PAP therapy and readmission rates after in-hospital laboratory titration polysomnography in patients with hypoventilation

Abstract

Study Objectives:

Hypoventilation associated with sleep-disordered breathing in inpatients is associated with higher risk of morbidity, hospitalizations, and death. In-hospital titration polysomnography qualifies patients for positive airway pressure (PAP) therapy and optimizes settings, but impact is unknown. This study describes a process for in-hospital sleep testing and evaluates subsequent PAP adherence and readmission.

Methods:

A retrospective cohort of patients with hypoventilation and in-hospital titration polysomnography with available PAP data were analyzed to determine whether PAP adherence was associated with 90-day readmission. Absolute differences were obtained using logistic regression models. Models were adjusted for body mass index, age, and Elixhauser index. PAP adherence and nonadherence were defined as ≥ 4 and < 4 hours of daily average use prior to readmission or first 90 days postdischarge.

Results:

Eighty-one patients, 50.6% male, with age (mean ± SD) 61.1 ± 13.5 years were included. Comorbid sleep disorders included 91.4% with obstructive sleep apnea and 23.5% with central sleep apnea. Twenty-eight of 52 (53.8%) nonadherent and 6 of 29 (20.7%) adherent patients had 90-day readmissions. Eleven (13.6%) patients (all nonadherent) were readmitted within 2 weeks of discharge. The adjusted model showed a 35.6% (95% confidence interval 15.9–55.2%) reduction in 90-day readmission in the adherent group compared with the nonadherent group (P = .004). Similar reductions in readmission were found with adherence of ≥ 50% and ≥ 70% of days ≥ 4 hours. Male sex, treatment with iVAPS (intelligent volume-assured pressure support), and highest CO₂ ≥ 60 mmHg on polysomnography were associated with the largest differences in readmission rates between adherent and nonadherent patients.

Conclusions:

Adherence to optimized PAP therapy after in-hospital titration polysomnography in patients with hypoventilation may decrease readmissions.

Citation:

Johnson KG, Rastegar V, Scuderi N, Johnson DC, Visintainer P. PAP therapy and readmission rates after in-hospital laboratory titration polysomnography in patients with hypoventilation. J Clin Sleep Med. 2022;18(7):1739–1748.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Patients with hypoventilation are often discharged from hospitals without appropriate positive airway pressure (PAP) therapy, which may contribute to their high readmission rate. Polysomnography with PAP titration during hospitalization can qualify hypercapneic patients for appropriate PAP therapy but is rarely done due to lack of reimbursement.

Study Impact: We present a model for selecting and studying high-risk patients with in-hospital sleep studies. Our finding that patients with hypoventilation who have in-hospital titration sleep studies and are subsequently adherent to PAP therapy have fewer readmissions than nonadherent patients suggests that discharging hypercapneic patients on appropriate PAP therapy may improve outcomes, especially if postdischarge PAP setup and adherence can be optimized.

INTRODUCTION

Hypoventilation is a common finding in inpatients and can be acute or chronic in nature. Chronic hypercapneic respiratory failure (CHRF) is most commonly caused by chronic obstructive pulmonary disease (COPD), obesity hypoventilation syndrome (OHS), cardiogenic pulmonary edema, and/or neuromuscular disorders. The 1-year readmission rate for patients with CHRF is 60–80%, with approximately 23% of patients within 30 days and 30–49% of patients die within 1 year of discharge.^1,2 COPD is the third most common cause of 30-day readmissions costing over 15 billion annually.²

Sleep often worsens hypoventilation through several mechanisms, including increased upper airway resistance leading to obstructive sleep apnea (OSA), decreased respiratory muscle activation especially in rapid eye movement sleep, decreased chemosensitivity to O₂ and CO₂, supine positioning leading to fluid shifts and increasing ventilation/perfusion (V/Q) mismatch.³ Obstructive sleep apnea (OSA) alone is an independent risk factor for readmission within 30 days.⁴ Overlap syndrome with COPD and OSA has been associated with an increase in 30-day readmissions.⁵ Treatment of breathing in sleep to control sleep-disordered breathing and sleep-related hypoventilation may be able to reduce readmissions by reducing the risk of respiratory failure.

Treatment of sleep-disordered breathing often involves positive airway pressure (PAP) modalities including continuous positive airway pressure, bilevel PAP (BPAP) with or without a backup rate, servoventilation, or volume-assured pressure support (VAPS), which can be delivered by noninvasive ventilation or home mechanical ventilation (HMV). We will use the term PAP to generally refer to all modes and device types, unless specifically notated.

