Acute Respiratory Distress Syndrome in Patients with Subarachnoid Hemorrhage: Incidence, Predictive Factors, and Impact on Mortality

Abstract

Introduction

Acute respiratory distress syndrome (ARDS) is a known predictor of poor outcomes in critically ill patients. We sought to examine the role ARDS plays in outcomes in subarachnoid hemorrhage (SAH) patients. Prior studies investigating the incidence of ARDS in SAH patients did not control for SAH severity. Hence, we sought to determine the incidence ARDS in patients diagnosed with aneurysmal SAH and investigate the predisposing risk factors and impact upon outcomes.

Methods

A retrospective cohort study was conducted using the National Inpatient Sample (NIS) database for the years 2008 to 2014. Multivariate stepwise regression analysis was performed to identify the risk factors and outcome associated with developing ARDS in the setting of SAH.

Results

We identified 170,869 patients with non-traumatic subarachnoid hemorrhage, of whom 6962 were diagnosed with ARDS and of those 4829 required mechanical ventilation. ARDS more frequently developed in high grade SAH patients (1.97 ± 0.05 vs. 1.15 ± 0.01; p < 0.0001). Neurologic predictors of ARDS included cerebral edema (OR 1.892, CI 1.180–3.034, p = 0.0035) and medical predictors included cardiac arrest (OR 4.642, CI 2.273–9.482, p < 0.0001) and cardiogenic shock (OR 2.984, CI 1.157–7.696, p = 0.0239). ARDS was associated with significantly worse outcomes (15.5% vs. 52.9% discharged home, 63.0% vs. 40.8% discharged to rehabilitation facility and 21.5% vs. 6.3% in-hospital mortality).

Conclusion

Patients with SAH who developed ARDS were less likely to be discharged home, more likely to need rehabilitation and had a significantly higher risk of mortality. The identification of risk factors contributing to ARDS is helpful for improving outcomes and resource utilization.

Introduction

Non-traumatic subarachnoid hemorrhage (SAH) is a devastating disease that affects over 30,000 people every year in the United States. ¹ Spontaneous SAH arising from a ruptured aneurysm or arteriovenous malformation can have catastrophic consequences. Even if the patient survives after initial treatment, many systemic complications contribute to mortality. Pulmonary complications are the most common non-neurological cause of death and of these, acute respiratory distress syndrome (ARDS) is the most frequently encountered.^2,3

ARDS is often devastating with a significantly high mortality rate and has been shown to predispose survivors to long-term cognitive deficits. ARDS affects 200,000 patients every year in the United States and accounts for 10.4% of all intensive care unit (ICU) admissions. ⁴ The incidence of ARDS in patients with SAH increased from 1993 to 2008, while mortality decreased over the same period. ⁵ The mechanism by which ARDS occurs in this patient population is likely multifactorial, but it is proposed to occur secondary to neurogenic pulmonary edema or systemic inflammatory response. ⁶

Studies have shown that patients with aneurysmal SAH who experience pulmonary complications also have a higher incidence of symptomatic vasospasm resulting in worse neurologic outcomes. ⁷ In addition to poor neurological outcomes, ARDS is associated with a significant increase in hospital mortality and ICU length of stay. ⁸

Understanding the risk factors predisposing SAH patients to the development of ARDS may help identify specific subsets of patients that benefit from a high index of suspicion and for whom early interventions may improve outcomes. It has been previously shown that predictors of ARDS in these patients were old age, status epilepticus, cardiac arrest, congestive heart failure, cardiovascular dysfunction, hypertension, chronic obstructive pulmonary disease, hematologic dysfunction, metabolic dysfunction, renal and neurologic dysfunction. ⁵

In this study, we utilized the National Inpatient Sample (NIS) database from 2008 to 2014 to identify risk factors and outcomes in patients with SAH who subsequently developed ARDS for the purpose of identifying predictors of ARDS not previously discussed in literature. We expanded to include the Charlson Comorbidity Index (CCI), a 10-year survival predictive formula based on multiple comorbidities. The CCI is valuable as multiple comorbidities often have a synergistic effect in portending risk for poor outcomes, and it can be useful for providers as a facile tool to stratify risk of ARDS at the time of admission for patients with SAH.^7,9 We also examined novel disease specific-predictors such as the occurrence of ischemic stroke, aneurysm clipping, and the NIS stoke severity score (NIS-SSS) all of which are related to the underlying disease process of SAH.

