The estimate of 0.4–1.0% prevalence of atrial fibrillation in the most recent American guidelines is based mainly on studies including patients with permanent atrial fibrillation (AF), although recent evidence shows that the stroke risk is similar with paroxysmal and persistent AF. Our objective was to determine the prevalence of AF in Sweden, irrespective of type and to what extent patients with AF receive adequate stroke prophylaxis.
Retrospective study of patients with a clinical diagnosis of atrial fibrillation between 2005 and 2010 in the national Swedish Patient Register matched with data from the National Prescribed Drugs Register.
We identified 307 476 individuals with a diagnosis of atrial fibrillation. Of these, 209 141 were still alive on the last day of the inclusion period, signifying a prevalence of clinically diagnosed AF in Sweden of 2.9% of the total adult (≥20 years) population. Only 42% of them had purchased an oral anticoagulant within 6 months of the first presentation with AF during the study period. Those at the highest risk of stroke were those least likely to receive anticoagulant treatment. Undertreatment was common amongst women and individuals >80 years, whilst overtreatment was common amongst young men without risk factors.
The prevalence of atrial fibrillation is at least 2.9% of the Swedish adult population, not counting ‘silent atrial fibrillation’. The official US figures probably underestimate the magnitude of the problem by a factor of 3–5. More than 80% had risk factors motivating anticoagulation therapy.
Atrial fibrillation (AF) is associated with a 4–5-fold increased risk of stroke [1, 2], a 2–3-fold increased risk of cardiac failure , almost a doubling of mortality [4, 5] and impaired quality of life . It is generally recognized as the most common significant dysrhythmia, although the prevalence estimates are highly diverging. The latest version of the American guideline document on AF says, ‘The estimated prevalence of AF is 0.4% to 1.0% in the general population’  whilst their European counterpart estimates the prevalence to be ‘1.5–2.0% of the general population’ . This variability and uncertainty about the prevalence is in part due to inclusion of different types of AF. Older studies mostly counted patients with permanent AF, which only constitute a minority of all patients with AF . For many years, the general belief was that paroxysmal AF is less hazardous than permanent AF. However, it has been shown that stroke rates are similar in both forms [10-12]. Thus, there is lack of information about the clinically relevant AF prevalence, which is needed for the assessment of its impact on, for example, stroke and heart failure and for planning of future health care.
The aim of this study was to determine the prevalence of clinically diagnosed AF in adults in Sweden, irrespective of type (paroxysmal, persistent or permanent) and to determine to what extent patients with AF are receiving adequate stroke prophylaxis.
We included all adults (≥20 years) with a primary or secondary diagnosis of AF between 1 July 2005 and 31 December 2010, listed in the national Swedish Patient Register. It covers all hospitals in the country since 1987 and provides complete lists of dates of admissions and discharges with diagnostic codes according to the 10th revision of the International Classification of Diseases (ICD-10), as well as codes for surgical and therapeutic procedures. The Patient Register also includes information about out-patient visits at hospitals and hospital-managed satellite centres, but not about visits in the primary care (general practitioners offices).
We used the ICD-10 code I489 to identify individuals with an AF diagnosis. This definition includes both atrial fibrillation and flutter because these dysrhythmias are closely related and also have a similar stroke risk . Each individual was only counted once and there were no exclusion criteria. The date of the first occasion with an AF diagnosis during the inclusion period was used as index date. Diagnoses given prior to index date were used for the characterization of previous and concurrent diseases.
We used this background information to calculate each patient's stroke risk score according to the CHADS2 scheme, which gives 2 points for a previous stroke, TIA or systemic emboli, and one point each for heart failure, hypertension, diabetes and age ≥75 years . We also calculated the newer CHA2DS2-VASc score, which in addition to the CHADS2 score gives 2 points for age ≥75, and 1 point each for age 65-74 years, vascular disease and female sex [1, 15]. The codes used to define these conditions are listed in Appendix S1.
Bleeding risk was calculated using a modified HAS-BLED score (16) counting points for hypertension, renal failure, liver disease, thromboembolism, previous bleeding, age ≥65 years, prescription of ASA or clopidogrel and alcohol abuse defined from a diagnostic code belonging to the ‘alcohol index’ used for statistical purposes by the National Board of Health and Welfare (see Appendix S1). We had no information about INR values for patients treated with warfarin and had omit giving points for that.
Dates of deaths were obtained from the Swedish Population Register. Information about the general population was obtained from the government agency Statistics Sweden. To make our results comparable with other studies in the field, we expressed prevalence as that in the adult population aged ≥20 years, which on 31 December 2010 was 7 232 006.
