Search tips
Search criteria 


Logo of brjgenpracRCGP homepageJ R Coll Gen Pract at PubMed CentralBJGP at RCGPBJGP at RCGP
Br J Gen Pract. 2012 October; 62(603): e710–e717.
Published online 2012 October 1. doi:  10.3399/bjgp12X656856
PMCID: PMC3459779

Risk of stroke and oral anticoagulant use in atrial fibrillation: a cross-sectional survey

Tim A Holt, PhD MRCP FRCGP, Academic Clinical Lecturer
Department of Primary Care Health Sciences, Oxford University, Oxford
Tina D Hunter, PhD, honorary senior researcher
Department of Primary Care Health Sciences, Oxford University, Oxford; executive director of outcomes research, S2 Statistical Solutions, Cincinnati, OH
Candace Gunnarsson, PhD, president
S2 Statistical Solutions, Cincinnati, OH
Nada Khan, PhD, research fellow
Department of Primary Care Health Sciences, Oxford University, Oxford
Paul Cload, PhD, Global Head of Cardiology & General Medicine
Global Marketing Organisation, GE Healthcare, Chalfont St Giles
Gregory YH Lip, MD, FRCP, DFM, FACC, FESC, Professor of Cardiovascular Medicine



Oral anticoagulants substantially reduce the risk of stroke in atrial fibrillation but are underutilised in current practice.


To measure the distribution of stroke risk in patients with atrial fibrillation (using the CHADS2 and CHA2DS2-VASc scores) and changes in oral anticoagulant use during 2007–2010.

Design and setting

Longitudinal series of cross-sectional survey in 583 UK practices linked to the QResearch® database providing 99 351 anonymised electronic records from people with atrial fibrillation.


The proportion of patients in each CHADS2 and CHA2DS2-VASc risk band in 2010 was calculated; for each of the years 2007–2010, the proportions with risk scores ≥2 that were using anticoagulants or antiplatelet agents were estimated. The proportions identified at high risk were re-estimated using alternative definitions of hypertension based on coded data. Finally, the prevalence of comorbid conditions in treated and untreated high-risk (CHADS2 ≥2) groups was derived.


The proportion at high risk of stroke in 2010 was 56.9% according to the CHADS2 ≥2 threshold, and 84.5% according to CHA2DS2-VASc ≥2 threshold. The proportions of these groups receiving anticoagulants were 53.0% and 50.7% respectively and increased during 2007–2010. The means of identifying the population of individuals with hypertension significantly influenced the estimated proportion at high risk. Comorbid conditions associated with avoidance of anticoagulants included history of falls, use of nonsteroidal anti-inflammatory drugs, and dementia.


Oral anticoagulant use in atrial fibrillation has increased in UK practice since 2007, but remains suboptimal. Improved coding of hypertension is required to support systematic identification of individuals at high risk of stroke and could be assisted by practice-based software.

Keywords: anticoagulants, atrial fibrillation, electronic health records, primary health care, stroke


Atrial fibrillation (AF) is the commonest disorder of heart rhythm and an important risk factor for thromboembolic stroke.1 An individual’s risk of stroke can be estimated using risk-stratification schemes that have been validated in various studies.25 In Europe, the CHADS2 and CHA2DS2-VASc scores are recommended (Box 1).6 CHA2DS2-VASc is significantly more inclusive of common stroke risk factors, while CHADS2 is simpler, easier to remember, and was proposed over 10 years ago, so has tended to be used more extensively in clinical practice. However, the limitations of the CHADS2 score have been highlighted recently.7,8 The CHA2DS2-VASc score is more effective at identifying the ‘truly low-risk’ population where (in patients with a CHA2DS2-VASc score = 0) no thromboprophylaxis is needed.9 For the higher-risk groups, oral anticoagulation significantly reduces the risk of stroke and mortality.10

Box 1. CHADS2 and CHA2DS2-VASc risk scores

Unless indicated otherwise, 1 point is allocated to each risk factor

CHADS2 (maximum 6 points)