Methods

Data source

The NIS database, a component of the Healthcare Cost and Utilization Project (HCUP), is the largest all-payer inpatient care database in the USA. It contains data on eight million discharges from more than 1000 hospitals each year, which approximates a 20% stratified sample of all US community hospitals. ¹⁰ The large sample size afforded by the NIS allows for substantive inquiry into healthcare utilization, access, charges, quality, and outcomes. Data elements include demographic characteristics, hospital and regional information, diagnoses, procedures, and discharge disposition for all documented patients.

Patient population

The patient sample was selected by using previously validated International Classification of Diseases, 9th and 10th Editions, Clinical Modification (ICD-9-CM and ICD-10-CM) codes.^11–14 Inclusion criteria was a diagnosis with nontraumatic SAH (ICD-9-CM 430, ICD-10-CM I60) and subsequent ARDS (ICD-9-CM 518.82, ICD-10-CM J80) during their admission. Patients were excluded if they were diagnosed with traumatic SAH or had incomplete or unavailable records.

Measures

We determined the incidence of ARDS in patients admitted with SAH from 2008–2014 and identified risk factors for developing ARDS. Primary endpoints included discharge to home, other than routine (OTR) discharge, and inpatient mortality.

Statistical analysis

Patient demographics, treatment characteristics, and comorbid conditions were assessed using standard descriptive statistics. A multivariate analysis was performed to determine statistically significant predictors of ARDS.

To determine the outcomes of patients with ARDS secondary to SAH, two separate multivariate analyses were done. The first model (Model 1) adjusted for patient-level characteristics – age, gender, NIS Subarachnoid Hemorrhage Severity Score, CCI, Takotsubo cardiomyopathy (TCM), cardiogenic shock, stress ulcer, hematological dysfunction, pulmonary embolism (PE), deep vein thrombosis (DVT), meningitis, encephalitis, pneumonia, cardiac arrest, ventriculitis, electrolyte disorders, decompressive hemicraniectomy, cerebral edema, intracerebral hemorrhage, seizure disorders, alcohol abuse, drug abuse, smoking history, hypertension, and type of aneurysmal repair (surgical, endovascular, or none). The NIS subarachnoid severity score (includes coma/AMS, stupor, hydrocephalus, ventriculostomy, aphasia, cranial nerve deficits, hemiplegia or paraplegia, and mechanical ventilation. The second model (Model 2) adjusted for the components of the first model as well as hospital characteristics.

All statistical analyses were performed using the Statistical Package for Social Sciences (SPSS, Version 24; SPSS, Inc., Armonk, NY).

Results

Patient demographics and characteristics

A total of 170869 patients met the eligibility criteria for SAH, 6962 (4.1%) of whom developed ARDS. The average age of SAH patients with ARDS was 57.87 ± 0.37 versus 59.03 ± 0.14 without (p = 0.0014). 64.07% of SAH patients with ARDS versus 62.17% without were female. Racial breakdown of these patients did not significantly differ (p > 0.05) with 52.88% versus 55.54% White, 13.67% versus 12.99% Black, 10.90% versus 10.39 Hispanic, 7.59% versus 8.64% other, with 14.95% versus 12.43% missing demographic information. Average subarachnoid hemorrhage severity score was 1.97 ± 0.04 versus 1.11 ± 0.01 in SAH patients with ARDS compared to those without. SAH patients with past medical histories of drug abuse (6.05% vs. 4.81%, p < 0.0321) and hypertension (68.15% vs. 64.51%, p = 0.0171), were more likely to develop ARDS on admission. (Table 1)

Table 1.

Baseline patient nand hospital characteristics of ARDS in nontraumatic SAH population 2008 − 2014.

		Total	ARDS	No ARDS	p value
Patient level characteristics
N	168736	4829	163907
Age, years	59.0 ± 0.1	57.9 ± 0.4	59.0 ± 0.2	0.001
Length of Stay	11.7 ± 0.2	20.1 ± 0.5	11.4 ± 0.2	<0.0001
Cost	$52506.0 ± 1047.9	$89744.0 ± 3402.7	$51075.0 ± 1061.1	<0.0001
Women (%)	62.3	64.1	62.2	0.146
Race (%)				0.095
White	55.4	52.9	55.5
Black	13.0	13.7	13.0
Hispanic	10.4	10.9	10.4
Other	8.6	7.6	8.6
Payer status (%)				0.001
Medicare	35.3	32.7	35.4
Medicaid	12.8	16.4	12.6
Private other	41.5	40.5	41.5
Self-pay/No charge	10.2	10.2	10.2
APR DRG Risk Mortality (%)				<0.0001
Mild to moderate	36.6	26.8	37.0
Major	29.2	27.5	29.3
Extreme	34.2	45.6	33.7
Disposition (%)				<0.0001
Routine	36.5	15.5	52.9
Discharge with services	43.1	63.0	40.8
Died	20.4	21.5	6.3
Hospital level Characteristics
Hospital bed size (%)				<0.0001
Small/medium	22.0	15.3	22.3
Large	78.0	84.7	77.7
Hospital teaching (%)				<0.0001
Nonteaching	21.5	15.2	21.7
Teaching	78.5	84.8	78.3
Hospital region (%)				0.539
Northeast	17.2	15.2	17.3
Midwest	21.8	21.6	21.8
South	37.2	37.6	37.2
West	23.8	25.6	23.7

All patient refined- diagnosis related group (APR-DRG).

