Doi:10.1016/j.amjmed.2005.02.022

The American Journal of Medicine (2005) 118, 612– 617 CLINICAL RESEARCH STUDY
Incidence of intracranial hemorrhage in patients with
atrial fibrillation who are prone to fall
Brian F. Gage, MD, MSc,a Elena Birman–Deych, MS,a Roger Kerzner, MD,b
Martha J. Radford, MD,c David S. Nilasena, MD, MSPH, MS,d Michael W. Rich, MDb
aDivision of General Medical Sciences, Washington University School of Medicine, andbDivision of Cardiology, Washington University School of Medicine, St. Louis, Missouri;cCenter for Outcomes Research and Evaluation at Yale New Haven Health System, and Division of Cardiology, YaleSchool of Medicine, New Haven, Connecticut; and thedCenters for Medicare and Medicaid Services, Dallas, Texas. KEYWORDS:
ABSTRACT
PURPOSE: Patients at high risk for falls are presumed to be at increased risk for intracranial
hemorrhage, and high risk for falls is cited as a contraindication to antithrombotic therapy. Data substantiating this concern are lacking.
METHODS: Quality improvement organizations identified 1245 Medicare beneficiaries who were
documented in the medical record to be at high risk of falls and 18 261 other patients with atrial fibrillation. The patients were elderly (mean 80 years), and 48% were prescribed warfarin at hospital discharge. The primary endpoint was subsequent hospitalization for an intracranial hemorrhage, basedon ICD-9 codes.
RESULTS: Rates (95% confidence interval [CI]) of intracranial hemorrhage per 100 patient-years
were 2.8 (1.9 – 4.1) in patients at high risk for falls and 1.1 (1.0 –1.3) in other patients. Rates (95% CI)of traumatic intracranial hemorrhage were 2.0 (1.3–3.1) in patients at high risk for falls and 0.34(0.27– 0.45) in other patients. Hazard ratios (95% CI) of other independent risk factors for intracranialhemorrhage were 1.4 (1.0 –3.1) for neuropsychiatric disease, 2.1 (1.6 –2.7) for prior stroke, and 1.9(1.4 –2.4) for prior major bleeding. Warfarin prescription was associated with intracranial hemorrhagemortality but not with intracranial hemorrhage occurrence. Ischemic stroke rates per 100 patient-yearswere 13.7 in patients at high risk for falls and 6.9 in other patients. Warfarin prescription in patientsprone to fall who had atrial fibrillation and multiple additional stroke risk factors appeared to protectagainst a composite endpoint of stroke, intracranial hemorrhage, myocardial infarction, and death.
CONCLUSION: Patients at high risk for falls with atrial fibrillation are at substantially increased risk
of intracranial hemorrhage, especially traumatic intracranial hemorrhage. However, because of theirhigh stroke rate, they appear to benefit from anticoagulant therapy if they have multiple stroke riskfactors.
2005 Elsevier Inc. All rights reserved.
0002-9343/$ -see front matter 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.amjmed.2005.02.022 Risk of Intracranial Hemorrhage in Patients Prone to Fall In atrial fibrillation trials of carefully selected partici- a random sample stratified by state from all Medicare ben- pants, antithrombotic therapy prevented strokes with an eficiaries who were hospitalized with an International Clas- acceptable rate of intracranial hemorrhages: 0.3 intracranial sification of Diseases, 9th Revision, Clinical Modification hemorrhage per 100 patient-years with aspirin and 0.4 – 0.5 (ICD-9) code for atrial fibrillation (427.31) in any diagnos- intracranial hemorrhage per 100 patient-years with warfa- tic position and who were discharged between April 1, 1998 However, elderly patients at high risk for falls were and March 31, 1999. We obtained inpatient (part A) and excluded from clinical and no longitudinal data outpatient (part B) Medicare records from 1995 through quantify the risk of intracranial hemorrhage and stroke in The structured abstraction of data from medical charts Perception of traumatic intracranial hemorrhage risk in- was performed by the Clinical Data Abstraction Centers, fluences selection of antithrombotic therapy. which confirmed the presence of atrial fibrillation during the and medical record demonstrate that physicians index admission. The remaining records were systemati- avoid antithrombotic therapy in elderly patients with atrial cally reviewed for risk of falls, stroke risk factors, and fibrillation who seem likely to fall and sustain an intracra- discharge medications. Chart abstractors were unaware of nial hemorrhage. Epidemiological studies have found that antithrombotic therapy can double the risk of intracranialespecially fatal Thus, accurate Outcomes assessment
knowledge of the rate of intracranial hemorrhage in patientsat high risk for falls with atrial fibrillation would help Intracranial hemorrhage and ischemic strokes in the fol- determine the optimal antithrombotic therapy.
