Secondary Logo

Journal Logo

Clinical Cardiovascular

Ventricular Assist Device Patients Have Different Clinical Outcomes and Altered Patterns of Bleeding with Intracranial Hemorrhage

Ahmed, Mustafa M.*; Rahman, Maryam; Neal, Dan; Aranda, Juan M. jr*; Klodell, Charles T.

Author Information
doi: 10.1097/MAT.0000000000000744
  • Free


The use of left ventricular assist devices (LVAD) for advanced heart failure continues to proliferate rapidly. Presently more than 140 centers implant over 2,500 patients annually with these life-saving devices.1 Furthermore, the use of LVADs as destination therapy has risen dramatically in each subsequent era of device support from 2006 to 2013,1 leading to a rising population of long-term mechanical support. While this has resulted in a growing awareness of device-related complications, relatively little is known about intracranial hemorrhage (ICH) in this patient population with only single-center data available in the published literature.2 The purpose of this manuscript is to describe patterns of ICH in patients with and without LVADs, as well as their clinical outcomes.

Materials and Methods

We queried the National Inpatient Sample (NIS) database, part of the Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project, for all hospital admissions between 2002 and 2012 with a diagnosis of ICH (ICD-9 codes 431 [intracerebral]; 430, 852.00-.06, 852.09-.16, 852.19 [subarachnoid]; 852.20-.26, 852.29-.36, 852.39, 432.1 [subdural]; 852.40-.46, 852.49-.56, 852.59, 432.0 [extradural]; 772.10-.14 [intraventricular]; 432, 432.9 853.00-.06, 853.09-.16, 853.19 [unspecified]). We classified these patients into three groups: 1) those who had hemorrhage but neither heart failure nor LVAD, 2) those who had hemorrhage and also had a diagnosis of heart failure (ICD-9 codes 402.01, 402.11, 402.91, 428.0, 428.1, 428.20, 428.21-.23, 428.40-.43, 428.9), 3) those who had hemorrhage and also an LVAD (ICD-9 diagnosis codes V43.21 or V43.22, or procedure code 37.66).

We also used ICD-9 diagnosis and procedure codes to identify patients with diabetes mellitus (diagnosis codes 250.00-.03), hypertension (401.0, 401.1, 401.9, 405.01, 405.09, 405.11, 405.19, 405.91, 405.99), perivascular disease (443.0-.2, 443.8, 443.9), patients who received anticoagulants (diagnosis code V58.61), as well as patients who had cranial surgery (intracranial pressure monitoring [procedure code 01.10], other craniotomy [01.24], craniotomy and craniectomy [01.21-.27], insertion of external ventricular drain [02.21], insertion of catheter via burr hole [01.28]). We used the NIS categories of discharge disposition to classify patients as experiencing “good outcome” (discharged home, transferred to short-term hospital, discharged home under home health care, left against medical advice, discharged to home IV provider, discharged to inpatient rehabilitation facility, discharged to another institution for outpatient care), or “poor outcome” (discharged to skilled nursing facility, discharged to intermediate care facility, died in hospital, discharged to home hospice, discharged to hospice facility, discharged to hospital-based Medicare-approved swing bed, discharged to Medicare certified long-term care hospital, discharged to Medicaid-certified nursing facility). This is a previously described outcome and has been utilized for neurosurgery patients in the literature.3,4

The SAS statistical software package (Cary, NC; V9.4) was used to calculate all descriptive statistics and conduct all statistical tests. We made pairwise comparisons across the three groups (ICH with heart failure, ICH with LVAD, or ICH without heart failure or LVAD) on patient characteristics, comorbidities, hemorrhage and surgery type, hospital factors, and outcomes. We adjusted hospital charges by 3% per year to account for inflation. We used Fisher’s exact test to compare the groups on categorical variables and Mann–Whitney tests to compare them on continuous variables. We used the same methods to compare year 2002–2006 admissions with 2007–2012 admissions within the LVAD group, and to assess the possible effects of anticoagulants within each of the three diagnosis groups.