C Congestive heart failure

H Hypertension

A Age >75 years

D Diabetes

S Stroke, transient ischaemic attack (TIA), or other thromboembolism history (2 points)

CHA2DS2–VASc (maximum 9 points)

C Congestive heart failure

H Hypertension

A Age >75 years (2 points)

D Diabetes

S Stroke, TIA, or other thromboembolism history (2 points)

V Vascular disease (coronary artery or peripheral vascular disease)

A Age 65–74 years

Sc Sex category (female 1 point, male 0 points)

People with a CHADS2 score of 2 or higher are recommended anticoagulation unless contraindicated. Those with CHADS2 score less than 2 can be assessed using CHA2DS2-VASc. Those with CHA2DS2-VASc =0 require no thromboprophylactic therapy.

Decisions over anticoagulation need to be tailored to the individual. However, suboptimal use and uptake in AF is a global phenomenon.1113 Opportunities are thereby missed to impact significantly on an important cause of cardiovascular morbidity and mortality.

Surveys of stroke risk in AF based on UK primary care data have been published previously.1416 The authors were interested not only in the patterns of risk and levels of anticoagulant use, but also in the ability of primary care data to support risk estimation. A software tool (GRASP-AF — Guidance on Risk Assessment and Stroke Prevention in Atrial Fibrillation) has been developed for UK general practices,17 which produces lists of high-risk but untreated individuals. The success of this approach depends on the data quality of the source systems, including the disease registers.

Specifically, the hypertension register is problematic. Clinicians may not always enter a hypertension code if the person is already on another disease register that has a blood pressure target. To do so may duplicate the processes of recall. For those with AF, exclusion of a patient with hypertension from the hypertension register runs the risk of underestimating stroke risk and of failing to identify the need for anticoagulation. While administrative registers may be useful for estimating the prevalence of hypertension at population level,18,19 their adequacy for supporting estimation of stroke risk has not been confirmed.

This study used the QResearch® database to determine the distribution of CHADS2 and CHA2DS2-VASc scores and changes in prescribing of oral anticoagulant and antiplatelet agents during the years 2007–2010, following publication of a National Institute for Health and Clinical Excellence (NICE) guideline on AF.20 This was the prevailing guideline for the period of the study. The impact on numbers identified of alternative definitions of hypertension was investigated. The study also measured the differences between treated and untreated populations in the prevalence of a range of prespecified comorbidity codes potentially relevant to anticoagulant prescribing.

How this fits in

Suboptimal use of oral anticoagulants in patients with atrial fibrillation is a missed opportunity to impact on an important cause of thromboembolic stroke. Systematic identification of those at high stroke risk requires consistent recording of risk factors during routine care. About one in seven will currently be missed if inclusion in the hypertension register is used to determine hypertensive status, highlighting the need for improved recording of hypertension in this group. A number of barriers must be overcome to optimise the safe and effective use of anticoagulants for stroke prevention in patients with atrial fibrillation.


Source of data

The QResearch database contains anonymous electronic data from over 600 UK practices. Records in this database from people with a diagnosis of AF were identified. Individual records were excluded if the person had been registered for less than 12 months, or was a temporary resident. No other exclusion criteria were applied, and all available eligible records were included in the study. QResearch has been validated using other nationally published sources, as broadly representative of the UK population.21

Statistical methods

Analyses were carried out using STATA (version 11) and SAS (version 9.2). The χ2 test was used to measure differences in proportions and to calculate P-values for significance.

Cross-sectional surveys

For each identified record, a stroke risk score was carried out using the CHADS2 and CHA2DS2-VASc algorithms. For each risk band, the proportion currently treated with anticoagulants or antiplatelet drugs was derived, based on the presence of a prescription for the drug in the record in the last 6 months. This cross-sectional survey was repeated for each of the years 2007 to 2010. The main outcomes of the study used the most recent (2010) cross-section.