low-up period were identified by recently validated ICD-9 Our primary goal was to quantify the incidence of intra- codes in Medicare Part A data. We identified ischemic cranial hemorrhage in Medicare beneficiaries with atrial stroke from ICD-9 codes 433.x1, 434.x1, 436, 437.1, and fibrillation who were at high risk of falls. Our secondary 437.9 (“x” represents any digit) in any position. As com- goals were to identify independent risk factors for intracra- pared to structured chart abstraction, these codes have a nial hemorrhage and to quantify the benefits of warfarin positive predictive value of We identified nontrau- therapy, if any, in patients at high risk for falls.
matic intracranial hemorrhages from codes 430x– 432x andtraumatic intracranial hemorrhage from codes 800.2x,800.3x, 800.7x, 800.8x, 801.2x, 801.3x, 801.7x, 801.8x, 803.2x, 803.3x, 803.7x, 803.7x, 804.2x, 804.3x, 804.7x,804.8x, 852, and 853. The positive predictive value of these The primary endpoint was hospitalization for an intra- ICD-9 codes for intracranial hemorrhage is cranial hemorrhage after the index hospital admission. The In a subgroup analysis we identified myocardial infarc- study was approved by the Washington University human tion from ICD-9 code 410.x and noncerebral hemorrhage based on ICD-9 codes validated by White et Weassessed predictive validity of the high-fall-risk designation Formation of the second national registry of
using Medicare Part A and B claims by assessing for falls atrial fibrillation
(E880.0-E886.9, E888) and fractures (800-829).
In the primary analysis we censored patients at the time The National Registry of Atrial Fibrillation II dataset of their last hospitalization if they died outside of the hos- was created from 23 657 anonymous patient records gath- pital because neither Medicare Part A or B claims nor the ered by Quality Improvement Organizations for the Na- Medicare Denominator File provides cause of death. Like- tional Stroke Project. The project was managed by the Iowa wise, those who died after the baseline hospitalization (be- Foundation for Medical Care. The dataset included both fore another hospitalization) were excluded (n ϭ 1824). In Medicare records and chart-abstracted data from 3586 hos- a secondary analysis, we included out-of-hospital deaths as pitals in all 50 US states. Medical records were selected as part of a composite endpoint that also included stroke,intracranial hemorrhage, and myocardial infarction. Forbeneficiaries who experienced multiple adverse events, we Grant Support: Supported by the American Heart Association
(0270099N). Disclaimer: The conclusions presented are solely those of the
excluded events and days of follow-up that occurred after authors and do not represent those of the Quality Improvement Organiza- tions, American Heart Association, or CMS. The content of this publicationdoes not necessarily reflect the views or policies of the Department ofHealth and Human Services, nor does mention of commercial products Cohort formation
imply endorsement of them by the U. S. Government. The authors assumefull responsibility for the accuracy and completeness of the ideas presented.
The only acceptable source of information for risk of Requests for reprints should be sent to: Brian F. Gage, MD, MSc, falls was physician documentation in the medical record.