Data on the impact of PAP treatment after hospitalization are limited. Nonadherence to PAP treatment is associated with increased 30-day readmission in patients with OSA.^6,7 In patients with COPD^8,9 and patients with OHS,^10–13 noninvasive ventilation has been shown to decrease mortality, and in patients with COPD to prolong time to readmission.⁸ Research suggests that treatment immediately on discharge in patients with OHS may improve mortality.¹⁰ The trials in nonobese patients with COPD that found benefit with PAP therapy on 12-month readmission and mortality, however, recruited patients with hypoventilation persisting 2 weeks after hospitalization, so the effect of immediate setup is unknown.^8,9 Adherence to empiric PAP therapy in patients with congestive heart failure (CHF) with suspected sleep-disordered breathing followed by postdischarge sleep studies is associated with a nonsignificant trend toward increased survival.¹⁴ Currently, there are no data comparing the efficacy of different PAP therapy types after hospitalization and current trials often exclude comorbidities and overlapping causes of hypoventilation that may make empiric therapy less successful in more complex patients.

Inpatient sleep studies can serve 2 important purposes: first, to qualify a patient to prevent a delay in starting therapy, and second, to determine optimal settings. They are rarely performed due to lack of reimbursement,¹⁵ but if they can prevent readmissions, it may be cost-effective. This study describes a process to perform sleep studies on inpatients. Using a retrospective cohort of patients with hypoventilation who underwent inpatient sleep titration studies, we aimed to see whether adherence to PAP therapy after hospitalization was associated with a reduction in 90-day readmission.

METHODS

This study was conducted as a retrospective cohort of consecutive inpatients who had an in-laboratory titration polysomnography (PSG) performed at Baystate Medical Center, a tertiary academic hospital in western Massachusetts. Determination for ordering inpatient PSG was typically made by a hospitalist or a pulmonary or cardiology consultant. Inpatient sleep medicine consultations were not performed, but all study requests are approved and protocoled by a sleep medicine physician. Instructions for ordering inpatient sleep studies at Baystate (“Hospitalist protocol for qualifying patients for inpatient sleep studies” in the supplemental material) and the decision flow chart (Figure S1 ^{(667.4KB, pdf)} in the supplemental material) for determining need for inpatient PSG that was in place at the time of the study are shown. The protocol was approved by the Institutional Review Board of Baystate Medical Center.

All inpatients with sleep studies between January 10, 2015, and January 4, 2018, were screened. Follow-up clinical data through May 1, 2018, were utilized for readmission data. Inclusion criteria included age ≥ 18 years, evidence of hypoventilation, in-laboratory titration, or split-night PSG. Because arterial blood gas (ABG) was not available on all patients, we utilized a broad definition of hypoventilation defined as ABG during index admission with partial pressure of carbon dioxide (PCO₂) ≥ 45 mmHg or end-tidal CO₂ (ETCO₂)/transcutaneous CO₂ (TCCO₂) during PSG ≥ 50 mmHg for at least 10 minutes. Patients were excluded if PAP therapy was not recommended or if postdischarge PAP use or nonuse could not be confirmed with adherence data or records. The index admission was considered the admission during which the first titration PSG was done. If the patient only had a titration study during admission because of a previous diagnosis, diagnostic sleep study data were obtained from the most recent baseline study prior to admission.

Clinical data were abstracted from electronic medical records. COPD diagnosis was determined by any of the following: pulmonary function testing with forced expiratory volume in 1 second/forced vital capacity (FEV₁/FVC) < 70%, emphysema on computed tomography scan, or if both of those were unavailable, ≥ 20 pack-year smoking history and pulmonary consultant reporting presumed COPD. OHS diagnosis was defined by a body mass index (BMI) ≥ 30 mg/kg², daytime ABG PCO₂ ≥ 45 mmHg, and lack of other severe pulmonary disease that could fully explain the hypoventilation. If daytime ABG was not available, nocturnal CO₂ ≥ 55 mmHg and/or restriction pattern on pulmonary function testing without other severe pulmonary disease were considered OHS. The BMI reported on the PSG was used. The Elixhauser index was chosen as a measure of burden of comorbidities.¹⁶ Data to calculate the Elixhauser index were pulled from the billing records from the index admission.

Patients had nocturnal in-laboratorysplit-night or titration studies. Some patients also had preceding baseline PSG or portable “home” sleep apnea testing with ApneaLink (ResMed, San Diego, CA) while an inpatient. Full standard PSGs were performed using NeuroWorks software (Natus, Pleasanton, CA) at Baystate Medical Center and analyzed manually according to American Academy of Sleep Medicine (AASM) criteria.¹⁷ TCCO₂ monitoring (Radiometer, Copenhagen, Denmark) was used if available; otherwise, ETCO₂ monitoring was used unless interfering with titration. OSA was defined as an obstructive apnea-hypopnea index (OAHI) ≥ 5 events/h and presence of any central sleep apnea was defined by central AHI (CAHI) ≥ 5 events/h. Treatment-emergent central sleep apnea was defined by an OAHI < 5 events/h when CAHI ≥ 5 events/h during titration in a patient who did not have central sleep apnea at baseline. Sleep-related hypoxemia was defined as oxygen saturation ≥ 5 minutes at or below 88% that could not be explained by desaturations due to respiratory events. Titration was primarily done with ResMed (ResMed, San Diego, CA) devices and intelligent VAPS (iVAPS) settings if VAPS was required. Figure S2 ^{(667.4KB, pdf)} in the supplemental material summarizes our inpatient sleep study protocol used for most of the studies. During the study period, BPAP with spontaneous timed back-up rate (BPAP ST) was sometimes tried prior to VAPS, although it is not in our current protocol.