Information about medication was obtained from the National Prescribed Drugs Register. This register automatically stores detailed information about every prescription that is handled in every pharmacy in the country since 1 July 2005 and is therefore almost 100% complete. Medication at baseline was defined as drugs that had been collected at a pharmacy within ±90 days of the index date. The only registered oral anticoagulant in Sweden during the study period was warfarin, with phenprocoumon as an alternative on special licence for a very small number of patients intolerant to warfarin.
Baseline characteristics were presented descriptively, and differences were tested with t-tests and chi-squared test. P-values <0.05 were considered significant. All analyses were performed in spss 20.0 (IBM SPSS Statistics, IBM Corporation, Route 100, Somers, NY 10589, USA).
Approval for the study was obtained from the regional ethical committee in Stockholm (EPN 2005/22–21/4, 2008/433–32).
During the 5.5-year inclusion period, 307 476 unique adult individuals received a hospital diagnosis of AF in Sweden; 98 335 of them died before the end of that period. Thus, on 31 December 2010, there were 209 141 living men and women with a diagnosis of AF corresponding to a prevalence of 2.9% of the adult Swedish population. The mean age was 75.2 ± 12.2 years at the index date (men 71.9 ± 12.3 years, women 82.2 ± 8.5 years). Clinical characteristics of those who survived, and of those who died before the end of the inclusion period, are presented in Table 1. The prevalence increased with age up to 14.3% (6168/43 237) at 84 years where after the proportion of individuals with AF appeared to decrease (Fig. 1). The prevalence was higher in men than in women in all age groups (Table 2) and higher in rural areas where the mean age of the population is higher. Thus, prevalence ranged from 2.5% in the Swedish capital Stockholm, where the mean age of the population is 39.0 years, up to 3.5% in the northernmost rural region of Norrbotten, where mean age of the population was 42.4 years.
Table 2. Atrial fibrillation prevalence by age and sex amongst 209 141 patients alive on 31 December 2010
All (n = 209 141) %
Men (n = 118 919) %
Women (n = 90 222) %
P Men vs. women
<60 years (n = 30 277)
60–69 years (n = 48 844)
70–79 years (n = 67 089)
80–89 years (n = 55 232)
≥90 years (n = 7699)
Over the years, the patients had accumulated a mean of 50.2 diagnoses prior to the index date. In most patients, a few diagnoses reappeared several times in conjunction with new healthcare contacts.
At the index date, >80% of the patients had comorbidities which would warrant consideration for stroke prophylaxis according to current American and European guideline recommendations (CHADS2 score ≥2 in 82% of patients, CHA2DS2-VASc score ≥2 in 83%) [7, 8]. About one-third (34.7%) had high bleeding risk defined as HAS-BLED score ≥3. Amongst the comorbidities in the risk stratification schemes, hypertension was the most common affecting 45%, followed by heart failure which was found in 31% (Table 3).
Table 3. Comorbidities in relation to age amongst all 307 476 patients with AF diagnosis
Thromboembolism (n = 63 118) %
Heart failure (n = 95 021) %
Diabetes (n = 51 223) %
Hypertension (n = 138 858) %
Vasc. disease (n = 71 244) %
AF, atrial fibrillation.
<60 (n = 31 952)
60–69 (n = 55 508)
70–79 (n = 88 886)
80–89 (n = 106 182)
≥90 (n = 24 948)
Warfarin, alone or in combination with ASA or clopidogrel, was used by 42% of the patients. ASA as single therapy was used by 34% and 20% of the patients had no therapy at all (Table 1). Warfarin use was inversely associated with stroke risk so that those at the highest risk of stroke were the least likely to receive it (Fig. 2). In contrast, ASA use increased in parallel with increasing stroke risk score (Fig. 3). After 80 years of age, warfarin use dropped rapidly, and at 90 and higher less than 10% received warfarin (Table 4, Fig. 4). Men were treated with warfarin more often than women (46% vs. 37%, P < 0.0001). Warfarin use was relatively common amongst young men with no risk points. In the very low-risk group with 0 points on CHA2DS2-VASc, 38% had warfarin. Cardioversion was not the primary reason for treatment in this group because only 11% of them (2347/20 899) were cardioverted during the study period.
Table 4. Warfarin use at the index date amongst all 307 476 patients with AF diagnosis
All (n = 307 476) %
Men (n = 168 630) %
Women (n = 138 846) %
P Men vs. women
AF, atrial fibrillation.