The baseline characteristics of the three groups were dissimilar (Table 1). Individuals with LVADs (group 3) were noted to be younger, with a greater male predominance, and more racially diverse than their counterparts. Varying rates of hypertension, diabetes mellitus, and peripheral vascular disease were also noted in the three groups, as well as total burden of comorbid conditions, as assessed by the Elixhauser comorbidity index. This index was developed as method of determining degree of comorbid illness in large populations based on ICD-9 diagnosis codes, specifically for use in administrative and other large datasets. It is comprised of 31 different categories and has been utilized to predict healthcare resource utilization and hospital mortality.5,6 By this metric, those individuals who suffered an ICH with the diagnosis of HF but no LVAD (group 2) were noted to be the most ill. Varying rates of anticoagulant use were also noted, with the highest with group 3. The breakdown of individual comorbid conditions per group is in Table 1. The total number of VAD recipients in our dataset was 5,747, of which 118 suffered ICH. This defines group 3 and provides a total incidence of ICH in this dataset as 2.05% (118 of 5,747).

Table 1.:
Baseline Demographics

Hospital factors were also collected and analyzed. The geographic distributions of each group were similar; however, patients in group 3 were more likely to receive care in large, urban teaching hospitals.

A great deal of variability was noted in mortality and other clinical outcomes and quality metrics (Table 2). Inpatient mortality was high across all three groups; however, group 3 was noted to have the worst survival with 39% of patients failing to survive to discharge, compared with 20% in group 2, and 13% in group 1 (p < 0.0001). Despite this apparent survival disadvantage for those individuals with LVADs, their dichotomized discharge disposition was statistically no different than those with HF and no LVAD, with a numerical trend toward being improved, 54.8% vs. 58.7% being discharged with a “poor disposition” (p = NS). This suggests that while patients with LVADs who suffer ICH have a worse inpatient mortality, those that do survive have no worse outcomes than patients with ICH and no LVAD. Length of stay and total charges were highest in group 3. The presence of an LVAD was not identified as a barrier to neurosurgical intervention, and patients in group 3 had the highest rates of intervention (5.7% vs. 4.0% and 3.2%, respectively; p = NS) when examining all interventions in aggregate (intracranial pressure monitoring [procedure code 01.10], other craniotomy [01.24], craniotomy and craniectomy [01.21-.27], insertion of external ventricular drain [02.21], insertion of catheter via burr hole [01.28]).

Table 2.:
Clinical Outcomes in All Groups

When examining outcomes by anticoagulation status (Table 3), we found that the use of chronic anticoagulants portended a worse outcome in group 1. This relationship, however, was not consistent in group 2, as the use of chronic anticoagulants conferred a worse survival, but better discharge disposition in group 2, with no statistically significant relationship in group 3.

Table 3.:
Clinical Outcomes Based on Anticoagulation Status

Patients were noted to have significantly different patterns of ICH (Figure 1). Those with LVADs were much more likely to have subarachnoid and intracerebral hemorrhages, while also being much less likely to have subdural hemorrhage.

Figure 1.:
Subtypes of intracranial hemorrhage.

Finally, a comparison across LVAD eras (2002–2006 vs. 2007–2012) yielded no statistically significant differences in patient and hospital factors, neurosurgical intervention rates, or clinical outcomes (Table 4).

Table 4.:
Comparison Across LVAD Eras


To our knowledge, this is the largest analysis of patterns of ICH in LVAD patients and their clinical outcomes. Previously, Wilson et al2 described their experience in 36 LVAD patients who suffered an ICH at their institution. The authors of this study found a greater a proportion of nontraumatic ICH, with an overall incidence of 11% for any ICH, and similar mortality of 42%. Our study confirms this pattern of ICH, and furthermore defines this pattern to be distinct from those individuals without an LVAD. This suggests that there may be an alternative pathophysiology of bleeding in the LVAD population. Subdural and epidural bleeding is largely related to trauma, while intracerebral bleeding is more related to abnormalities of cerebral vasculature, such as arteriovenous malformations and aneurysms, as well as septic arteritis.7–10