Adherence data

Adherence data were abstracted from either Care Orchestrator (Philips, Murrayville, PA) or AirView (ResMed, San Diego, CA). Given that studies have suggested PAP usage for < 4 hours on 70% of nights may be beneficial,^18,19 we chose to look for an effect of different levels of use. Average use of PAP per day was calculated by total number of hours used divided by the total number of days from the date of discharge of the index admission to either the day before the first readmission or 90 days if no readmission. We also calculated adherence based on having ≥ 50% and ≥ 70% of days with ≥ 4 hours of use and ≥ 50% of days with any use and any use at all in the same time period. Due to the limited sample size and large number of patients with no usage, we dichotomized average PAP usage to adherent (≥ 4 hours of average use) and nonadherent (< 4 hours of average use) patients. For the Kaplan-Meier curve, to define groups prior to follow-up for outcome, we calculated adherence based on having ≥ 4 hours of average use in the first 3 days postdischarge and follow-up time began on day 4. Four patients were excluded from this analysis due to readmission within 3 days of discharge.

Outcomes

The primary outcome was 90-day all-cause readmission after discharge. Readmission data were limited to the Baystate Health System, which includes Baystate Medical Center and 2 local hospitals.

Statistical analysis

Baseline characteristics are reported as means ± standard deviation or medians (interquartile interval) for continuous variables and as counts and percentages for categorical variables. Unadjusted and adjusted logistic regression models were used to determine the association of average hours and percentage of daily PAP usage. Average daily use of PAP ≥ and < 4 hours was used as the primary adherence cutoff. The models were adjusted for Elixhauser comorbidity index, age, and BMI. Using an interaction term in the adjusted and unadjusted models, we looked individually at possible effect modifiers, including COPD status, BMI (< 35 or ≥ 35 kg/m²), heart failure with preserved ejection fraction, treatment type (BPAP vs iVAPs only), sex (male vs female), highest CO₂ on sleep study (< 60 vs ≥ 60 mmHg). The cutoffs for these variables were based on clinically meaningful measurements and limited sample size. We utilized the Stata command “–margins–” to estimate the absolute differences between groups and associated 95% confidence intervals (CIs). All data management and statistical analyses were performed using Stata version 16.1 (StataCorp, College Station, TX).

RESULTS

Patients

A total of 81 patients (50.6% male, average age 61.1 ± 13.5 years) with hypoventilation who underwent an inpatient titration PSG with available PAP adherence data and were discharged between January 2015 and January 2018 were included. Figure 1 shows the selection of 81 (24.8%) patients out of 326 adults who had inpatient sleep studies. The primary exclusion reasons were lack of hypoventilation (112; 34.4%) and unknown adherence status (129; 39.5%). Patients discharged to post–acute care were more likely (87.3%) to lack adherence data compared with patients discharged to home (48.2%).

Figure 1. Patient selection.

Table 1 and Table 2 describe the patient characteristics and PSG data by adherence status. Sleep-related hypoventilation by AASM criteria was present in 71 (87.7%) patients. Seventy-four (91%) patients had OSA. Overall apnea-hypopnea index (AHI) was < 5 events/h in 5 (6%) patients, 5 to < 15 events/h in 14 (17%) patients, 15 to < 30 events/h in 17 (21%) patients, and ≥ 30 events/h in 45 (56%) patients. Of the 19 patients with CAHI ≥ 5 events/h on the diagnostic study, 5 (6.2%) patients had central predominant disease with CAHI > OAHI and 15 (18.5%) patients had Cheyne Stokes respiration. Ninety-six percent of patients had at least 1 of the following diagnoses: OHS, CHF, or COPD. The most common combination was OHS and CHF without COPD, which was present in one-third of patients, followed by OHS alone in 15% of patients. All patients with OHS had OSA.

Table 1.

Patient characteristics.