Influence of different definitions of hypertension

The cross-sectional surveys required a coded diagnosis of hypertension for the risk score, and a range of Read Codes were used, similar to those supporting the Quality and Outcomes Framework (QOF).22 These are all ‘child’ codes of the G2 Read Code group. More inclusive means of identifying the population of individuals with hypertension (Box 2) were then used, based on use of antihypertensive medication and blood pressure values. The most recent cross-sectional survey (for 2010) was repeated using these alternative definitions, to measure the impact on the numbers identified.

Box 2. Definitions of hypertension based on primary care data (details available from the authors)

Definition A: Requires a specific G2 Read code for hypertension, as in the Quality and Outcomes Framework (QOF)

Definition B: As for A, but includes also those with no specific hypertension code, but using an antihypertensive agent

Definition C: As for B, but excludes those with no hypertension code that have another possible reason for using an antihypertensive agent (for example, coronary heart disease if taking a calcium channel blocker, heart failure if taking an angiotensin-converting enzyme inhibitor)

Definition D: As for C, but includes also those with evidence for hypertension based on recorded blood pressure values (mean of the last three measurements if available ≥150 mmHg systolic or ≥90 mmHg diastolic, even for another possible reason) or those not on treatment for hypertension but with blood pressure ≥160 mmHg systolic or ≥100 mmHg diastolic

For the main cross-sectional surveys 2007–2010, definition A was used.

Comorbidity codes associated with lack of use of anticoagulation

The prevalence of a range of comorbidity codes was examined in the treated and untreated high-risk populations recommended anticoagulation; χ2 tests were used to measure the significance of differences in prevalence.


The sample included 583 UK practices, and 99 351 patient records. Of these, 59 804 were available for the main cross-section of 1 October 2010; the remainder had died, left the practice, or were in a practice whose most recent data upload was before 1 October 2010. Table 1 gives the patient characteristics for the 2010 cross-section. The median age in 2010 was 80.0 years with interquartile range 71.0 to 87.0 years. The median age at AF diagnosis was 73.0 years.

Table 1
Characteristics of the population with atrial fibrillation (AF)

Table 2 shows the distribution of risk for the main survey of 2010 and the proportions treated with anticoagulants, antiplatelet agents, or both. An estimated 56.9% of people with AF had a CHADS2 score ≥2, and 84.5% had a CHA2DS2-VASc score ≥2. Only 53.0% and 50.7% of these groups respectively were using oral anticoagulants, with the majority of the remainder receiving an antiplatelet agent.

Table 2
Distribution of stroke risk for the 2010 cross-sectional survey (n = 59 804) by use of anticoagulants and antiplatelet agents

Table 3 and Figure 1 report the changes in numbers identified and proportions treated for the years 2007–2010. The proportion with CHADS2 score ≥2 remained stable and use of anticoagulants in this group increased from 49.7% in 2007 to 53.0% in 2010 (P<0.001).

Figure 1
Trends in prescribing of anticoagulants in those with CHADS2≥2, 2007-2010.
Table 3
Trends in prescribing of anticoagulants in those with CHADS2 ≥2 and CHA2DS2-VASc ≥2, 2007–2010a

Table 4 gives the numbers identified in each risk band, according to the alternative definitions of hypertension A–D for the 2010 cross-section only. Definition A requires a G2 Read Code for hypertension and is the basis for the estimated 56.9% of AF patients with CHADS2 ≥2 in 2010. This figure rises to 67.2% using definition B (which also includes people with no G2 code but taking an antihypertensive drug). When those with a possible alternative reason to be taking the drug, such as heart failure or angina (definition C), are removed, the proportion is 65.9%. Adding to this, people who do not have a coded diagnosis of hypertension but do have evidence of raised blood pressure (definition D), the figure is 66.1%.

Table 4
Influence of alternative definitions of hypertension on the numbers identified as requiring anticoagulation for the latest survey, 2010a

Table 5 gives the prevalence of relevant comorbidity codes in the treated and untreated populations for those recommended anticoagulants according to the CHADS2 ≥2 threshold, and the significance estimates for the differences. The difference is significant for most comorbidity codes, but those particularly relevant for clinical practice include a history of falls, peptic ulceration, and other upper gastrointestinal (GI) tract disorders; use of drugs for dyspepsia; and use of nonsteroidal anti-inflammatory drugs. These may be associated with avoidance of oral anticoagulation therapy.