Division of General Medical Sciences, Washington University School of The terms “frequent falls,” “history of falls,” “multiple Medicine Campus Box 8005, 660 S. Euclid Ave., St. Louis, MO 63110.
falls,” or “tendency for falls” were considered synonymous The American Journal of Medicine, Vol 118, No 6, June 2005 with high risk for falls; a single fall was insufficient docu- Demographic and clinical factors of study cohorts mentation for this designation. We also identified a trial-likecohort comprising patients who lacked the following co- morbid conditions: high risk for falls, age 80 or older, chronic renal disease, uncontrolled hypertension, malig- nancy, alcoholism, history of bleeding, neuropsychiatric impairment, prior ischemic stroke or transient ischemic at- tack (TIA), anemia, bleeding disorder, and combination therapy with warfarin and aspirin at discharge.
Statistical analyses
We used time-to-event analyses to test our hypotheses that the risks of intracranial hemorrhage and of stroke were greater in patients at high risk for falls. We used backward elimination to develop parsimonious Cox models. We ver- ified the proportionality assumption graphically and by time-dependent covariates. In the intracranial hemorrhage model, we tested for effects of history of ischemic stroke or TIA, age, sex, race, alcoholism, history of bleeding, a bleed- ing disorder (eg, hemophilia or leukemia), nursing home residence, neuropsychiatric impairment (schizophrenia, de- mentia, or Parkinson’s disease), and antithrombotic therapy.
In the ischemic stroke Cox model, we tested for effects of sex, nursing home residence, antithrombotic therapy, and stroke factors. We quantified the risk of stroke using a clinical prediction (CHADS ) that assigns 1 point for the presence of Congestive heart failure, Hypertension, Age Ͼ 75, or Diabetes mellitus, and 2 points for a prior Stroke or In a subgroup analysis, we initially stratified the analysis into CHADS scores of 0 –1, 2, or 3– 6, because the benefit of warfarin therapy is greater in patients with atrialfibrillation at greater risk of Because the apparentbenefit of warfarin was similar in patients with 2 points and were coded after the baseline hospitalization. Compared to with 3– 6 points, we combined these two cohorts.
other patients, patients at high risk for falls were 2.0 (95% Statistical tests were two-tailed. We performed statistical CI: 1.6 –2.4) times more likely to fall. The rates of fractures analyses in SAS version 9.0 (SAS Institute Inc; Cary, NC).
per 100 patient-years were 27.9 in patients at high risk forfalls and 12.0 in other patients.
Intracranial hemorrhage
Cohort descriptions
The rates of intracranial hemorrhage per 100 patient-years were 2.8 (95% CI: 1.9 – 4.1) in patients at high Subjects at high risk for falls were older and had more risk for falls and 1.1 (95% CI: 1.0 –1.3) in other patients (P comorbidities than other patients. In addition, they were Ͻ 0.0001, log-rank test). In the subset of the trial-like significantly less likely to receive warfarin or aspirin ther- patients, the rate of intracranial hemorrhage was 0.53 (95% CI: 0.3– 0.8). The rates of traumatic intracranial hemorrhage Trial-like patients (n ϭ 3236) were younger (mean age, per 100 patient-years were 2.0 (95% CI: 1.3–3.1) in patients 73 years) and healthier (mean number of bleeding risk at high risk for falls and 0.34 (95% CI: 0.27– 0.45) in other factors, 0.6). Most of them were prescribed antithrombotic patients (P Ͻ 0.0001, log-rank test).
therapy (53.7% warfarin; 23.1% aspirin).
In the multivariate Cox model patients at high risk for falls were 1.9 (95% CI 1.3–2.9) times more likely to Data integrity and validation
have any intracranial hemorrhage and 4.1 (95% CI 2.4 –7.1)times more likely to have a traumatic intracranial hemor- We validated the fall-risk designation by examining rhage than other patients. Prior stroke (HR 2.2), prior major ICD-9 codes for falling and for nonpathologic fractures that bleeding (HR 1.8), and neuropsychiatric impairment (HR Risk of Intracranial Hemorrhage in Patients Prone to Fall 1.4) also were independently associated with any intracra- Multivariate Cox regression showing hazard ratios nial hemorrhage Prescription of warfarin or as- (HR) of independent predictors of intracranial hemorrhage pirin at baseline did not significantly affect risk of intracra-nial hemorrhage: the HR for warfarin was 1.0 (95% CI 0.8 –1.4) and the HR for aspirin was 1.1 (95% CI 0.8 –1.4).