<60 (n = 31 952)
60–69 (n = 55 508)
70–79 (n = 88 886)
80–89 (n = 106 182)
≥90 (n = 24 948)
The present prevalence estimate of at least 2.9% is considerably higher than the current official estimate of 0.4–1.0% in the 2011 guideline update from the American College of Cardiology/American Heart Association/Heart Rhythm Society . The American guidelines refer to two sources. One is a meta-analysis from 1995  where the authors admit that there were so few elderly patients in the studies that the they had to ‘arbitrarily chose a prevalence rate of 10% for all persons older than 80 years’, which is obviously too low. The lack of data on elderly patients was due to upper age limits for inclusions in the studies. The second reference is the ATRIA study, which only counted patients ‘with nontransient atrial fibrillation’, that is, permanent AF, which also had to be the principal diagnosis . In the Euro Heart Survey  only 30% of the patients had permanent AF and would have fulfilled the ATRIA criteria. Furthermore, the criterion that AF had to be the principal diagnosis presumably resulted in the exclusion of patients with many comorbidities who are likely to receive some other more urgent diagnosis at discharge, rather than AF. Thus, many patients with AF were not counted in these earlier prevalence studies performed at a time when nonpermanent AF was believed to carry a low risk for stroke.
Observations on stroke prophylaxis
Only 42% of the patients had purchased warfarin at least once during a 6-month period framing the index date, although 83% of them had a risk score indicating that they would have benefited from anticoagulation therapy, unless there were strong reasons against it [19, 20]. Paradoxically, those at the highest risk of stroke were those least likely to receive warfarin treatment. There was an almost linear decrease in the likelihood of receiving warfarin with increasing stroke risk. In contrast, the likelihood for receiving ASA increased almost linearly with increasing risk. One way to interpret this is that prescribing doctors recognize the high stroke risk in patients with high-risk scores, but that they consider them too frail for anticoagulants and chose ASA instead. ASA is often perceived as a ‘milder’ treatment option, although the protective effect is very weak and the bleeding risk is about the same as with well-managed warfarin therapy [8, 21].
Warfarin treatment rapidly declined after the age of 80, although it is well recognized that advanced age is one of the most important stroke risk factors. Women were less often treated with warfarin than men at all ages (37% vs. 46%), despite their higher stroke risk [22, 23]. The reason for this is unclear and calls for further investigation and measures to improve current practices.
Generalizability of the results
The prevalence of AF within a population is clearly related to the proportion of elderly individuals. In our study, we found that the regional prevalence varied between 2.5% and 3.5% between two regions where the mean ages of the populations only differed by 3.4 years. In countries with younger population structures than the Swedish, the overall AF prevalence will probably be lower, but age and sex stratified prevalence may still be similar. Considering that many countries have rapidly ageing populations the information about the age and sex stratified prevalence is relevant for the dimensioning of the future healthcare system, especially with regard to the association with stroke and heart failure, which are the major cost drivers .
It has been suggested that AF prevalence may vary according to the ethnic background, which is not registered for Swedish residents, unlike in the United States and many other countries. Although Sweden has become more multicultural in recent decades, the Swedish population still consists mainly of Caucasians. According to the official statistical agency Statistics, Sweden only 14% of the inhabitants were born abroad, and of these, the majority were born in Europe . The results of our study may therefore not be directly applicable to a country with a predominantly non-Caucasian population.
It has not been possible for us to verify the AF diagnoses by ECG recordings. However, a recent validation study of the Swedish Patient Register could confirm AF in 97% of a random sample . The extent of under diagnosis is not known and would require population screening to be determined. However, in the greater Gothenburg area (Vaestra Goetaland County) of 1.2 million adults, 22% of the patients with AF were cared for only in primary care (Staffan Björck, personal communication), which extrapolated to the whole country suggests a prevalence of 3.5%. Patients with several more urgent or serious diagnoses may have been left without an AF diagnosis at discharge.
Underreporting of concomitant and previous diseases may have occurred for similar reasons. This may have affected the risk scores for ischaemic stroke, CHADS2 and CHA2DS2-VASc, giving patients falsely lower risk scores than they actually should have had.
The prevalence of AF is at least 2.9% of the Swedish adult population. The current US guidelines probably underestimate the magnitude of the problem by a factor 3–5 and therefore also the fraction of individuals at risk of AF-related stroke, who might benefit from prophylactic therapy with oral anticoagulants.
Conflict of interest statement
Both authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: the submitted work was supported by The Swedish Heart and Lung Foundation and The Stockholm County Council; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; Outside of the submitted work, Karolinska Institute received grants in support of LF's research from Boehringer-Ingelheim, Sanofi-Aventis, Bristol-Myers-Squibb and Bayer. LF has participated in advisory boards with Boehringer-Ingelheim and Sanofi-Aventis. LB has participated in advisory boards with Sanofi-Aventis, Boehringer-Ingelheim and MSD, and given lectures supported by Sanofi-Aventis and MSD.
Both authors participated in the drafting of the study and in revisions of the manuscript. LF prepared the data file, made the statistical analyses and wrote the first manuscript draft. LF takes full responsibility for the accuracy of the statistical analyses.
The Swedish Heart and Lung Foundation, and the Stockholm County Council. The manuscript conforms to the STROBE statement for cross-sectional studies.