Table 5
Prevalence of comorbidity codes in the anticoagulant treated and untreated populations of those in whom anticoagulation is recommended by CHADS2 ≥2, (n = 34 041)



Only 53.0% of patients in the study at high risk of stroke were using oral anticoagulants. This proportion increased only slightly in the years 2007–2010. At the same time, higher than expected usage was found in the low-risk groups: 32.1% of people with CHADS2 = 0 and 23.0% with CHA2DS2VASc = 0. While anticoagulation may be appropriate for some of these individuals (for example, those with valvular disease), this suggests that use of these algorithms has still to become established. CHADS2 was first proposed over a decade ago, while CHA2DS2-VASc was introduced much more recently.

The estimated high-risk population increases when more inclusive definitions of hypertension are used. The authors considered the safest to be ‘C’, which increases the proportion identified from 56.9% to 65.9%. The additional inclusion of those with raised blood pressure levels (definition D) may be unreliable, particularly given the recent emphasis on home, rather than office-based measurements for diagnosis.23 About one in seven people at high risk of stroke (CHADS2 ≥2) will have their risk estimated as CHADS2 = 1 and therefore not be recognised as requiring anticoagulant therapy by this decision rule if the presence of a code for hypertension (G2) is required to confirm hypertensive status. Improved coding of hypertension in primary care could readily be assisted by practice-based software.

Certain comorbidity codes were found to be more prevalent in the records of untreated compared with treated people in the CHADS2 ≥2 population. This analysis is observational and needs to be interpreted with caution. The increased recording of adverse reactions to anticoagulants in the treated population may reflect the fact that such recording only occurs in people exposed to treatment and that this does not always require withdrawal of the drug. Similarly, the recording of a bleeding tendency may be noted as a result of treatment and is more common in the treated population. Alcohol overuse and chronic liver disease do not, perhaps surprisingly, appear to be barriers to anticoagulation, but other findings confirmed previously suspected obstacles, including a history of falls. Each person needs to be assessed individually, as some are at much higher risk from this than others, but the risk from falls (compared to the benefits of anticoagulation) may be overestimated by clinicians managing people with AF and at risk of stroke.24 Drugs for dyspepsia are used more commonly in untreated individuals, who are also more likely to have a record of peptic ulcer or other upper GI disorders. Whether this history is a sufficient basis for avoidance is again an individual matter, and may depend on how recently the upper GI pathology occurred. The present survey also suggests avoidance of anticoagulants in people using nonsteroidal anti-inflammatory drugs, which interact with warfarin but are not an absolute contraindication in most cases. Those with dementia may be less able to concur with treatment and monitoring, and less able to consent to treatment, so in some cases avoidance in such people may be appropriate. Stroke is more often disabling than fatal, and individuals vary widely in their anticipated quality of life utility associated both with stroke and with antithrombotic therapies.25,26

Use of anticoagulants in the CHADS2 and CHA2DS2-VASc risk groups above the level of 2 tends to level off rather than increase with stroke risk (Table 2). This may imply that comorbidity itself offsets rather than increases the tendency to prescribe. Those at highest risk of thromboembolism are also at highest risk of bleeding complications, a fact that may influence clinicians. Haemorrhages related to warfarin are a significant problem and may be more common in ‘real-world’ practice27 than under the more careful monitoring arrangements of randomised trials.28 Despite this, the risk–benefit ratio is particularly in favour of anticoagulant therapy in the higher-risk groups,29 provided it is carefully monitored.

