	0.19	0.15
Obesity/bariatric surgery	≤ 5 (0.1)	93 (1.0)	1,280 (0.2)	0.12	0.01	0.11
Supplemental oxygen use	150 (20.1)	778 (8.4)	2,346 (0.3)	0.34	0.69	0.40
In the past 2 y
Nonpsychotic mood and anxiety disorders	209 (27.9)	2,563 (27.5)	226,305 (31.0)	0.01	0.07	0.08
In the past year
Opioid use	283 (37.8)	3,845 (41.3)	141,550 (19.4)	0.07	0.42	0.49
No. of outpatient physician visits	10 (5-18)	11 (6-18)	7.00 (4-12)	0.06	0.40	0.53
Prior SDB-relevant characteristics, past 5 y
Any PSGs	161 (21.5)	2,052 (22.0)	55,608 (7.6)	0.01	0.4	0.41
SDB diagnosis at any physician visit	590 (78.9)	6,354 (68.2)	426,508 (58.3)	0.24	0.45	0.21
Any PAP claims	46 (6.1)	795 (8.5)	37,066 (5.1)	0.09	0.05	0.14
SDB-relevant characteristics at the index PSG date
Predicted probability of AHI > 30	0.16 (0.10-0.30)	0.17 (0.08-0.37)	0.09 (0.04-0.29)	0.03	0.56	0.53
AHI > 30 (> 50% probability)	68 (9.1)	1,233 (13.2)	33,126 (4.5)	0.13	0.18	0.31
PSG type
Incomplete (< 2 h of recording)	12 (1.6)	136 (1.5)	6,153 (0.8)	0.01	0.07	0.06
Diagnostic	551 (73.7)	7,086 (76.1)	638,887 (87.4)	0.06	0.35	0.30
Therapeutic	185 (24.7)	2,088 (22.4)	85,927 (11.8)	0.05	0.34	0.29

Data are presented as No. (%), unless otherwise noted. Nos. may not add to the total because of missing values. AHI = apnea-hypopnea index; CHF = chronic heart failure; CV = cardiovascular; ON-Marg = Ontario Marginalization Index; PAP = positive airway pressure; PSG = polysomnography; Q = quintile; SDB = sleep-disordered breathing.

^aAll baseline variables (additionally including geographic regions, year of index PSG, and some comorbidities) with missing values and more details on the distribution are presented in e-Table 2.

^bON-Marg measures multiple aspects of marginalization in Ontario and combines 4 distinct dimensions of marginalization assessed at a standard census geographic area of 400 to 700 people, including households and dwellings, material resources, age and labor force, and racialized and newcomer populations (percent of recent immigrants and percent self-identified as a visible minority).²⁷ Please also see details in E-Text 2.

Among hospitalized individuals, in multivariable analysis, female patients and those with a higher CCI, prior COPD, CHF, dementia, on supplemental oxygen at home, and seen for SDB diagnosis as outpatients or underwent any PSG in the last 5 years had higher odds of an inpatient PSG (Table 2). Individuals with prior CVD conditions and previously claimed PAP therapy were less likely to have an inpatient PSG. Higher odds of prolonged LOS were associated with the inpatient group compared with the delayed PSG group.

Table 2.

Estimates From Multivariable^a Logistic Regression to Identify Baseline Characteristics Associated With Having Inpatient PSG vs Delayed PSG (Objective 1)

