Outcomes of Hemorrhagic Stroke Patients with Atrial Fibrillation or Flutter

Sadaf Farasat1, Asa Tsoi2, Kyulim Lee 2, Craig I Coleman 2, Elaine Nguyen3

1Division of Geriatrics Medicine, University of California San Diego, La Jolla, CA, USA.2University of Connecticut School of Pharmacy, Storrs, CT, USA.3Idaho State University College of Pharmacy, Meridian, ID, USA.

Abstract

Introduction

Hemorrhagic stroke is a life-threatening complication, and may be particularly prevalent in patients with atrial fibrillation/flutter (AF/AFL) due to their need for anticoagulation. We sought to estimate in-hospital mortality, length of stay (LOS) and in-patient treatment cost of patients with AF/AFL admitted with hemorrhagic stroke.

Methods

We used the 2008-2011 National Inpatient Sample to identify adult AF/AFL patients with a principle discharge diagnosis of subarachnoid or intracerebral hemorrhage. Endpoints of interest included in-hospital mortality, LOS, and hospital treatment costs (2015 US$).

Results

In-hospital mortality for subarachnoid and intracerebral hemorrhage was 31.3% and 31.8%, respectively. Patients had a median LOS of 5 days and 25% of patients accrued costs in excess of $24,107/stay.

Conclusions

Admissions for hemorrhagic stroke among patients with AF/AFL were associated with substantial in-hospital mortality and treatment costs.

Key Words : Hemorrhagic stroke, Atrial fibrillation, Mortality, Length of stay, Cost.

Correspondence to: Elaine Nguyen, PharmD, MPH Idaho State University College of Pharmacy 1311 E Central Drive Meridian, ID 83642

Introduction

Atrial fibrillation/flutter (AF/AFL) is a highly prevalent and important risk factor for stroke. Oral anticoagulation offers significant protection against AF/AFL-related thromboembolic events, but severe adverse events such as hemorrhagic stroke (i.e., subarachnoid and intracerebral hemorrhage) may complicate its use [1,2]. There is a paucity of recent data describing costs and consequences of such events. Thus, we sought to estimate in-hospital mortality, length of stay (LOS) and hospital treatment costs for hemorrhagic stroke admissions among United States (US) AF/AFL patients.

Material and Methods

This study used the Agency for Healthcare Research and Quality’s (AHRQ’s) National Inpatient Sample (NIS) database for the years 2008-2011. The NIS provides a nationally representative ~20% sample of US hospitals and their admissions [3]. We identified adult AF/AFL patients with an International Classification of Diseases, ninth-edition (ICD-9) code of 427.31 or 427.32 (any position) and a primary diagnosis code of 431 or 430 for subarachnoid or intracerebral hemorrhage, respectively. Patients not admitted through the emergency department or transfers from an outside facility were excluded. Endpoints of interest included in-hospital mortality, LOS and hospital treatment costs (in 2015 US dollars). Since all data were de-identified the study did not require institutional review board oversight.

Baseline patient (age, sex, race, primary payer, median household income for zip code and comorbidity status) and hospital characteristics (size, type and region) were descriptively summarized. Stroke and bleeding risk were calculated using congestive heart failure, hypertension, age, diabetes mellitus, prior stroke (CHADS₂) [4], and anticoagulation and risk factors in AF (ATRIA) [5] scores, respectively. Categorical data were reported as percentages, ordinal data as medians (25%, 75% range) and continuous data as means ± standard deviations (SD). We estimated the percentage (95% confidence interval) of patients with in-hospital mortality and median (25%, 75% range) LOS and costs for subarachnoid and intracerebral hemorrhage combined and separately. Charges were converted to cost using charge-to-cost ratios and inflated to 2015 US dollars using the Consumer Price Index for Medical Care. Descriptive statistics were performed using SPSS version 22 (IBM Corp., Armonk, NY).

Results

Among the 3,034,569 admissions for patients with a diagnosis for AF/AFL in the 2008-2011 NIS database, 7,897 (0.26%) had a primary diagnosis code for hemorrhagic stroke and met our inclusion criteria [Table 1]. Patients with AF/AFL experiencing a hemorrhagic stroke (70% white; 51% female) had a mean ± standard deviation age of 77.5 ± 11.0 years; median (25%, 75% range) CHADS2 score of 2 [2,3] and ATRIA score of 4 [3,4]. Most hemorrhagic stroke admissions (87.1%) were for intracerebral hemorrhage.