Strengths and limitations

This is a large survey of a representative population of people with AF diagnosed and managed in both primary and secondary care settings. It includes people with all forms of AF, including both paroxysmal as well as chronic AF. The study may have not detected some individuals receiving anticoagulant prescriptions in secondary care. The National Patient Safety Agency has emphasised the importance of good communication between different bodies sharing responsibility for prescribing potentially interacting medication,30 and this has increased the use of codes in primary care to maintain awareness of anticoagulant therapy prescribed elsewhere. Identifying people taking both anticoagulants and antiplatelet agents is difficult in surveys of this kind, because a person changing from one treatment to the other may have evidence of both in their recent record and this proportion may be slightly overestimated. The prevalence of falls in the population with AF is higher than reported in this study because a limited code set was used to identify them. A limited set of ‘chronic kidney disease’ codes was chosen in order to identify those with more significant forms of renal disease.

Comparison with existing literature

The findings of this study support other studies reporting suboptimal uptake of oral anticoagulants in AF.11 The figure for the proportion of individuals with CHADS2 ≥2 treated with oral anticoagulants (53%) is very similar to that derived from a Canadian study published in 2011.13 This study has confirmed the lower usage in people with a history of falls, and those with dementia, previously reported in 2008.15

Implications for practice and research

There is still huge potential for reducing the stroke risk of the population with AF by identifying people in primary care requiring anticoagulant treatment. Routinely collected data are able to support this process but consistent coding of hypertension is important. Several factors may have contributed to the minor improvements demonstrated in this study. The NICE guideline of 2006 may have raised awareness of the benefits of this treatment.20 The BAFTA (Birmingham Atrial Fibrillation Treatment of the Aged) trial was published in 2007 and was reassuring over the risks/benefits of warfarin in older people with AF.31 A further factor was the introduction in 2006 of AF registers as part of the UK QOF, involving payments based on the proportion of people with AF treated with either anticoagulants or antiplatelet agents (irrespective of stroke risk level). From April 2012, the targets include the proportion of those with identified CHADS2 score ≥2 who are receiving anticoagulant therapy.32

However, it is increasingly recognised that some patients with a CHADS2 score of 0–1 are at significant risk and would benefit from anticoagulation. Among those with CHADS2 = 0, the stroke/thromboembolism risk can vary between 0.84%/year (if the CHA2DS2-VASc score = 0) to 3.2%/year (if the CHA2DS2-VASc score = 3).33 Thus, a CHADS2 score of 0–1 is not necessarily ‘low risk’ and only the CHA2DS2-VASc score can identify the truly low-risk group. The CHADS2 ≥2 threshold for determining anticoagulant uptake under the QOF is therefore an audit standard and not a definition of best practice.

A recent consensus statement from the Royal College of Physicians of Edinburgh (RCPE) supports this conclusion, arguing for anticoagulation in all but the truly low-risk groups.34 According to the present survey results and their threshold, this would mean that around 91.6% of people with AF should be considered for this treatment. Strokes related to AF are associated with higher case fatality and more profound disability in survivors than non-AF related strokes,35 but could be minimised by this approach. As recommended in the European Society for Cardiology guidelines,6 risk scores have evolved to become more useful for identifying those at low risk (not requiring treatment) than those at high risk.36 The RCPE also recommends that, as aspirin is ineffective as thromboprophylaxis in AF, it should not be used for this reason alone. These developments have simplified decision-making policy and are likely to change practice significantly towards anticoagulation. Newer oral anticoagulants may potentially improve uptake still further. Future studies need to explore the remaining barriers, including the qualitative issues influencing patient-centred decision making over anticoagulation.


We acknowledge the contribution of EMIS practices who contribute to QResearch, the University of Nottingham, and EMIS for establishing, developing, and supporting the database and supplying the data for this research project.



The study was funded by GE Healthcare (reference number 2580270919).

Ethics committee

QResearch studies are ethically approved by Trent MREC 03/04/021.


Freely submitted; externally peer reviewed.

Competing interests

Paul Cload was an employee of GE Healthcare during this project. Candace Gunnarsson had financial support from GE Healthcare for the study. There are no other competing interests.