Alcohol abuse, non-Caucasian race, bleeding disorder, and renal disease were too infrequent in the dataset to quantify After an intracranial hemorrhage, 30-day mortality was 42% in patients at high risk of falls and 48.2% in otherpatients (P Ͼ 0.2). The intracranial hemorrhage 30-day Risks and benefits of warfarin therapy in patients
mortality was 51.8% in patients who had been prescribed at high risk for falls
warfarin, and 33.6% in patients who had not been pre-scribed warfarin after the baseline hospitalization (P ϭ To determine the potential benefit of prescribing warfa- 0.007). In a stepwise logistic regression model, the only rin in patients at high risk for falls, we quantified the independent predictors of 30-day mortality post intracranial association between warfarin use and the composite out- hemorrhage were prior prescription of warfarin (odds ratio come of hospitalization for stroke, any hemorrhage (includ- 2.5; 95% CI 1.4 – 4.5, P ϭ 0.002) and nursing home resi- ing intracranial hemorrhage), myocardial infarction, or out- dency (odds ratio 3.3; 95% CI 1.6 – 6.8, P ϭ 0.0009).
of-hospital death (except in patients with terminal disease).
In the Cox model that controlled for bleeding risk factors,aspirin prescription, nursing home residency, and sex, war- Ischemic stroke
farin was significantly protective in 1086 patients with 2 ormore CHADS points (HR 0.75), but not protective in 159 Stroke rates (95% CI) per 100 patient-years were 13.7 patients with 0 or 1 points in (HR 0.98) The (11.6 –16.3) in patients at high risk for falls and 6.9 (6.5– P-value for interaction between warfarin and stroke risk was 7.3) in other patients. In the subset of the trial-like patients, the stroke rate was 2.8 (95% CI: 2.3–3.4). Compared toother patients, patients at high risk for falls had a 1.3-fold(95% CI: 1.1–1.6) increased risk of stroke (P ϭ 0.002).
Discussion
Each 1-point increase in the CHADS score increased the risk of stroke by a factor of 1.42 (95% CI: 1.37–1.47, P Ͻ Despite their low use of warfarin (33.5%), patients at 0.0001). Hazard ratios (HR) and 95% CI of other indepen- high risk for falls suffered 2.8 intracranial hemorrhages per dent stroke risk factors were neuropsychiatric impairment 100 patient-years, more than twice the 1.1 intracranial hem- 1.22 (1.06 –1.40, P ϭ 0.05) and nursing home residence orrhage rate of other participants and more than 5 times the 1.45 (1.29 –1.64, P Ͻ0.0001); warfarin prescription at hos- 0.5 rate of trial-like participants. The increased risk of pital discharge had a modestly protective effect 0.78 (0.70 – intracranial hemorrhage in patients at high risk for falls was 0.86, P Ͻ0.0001). After stroke, 30-day mortality was 34.4% due to their increased incidence of traumatic intracranial among high-fall-risk subjects, 27.8% in other patients, and hemorrhage, which was increased four-fold compared to other patients, even after adjusting for the covariates. The30-day mortality after an intracranial hemorrhage was sig- Hazard ratio of warfarin for composite outcome— Rates of intracranial hemorrhage, stratified by out-of-hospital death or hospitalization for stroke, myocardial infarction, or hemorrhage—in 1245 patients athigh risk for falls Intracranial hemorrhage rate (95% CI) per 100 patient-years CHADS stroke score was calculated by adding 1 point for each of the following conditions: congestive heart failure, hypertension, ageϾ 75 years, or diabetes and 2 points for a prior stroke or transient *P Ͻ 0.0001 High-fall vs other patients.
ischemic attack. CI indicates confidence interval.