Regardless of hemorrhage subtype, nearly one-third of patients died prior to discharge [Table 2]. The incidence of in-hospital mortality was 31.3% for subarachnoid and 31.8% for intracerebral hemorrhage. Patients had a median LOS of 5 days and 25% of patients accrued costs in excess of $24,107/stay.

Table 1. Characteristics of atrial fibrillation/flutter patients admitted with hemorrhagic stroke
Characteristic Total, n (%) Subarachnoid hemorrhage, n (%) Intracerebral hemorrhage, n (%)
N = 7,897 N = 1,022 N = 6,875
Age, years, mean ± SD 77.5 ± 11.0 72.9 ± 13.9 78.2 ± 10.4
Age, years (categories)
18 – 40 37 (0.5) 22 (2.2) 15 (0.2)
41 – 65 1,040 (13.2) 251 (24.6) 789 (11.5)
66 – 75 1,692 (21.4) 228 (22.3) 1,464 (21.3)
> 76 5,128 (64.9) 521 (51.0) 4,607 (67.0)
Male sex 3,869 (49.0) 466 (45.6) 3,403 (49.5)
Race
White 5,500 (69.6) 688 (67.3) 4,812 (70.0)
Black 523 (6.6) 67 (6.6) 456 (6.6)
Hispanic 464 (5.9) 53 (5.2) 411 (6.0)
Asian/Pacific Islander 295 (3.7) 44 (4.3) 251 (3.7)
Other 214 (2.7) 32 (3.1) 182 (2.6)
Not reported 901 (11.4) 138 (13.5) 763 (11.1)
CHADS2 score, median (25%, 75% range) 2 (2, 3) 2 (1, 3) 2 (2, 3)
ATRIA score, median (25%, 75% range) 4 (3, 4) 4 (2, 4) 4 (3, 4)
Number of AHRQ-29 comorbidities
0 249 (3.2) 42 (4.1) 207 (3.0)
1 1,378 (17.4) 189 (18.5) 1,189 (17.3)
2 2,046 (25.9) 245 (24.0) 1,801 (26.2)
3 1,800 (22.8) 218 (21.3) 1,582 (23.0)
4 1,151 (14.6) 133 (13.0) 1,018 (14.8)
≥ 5 1,273 (16.1) 195 (19.1) 1,078 (15.7)
Primary Payer
Medicare 6,350 (80.4) 709 (69.4) 5,641 (82.1)
Medicaid 281 (3.6) 53 (5.2) 228 (3.3)
Private (including HMO) 991 (12.5) 199 (19.5) 792 (11.5)
Self-pay 161 (2.0) 35 (3.4) 126 (1.8)
Other/Not reported 114 (1.4) 26 (2.5) 88 (1.3)
Median household income for ZIP code, $
1 – 38,999 1,711 (21.7) 227 (22.2) 1,484 (21.6)
39,000 – 47,999 1,884 (23.9) 252 (24.7) 1,632 (23.7)
48,000 – 62,999 1,984 (25.1) 279 (27.3) 1,705 (24.8)
> 63,000 2,159 (27.3) 248 (24.3) 1,911 (27.8)
Not reported 159 (2.0) 16 (1.6) 143 (2.1)
Hospital size
Small 619 (7.8) 70 (6.8) 549 (8.0)
Medium 1,667 (21.1) 185 (18.1) 1,482 (21.6)
Large 5,502 (69.7) 741 (72.5) 4,761 (69.3)
Not reported 109 (1.4) 26 (2.5) 83 (1.2)
Hospital type
Rural 559 (7.1) 65 (6.4) 494 (7.2)
Urban non-teaching 3,257 (41.2) 347 (34.0) 2,910 (42.3)
Urban teaching 3,972 (50.3) 584 (57.1) 3,388 (49.3)
Not reported 109 (1.4) 26 (2.5) 83 (1.2)
Hospital region
Northeast 1,689 (21.4) 203 (19.9) 1,486 (21.6)
Midwest 1,468 (18.6) 209 (20.5) 1,259 (18.3)
South 2,991 (37.9) 388 (38.0) 2,603 (37.9)
West 1,749 (22.1) 222 (21.7) 1,527 (22.2)

AHRQ=Agency for Healthcare Research and Quality; ATRIA=anticoagulation and risk factors in atrial fibrillation; CHADS2=congestive heart failure, hypertension, age, diabetes mellitus, prior stroke; HMO=health maintenance organization; SD=standard deviation