Discuss this article

Contribute and read comments about this article on the Discussion Forum:


1. Go AS, Hylek EM, Phillips KA, et al. Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study. JAMA. 2001;285(18):2370–2375. [PubMed]
2. Gage BF, Waterman AD, Shannon W, et al. Validation of clinical classification schemes for predicting stroke. Results from the national registry of atrial fibrillation. JAMA. 2001;285(22):2864–2870. [PubMed]
3. Olesen JB, Lip GY, Hansen ML, et al. Validation of risk stratification schemes for predicting stroke and thromboembolism in patients with atrial fibrillation: nationwide cohort study. BMJ. 2001;342:d124. [PubMed]
4. Stroke Risk in Atrial Fibrillation Working Group. Comparison of 12 risk stratification schemes to predict stroke in patients with nonvalvular atrial fibrillation. Stroke. 2008;39(6):1901–1910. [PubMed]
5. Lip GY, Nieuwlaat R, Pisters R, et al. Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the euro heart survey on atrial fibrillation. Chest. 2010;137(2):263–272. [PubMed]
6. Camm AJ, Kirchhof P, Lip GYH, et al. for the European Society for Cardiology Guidelines for the management of atrial fibrillation. Eur Heart J. 2010;31(19):2369–2429. [PubMed]
7. Karthikeyan G, Eikelboom JW. The CHADS2 score for stroke risk stratification in atrial fibrillation — friend or foe? Thromb Haemost. 2010;104(1):45–48. [PubMed]
8. Keogh C, Wallace E, Dillon C, et al. Validation of the CHADS2 clinical prediction rule to predict ischaemic stroke. A systematic review and meta-analysis. Thromb Haemost. 2011;106(3):528–538. [PubMed]
9. Lip GYH, Halperin JL. Improving stroke risk stratification in atrial fibrillation. Am J Med. 2010;123(6):484–488. [PubMed]
10. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have non-valvular atrial fibrillation. Ann Intern Med. 2007;146(12):857–867. [PubMed]
11. Ogilvie IM, Newton N, Welner SA, et al. Underuse of oral anticoagulants in atrial fibrillation: a systematic review. Am J Med. 2010;123(7):638–645. e4. [PubMed]
12. Nieuwlaat R, Capucci A, Lip GYH, et al. on behalf of the Euro Heart Survey Investigators Antithrombotic treatment in real-life atrial fibrillation patients: a report from the Euro Heart Survey on Atrial Fibrillation. Eur Heart J. 2006;27(24):3018–3026. [PubMed]
13. Sandhu RK, Bakal JA, Ezekowitz JA, McAlister FA. Risk stratification schemes, anticoagulation use and outcomes: the risk-treatment paradox in patients with newly diagnosed non-valvular atrial fibrillation. Heart. 2011;97(24):2046–2050. [PubMed]
14. Majeed A, Moser K, Carroll K. Trends in the prevalence and management of atrial fibrillation in general practice in England and Wales, 1994–1998: analysis of data from the general practice research database. Heart. 2001;86(3):284–288. [PMC free article] [PubMed]
15. Gallagher AM, Rietbrock S, Plumb J, van Staa TP. Initiation and persistence of warfarin or aspirin in patients with chronic atrial fibrillation in general practice: do the appropriate patients receive stroke prophylaxis? J Thrombosis Haemost. 2008;6(9):1500–1506. [PubMed]
16. Van Staa TP, Setakis E, Di Tanna GL, et al. A comparison of risk stratification schemes for stroke in 79 884 atrial fibrillation patients in general practice. J Thromb Haemost. 2011;9(1):39–48. [PubMed]
17. NHS Improvement. Guidance on Risk Assessment and Stroke Prevention for Atrial Fibrillation (GRASP-AF) (accessed 29 Aug 2012)
18. Quan H, Khan N, Hemmelgarn BR, et al. Validation of a case definition to define hypertension using administrative data. Hypertension. 2009;54(6):1423e8. [PubMed]