†P ϭ 0.0005 High-fall vs other patients.
The American Journal of Medicine, Vol 118, No 6, June 2005 nificantly greater in patients who had been prescribed war- what antithrombotic therapy was prescribed at hospital dis- farin after the baseline hospitalization (51.8%) than in pa- charge but did not capture subsequent initiation or discon- tients who had not been prescribed warfarin (33.6%). These tinuation of therapy. Thus, the observed lack of association observations highlight the substantial risk and mortality of between warfarin and intracranial hemorrhage in this study intracranial hemorrhage in populations who are older and does not refute such an A third limitation is frailer than carefully selected trial participants.
that we had to exclude patients who died after the baseline Despite the significant association between intracranial hospitalization and without another hospitalization because hemorrhage and fall risk, the findings support the use of they had no follow-up data. A minor limitation was that we anticoagulants in patients at high risk for falls who are at could not evaluate risk factors for intracranial hemorrhage moderate to high risk of stroke. Prescribing warfarin in that were unavailable (eg, leukoaraiosis on brain imaging or patients at high risk for falls with 2 or more CHADS points tobacco use). Likewise, we used physician documentation was associated was a 25% relative risk reduction (HR 0.75) to ascertain risk of falls and were unable to evaluate the risk in the composite outcome When prescribing of specific impairments in strength, balance, vision, or or- warfarin to these patients, providers could instruct them to take precautions to limit their risk of falling: wear stable These limitations are offset by several strengths. First, the cohort design allowed us to calculate incidence rates and aids,and discontinue unnecessary medications.
to avoid recall bias. Second, we were able to use structured In contrast to the potential benefits of warfarin in patients medical record abstraction to adjust for potential confound- with greater CHADS scores, prescribing warfarin in pa- ing factors. Third, the large size and national sample of the tients at high risk for falls with 0 or 1 CHADS points was dataset provide great generalizability. Fourth, the designa- associated with a nonsignificant reduction in the composite tion “high-fall-risk” predicted subsequent fractures and outcome Because the 95% confidence interval of falls, validating the designation. The specificity between therelationship between high-fall-risk and traumatic intracra- the HR (0.98) was wide (0.59 –1.72), warfarin could either nial hemorrhage validates the association.
be beneficial or harmful in this population. Given this un- In summary, patients at high risk for falls are at substan- certainty and the known expense, hassle, and risks of war- tially increased risk of intracranial hemorrhage, especially farin therapy, we recommend aspirin or no antithrombotic traumatic intracranial hemorrhage. However, because of their increased risk of stroke, they appear to benefit from Besides fall risk, prior stroke, prior bleeding, and neuro- anticoagulant therapy if they have atrial fibrillation and at psychiatric impairment were associated with incident intra- cranial hemorrhage. Others also have found an association between prior stroke and intracranial Themechanism behind this association may be loss of micro-vascular integrity or disruption of neurovascular homeosta- Acknowledgment
sis. Patients with a history of epistaxis are more likely tohave an intracranial and patients with a prior The authors appreciate the assistance of the Iowa Foun- intracranial hemorrhage, especially a primary lobar intra- dation for Medical Care, the other Quality Improvement cranial hemorrhage, are at increased risk of Organizations, and the Centers for Medicare & Medicaid The association between intracranial hemorrhage and neu- Services (CMS) in providing data that made this research ropsychiatric disease may depend on the type of neuropsy- chiatric disease and the degree of impairment in perfor-mance Psychiatric disease may cause falls becauseof treatment with psychotropic associated References
alcohol use, or poor compliance. Patients with Parkinson’sdisease can fall because of a festinating Patients with 1. Van Walraven C, Hart RG, Singer DE, et al. Oral anticoagulants vs Alzheimer’s dementia may be predisposed to intracranial aspirin in nonvalvular atrial fibrillation: an individual patient meta- hemorrhage if they have cerebral amyloid angiopathy or analysis. JAMA. 2002;288:2441–2448.