Complication and associations

ARDS was more likely to occur in SAH patients with higher NIS-SSS on admission (1.97 ± 0.05 vs. 1.15 ± 0.01; p < 0.0001), associated stupor (2.37% vs. 1.66%, p = 0.042), hydrocephalus (48.34% vs. 28.43%; p < 0.0001), ventriculostomy (54.17% vs. 27.22%, p < 0.0001), aphasia (10.46% vs. 6.16%, p < 0.0001), cranial nerve deficits (3.72% vs. 2.30%, p < 0.0001), mechanical ventilation, (69.36% vs. 36.28%, p < 0.0001), PEG tube placement (18.04% vs. 7.24%, p < 0.0001), tracheostomy (17.66% vs. 6.50%, p < 0.0001), cerebral edema (22.09% vs. 13.58%, p < 0.0001), meningitis (4.11% vs. 2.23%, p < 0.0001), ischemic stroke (13.56% vs. 5.74%, p < 0.0001), intracerebral hemorrhage (8.87% vs. 6.82%, p = 0.0054), ventriculitis (3.38% vs. 1.70%, p < 0.0001), electrolyte abnormalities (62.30% vs. 39.20%, p < 0.0001), craniotomy (10.12% vs. 3.64%, p < 0.0001), takotsubo cardiomyopathy (1.32% vs. 0.76%, p = 0.0166), cardiogenic shock (1.79% vs. 0.79%, p = 0.0004), hematologic dysfunction (7.63% vs. 5.41%, p = 0.0003), pulmonary embolism (2.00% vs. 0.92%, p < 0.0001), and deep vein thrombosis (11.60% vs. 6.81%, p < 0.0001) (Table 2).

Table 2.

Baseline SAH comorbidities, complications and treatment of ARDS in nontraumatic SAH population 2008-2014.

	Total	ARDS	No ARDS	p value
NIS-SSS Mean ± SE	1.15 ± 0.01	1.97 ± 0.04	1.11 ± 0.01	<0.0001
Coma/Altered Mental Status	7.8	8.1	7.8	0.7277
Stupor	1.7	2.4	1.7	0.042
Hydrocephalus	29.2	48.3	28.4	<0.0001
Ventriculostomy	28.3	54.2	27.2	<0.0001
Aphasia	6.3	10.5	6.2	<0.0001
Cranial nerve deficits	2.4	3.7	2.3	0.0003
Hemiplegia or paraplegia	1.2	0.8	1.2	0.1541
Mechanical ventilation	37.6	69.4	36.3	<0.0001
PEG	7.7	18.0	7.2	<0.0001
Tracheostomy	7.0	17.7	6.5	<0.0001
Cerebral edema	13.9	22.1	13.6	<0.0001
Meningitis	2.3	4.1	2.2	<0.0001
Ischemic stroke	6.1	13.6	5.7	<0.0001
Intra cerebral hemorrhage	6.9	8.8	6.8	0.0054
Seizures	10.7	11.1	10.7	0.6899
Charlson comorbidity index	1.85 ± 0.01	2.01 ± 0.04	1.84 ± 0.01	<0.0001
Ventriculitis	1.8	3.4	1.7	<0.0001
Electrolyte abnormalities	40.1	62.3	39.2	<0.0001
Craniotomy	3.9	10.1	3.6	<0.0001
Takotsubo cardiomyopathy	0.8	1.3	0.8	0.0166
Cardiogenic shock	0.8	1.8	0.8	0.0004
Cardiac arrest	3.2	2.9	3.2	0.5517
Hematological dysfunction	5.5	7.6	5.4	0.0003
Pulmonary Embolism	1.0	2.0	0.9	<0.0001
Deep vein Thrombosis	7.0	11.6	6.8	<0.0001
Alcohol abuse	5.6	6.4	5.5	0.2045
Drug Abuse	4.9	6.0	4.8	0.0321
Tobacco use	29.6	30.8	29.6	0.318
Hypertension	64.7	68.2	64.5	0.0171
Treatments (%)				<0.0001
Clipping	15.0	38.8	14.0
Coiling	24.9	36.1	24.5
Non-documented	60.1	25.1	61.6

NIS-SSS (Nationwide Inpatient Sample-Subarachnoid Hemorrhage Severity Score), Acute respiratory distress syndrome (ARDS), Percutaneous endoscopic gastrostomy (PEG).

