EDITORIAL
Europace (2014) 16, 1701–1702 doi:10.1093/europace/euu173
The atrial fibrillation epidemic: a validated diagnosis, or not? Jonas Bjerring Olesen 1, Gregory Y.H. Lip 2, and Laurent Fauchier3* 1 Department of Cardiology, Copenhagen University Hospital Gentofte, Hellerup 2900, Denmark; 2University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham B18 7QH, UK; and 3Service de Cardiologie, Poˆle Coeur Thorax Vasculaire, Centre Hospitalier Universitaire Trousseau et Faculte´ de Me´decine, Universite´ Franc¸ois Rabelais, Tours, France
Online publish-ahead-of-print 10 July 2014
This editorial refers to ‘Atrial fibrillation in patients with ischaemic stroke in the Swedish national patient registers: how much do we miss?’ by M.A. Baturova et al., on page 1714–1719.
The opinions expressed in this article are not necessarily those of the Editors of Europace or of the European Society of Cardiology.
* Corresponding author. E-mail address: [email protected] Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2014. For permissions please email: [email protected].
Downloaded from by guest on December 21, 2014
In Scandinavia, all residents have a personal permanent civil registration number which enables linkage of data on a personal level across registries, including cross-linkage of national health care data on, e.g. hospital diagnoses, claimed prescriptions, and causes of death. These data are often used for register-based epidemiological studies as they provide unique possibilities for investigating associations not accessible or feasible in clinical trials, which by nature have a selected population based on the respective trial’s inclusion/ exclusion criteria. Using the Swedish or Danish patient registries, many studies have been published on atrial fibrillation (AF).1,2 However, the validity of the registries and the used diagnoses have not always been validated prior to their use. Crudely speaking, it is well known that if garbage data come into the register, only garbage results will come out. Hence, validation studies are essential for the later interpretation of register-based observational cohort studies. In general, the prior studies of diagnosis validations have very much supported that the ‘all-comers’ (or ‘real world’) Swedish and Danish nationwide patient registries can be used for valid studies on diagnosis associations.3 It is important to underline that different levels of diagnosis accuracy are needed for different types of study design. Some attention is also needed to the inclusion/exclusion criteria of some observational cohort studies, which may confer selection bias. For example, one notable cohort study even excluded patients with no health plan membership after diagnosis of AF or those with no outpatient care during the 12 months after index date.4 For studies investigating a specific patient population, e.g. a cohort of patients with AF, the key issue should be a high positive predictive value for the diagnosis of AF, whereas the sensitivity of the AF diagnosis is less important as long as the patients with a registered diagnosis are representative. With outcome studies, for example, the outcome of stroke in a cohort of patients with AF, a high sensitivity
is much more important since even an equal distribution of underreporting will cause underestimations of event rates and a lower likelihood of finding important differences. There are also particular additional issues pertaining to AF. First, it is difficult for many physicians (particularly non cardiologists) to recognize that AF should be considered as a general ‘long term’ or chronic condition (i.e. a comorbidity when you had a history of episodes of AF a long time ago), and not only an acute condition. This is a major issue for any registry on AF based on administrative records, where long-term secondary diagnoses may not be coded. Indeed, we all know that an episode of paroxysmal AF 5 years ago should be treated on the long-term with antithrombotic therapy for most patients, in the presence of comorbidities. By contrast, the purpose of hospital discharge reports is most often to establish an evaluation for cost and reimbursement for the actual condition during hospital stay and administrators in charge of diagnostic coding may not keep the diagnosis of AF especially for a history of previous long-standing AF, as they would do for diabetes, kidney disease, coronary artery disease, etc.—particularly if new onset and a procedure is performed. Finally, the relevance of such large registries on AF may vary if one discusses clinical implications for antithrombotic management (which is relatively well established with simple algorithms in current guidelines), or for rhythm management which is still a matter of debate on many points and for which the clinical judgement plays a major role while being more difficult to capture with International classification of diseases codes or basic administrative records. In the current issue of Europace, Baturova et al. 5 investigated the validity of the diagnosis of AF in the Swedish Patient Register. The population used for the validation was 336 patients with first-time ischaemic stroke and 336 sex and gender-matched control subjects without stroke. Median age was 76 years and thus relevant for investigating patients with AF. For the patients with ischaemic stroke, the following results were observed: positive predictive value 85%, negative predictive value 89%, sensitivity 82% and specificity 91%. For the control subjects, who to a higher degree resembled the general patient in the Swedish Patient Register, the corresponding results
1702
only analysed a small population where one half had stroke and the other half were control subjects. Thus, their findings may not be true for the general Swedish Patient Register. The fact that no significant difference was found between the stroke patients and control subjects do not justify that their overall results are systematically generalized. The study also highlights that using AF as an outcome event should only be done with great caution since the timing of the diagnosis is not perfectly correlated with the first-time occurrence of the disease and sensitivity was only 76%. Finally, in the context of new antithrombotic therapies for patients with AF, many countries try to evaluate if the findings in randomized trials can be translated to all ‘real world’ AF patients. Analyses of national registries based on administrative records are probably the best tools to make such evaluation in view of the comprehensive information that they may provide. The study by Baturova et al. was performed in a specific population in Sweden and some of the results might not be extrapolated to other registries. It is possible that similar analysis in other settings and countries would have a similar reliability. Conflict of interest: none declared.