2. Olsson SB, for the Executive Steering Committee on behalf of the This study has limitations inherent to use of administra- SPORTIF III Investigators. Stroke prevention with the oral direct tive data. First, because all participants were deidentified, thrombin inhibitor ximelagatran compared with warfarin in patientswith non-valvular atrial fibrillation (SPORTIF III): randomized con- we were not able to validate the diagnosis of intracranial trolled trial. Lancet. 2003;362:1691–1698.
hemorrhage by reviewing brain-imaging studies. However, 3. Albers GW, Diener HC, Frison L, et al. Ximelagatran vs warfarin for the ICD-9 codes we used to identify intracranial hemor- stroke prevention in patients with nonvalvular atrial fibrillation: a rhage have a specificity of over Had we corrected randomized trial. JAMA. 2005;293:690 – 698.
4. The Boston Area Anticoagulation Trial for Atrial Fibrillation Investi- for the known misclassification rate of ICD-9 codes for gators. The effect of low-dose warfarin on the risk of stroke in patients intracranial hemorrhage, incidence rates would be increased with nonrheumatic atrial fibrillation. N Engl J Med. 1990;323: by a factor of 1.26. A second limitation is that we knew Risk of Intracranial Hemorrhage in Patients Prone to Fall 5. Connolly SJ. Canadian Atrial Fibrillation Anticoagulation (CAFA) 21. Lord SR, Castell S, Corcoran J, et al. The effect of group exercise on Study. J Am Coll Cardiol. 1991;18:349 –355.
physical functioning and falls in frail older people living in retirement 6. Stroke Prevention in Atrial Fibrillation Investigators. Stroke Preven- villages: a randomized, controlled trial. J Am Geriatr Soc. 2003;51: tion in Atrial Fibrillation (SPAF) Study. Final results. Circulation.
22. Bischoff-Ferrari HA, Dawson-Hughes B, Willett WC, et al. Effect of 7. Stroke Prevention Atrial Fibrillation III Writing Committee. Patients vitamin D on falls: a meta-analysis. JAMA. 2004;291:1999 –2006.
with nonvalvular atrial fibrillation at low risk of stroke during treat- 23. Jensen J, Lundin-Olsson L, Nyberg L, Gustafson Y. Fall and injury ment with aspirin: Stroke Prevention in Atrial Fibrillation III Study.
prevention in older people living in residential care facilities. a cluster randomized trial. Ann Intern Med. 2002;136:733–741.
8. Gulløv AL, Koefoed BG, Petersen P, et al. Fixed minidose warfarin 24. Kallin K, Lundin-Olsson L, Jensen J, et al. Predisposing and precipi- and aspirin alone and in combination vs adjusted-dose warfarin for tating factors for falls among older people in residential care. Public stroke prevention in atrial fibrillation: Second Copenhagen Atrial Fi- brillation, Aspirin, and Anticoagulation Study. Arch Intern Med. 1998; 25. Tinetti ME. Clinical practice: preventing falls in elderly persons.
N Engl J Med. 2003;348:42– 49.
9. Pengo V, Zasso A, Barbero F, et al. Effectiveness of fixed minidose 26. Woo D, Sauerbeck LR, Kissela BM, et al. Genetic and environmental warfarin in the prevention of thromboembolism and vascular death in risk factors for intracerebral hemorrhage: preliminary results of a nonrheumatic atrial fibrillation. Am J Cardiol. 1998;82:433– 437.
10. Hellemons BS, Langenberg M, Lodder J, et al. Primary prevention of population-based study. Stroke 2002;33:1190 –1196; discussion, 1190- arterial thromboembolism in non-rheumatic atrial fibrillation in pri- mary care: randomized controlled trial comparing two intensities of 27. Berwaerts J, Webster J. Analysis of risk factors involved in oral-antico- coumarin with aspirin. BMJ. 1999;319:958 –964.
agulant-related intracranial haemorrhages. Qjm 2000;93:513–521.