Treatment modalities of clipping and coiling were both more frequent in SAH patients with ARDS (38.84% vs. 13.95%; p < 0.0001, and 36.08% vs. 24.94%; p < 0.0001), respectively. Lack of ARDS development was more common in patients for whom surgical treatment was not documented (25.08% vs. 61.58%; p < 0.0001) (Table 2).

Predictors of ARDS

Factors found to be independently associated with the development of ARDS include age (57.9 ± 0.37 vs. 59.0 ± 0.15; p = 0.0014), length of stay (LOS) (20.1 ± 0.47vs 11.4 ± 0.15; p < 0.001), cost of treatment ($89744 ± 3402.7 vs. $51075 ± 1061.1; p < 0.001), payer status (Medicare, Medicaid, Private Insurance/Other, Self-pay/No-charge) (p = 0.0007). (Table 1)

Multivariate analysis shows that independent predictors of ARDS in patients admitted with SAH were cardiac arrest (OR 4.642, 95% CI 2.273−9.482, p-value <0.0001), cardiogenic shock (OR 2.984, 95% CI 1.157−7.696, 0.0239), stress ulcers (OR 2.732, 95% CI 1.395−5.351, 0.0035), cerebral edema (OR 1.892, 95% CI 1.180−3.034, p-value 0.0083), and age (OR 1.023, 95% CI 1.009−1.037, p = 0.0017) (Table 3).

Table 3.

Logistic regression model evaluating odds of ARDS in nontraumatic SAH population 2008-2014.

Predictors variables	Odds Ratio	95% Confidence Interval	p value¥
Cardiac Arrest	4.642	2.273-9.482	<0.0001
Cardiogenic Shock	2.984	1.157-7.696	0.0239
Craniotomy	2.066	1.081-3.947	0.0282
Cerebral Edema	1.892	1.18-3.034	0.0083
Age	1.023	1.009-1.037	0.0017

Outcomes

After controlling for SAH severity, patients with ARDS were less likely to be discharged home (15.5% vs. 52.9%; p < 0.0001), more likely to be discharged to a rehab facility (63.0% vs. 40.8%; p < 0.0001), and experienced a higher mortality rate than patients without ARDS (21.51% vs. 6.29%; p = 0.0012) (Table 4). Results of the multivariate model were found to be independent of factors including hospital bed size, hospital region, and academic teaching location, suggesting that the effect of ARDS upon outcomes in SAH patients did not differ between location, size, or type of hospital.

Table 4.

Outcomes of patients with ARDS in SAH admissions (2008–2014).

	ARDS (4829)	No ARDS (163907)	Odds ratios	95% Confidence Interval	p value
Discharge to home (%)
Multivariate model 2	15.5	52.9	0.2928	0.2722-0.3150	<0.0001
Other than routine discharge (OTR) (%)
Multivariate model 2	63.0	40.8	1.5429	1.4535-1.6378	<0.0001
In-hospital mortality (%)
Multivariate model 2	21.5	6.3	3.4206	3.1799-3.6795	0.0012

NIS-SSS (Nationwide Inpatient Sample-Subarachnoid Hemorrhage Severity Score), Acute respiratory distress syndrome (ARDS).

Discussion

ARDS poses additional challenges to the recovery of patients with SAH. ³ It is not clear as to why the incidence of ARDS was declining in this study period between 2008 and 2014, when compared to previous years. It may be possible that a greater general awareness and understanding of ARDS as a disease entity, and ensuing changes in institutional and societal guidelines pertaining to fluid and lung management have resulted in an overall decrease over time. However, given that ARDS continues to occur with some frequency and is associated with significant morbidity and cost, it is important to understand the population at risk, the methods by which risk can be assessed, and strategies to potentially mitigate such risk.