	All (n = 81)	Nonadherent (n = 52)	Adherent (n = 29)
Age (years)	61.1 ± 13.5	61.7 ± 13.7	60.1 ± 13.2
Male sex	41 (50.6)	26 (50.0)	15 (51.7)
BMI (kg/m2)	40.8 ± 14.1	42.3 ± 14.8	38.0 ± 12.5
BMI category
< 30 kg/m2	19 (23.5)	11 (21.2)	8 (27.6)
30–39.9 kg/m2	45 (55.6)	27 (51.9)	18 (62.1)
≥ 40 kg/m2	17 (21.0)	14 (26.9)	3 (10.3)
Race
White	66 (81.5)	39 (75.0)	27 (93.1)
Black	5 (6.2)	4 (7.7)	1 (3.4)
Hispanic/Latino	6 (7.4)	6 (11.5)	0 (0.0)
Asian	1 (1.2)	0 (0.0)	1 (3.4)
Other	3 (3.7)	3 (5.8)	0 (0.0)
Comorbidities
COPD	28 (34.6)	18 (34.6)	10 (34.5)
OHS	48 (59.3)	32 (61.5)	16 (55.2)
HFpEF	46 (56.8)	32 (61.5)	14 (48.3)
HFrEF	9 (11.1)	7 (13.5)	2 (6.9)
ILD	12 (14.8)	7 (13.5)	5 (17.2)
Pulmonary HTN	31 (38.3)	13 (44.8)	31 (38.3)
Comorbidity combinations
− COPD, + OHS, + CHF	27 (33.3)	19 (36.5)	8 (27.6)
− COPD, + OHS, − CHF	12 (14.8)	6 (11.5)	6 (20.7)
− COPD, − OHS, + CHF	11 (13.5)	8 (15.4)	3 (10.3)
+ COPD, − OHS, − CHF	10 (12.3)	6 (11.5)	4 (13.8)
+ COPD, − OHS, + CHF	9 (11.1)	5 (9.6)	4 (13.8)
+ COPD, + OHS, + CHF	8 (9.9)	7 (13.5)	1 (3.4)
− COPD, − OHS, − CHF	3 (3.7)	1 (1.9)	2 (6.9)
+ COPD, + OHS, − CHF	1 (1.2)	0 (0)	1 (3.4)
Primary reason for admission
COPD exacerbation	18 (22.2)	13 (25.0)	5 (17.2)
Other hypercarbic RF	17 (21.0)	12 (23.1)	5 (17.2)
Other respiratory	7 (8.6)	3 (5.8)	4 (13.8)
Diastolic CHF	23 (28.4)	14 (26.9)	9 (31.0)
Systolic CHF	5 (6.2)	4 (7.7)	1 (3.4)
Other	11 (13.6)	6 (11.5)	5 (17.2)
Risk stratification
Elixhauser index	5.7 ± 1.7	5.9 ± 1.6	5.3 ± 1.9
Prior 1-year hospital days	2.8 (0.0, 14.9)	3.8 (0.0, 17.6)	0.0 (0.0, 10.4)
Prior 1-year admissions	1.0 (0.0, 3.0)	1.0 (0.0, 3.8)	0.0 (0.0, 1.0)
Prior home O2 use	21 (25.9)	12 (23.1)	9 (31.0)
Highest PaCO2 (mmHg) (n = 59)	78.0 ± 21.9	79.1 ± 23.2 (n = 36)	76.3 ± 20.1 (n = 23)
PaCO2 before PSG (mmHg) (n = 59)	63.5 ± 14.8	65.8 ± 15.3	60.1 ± 13.9

Data are presented as n (%), mean ± SD, or median (lower quartile; upper quartile). COPD was considered the primary reason for admission if multiple causes of respiratory failure in patients with COPD. Other respiratory causes included pneumonia (4), pulmonary embolism (1), sarcoidosis (1), asthma (1). BMI = body mass index, CHF = congestive heart failure, COPD = chronic obstructive pulmonary disease, HFpEF = heart failure with preserved ejection fraction, HFrEF = heart failure with reduced ejection fraction, HTN = hypertension, ILD = interstitial lung disease, OHS = obesity hypoventilation, PaCO₂ = arterial partial pressure of carbon dioxide, PSG = polysomnography, RF = respiratory failure, SD = standard deviation.

Table 2.

Sleep study data.

	All (n = 81)	Nonadherent (n = 52)	Adherent (n = 29)
Baseline sleep study data
AHI (events/h)	44.5 ± 39.2	41.4 ± 35.0	50.1 ± 45.9
Lowest O2 (%)	75.4 ± 10.8	75.8 ± 9.8	74.6 ± 12.5
Highest CO2 (mmHg)	64.1 ± 13.5	63.9 ± 14.5	64.5 ± 11.7
Baseline sleep study diagnoses
OSA	74 (91.4)	49 (94.2)	25 (86.2)
Baseline CAHI ≥ 5 events/h	19 (23.5)	12 (23.1)	7 (24.1)
CSR	15 (18.5)	10 (19.2)	5 (17.2)
Hypoxia	70 (86.4)	44 (84.6)	26 (89.7)
Sleep hypoventilation	71 (87.7)	43 (82.7)	28 (96.6)
Additional sleep study diagnoses
TE-CSA on PAP	22 (27.2)	13 (25.0)	9 (31.0)
PAP mode recommended
CPAP	12 (14.8)	11 (21.2)	1 (3.4)
BPAP	21 (25.9)	12 (23.1)	9 (31.0)
BPAP ST	2 (2.5)	2 (3.8)	0 (0.0)
ASV	4 (4.9)	2 (3.8)	2 (6.9)
iVAPS	42 (51.9)	25 (48.1)	17 (58.6)
Oxygen therapy
Recommended	59 (72.8%)	38 (73.1)	21 (72.4)
Treatment response
Residual AHI	4.8 ± 7.3	5.7 ± 9.0	3.5 ± 2.6
Average use (minutes)	142.4 ± 201.8	46.4 ± 97.5	314.4 ± 226.5
Percentage of days > 4 hours	59.3 ± 39.3	41.9 ± 38.5	90.3 ± 11.8
CO2 resolved or improved on PAP*	48 (59.3)	31 (59.6)	17 (58.6)