Baseline Characteristics	OR Estimate	Lower 95% Confidence Limit for OR	Upper 95% Confidence Limit for OR
Sociodemographics
Age, y, per year increase	1.00	0.99	1.01
Sex, F vs M	1.25	1.04	1.51
Neighborhood income quintiles
Q1 vs Q5	1.00	0.67	1.50
Q2 vs Q5	0.90	0.63	1.29
Q3 vs Q5	0.86	0.62	1.19
Q4 vs Q5	0.89	0.65	1.22
Rurality: N vs Y	0.97	0.74	1.26
Components of Ontario Marginalization Index, considered separately per quintile increase
Age and labor force quintile	0.93	0.87	1.01
Material resources quintile	1.00	0.91	1.09
Racialized and newcomer populations quintile	1.03	0.94	1.13
Households and dwellings quintile	1.05	0.97	1.14
Comorbidities
Charlson Comorbidity Index, per 1 score (comorbidity) increase	1.34	1.24	1.43
Prevalent: Y vs N
COPD	1.48	1.20	1.81
Cancer	0.88	0.68	1.15
Asthma	1.11	0.90	1.38
Hypertension	1.00	0.79	1.26
AMI	0.91	0.68	1.22
CHF	2.04	1.64	2.53
Diabetes	0.97	0.78	1.22
Dementia	1.86	1.27	2.73
In the past 5 y: Y vs N
Neuromuscular disease	1.16	0.87	1.54
End-stage renal disease/hemodialysis	1.31	0.95	1.80
Any CV hospitalization	0.79	0.65	0.98
Alcohol dependence/intoxication	1.20	0.82	1.77
Obesity/bariatric surgery	0.22	0.03	1.67
Supplemental oxygen use	1.39	1.08	1.79
In the past 2 y: Y vs N
Nonpsychotic mood and anxiety disorders	0.95	0.76	1.17
In the past year
Opioid use: Y vs N	0.92	0.76	1.11
No. of outpatient physician visits past year, per 1-unit increase	0.99	0.98	1.00
No. of hospitalizations/ED visits past year, per 1-unit increase	0.98	0.95	1.02
Prior SDB-relevant characteristics, past 5 y
Any PSGs	1.35	1.06	1.71
Outpatient SDB diagnosis at any physician visit	1.54	1.23	1.94
Any PAP claim	0.55	0.37	0.82
SDB-relevant characteristics at the index PSG date
Year of the index PSG, per year increase	0.95	0.90	0.99
Predicted probability of severe SDB (AHI > 30), per 5% increase	0.96	0.93	0.98
Index hospitalization
LOS at the index hospitalization, per 1-d increase	1.06	1.05	1.07

Estimates are presented as ORs and 95% CIs to identify odds of having PSG as an inpatient compared with delayed PSG after discharge. Statistically significant values are in boldface font. AHI = apnea-hypopnea index; AMI = acute myocardial infarction; CHF = chronic heart failure; CV = cardiovascular; ED = emergency department; F = female; LOS = length of stay; M = male; N = no; PAP = positive airway pressure; PSG = polysomnography; Q = quintile; SDB = sleep-disordered breathing; Y = yes.

^aAll variables listed were included in the statistical model; the model was also additionally adjusted for 14 Ontario health care-relevant geographic regions (estimates are not shown).

Sleep Clinic Visits and PAP Initiation Within the First Year After the Index PSG (Objective 2)

In unadjusted analysis (Table 3), within the first year after the index PSG, only 390 (52.1%) individuals in the inpatient PSG group were seen in a sleep clinic vs 6,908 (74.2%) in the delayed PSG group and 561,744 (76.8%) in the outpatient PSG group (P < .001) (Fig 1). The time for the first physician assessment in a sleep clinic after an index PSG was also longer in the inpatient PSG group compared with the delayed or outpatient PSG groups (median, 58; IQR, 17-105 vs median, 36; IQR, 18-73 vs median, 40; IQR, 22-77 days; P < .001). The results of the PAP therapy initiation were driven by PAP modes (Fig 2, Table 3). Although in total, the number of PAP therapy initiations within the first year after the index PSG was lower in the inpatient PSG compared with the delayed or outpatient PSG groups (248 [33.2%] vs 4,178 [44.9%] vs 270,703 [37.0%]; P < .001), the time to first PAP claim was shorter in the inpatient PSG group (median, 63; IQR, 16-133 vs median, 81; IQR, 40-143 days in the delayed PSG group and median, 100; IQR, 56-165 days in the outpatient PSG group; P < .001). In contrast with CPAP or APAP modes, BPAP initiation was the most frequent in the inpatient PSG group compared with the delayed or outpatient PSG groups, with the shortest time to the first claim.

Table 3.

Sleep Medicine-Related Services Within the First Year Since the Index PSG Date by PSG Groups (Objective 2)