19. Tu K, Campbell NRC, Chen Z-L, Cauch-Dudek KJ, McAlister FA. Accuracy of administrative databases in identifying patients with hypertension. Open Med. 2007;1(1):e18–e26. [PMC free article] [PubMed]
20. National Institute for Health and Clinical Excellence. CG36: Atrial fibrillation. London: National Institute for Health and Clinical Excellence; 2006.
21. Hippisley-Cox J, Vinogradova Y, Coupland C, Pringle M. Comparison of key practice characteristics between general practices in England and Wales and general practices in the QResearch data. Report to the Health and Social Care Information Centre. Nottingham: University of Nottingham; 2005.
22. NHS The Information Centre. Quality and Outcomes Framework. (accessed 29 Aug 2012)
23. National Institute for Health and Clinical Excellence. CG127: Hypertension. The clinical management of primary hypertension in adults. London: National Institute for Health and Clinical Excellence; 2011.
24. Man-Son-Hing M, Nichol G, Lau A, Laupacis A. Choosing antithrombotic therapy for elderly patients with atrial fibrillation who are at risk for falls. Arch Intern Med. 1999;159(7):677–685. [PubMed]
25. Gage BF, Cardinalli AB, Owens DK. The effect of stroke and stroke prophylaxis with aspirin or warfarin on quality of life. Arch Intern Med. 1996;156(16):1829–1836. [PubMed]
26. MacLean S, Mulla S, Akl EA, Jankowski M, et al. Patient values and preferences in decision making for antithrombotic therapy: a systematic review: Antithrombotic Therapy and Prevention of Thrombosis, 9th edn: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest. 2012;141(2 suppl):e1S–23S. [PubMed]
27. Hollowell J, Ruigomez A, Johansson S, et al. The incidence of bleeding complications associated with warfarin treatment in general practice in the United Kingdom. Br J Gen Pract. 2003;53(489):312–314. [PMC free article] [PubMed]
28. Agarwal S, Hachamovitch R, Menon V. Current trial-associated outcomes with warfarin in prevention of stroke in patients with nonvalvular atrial fibrillation: a meta-analysis. Arch Intern Med. 2012;172(8):623–631. [PubMed]
29. Olesen JB, Lip GYH, Lindhardsen J, et al. Risks of thromboembolism and bleeding with thromboprophylaxis in patients with atrial fibrillation: a net clinical benefit analysis using a ‘real world’ nationwide cohort study. Thromb Haemost. 2011;106(4):739–749. [PubMed]
30. Baglin TP, Cousins D, Keeling DM, et al. Recommendations from the British Committee for Standards in Haematology and National Patient Safety Agency. Br J Haematol. 2006;136(1):26–29. [PubMed]
31. Mant J, Hobbs FDR, Fletcher K, et al. Warfarin versus aspirin for stroke prevention in atrial fibrillation in the elderly community population: the Birmingham Atrial Fibrillation Treatment of the Aged Study (BAFTA), a randomised controlled trial. Lancet. 2007;370(9586):493–503. [PubMed]
33. Olesen JB, Torp-Pedersen C, Hansen ML, Lip GY. The value of the CHA2DS2-VASc score for refining stroke risk stratification in patients with atrial fibrillation with a CHADS2 score 0–1: a nationwide cohort study. Thromb Haemost. 2012;107(6):1172–1179. [PubMed]
34. Stott DJ, Dewar RI, Garratt CJ, et al. RCPE UK Consensus Conference on ‘approaching the comprehensive management of atrial fibrillation: evolution or revolution?’ J R Coll Physicians Edinb. 2012;42(1):34–35. [PubMed]
35. Lin H-J, Wolf PA, Kelly-Hayes M, et al. Stroke severity in atrial fibrillation: the Framingham study. Stroke. 1996;27(10):1760–1764. [PubMed]
36. James MA, Campbell JL. Better prevention of stroke through screening for atrial fibrillation. Br J Gen Pract. 2012;62(598):234–235. [PMC free article] [PubMed]

Articles from The British Journal of General Practice are provided here courtesy of Royal College of General Practitioners