Data are presented as mean ± SD or n (%). Patient could have > 1 sleep diagnosis. CSA defined as CAHI > 5 events/h. OSA defined by OAHI > 5 events/h. TE-CSA defined by CAHI > 5 events/h with OAHI < 5 events/h during titration when CSA was not present at baseline. Sleep hypoventilation defined by CO₂ ≥ 55 mmHg for 10 minutes or ≥ 50 mmHg for 10 minutes with a 10-mmHg increase from baseline. *CO₂ was considered resolved if ETCO₂ or TCCO₂ improved to < 50 mmHg or improved if decreased by at least 5 mmHg with PAP therapy during sleep study. AHI = apnea-hypopnea index, ASV = ResMed adaptive servoventilation, BPAP = bilevel positive airway pressure without a back-up rate, BPAP ST = BPAP with spontaneous timed back-up rate, CAHI = central apnea-hypopnea index, CPAP = continuous positive airway pressure, CSR = Cheyne Stokes Respiration, ETCO₂ = end-tidal CO₂, iVAPS = ResMed intelligent volume-assured pressure support, OAHI = obstructive apnea-hypopnea index, OSA = obstructive sleep apnea, SD = standard deviation, TCCO₂ = transcutaneous CO₂, TE-CSA = treatment-emergent central sleep apnea.

Post-hospital disposition included 72 (88.9%) patients discharged to home, with 9 (11.1%) to post–acute care, with no difference between adherent (26, 89.7%) and nonadherent (46, 88.5%) patients discharged to home. Five (6.2%) patients died in the year after the admission, including 4 (7.7%) nonadherent patients and 1 (3.4%) adherent patient (P = .45) who died 38 days after discharge after being readmitted. The other deaths were after 90 days.

Table S1 ^{(667.4KB, pdf)} in the supplemental material shows characteristics of the excluded patients for lack of adherence data. Table S2 ^{(667.4KB, pdf)} shows patient characteristics for continuous PAP, BPAP, and iVAPS patients. Sixty-four of 81 patients had ABG data and 5 patients did not have TCCO₂ monitoring on any study, including 1 patient with COPD.

There was PAP use in 13 (16.0%) patients prior to the index admission, including 4 (4.9%) on continuous PAP (1 switched to BPAP, 3 to iVAPS), 7 (8.6%) on BPAP (3 remained on BPAP, 4 switched to iVAPS), and 2 (2.5%) on iVAPS remained on iVAPS.

Adherence

Adherence rates prior to readmission or the first 90 days after discharge varied depending on the measure used. Figure 2 shows the adjusted probability of 90-day readmission by each criterion. For the primary analysis, we used ≥ 4 hours of average use to define adherence, although we found similar results using ≥ 70% and ≥ 50% of days with over 4 hours of use (data not shown). All readmitted adherent patients were using their machine regularly until the day prior to admission, so readmission could not be explained by subsequent nonadherence.

Figure 2. Proportion of patients with 90-day readmission by adherence criteria.

Adjusted proportion with 90-day readmission with 95% confidence interval is shown for each criterion. Adjusted for Elixhauser comorbidity index, age, and body mass index. *Adherence defined in days before readmission or 90 days if no readmission. #Excluded 4 patients with admission prior to day 4 (n = 77). CI = confidence interval, n (%) = percentage of patients adherent or nonadherent to each criterion.

90-day readmission

Readmission within 90 days occurred in 15 of 27 (55.6%) patients with no PAP use, including 10 patients readmitted within 14 days. The adjusted probability of 90-day readmission was 55.5% (95% CI: 41.6–68.0) of nonadherent patients and 19.5% (95% CI: 5.4–33.5) of adherent patients (P = .004). There were no significant differences between unadjusted and adjusted models. In a further analysis to determine how different variables affected readmission rates, 27.5% lower adjusted predicted readmission rates were found in males compared with females (95% CI: –12.6%, 67.7%), 29.8% lower in patients treated with iVAPS compared with BPAP (95% CI: –75.4%,15.8%), and 23% lower in patients with the highest PSG CO₂ ≥ 60 compared with < 60 (95% CI: –67.7%, 21.7%) (Table 3).

Table 3.