	Unadjusteda Analyses			Adjustedb Analyses
	Inpatient PSG (n = 748)	Delayed PSG (n = 9,310)	Outpatient PSG (n = 730,967)	Inpatient vs Outpatient PSG	Delayed vs Outpatient PSG	Inpatient vs Delayed PSG
SDB-Relevant Outpatient Visits With a Physician Registered With ADP				Rate Ratios (Poisson Regression, Count) Hazard Ratios (Cox Regression, Time to Event)
No. (%)	390 (52.1)	6,908 (74.2)	561,744 (76.8)	0.85 (0.79-0.92)	1.00 (0.98-1.02)	0.85 (0.79-0.92)
Mean [SD]	0.9 [1.4]	1.4 [1.3]	1.3 [1.1]
Median (IQR)	1.0 (0.0-1.0)	1.0 (0.0-2.0)	1.0 (1.0-2.0)
No. of visits (%)
0	358 (47.9)	2,402 (25.8)	169,223 (23.2)
1	234 (31.3)	3,501 (37.6)	324,316 (44.4)
≥ 2	156 (20.9)	3,407 (36.6)	237,428 (32.5)
Time to first visit, d, mean [SD]	76.8 [77.3]	57.5 [62.2]	61.3 [62.0]	0.79 (0.71-0.87)	1.00 (0.98-1.03)	0.79 (0.71-0.87)
Time to first visit, d, median (IQR)	58.0 (17.0-105.0)	36.0 (18.0-73.0)	40.0 (22.0-77.0)
Any PAP Claims From the ADP's Respiratory Equipment Pool (First Device Per Person)c				Hazard Ratios (Cox Regression, Time to Event)
No. (%)	248 (33.2)	4,178 (44.9)	270,703 (37.0)	1.27 (1.12-1.44)	0.94 (0.90-0.97)	1.36 (1.19-1.54)
On social assistance, No. (%)	54 (7.2)	466 (5.0)	18,183 (2.5)
Time to first claim, d, mean [SD]	89.7 [89.2]	102.3 [81.1]	119.1 [82.1]
Time to first claim, d, median (IQR)	63.0 (16.0-133.0)	81.0 (40.0-143.0)	100.0 (56.0-165.0)
Any CPAP Claims
No. (%)	148 (19.8)	3,474 (37.4)	241,682 (33.1)	0.69 (0.59-0.82)	0.87 (0.84-0.91)	0.79 (0.67-0.94)
Time to first claim, d, mean [SD]	89.2 [79.5]	99.0 [78.6]	117.0 [81.5]
Time to first claim, d, median (IQR)	66.0 (22.5-130.5)	79.0 (39.0-138.0)	98.0 (54.0-162.0)
Any APAP Claims
No. (%)	11 (1.7)	343 (3.7)	21,753 (3.0)	0.47 (0.27-0.82)	0.90 (0.80-1.02)	0.52 (0.30-0.91)
Time to first claim, d, mean [SD]	134.5 [94.7]	112.5 [85.2]	128.5 [81.7]
Time to first claim, d, median (IQR)	109.0 (89.0-199.0)	89.0 (44.0-157.0)	112.0 (67.0-175.0)
Any BPAP Claims (Without BUR)
No. (%)	89 (11.9)	361 (3.9)	7,268 (1.0)	3.39 (2.68-4.30)	1.45 (1.28-1.65)	2.34 (1.84-2.96)
Time to first claim, d, mean [SD]	113.2 [101.8]	131.6 [89.4]	161.9 [89.9]
Time to first claim, d, median (IQR)	87.0 (25.0-165.0)	112.0 (60.0-193.0)	151.0 (90.0-226.0)
Any BPAP-Vent Claims (With BUR) From the Ventilator Equipment Poolc
No. (%)	73 (9.8)	153 (1.6)	1,186 (0.2)	10.03 (7.30-13.77)	3.82 (3.09-4.73)	2.62 (1.94-3.54)
Time to first claim, d, mean [SD]	65.0 [91.8]	100.6 [99.4]	138.5 [98.7]
Time to first claim, d, median (IQR)	22.0 (5.0-84.0)	64.0 (21.0-176.0)	125.5 (55.0-212.0)

ADP = Assistive Devices Program; BPAP = bilevel positive airway pressure; BUR = backup rate; IQR = interquartile range; PAP = positive airway pressure; PSG = polysomnography; SDB = sleep-disordered breathing. Statistically significant values are in boldface font.

^aIn adjusted analyses, all P values for PSG group comparisons were ≤ .001.

^bAll variables listed in Table 2 were included in the statistical model; the model was also additionally adjusted for 14 Ontario health care-relevant geographic regions.

^cThe devices from the ADP's Respiratory Equipment Pool and the Ventilator Equipment Pool were reported separately because of the differences in coverage and indications.

Figure 2.

A-C, Cumulative incidence function plots by PSG group: inpatient, delayed, or outpatient PSG. The outcome of interest within the first year since the index PSG date: (A) CPAP therapy initiation (claim), (B) BPAP without backup rate therapy initiation (claim), and (C) sleep-disordered breathing-relevant follow-up appointments in a sleep clinic with a physician registered with the Assistive Devices Program. BPAP = bilevel positive airway pressure; PSG = polysomnography.