References 1. Friberg L, Rosenqvist M, Lip GY. Net clinical benefit of warfarin in patients with atrial fibrillation: a report from the Swedish atrial fibrillation cohort study. Circulation 2012; 125:2298 –307. 2. Olesen JB, Lip GY, Kamper AL, Hommel K, Kober L, Lane DA et al. Stroke and bleeding in atrial fibrillation with chronic kidney disease. N Engl J Med 2012;367:625 –35. 3. Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull 1999;46:263 –8. 4. Go AS, Hylek EM, Borowsky LH, Phillips KA, Selby JV, Singer DE. Warfarin use among ambulatory patients with nonvalvular atrial fibrillation: the anticoagulation and risk factors in atrial fibrillation (ATRIA) study. Ann Intern Med 1999;131:927 –34. 5. Baturova MA, Lindgren A, Carlson J, Shubik YV, Olsson SB, Platonov PG. Atrial fibrillation in patients with ischemic stroke in the Swedish national patient registers: how much do we miss? Europace 2014;16:1714 –9. 6. Frost L, Andersen LV, Vestergaard P, Husted S, Mortensen LS. Trend in mortality after stroke with atrial fibrillation. Am J Med 2007;120:47 –53. 7. Rix TA, Riahi S, Overvad K, Lundbye-Christensen S, Schmidt EB, Joensen AM. Validity of the diagnoses atrial fibrillation and atrial flutter in a Danish patient registry. Scand Cardiovasc J 2012;46:149 – 53. 8. Smith JG, Platonov PG, Hedblad B, Engstrom G, Melander O. Atrial fibrillation in the Malmo Diet and Cancer study: a study of occurrence, risk factors and diagnostic validity. Eur J Epidemiol 2010;25:95 –102. 9. Lowres N, Neubeck L, Salkeld G, Krass I, McLachlan AJ, Redfern J et al. Feasibility and cost-effectiveness of stroke prevention through community screening for atrial fibrillation using iPhone ECG in pharmacies. The SEARCH-AF study. Thromb Haemost 2014;111:1167 –76. 10. Lowres N, Neubeck L, Redfern J, Freedman SB. Screening to identify unknown atrial fibrillation. A systematic review. Thromb Haemost 2013;110:213 –22. 11. Banerjee A, Taillandier S, Olesen JB, Lane DA, Lallemand B, Lip GY et al. Pattern of atrial fibrillation and risk of outcomes: the Loire Valley Atrial Fibrillation Project. Int J Cardiol 2013;167:2682 –7.
Downloaded from by guest on December 21, 2014
were positive predictive value 74%, negative predictive value 95%, sensitivity 76% and specificity 94%. The positive predictive value of 74% is much lower than the values previously reported from the Danish Patient Register, where (for example) in 2007 Frost et al. found a positive predictive value of 99% using information from electrocardiograms, printouts from telemetry, Holter recordings and event recordings,6 while in 2012 Rix et al. found a positive predictive value of 92.6% using similar information.7 Notably, one of the co-authors from the Baturova et al. study published a study in 2010 reporting a positive predictive value of the AF diagnosis of 95% in the Swedish Patient Register.8 However, the difference between the predictive values found by Baturova et al. in the Swedish Patient Register and the previous Swedish and Danish results may be due to different validation techniques, for example, Baturova et al. only used information from electrocardiograms and cases with paroxysmal AF may have been missed. Indeed, screening for AF often yields many patients with previously undiagnosed AF, with a higher pickup rate if more prolonged monitoring is performed.9,10 The same can be said for associated comorbidities, where accuracy and completeness of diagnosis may vary. Indeed, the authors do not provide information on AF pattern, possibly paroxysmal, persistent or permanent. We do not know which pattern was the most accurately diagnosed, but it is very likely that permanent AF was easier to confirm on ECG records. By contrast, paroxysmal AF is not so easy to retrieve in ECG records, and a physician who diagnosed arrhythmia in an outpatient visit may have had the right diagnosis, while the authors had a wrong negative finding in their analysis. This once again may have more consequences if one considers implications for rhythm management than for antithrombotic management, the latter decision being unaffected by AF pattern or arrhythmia burden.11 Indeed, what is low AF burden today is not necessarily ‘low’ in the future. However, it was not the intention of the authors to make many statements concerning their ‘wrong’ AF diagnosis. Instead, the main finding of the paper was the significant underestimation of AF diagnosis in the patient registry, i.e. the lack of diagnosis while a diagnostic ECG was available. In that sense, the conclusion is reasonable when it states that we may underestimate prevalent and incident AF cases if only register data are used for identification of subjects with AF in epidemiology studies. We commend Baturova et al. for taking the important effort to validate one of the diagnoses so often used in the register-based studies. The validity of the diagnosis of AF was surprisingly low, and lower than previously reported, and this may be caused by a conservative validation technique. Certainly, Swedish studies on AF using the Swedish Patient Register should not be disregarded. The authors
Editorial
Europace (2014) 16, 1701–1702 doi:10.1093/europace/euu173
The atrial fibrillation epidemic: a validated diagnosis, or not? Jonas Bjerring Olesen 1, Gregory Y.H. Lip 2, and Laurent Fauchier3* 1 Department of Cardiology, Copenhagen University Hospital Gentofte, Hellerup 2900, Denmark; 2University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham B18 7QH, UK; and 3Service de Cardiologie, Poˆle Coeur Thorax Vasculaire, Centre Hospitalier Universitaire Trousseau et Faculte´ de Me´decine, Universite´ Franc¸ois Rabelais, Tours, France
Online publish-ahead-of-print 10 July 2014
This editorial refers to ‘Atrial fibrillation in patients with ischaemic stroke in the Swedish national patient registers: how much do we miss?’ by M.A. Baturova et al., on page 1714–1719.