Our assertion is furthered by a recent analysis demonstrating that patients with LVAD who had persistent bloodstream infections were more likely to have a cerebrovascular accident (CVA) than those with transient infections and those without any bacteremia. Neither mean arterial blood pressure nor international normalized ratio (INR) were associated with CVA on multivariate analysis.11

The incidence of nontraumatic intracerebral hemorrhage is 10–30 per 100,00012,13 and has not been shown to be directly attributable to the taking of oral anticoagulants in other studies.11,14 Therefore, the higher overall incidence of ICH in LVAD patients, 2.05%, in this analysis cannot be explained only by the use of anticoagulation. While prior single-center data have demonstrated a higher incidence of ICH,2 the vast majority of the literature reports the incidence of hemorrhagic strokes. While this is often thought of as synonymous with ICH, they in fact represent distinct clinical entities. The distinction may explain why the rate of ICH noted in our study is incongruent with other reports of neurologic complications of mechanical circulatory support. Despite this difference in semantics and definitions, it should be noted that the incidence of hemorrhagic stroke in the recently published prevention of heartmate II pump thrombosis through clinical management (PREVENT) study was 2.7% and that in the control arm of the the multicenter study of maglev technology in patients undergoing mechanical circulatory support therapy with heartmate 3 (MOMENTUM 3) trial was 5.8%,15,16 while that in the control arm of the a prospective, randomized, controlled, un-blinded, multi-center clinical trial to evaluate the heartware ventricular assist system for destination therapy of advanced heart failure (ENDURANCE) study was 4.0%.17 Furthermore, data from the interagency registry for mechanically assisted circulatory support (INTERMACS) registry places the incidence for “neurologic dysfunction,” which should be inclusive of hemorrhagic stroke and ICH at an even lower rate. This would suggest that the true incidence of these complications is poorly defined and requires more rigorous study with universal definitions.

While a great deal of interest has been devoted to the loss of pulsatile flow in the gastrointestinal system and how this affects vasculature, far less is known about the effects of an LVAD on the cerebrovascular system and how it may contribute to bleeding. It is known, however, that endothelial cell dysfunction is a risk factor for ICH, an abnormality seen in those with HF.18,19 Additionally, LVAD recipients have been demonstrated to have altered activation of endothelial and coagulation systems in a fashion opposite to the effects of the pulsatile augmentation of an intra-aortic balloon pump.20,21 Furthermore, models of variable levels of shear stress have established a link with endothelial cell function.22 These data may provide the opportunity for future research to more keenly identify risk factors for ICH in LVAD patients.

As expected, in our study, patients with LVADs who suffered ICH had significantly longer length of stay and associated hospital charges. What we were unable to determine from these data, however, is what portion of length of stay was spent in the intensive care unit, and if this was a neurointensive care or cardiac intensive care unit. Due to the specialized nature of LVAD care, many institutions restrict housing of these patients to certain wards, staffed by certified nurses and physicians. This may prolong intensive care unit stays or impede care. A closer examination of LVAD patient flow within and across institutions may allow for streamlined care and decreased intensive care unit stays and resultant hospital charges. Additionally, the need for short-term rehabilitation placement may have also led to a lengthier stay. The limited availability of rehabilitation centers with LVAD competency is a potential target for improvement in this regard. Left ventricular assist device patients have been demonstrated to have improvements in functional independence measure scores during inpatient rehabilitation, with improvement in both functional and cognitive measures23 despite their medical complexity.

While identifying targets for improvement and areas for the potential creation of LVAD-related health care disparities, it is important to note the presence of an LVAD does not represent a barrier to neurosurgical intervention. In our study, patients with LVADs demonstrated a nonsignificant but higher numerical rate of neurosurgical intervention. The development and implementation of rapid acting, multidisciplinary teams to care for this patient population, as outlined by Willey et al,24 should allow for better outcomes.