One clinical trial showed that in patients with severe head injury, a history of pre-existing hypertension was associated with the development of ARDS, a finding we corroborated. ¹⁵ A 2014 analysis of the NIS database of ARDS in patients with SAH showed that predictors of ARDS were old age, status epilepticus, cardiac arrest, sepsis, congestive heart failure, cardiovascular dysfunction, hypertension, chronic obstructive pulmonary disease, hematologic dysfunction, metabolic dysfunction, renal and neurologic dysfunction. ⁵ This study however did not control for SAH severity so one could not rule out that the increased mortality was secondary to the patients having a higher grade hemorrhage, and hence more likely to have poor outcome. We also found craniotomy/craniectomy to be statistically significant predictors for ARDS which was present but not statistically significant in the previous study. ⁵ Our study also suggests that SAH patients who undergo clipping or coiling for treatment have lower odds of developing ARDS in their hospital admission than those managed medically or with only decompression, a finding not previously seen. It has been shown previously in the literature that lower grade SAH has lower incidence of recurrence and complications including vasospasm, delayed cerebral ischemia, as well as ARDS. ¹⁶ Indeed, when we controlled for SAH severity this relationship disappeared suggesting the association of treatment modality and ARDS can be explained by confounding SAH severity.

Our study showed that patients who developed ARDS during their hospital stay were less likely to be discharged home, required inpatient rehabilitation more often, and had higher rates of mortality. These adverse effects of ARDS upon patient outcome were independent of hospital location, size, or teaching status. These findings are supported by previous literature that found long-term physical decline in patients with ARDS. ¹⁷ The higher mortality rate in patients developing ARDS in our study agrees with literature demonstrating SAH patients developing ARDS have an increased mortality rate. ⁸

The incidence rate of ARDS in patients admitted with SAH in the NIS database was 4.1% in 2008 and 3.3% in 2014. A variety of factors - demographics, comorbidities, and procedures performed during the admission - appear to be independently associated with the development of ARDS in patients admitted with SAH. Additionally, our study confirmed the contribution of factors such as procedures performed that were proposed in previous studies. ⁵ Lastly, we found that patients who developed ARDS in the setting of SAH had a higher rate of in-hospital mortality and a higher rate of adverse discharge disposition (discharge to functionally dependent settings such as inpatient rehabilitation).

Recent literature has suggested that the prognosis of patients with ARDS can be established within 48 h after intubation. ¹⁸ With the information provided by our study, identifying patients with SAH who have the highest likelihood of developing ARDS can allow for early treatment and potentially a better outcome for these patients.

Limitations

The major limitation of this study is its retrospective design and the data source used. Differences in coding between centers that report data to the NIS database can result in either underreporting or overreporting of ARDS during SAH admissions. Moreover, the database does not allow for analyzing the long-term outcomes of these patients after discharge, which could provide more information about the challenges these patients face. While there are variances in diagnostic coding practices across the country, the NIS database has been previously validated as an accurate estimator of national inpatient admission diagnosis for SAH as well as ARDS.^10–14

Another limitation of this study is that ARDS severity and management could not be analyzed. The treatment of pulmonary syndromes associated with brain injury is a balancing act, and must be carefully managed. The hallmark of ARDS management has centered on lung protective ventilation strategies with low tidal volume and low plateau pressures. ¹⁹ This has since evolved to the development of new strategies at using alveolar recruitment with high peak-end expiratory pressures (PEEP), and inversion of inhalation:expiration ratios resulting in permissive hypercapnea. Though these strategies have led to substantial improvement in outcomes for patients with ARDS, they pose specific concerns in patients with intracranial pathology such as SAH. The increased transmittance of intrathoracic pressures to the intracranial cavity when using high PEEP and the cerebrovascular dilitation arising from permissive hypercapnea are both problematic in patients with SAH. A retrospective cohort study of 12 patients with SAH and ARDS found that these lung-protective ventilation practices were not contraindicated in patients with SAH but cautioned that continuous monitoring was required to avoid deleterious effects. ²⁰ Although evaluating the management of ARDS secondary to SAH was outside the scope of this investigation, future work looking at the effectiveness of managing ARDS in patients with SAH can be the foundation of treatment guidelines mitigating poor outcomes for these patients.

Conclusion

Our analysis determined that while the incidence of ARDS in patients admitted with SAH has decreased from 2008 to 2014, it still affects a significant percentage of patients. We were able to determine statistically significant baseline characteristics and predictive factors associated with those who developed ARDS. This information can be used to identify patients who are at risk of developing ARDS for the purpose of risk stratification, resource utilization, and early prevention and treatment.

Abbreviations

(SAH)

subarachnoid hemorrhage

(ARDS)

acute respiratory distress syndrome

(ICU)

intensive care unit

(NIS)

National Inpatient Sample

(CCI)

Charlson Comorbidity Index

(NIS-SSS)

NIS stoke severity score

(HCUP)

Healthcare Cost and Utilization Project

(OTR)

other than routine

(TCM)

Takotsubo cardiomyopathy

(PE)

pulmonary embolism

(DVT)

deep vein thrombosis