Effect modifier analysis on differences in predictive margins for 90-day readmission between adherent and nonadherent patients.

	n	Unadjusted	P	Adjusted*	P
Overall	81	−33.2 (−53.2, −13.1)	.0012	−35.6 (−55.2, −15.9)	.0004
COPD
Yes	28	−35.6 (−69.3, −1.77)	.0392	−36.5 (−67.8, −5.24)	.0061
No	53	−31.9 (−56.7, −7.04)	.0119	−35.0 (−60.1, −9.98)	.0221
BMI
< 35 kg/m2	34	−37.9 (−67.9, −7.85)	.0134	−34.6 (−63.2, −6.00)	.0177
≥ 35 kg/m2	47	−30.1 (−57.5, −2.65)	.0316	−33.9 (−63.0, −4.75)	.0226
HFpEF
Yes	46	−32.6 (−57.8, −7.39)	.0112	−34.7 (−60.0, −9.43)	.0071
No	35	−38.3 (−69.0, −7.71)	.0141	−38.9 (−69.4, −8.51)	.0121
Treatment type
BPAP	25	−20.0 (−55.1, 15.1)	.2636	−17.6 (−54.5, 19.4)	.3511
iVAPS	37	−43.2 (−71.4, −14.9)	.0027	−47.4 (−73.9, −20.8)	.0005
Sleep diagnosis
OSA no CSA	46	−32.6 (−57.8, −7.39)	.0112	−34.0 (−58.8, −9.29)	.007
Any CSA#	31	−38.0 (−70.2, −5.91)	.0203	−41.7 (−72.6, −10.7)	.0083
Sex
Male	41	−48.2 (−73.6, −22.8)	.0002	−48.4 (−73.1, −23.7)	.0001
Female	40	−17.6 (−48.3, 12.9)	.2577	−20.8 (−52.2, 10.5)	.193
Highest CO2 on sleep study
< 60 mmHg	30	−23.8 (−61.2, 13.6)	.2118	−19.5 (−57.1, 18.0)	.3085
≥ 60 mmHg	51	−36.6 (−60.2, −13.1)	.0023	−42.5 (−65.7, −19.3)	.0003

Data presented are the difference in predicted margins between patients with average use ≥ 4 hours and < 4 hours and 95% confidence interval. *Adjusted for age, Elixhauser index, and BMI. #Any CSA is the presence of CAHI ≥ 5 events/h on either baseline or titration study. BMI = body mass index, BPAP = bilevel positive airway pressure without a back-up rate, CAHI = central apnea-hypopnea index, COPD = chronic obstructive pulmonary disease, CSA = central sleep apnea, HFpEF = heart failure with preserved ejection fraction, iVAPS = ResMed intelligent volume-assured pressure support, OSA = obstructive sleep apnea.

In 13 patients, no PAP use was due to readmission occurring prior to home PAP setup. Eleven patients were started on PAP therapy after their first or second readmission; 6 patients subsequently became adherent and 5 were nonadherent. Readmission within 90 days of the subsequent hospital discharge occurred in 1 (16.7%) of the adherent patients and 3 (60%) of the nonadherent patients. When including data after discharge from subsequent admissions, 90-day readmission occurred in 6 of 35 (17.1%) initially or subsequently adherent patients and 30 of 46 (65.2%) never-adherent patients.

Of 13 patents utilizing PAP therapy prior to the index admission, 2 of 6 (33.3%) adherent patients and 5 of 7 (71.4%) nonadherent patients had 90-day readmissions. Readmission within 2 weeks occurred in 11 of 81 (13.6%) patients, including 4 of 28 (14.3%) patients with COPD, none of whom were adherent.

Four patients were readmitted in the first 3 days after discharge, all of whom were not adherent prior to readmission. The Kaplan-Meier estimates of the remaining 77 patients (Figure 3) shows early separation of curves consistent with early readmissions in patients who were never set up with PAP therapy or who were set up but were nonadherent in the first 3 days after discharge (Table S3 ^{(667.4KB, pdf)} shows patient characteristics based on these adherence criteria).

Figure 3. Kaplan-Meier estimates of no readmission based on PAP therapy use.

Adherence was defined by average use from the day after discharge to 3 days later. The not-set-up group did not receive a home PAP device by day 3 after discharge. Day 0 is 4 days after discharge from the index admission. For 4 patients who were readmitted prior to day 4, the date of second readmission was used if within 90 days. hr = hours, PAP = positive airway pressure.

DISCUSSION

Our retrospective analysis of patients with evidence of hypoventilation during admission who underwent inpatient PSG and started on PAP therapy found that the adjusted probability of 90-day readmission was 19.5% in patients who were adherent to PAP therapy vs 55.5% in nonadherent patients (a 65% reduction). The first 2 weeks after discharge accounted for much of the reduction, supporting initiation of treatment immediately on discharge. There were no significant differences with adjustment for age, Elixhauser comorbidity index, and BMI. We found improved readmission rates in adherent vs nonadherent patients, regardless of underlying disease.

As a sensitivity analysis to determine whether our readmission rates were representative of other inpatients with hypoventilation, we found 20 of 45 (44.4%) excluded patients with hypoventilation had 90-day readmission. The similar readmission rates to those included in the study (34/81, 42.0%) suggest that performance of inpatient titration alone is inadequate to influence readmission rates, and that protocols to ensure rapid setup and adherence are critical. This is also supported by the 48% initial readmission rate in patients who were not set up with PAP therapy on the index admission and the 17% subsequent readmission rate in those adherent to PAP therapy after a future admission. It is likely that the initial readmission could have been prevented if setup and adherence had not been delayed.