Controlling for confounders, individuals in the inpatient PSG group were less likely to be seen and less frequently seen in the sleep clinic within the first year since the index date compared with the delayed or outpatient PSG groups (rate ratio [RR], 0.85; 95% CI, 0.79-0.92; hazard ratio [HR], 0.79; 95% CI, 0.71-0.87) (Table 3). The difference between the delayed and outpatient PSG groups was not statistically significant. Although individuals in the inpatient PSG group were more likely to initiate PAP therapy sooner (ie, time to first PAP claim was shorter) than those in the delayed or outpatient PSG group, similar to unadjusted analysis, the results were driven by a PAP mode (Table 3). Specifically, compared with the delayed or outpatient PSG groups, those in the inpatient PSG group were less likely to initiate CPAP or APAP, but more likely to initiate BPAP without BUR (HR vs outpatient group, 3.39; 95% CI, 2.68-4.30; HR vs delayed group, 2.34; 95% CI, 1.84-2.96) or with BUR (HR vs outpatient group, 10.03; 95% CI, 7.30-13.77; HR vs delayed group, 2.62; 95% CI, 1.94-3.54). Compared with the outpatient PSG, those in the delayed PSG group were also less likely to initiate CPAP and more likely to initiate BPAP without or with BUR, but to a lesser degree.

We received similar results for secondary definitions; thus, we presented them only in e-Tables 3-4.

Discussion

To our knowledge, this is the first population-based study to describe sleep medicine health services in inpatient vs outpatient individuals. In this study, we confirmed our hypothesis that individuals who underwent an inpatient vs outpatient PSG were more likely to have greater morbidity, be marginalized, and live in rural areas. For these reasons (patient-borne costs, travel, poor access to health care), these individuals may be less likely to complete an outpatient assessment. Interestingly, in multivariable analyses, among hospitalized individuals, only sex, a higher CCI, specific comorbidities (COPD, CHF, dementia), and previous exposure to sleep medicine services were associated with increased odds of an inpatient PSG, but not measures of marginalization.

We also confirmed our hypothesis that individuals who underwent inpatient PSG have decreased follow-up rates in the sleep clinic. Although we found that those in the inpatient PSG group were less likely to initiate CPAP or APAP, they were more likely to initiate BPAP, especially with a BUR, which can be explained by several factors, including underlying comorbid conditions, such as COPD⁴¹ and CHF (ie, prescription was likely driven by respiratory failure), and being fully funded by the government. In addition, as a part of the ventilatory program in Ontario, BPAP could be delivered to a hospital or patient’s home with a prescription signed before the discharge by any physician, which may explain the shorter time to get the device and decrease follow-up rates.

The low number of inpatient PSGs could be explained by the logistical challenges and lack of provincial funding for technical components, making it financially difficult for sleep laboratories to conduct these studies. As expected, likely due to medical complexity, we found a prolonged LOS in those who underwent inpatient PSG vs delayed PSG after discharge. Given hospital overcapacity concerns in Ontario,⁴² a need for an increased LOS would also limit the use of inpatient PSG.

Our findings also suggest that the opportunity for intervention for inpatient SDB lies within the delayed PSG group, given that delayed PSG did not extend LOS and was associated with better sleep clinic follow-up than the inpatient PSG group. Increasing evidence suggests that individuals identified with SDB during hospitalization and who undergo a postdischarge PSG may experience reduced hospital readmissions and improved outcomes.^14,43,44 Additionally, adherence to PAP therapy in individuals who underwent postdischarge PSG was comparable with those undergoing outpatient PSG.⁴⁵

We confirmed our previous single-center study findings that individuals who underwent inpatient PSG have higher multimorbidity and greater social disadvantage.¹⁷ In our study, medically complex individuals were more likely to receive an inpatient PSG, but not sleep medicine follow-up after discharge compared with those who underwent postdischarge SDB sleep testing consistent with a smaller study.¹⁸ The lack of follow-up could impact PAP therapy adherence, as shown in patients after stroke who had difficulty tolerating APAP initiated while hospitalized.¹⁹ It has also been shown that despite proactive screening for SDB and patient education during hospitalization, only 4% of individuals returned after discharge for formal PSG.¹² Hospitalized patients often face care gaps during transitions, causing therapy interruptions and missed outpatient appointments or tests. This can undermine the potential benefits of the inpatient PSG and initiating treatment sooner and highlights the importance of an integrated inpatient/outpatient care system.