11. Monette J, Gurwitz JH, Rochon PA, Avorn J. Physician attitudes 28. Saloheimo P, Juvela S, Hillbom M. Use of aspirin, epistaxis, and concerning warfarin for stroke prevention in atrial fibrillation: results untreated hypertension as risk factors for primary intracerebral hem- of a survey of long-term care practitioners. J Am Geriatr Soc 1997; orrhage in middle-aged and elderly people. Stroke 2001;32:399 – 404.
29. Bailey RD, Hart RG, Benavente O, Pearce LA. Recurrent brain hem- 12. Weisbord SD, Whittle J, Brooks RC. Is warfarin really underused in orrhage is more frequent than ischemic stroke after intracranial hem- patients with atrial fibrillation? J Gen Intern Med. 2001;16:743–749.
orrhage. Neurology. 2001;56:773–777.
13. Go AS, Hylek EM, Borowsky LH, et al. Warfarin use among ambu- 30. Nieuwenhuis HK, Albada J, Banga JD, Sixma JJ. Identification of risk latory patients with nonvalvular atrial fibrillation: the anticoagulation factors for bleeding during treatment of acute venous thromboembo- and risk factors in atrial fibrillation (ATRIA) study. Ann Intern Med.
lism with heparin or low molecular weight heparin. Blood. 1991;78: 14. White RH, Beyth RJ, Zhou H, Romano PS. Major bleeding after 31. Rubenstein LZ, Josephson KR, Robbins AS. Falls in the nursing home.
hospitalization for deep-venous thrombosis. Am J Med. 1999;107: Ann Intern Med. 1994;121:442– 451.
32. Rosand J, Hylek EM, O’Donnell HC, Greenberg SM. Warfarin-asso- 15. He J, Whelton PK, Vu B, Klag MJ. Aspirin and risk of hemorrhagic ciated hemorrhage and cerebral amyloid angiopathy: a genetic and stroke: a meta-analysis of randomized controlled trials. JAMA. 1998; pathologic study. Neurology. 2000;55:947–951.
33. O’Donnell HC, Rosand J, Knudsen KA, et al. Apolipoprotein E ge- 16. Rosand J, Eckman MH, Knudsen KA, et al. The effect of warfarin and notype and the risk of recurrent lobar intracerebral hemorrhage. N Engl intensity of anticoagulation on outcome of intracerebral hemorrhage.
Arch Intern Med. 2004;164:880 – 884.
34. McCarron MO, Nicoll JA, Ironside JW, et al. Cerebral amyloid angi- 17. Birman-Deych E, Waterman AD, Yan Y, et al. Accuracy of ICD-9 opathy-related hemorrhage: interaction of APOE epsilon2 with puta- codes for identifying stroke and cardiovascular risk factors. Med Care.
2005;43:480-485.
tive clinical risk factors. Stroke. 1999;30:1643–1646.
18. Gage BF, Waterman AD, Shannon W, et al. Validation of clinical 35. Li J, Brown J, Levine M. Mild head injury, anticoagulants, and risk of classification schemes for predicting stroke: results from the National intracranial injury. Lancet. 2001;357:771–772.
Registry of Atrial Fibrillation. JAMA. 2001;285:2864 –2870.
36. Karni A, Holtzman R, Bass T, et al. Traumatic head injury in the 19. Gage BF, van Walraven C, Pearce L, et al. Validation of stroke-risk anticoagulated elderly patient: a lethal combination. Am Surg. 2001; stratification schemes. Circulation. 2004;110:2287-2292.
20. Hart RG, Halperin JL, Pearce AC, et al. Lessons from the stroke 37. Fang MC, Chang Y, Hylek EM, et al. Advanced age, anticoagulation prevention in atrial fibrillation trials. Ann Intern Med. 2003;138:831– intensity, and risk for intracranial hemorrhage among patients taking warfarin for atrial fibrillation. Ann Intern Med. 2004;141:745–752.

Source: http://www.scanlongtermcare.com/documents/cme/clinical-guidelines/Incidence%20of%20intracranial%20hemorrhage%20in%20patients.pdf