Although patients with LVADs who suffer an ICH have a worse mortality, it should be noted that nearly two-thirds of such patients survive their event. That the majority of such individuals can be managed successfully speaks to the growth in expertise and infrastructure required to care for LVAD patients. Additionally, nearly half of LVAD patients with ICH have a meaningful survival, with a discharge to home or to a short-term care facility. This compares favorably with those individuals with HF alone and an ICH.

The major limitations of this study relate to its retrospective and ICD-9 code driven analysis. While it is unclear to what extent coding variances and documentation errors may have affected our results, the low level of patients identified to be on anticoagulation is almost certainly due to the underutilization of the associated ICD-9 code. This may explain the variable and uncertain anticoagulant use relationship seen. However, this coding discrepancy is likely to have affected the groups equally. Additionally, we were unable to determine if a primary LVAD complication, such as pump thrombosis and its subsequent treatment, led to the ICH, nor do we have any data regarding LVAD parameters at the time of or leading toward the neurologic event. The timing of the ICH in relation to the date of LVAD implantation is also not ascertainable in this analysis. Type of LVAD support also cannot be determined in this analysis; however, the ICD-9 code utilized excludes percutaneous devices and extracorporeal life support. Additionally, we are unable to account for over-anticoagulation resulting in ICH as INR data were not available, as well as antiplatelet therapy use and the variances in this regard across implanting centers. Our inability to account for blood pressure at the time of ICH development is also a significant limitation given the importance of blood pressure control and risk of ICH.25,26 This risk may be more pronounced in the LVAD population.27

The lack of patient level data further limits the ability to define outcomes metrics more precisely, specifically as it relates to the degree of neurologic and cognitive impairment at time of discharge and later follow-up. Furthermore, as individual patients as the NIS database is linked to hospital encounters, individual patients cannot be followed over time, precluding analysis of readmission or the need for future procedures beyond the index hospitalization. An additional area of limitation comes from the nature of our comparison groups, their disparate sizes, and inherent heterogeneity.


Patients with an LVAD have a distinct pattern of ICH, with significantly more intracerebral bleeding, suggestive of an alternative and vascular pathophysiology of bleeding in these individuals. While LVAD patients who suffer an ICH have worse inpatient mortality rates than other groups, the majority of these patients survive, with nearly half achieving a meaningful discharge. Although the presence of an LVAD is not a barrier to neurosurgical intervention, more attention to LVAD patient flow and partnerships with rehabilitation may improve outcomes as it pertains to length of stay and hospital charges. Additional study is needed to better define the risk for development of an ICH and optimal management of ICH in this population.