Because of evidence that lower levels of adherence than the typical 70% of days with over 4 hours of use may be beneficial, we chose to look at different markers of adherence. There was significant improvement in readmissions whether adherence was defined as ≥ 50% of days with any use, ≥ 50% of days with ≥ 4 hours of use, or ≥ 4 hours of average use. This suggests that current requirements for ≥ 70% days with over 4 hours of use are too stringent and ending coverage of devices for shorter hours may be detrimental. This may be especially the case for inpatients with multiple comorbidities that may affect their sleep quality and duration. Unfortunately, we did not have consistent data on sleep architecture, duration, arousals, existence of other sleep comorbidities such as insomnia, or sleep medication use so this was unable to be evaluated.

In our experience, the more comorbidities a patient has, the more likely that empiric therapy will be inadequate and titration studies to guide settings are needed. While our heterogeneous population and multiple treatment modalities are a limitation to generalizing the effect of a given PAP therapy on an outcome in a specific population, we feel our study realistically represents the types of patients for whom inpatient PSG may be beneficial to allow for prompt initiation of PAP therapy on discharge. The frequent overlap of COPD, CHF, and OHS in inpatients is significant since BPAP can inadequately control hypoventilation or induce central sleep apnea,²⁰ the latter which was found in 27% of our population.

In contrast, the COPD studies by Murphy et al⁸ and Kohnlein et al,⁹ which demonstrated mortality benefit with BPAP ST, excluded obese patients and other causes of hypoventilation including heart failure, so represent a different patient population than ours. Half of our population was treated with iVAPS, which has been shown in small studies to have benefit of VAPS over BPAP ST, with outcomes including improvement in quality of life, sleep quality, and CO₂ levels, but none assessing readmission rates.^21–26 Our study had a trend toward lower readmission rates in those treated with iVAPS compared with BPAP. Overall, however, randomized trials comparing modalities in populations with multiple comorbidities who are likely to benefit the most from adaptive settings are needed.

There are some notable limitations to this study. First, this was a single-center retrospective study and is susceptible to indication and selection bias for which patients were ordered PSG, PAP therapy, and had available data. The decision to order inpatient PSG may have been affected by provider preference, patient willingness, ability of patient to follow up for outpatient testing, comorbidities, severity of illness, length of hospitalization, history of readmissions, inability to discharge to a facility without testing, and other factors. There was a nonsignificant trend toward adherent patients having milder disease at baseline than nonadherent patients with a lower Elixhauser index and fewer admissions in the prior year, which may have made them less likely to be readmitted regardless of PAP therapy. PAP-adherent patients may also be more adherent to other treatments affecting their admission rates. The lower readmission rate in patients who only became adherent patients after subsequent admission suggests that the benefit of adherence with PAP therapy is real.

Due to the retrospective nature of the study, and missing ABG data on some patients, we balanced our inclusion criteria for hypoventilation to be narrow enough for patients to have a similar high readmission risk, while including all patients who were referred for an inpatient sleep study for concern of hypoventilation. For this reason, our inclusion definition did not match AASM or International Classification of Sleep Disorders, third edition criteria for sleep-related hypoventilation. Limiting this analysis to only inpatients with hypoventilation did not allow us to evaluate possible benefits of PAP in inpatients with only OSA, which future studies can study.

Additionally, while we generally used data to confirm diagnoses of COPD, CHF, and OHS, not every patient had pulmonary function testing, ABG, or echocardiograms and 5 patients had studies without TCCO₂ monitoring. Other potential data limitations included possible underestimation of readmissions due to admission to other facilities or deaths that were not updated in our electronic medical record; however, our hospital is the primary tertiary care hospital in the region and our dataset includes readmissions to 2 affiliated hospitals. We did not have data on medication compliance. We also did not have data on sleep quality before and after PAP therapy or presence of other sleep disorders, including insomnia, that may also be affected by PAP therapy.

Another limitation is that some of the most severe inpatients with hypoventilation were not included in this study. Patients with COPD, who were able to be started on HMV without PSG, were not included in this study if it was not felt that inpatient titration was essential for optimizing treatment. Additionally, approximately 50% of patients with inpatient PSG were discharged to post–acute care, but because adherence data were unavailable on most of these patients, our sample only included 11% of patients discharged to post–acute care, likely excluding some of the sickest patients.

While our population primarily consisted of patients with CHRF, 10 patients did not meet AASM criteria for sleep-related hypoventilation. This may be due to acute hypercapnic respiratory failure that resolved by the time of the PSG or may have been underestimated in some patients who lacked rapid eye movement sleep on their baseline study or used ETCO₂ rather than TCCO₂ monitoring. The inclusion of patients without CHRF may affect the ability for PAP therapy to affect readmission rate or be associated with a lower readmission risk. We did find larger reductions in readmissions in those who had the highest CO₂ levels (≥ 60) on their sleep study, which is consistent with respiratory failure being a primary driver of readmission in these patients.