Table 2. Outcomes of atrial fibrillation/flutter patients admitted with hemorrhagic stroke
Mortality, % (95%CI) Length of stay, days median (25%, 75% range) Cost, 2015 US$* median (25%, 75% range)
Any hemorrhagic stroke (N=7,897) 31.7 (30.7 – 32.8) 5 (2, 9) $12,036 ($6,815, $24,107)
Subarachnoid hemorrhage (N=1,022) 31.3 (28.5 – 34.2) 6 (2, 14) $17,314 ($7,945, $51,535)
Intracerebral hemorrhage (N=6,875) 31.8 (30.7 – 32.9) 5 (2, 9) $11,677 ($6,715, $22,067)

CI=confidence interval *Charges were converted to cost using charge-to-cost ratios and inflated to 2015 US$ using the Consumer Price Index for Medical Care

Discussion

In our analysis of 2008-2011 NIS data, a small percentage of patients with AF/AFL had a primary diagnosis code for hemorrhagic stroke, with the majority of these patients having an intracerebral hemorrhage. In hemorrhagic stroke patients, in-hospital mortality was about one-third, median LOS and cost of hospitalization were 5 days and ~$12,000, respectively. Our analysis has shown that hemorrhagic stroke can be a fatal and costly complication in patients with AF/AFL, presumably on oral anticoagulation. As such, the risk and outcomes of this complication should be considered when selecting therapy for stroke prevention in atrial fibrillation (SPAF).

Alonso and colleagues [6] performed a similar analysis utilizing Market Scan 2009-2012 data to evaluate in-hospital mortality in AF patients on anticoagulant therapy (i.e., warfarin or dabigatran) with a primary diagnosis of intracranial hemorrhage (ICD-9 codes 430-432.x and 852-853.x). Patients were classified by the bleeding subtypes of subarachnoid/intracranial bleeding not otherwise specified (NOS) (n=410), intracerebral (n=748), and subdural (n=1,233) hemorrhage. In-hospital death occurred in 23% of subarachnoid/intracranial bleeding, 34% of intracerebral, and 15% of subdural hemorrhage patients. While our intracerebral hemorrhage mortality rates are in agreement with Alonso et al. (32% vs. 34%, respectively), it is possible that our difference in subarachnoid hemorrhage mortality is a result of our larger sample size and exclusion of patients with intracranial bleeding NOS (31% vs. 23%, respectively). Our analysis was restricted to hemorrhagic stroke and subdural hemorrhage mortality was not evaluated.

Pharmacologic options for SPAF are often compared in cost-effectiveness models with the validity of such models dependent on their underlying data [7]. In a systematic review of 30 pharmacologic SPAF cost-effectiveness models, Limone and colleagues found that cost and event rate data were often from common sources and utilized data from the 1990s [7]. This concerning finding led Limone et al. to recommend that investigators use the most up-to-date inputs to avoid common flaws in SPAF models [7]. Our data can be used to provide contemporary inputs and to better predict outcomes of hemorrhagic stroke in economic models.

Although the NIS provides a large and nationally representative sample of data, its use has limitations. The NIS does not provide clinical data to calculate stroke severity or prescription use to confirm patients were on oral anticoagulation during the time of hemorrhagic stroke. We chose to avoid use of ICD9-code V58.61 for long-term (current) use of anticoagulants because we do not believe it to be accurate for our purposes. However, it is likely that a majority of patients in our evaluation were utilizing anticoagulant therapy. In an observational registry of newly diagnosed AF patients with at least one additional stroke risk factor, ~70% of patients receive anticoagulation and >80% of patients are on antithrombotic treatment (i.e., anticoagulants and/or antiplatelets) [8]. While the use of ICD-9 billing codes can cause misclassification and other biases if inaccurate, we feel confident in the hemorrhagic stroke codes used in our study as they have previously shown high specificity (>0.99) and positive predictive values (0.77) for event identification [9].

Conflicts of interest

All authors (SF, KL, AT, CIC, EN) have no conflicts germane to this manuscript to report.

Conclusions

We found that admissions for hemorrhagic stroke among patients with AF/AFL were associated with substantial in-hospital mortality and treatment costs. Our results provide data that can be used to inform economic models and decision makers.

References

  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]
  • [ PubMed ]