1. Kirklin JK, Naftel DC, Pagani FD, et al.Sixth INTERMACS annual report: A 10,000-patient database. J Heart Lung Transplant 2014.33: 555564,
2. Wilson TJ, Stetler WR Jr, Al-Holou WN, Sullivan SE, Fletcher JJManagement of intracranial hemorrhage in patients with left ventricular assist devices. J Neurosurg 2013.118: 10631068,
3. Fargen KM, Neal D, Blackburn SL, Hoh BL, Rahman MHealth disparities and stroke: The influence of insurance status on the prevalence of patient safety indicators and hospital-acquired conditions. J Neurosurg 2015.122: 870875,
4. Lawson MF, Neal DW, Mocco J, Hoh BLRationale for treating unruptured intracranial aneurysms: Actuarial analysis of natural history risk versus treatment risk for coiling or clipping based on 14,050 patients in the Nationwide Inpatient Sample database. World Neurosurg 2013.79: 472478,
5. Garland A, Fransoo R, Olafson K, et alThe Epidemiology and Outcomes of Critical Illness in Manitoba. 2012.Winnipeg, MB: Manitoba Centre for Health Policy,
6. van Walraven C, Austin PC, Jennings A, Quan H, Forster AJA modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care 2009.47: 626633,
7. Regelsberger J, Elsayed A, Matschke J, et al.Diagnostic and therapeutic considerations for “mycotic” cerebral aneurysms: 2 case reports and review of the literature. Cent Eur Neurosurg 2011.72: 138143,
8. Zacko JC, Harris L, Bullock MRWinn HRSurgical management of traumatic brain injury. in Youmans Neurological Surgery, 2011.6th ed. Philadelphia, PA: Elsevier Saunders,
9. Farhat H, Kretschmer T, Morcos JJWinn HRNonlesional spontaneous intracerebral hemorrhage. in Youmans Neurological Surgery, 2011.6th ed. Philadelphia, PA: Elsevier Saunders,
10. Fischbein NJ, Wijman CANontraumatic intracranial hemorrhage. Neuroimaging Clin N Am 2010.20: 469492,
11. Trachtenberg BH, Cordero-Reyes AM, Aldeiri M, et al.Persistent blood stream infection in patients supported with a continuous-flow left ventricular assist device is associated with an increased risk of cerebrovascular accidents. J Card Fail 2015.21: 119125,
12. Labovitz DL, Halim A, Boden-Albala B, Hauser WA, Sacco RLThe incidence of deep and lobar intracerebral hemorrhage in Whites, Blacks, and Hispanics. Neurology 2005.65: 518522,
13. Qureshi AI, Tuhrim S, Broderick JP, Batjer HH, Hondo H, Hanley DFSpontaneous intracerebral hemorrhage. N Engl J Med 2001.344: 14501460,
14. Hart RGWhat causes intracerebral hemorrhage during warfarin therapy? Neurology 2000.55: 907908,
15. Maltais S, Kilic A, Nathan S, et al.PREVENtion of HeartMate II pump thrombosis through clinical management: The PREVENT multi-center study. J Heart Lung Transplant. 2017.36: 112,
16. Mehra MR, Naka Y, Uriel N, et al.A fully magnetically levitated circulatory pump for advanced heart failure. N Engl J Med. 2017.376: 440450,
17. Rogers JG, Pagani FD, Tatooles AJ, et al.Intrapericardial left ventricular assist device for advanced heart failure. N Engl J Med 2017.376: 451460,
18. Ganda A, Onat D, Demmer RT, et al.Venous congestion and endothelial cell activation in acute decompensated heart failure. Curr Heart Fail Rep 2010.7: 6674,
19. Keep RF, Hua Y, Xi GIntracerebral haemorrhage: Mechanisms of injury and therapeutic targets. Lancet Neurol 2012.11: 720731,
20. John R, Panch S, Hrabe J, et al.Activation of endothelial and coagulation systems in left ventricular assist device recipients. Ann Thorac Surg 2009.88: 11711179,
21. Serraino GF, Marsico R, Musolino G, et al.Pulsatile cardiopulmonary bypass with intra-aortic balloon pump improves organ function and reduces endothelial activation. Circ J 2012.76: 11211129,
22. Uzarski JS, Scott EW, McFetridge PSAdaptation of endothelial cells to physiologically-modeled, variable shear stress. PLoS One 2013.8: e57004,
23. Alsara O, Reeves RK, Pyfferoen MD, et al.Inpatient rehabilitation outcomes for patients receiving left ventricular assist device. Am J Phys Med Rehabil 2014.93: 860868,
24. Willey JZ, Demmer RT, Takayama H, Colombo PC, Lazar RMCerebrovascular disease in the era of left ventricular assist devices with continuous flow: Risk factors, diagnosis, and treatment. J Heart Lung Transplant 2014.33: 878887,
25. Biffi A, Anderson CD, Battey TW, et al.Association between blood pressure control and risk of recurrent intracerebral hemorrhage. JAMA 2015.314: 904912,
26. Koivunen RJ, Satopaa J, Meretoja A, et al.Incidence, risk factors, etiology, severity and short-term outcome of non-traumatic intracerebral hemorrhage in young adults. Eur J Neurol. 2015.22: 123132,
27. Saeed O, Jermyn R, Kargoli F, et al.Blood pressure and adverse events during continuous flow left ventricular assist device support. Circ Heart Fail 2015.8: 551556,

complications; heart failure; intracranial hemorrhage; ventricular assist device

Copyright © 2018 by the ASAIO