Many variables, including comorbidities, PAP adherence, PAP modality, and time to set up, can influence whether PAP therapy can impact readmission. In order to limit variables with the available retrospective data, we chose to compare the effect of adherence status to determine whether there may be an impact from PAP therapy in this setting. The positive results of this study show an association of PAP therapy with reduction in readmissions that should be clarified with future studies to more directly compare (1) empiric PAP therapy vs PAP therapy recommended by titration, (2) inpatient sleep study vs outpatient sleep study, (3) setup on discharge vs delayed setup, (4) patients with a single comorbidity vs multiple, and (5) comparing different PAP therapy modalities. The high rate of patients not set up on discharge or lost to follow-up despite inpatient PSG suggests that further research is needed into barriers to care and measures to improve continuity of care.

Given that the average cost of COPD, CHF, and respiratory failure admissions are approximately $9,400–$18,400,²⁷ it may be beneficial for hospitals to provide inpatient PSG if reduced readmissions or improvements in morbidity and mortality can be confirmed. Inpatients face considerable barriers to being discharged on PAP devices. Because most patients cannot qualify for a PAP/noninvasive ventilation without a PSG, inpatients with hypoventilation are often empirically prescribed HMV for CHRF or are discharged without therapy. HMV devices are expensive (∼ $15,000 for HMV with lifetime rental fees compared with $3,000 for PAP with backup or $1,500 for BPAP). Additionally, HMV devices are supposed to be discontinued once respiratory failure stabilizes, are bulkier and less patient friendly, have alarms that disrupt sleep, and until recently, have not had remote data and adjustment capabilities. An expert panel is currently advocating that Medicare remove qualification barriers for inpatients with hypoventilation,²⁸ which, if successful, may eliminate this reason for inpatient PSG.

Patients being discharged to post–acute care facilities have additional barriers. Many facilities do not accept patients on HMV, and may refuse those requiring outpatient PSG or VAPS due to additional costs without extra reimbursement. Additionally, because patients typically use devices owned or rented by the facility, providers often lack access to data to confirm or optimize therapy, consistent with the low rates of available PAP data in our study population. On discharge from post--acute care, if home setup is not arranged, lapses in treatment leave the patient at risk for readmission. Finally, once a patient gets a home device, required face-to-face follow-up in 60–90 days after setup may not be scheduled, which can result in the device being taken away. Additionally, settings may change after acute hospitalization so clinical follow-up is essential to ensure appropriate long-term treatment.

There are several barriers to providing in-laboratory PSG for inpatients. Our laboratory is located within the hospital, but other centers must bring equipment and technologists to the patient.²⁹ Second, TCCO₂ monitoring availability can also be a limiting factor. We find that TCCO₂ monitoring is critical to optimizing therapy for patients with hypoventilation since ETCO₂ often underestimates arterial PCO₂ in patients on PAP therapy and with comorbidities that increase CO₂ including OHS and COPD.³⁰ In our experience, we often further titrate patients based on TCCO₂ values after obstructive and central respiratory events are eliminated and hypoxia is normalized leading to change in recommended mode. Third, reimbursement for the study is combined in the overall hospitalization reimbursement, so may adversely affect the sleep laboratory budget. Fourth, waiting for a patient to be stable enough to perform the PSG can delay discharge; however, this can be offset by patients who are unable to be discharged because they are unstable off therapy.

In summary, inpatient titration PSG with subsequent optimized outpatient PAP therapy is feasible and associated with reduction in readmission but has many challenges logistically and financially. Because of study limitations, we are unable to make any definitive conclusions about benefit on readmission but feel the positive associations in this paper highlight the need to further study the role of inpatient PSG, especially in patients with hypoventilation. Future studies should specifically evaluate whether inpatient titration PSG to individualize treatment in complex inpatients results in better outcomes than empiric PAP settings, including patients with overlapping causes of respiratory failure. The large proportion of patients who were not set up with therapy on discharge or did not become adherent highlights the need for research into processes to facilitate transition of care from the inpatient to outpatient settings in patients requiring PAP therapy. Finally, understanding the heterogeneity of typical inpatients will also be important for designing target populations for future randomized trials as well as determining whether results of studies in specific disease populations can be applied more generally to inpatients.

ABBREVIATIONS

ABG

arterial blood gas

AHI

apnea-hypopnea index

BMI

body mass index

BPAP

bilevel positive airway pressure

CAHI

central apnea-hypopnea index

CHF

congestive heart failure

CHRF

chronic hypercapneic respiratory failure

CI

confidence interval

COPD

chronic obstructive pulmonary disease

ETCO₂

end-tidal CO₂

HMV

home mechanical ventilation

iVAPS

intelligent volume-assured pressure support

OAHI

obstructive apnea-hypopnea index

OHS

obesity hypoventilation syndrome

OSA

obstructive sleep apnea

PAP

positive airway pressure

PSG

polysomnography

TCCO₂

transcutaneous CO₂

VAPS

volume-assured pressure support

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at Baystate Medical Center, Springfield, Massachusetts. The authors report no conflicts of interest.