Future studies are required to evaluate and compare health care utilization of inpatient PSG and, through cost-effectiveness analysis, to identify the potential for cost savings and individuals most likely to benefit from inpatient PSG. Concerningly, our study demonstrated that in the inpatient PSG group, 79% of individuals were seen for SDB diagnosis, 22% underwent any PSG, and 6% claimed any PAP therapy within 5 years before hospitalization. This suggests missed opportunities to support patients in getting diagnosed, initiating PAP therapy, and improving PAP adherence or considering alternatives. However, it is also possible that some of these patients had already disengaged from treatment, raising concerns about the effectiveness of allocating limited inpatient resources for repeat testing. Thus, optimizing outpatient resource allocation and follow-up is critical to better support individuals with complex health needs at risk of hospitalization.

As a strength of our study, using health administrative databases, given the current health care model and coverage in Ontario, we have a unique opportunity to assess and compare inpatient and outpatient sleep medicine services in Ontario at the population level, with almost complete follow-up.

However, our study has several limitations. We cannot exclude the impact of unmeasured confounders and misclassification biases inherent to health administrative data (eg, lack of data on BMI and arterial blood gas measurements, reliance on diagnostic codes as proxies for obesity/hypoventilation, lack of symptom-level data that may influence acceptance of PAP therapy).⁴⁶ We cannot exclude that PSGs performed in the first month after discharge were scheduled before hospitalization. To capture individuals whose outpatient PSGs may be postponed, we did a sensitivity analysis using secondary definitions for the delayed PSG group (within 3 and 6 months after discharge) and demonstrated similar results. Despite lacking detailed SDB severity data in administrative records, relying solely on a single-center study design is inadequate due to the small sample size and the possibility of patients seeking follow-up evaluations at other sleep centers. We also included the probability of severe SDB³² in the statistical model. Next, we could not exclude the possibility that some individuals may acquire a PAP device without accessing funding through the ADP program and therefore are not registered in the ADP. However, this would be highly unusual in Ontario because all sleep physicians and CPAP vendors are legally required to inform of the ADP. Although information on PAP adherence and effectiveness is not available in administrative databases, this was not the focus of our study. Finally, although the generalizability of our study can be affected by sleep medicine-specific care and coverage in Ontario, our results are consistent with published smaller studies.

Intepretation

In this population-based study, individuals who underwent inpatient vs outpatient sleep studies had greater multimorbidity and higher levels of marginalization. Individuals in the inpatient PSG group were also less likely to be seen in a follow-up sleep clinic and more likely to initiate BPAP, but less likely to initiate CPAP or APAP. The delayed PSG group, in whom LOS was not extended but sleep clinic follow-up rates were greater than the inpatient group, may represent an optimal model of care. This study highlights gaps in care in sleep medicine in the inpatient and outpatient sleep environments and directions for future prospective studies.

Funding/Support

This study was supported by the 2023 CHEST and AASM (American Academy of Sleep Medicine) Foundation Research Grant in Sleep Medicine (United States), and by ICES (Ontario, Canada), which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC).

Financial/Nonfinancial Disclosures

The authors have reported to CHEST Pulmonary the following: T. K. is supported by the PSI (Physicians' Services Incorporated) Foundation: The 2020 PSI Graham Farquharson Knowledge Translation Fellowship; and reports CIHR Foundation support, outside of the submitted work. S. R. P. has received a sponsorship grant and honoraria from Jazz Pharmaceuticals, outside the submitted work. Outside the submitted work, M. I. B. reports consulting fees and honoraria from Jazz Pharmaceuticals, Paladin Labs, Eisai, Precision AQ, Lundbeck, and the OntarioMD Peer Leader Program; travel support from McGill University; and receipt of sleep equipment or research support from Zoll Itamar, Braebon Medical Corp, CGX, The Mahaffy Family Research Fund, and Green Mountain. Outside the submitted work, M. S. B. M. has received grants from ResMed, Paladin, Eisai, and Jazz; speaker’s honoraria from Eisai, Paladin, Idorsia, and Jazz; and is a shareholder of ResMed and Eli Lilly. Outside the submitted work, M. P. reports a grant from Paladin in support of the sleep fellowship program, contract research from Jazz Pharmaceuticals, and honoraria from AstraZeneca. Outside the submitted work, G. C. reports honoraria and advisory board fees from Merck, Gossamer Bio, and Janssen Pharmaceuticals; and clinical trials with Gossamer, Janssen Pharmaceuticals, and AstraZeneca. None declared (R. T., V. L., S. M., K. T.).