The opinions expressed in this article are not necessarily those of the Editors of Europace or of the European Society of Cardiology.
* Corresponding author. E-mail address: [email protected] Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2014. For permissions please email: [email protected].
Downloaded from by guest on December 21, 2014
In Scandinavia, all residents have a personal permanent civil registration number which enables linkage of data on a personal level across registries, including cross-linkage of national health care data on, e.g. hospital diagnoses, claimed prescriptions, and causes of death. These data are often used for register-based epidemiological studies as they provide unique possibilities for investigating associations not accessible or feasible in clinical trials, which by nature have a selected population based on the respective trial’s inclusion/ exclusion criteria. Using the Swedish or Danish patient registries, many studies have been published on atrial fibrillation (AF).1,2 However, the validity of the registries and the used diagnoses have not always been validated prior to their use. Crudely speaking, it is well known that if garbage data come into the register, only garbage results will come out. Hence, validation studies are essential for the later interpretation of register-based observational cohort studies. In general, the prior studies of diagnosis validations have very much supported that the ‘all-comers’ (or ‘real world’) Swedish and Danish nationwide patient registries can be used for valid studies on diagnosis associations.3 It is important to underline that different levels of diagnosis accuracy are needed for different types of study design. Some attention is also needed to the inclusion/exclusion criteria of some observational cohort studies, which may confer selection bias. For example, one notable cohort study even excluded patients with no health plan membership after diagnosis of AF or those with no outpatient care during the 12 months after index date.4 For studies investigating a specific patient population, e.g. a cohort of patients with AF, the key issue should be a high positive predictive value for the diagnosis of AF, whereas the sensitivity of the AF diagnosis is less important as long as the patients with a registered diagnosis are representative. With outcome studies, for example, the outcome of stroke in a cohort of patients with AF, a high sensitivity
is much more important since even an equal distribution of underreporting will cause underestimations of event rates and a lower likelihood of finding important differences. There are also particular additional issues pertaining to AF. First, it is difficult for many physicians (particularly non cardiologists) to recognize that AF should be considered as a general ‘long term’ or chronic condition (i.e. a comorbidity when you had a history of episodes of AF a long time ago), and not only an acute condition. This is a major issue for any registry on AF based on administrative records, where long-term secondary diagnoses may not be coded. Indeed, we all know that an episode of paroxysmal AF 5 years ago should be treated on the long-term with antithrombotic therapy for most patients, in the presence of comorbidities. By contrast, the purpose of hospital discharge reports is most often to establish an evaluation for cost and reimbursement for the actual condition during hospital stay and administrators in charge of diagnostic coding may not keep the diagnosis of AF especially for a history of previous long-standing AF, as they would do for diabetes, kidney disease, coronary artery disease, etc.—particularly if new onset and a procedure is performed. Finally, the relevance of such large registries on AF may vary if one discusses clinical implications for antithrombotic management (which is relatively well established with simple algorithms in current guidelines), or for rhythm management which is still a matter of debate on many points and for which the clinical judgement plays a major role while being more difficult to capture with International classification of diseases codes or basic administrative records. In the current issue of Europace, Baturova et al. 5 investigated the validity of the diagnosis of AF in the Swedish Patient Register. The population used for the validation was 336 patients with first-time ischaemic stroke and 336 sex and gender-matched control subjects without stroke. Median age was 76 years and thus relevant for investigating patients with AF. For the patients with ischaemic stroke, the following results were observed: positive predictive value 85%, negative predictive value 89%, sensitivity 82% and specificity 91%. For the control subjects, who to a higher degree resembled the general patient in the Swedish Patient Register, the corresponding results
1702
only analysed a small population where one half had stroke and the other half were control subjects. Thus, their findings may not be true for the general Swedish Patient Register. The fact that no significant difference was found between the stroke patients and control subjects do not justify that their overall results are systematically generalized. The study also highlights that using AF as an outcome event should only be done with great caution since the timing of the diagnosis is not perfectly correlated with the first-time occurrence of the disease and sensitivity was only 76%. Finally, in the context of new antithrombotic therapies for patients with AF, many countries try to evaluate if the findings in randomized trials can be translated to all ‘real world’ AF patients. Analyses of national registries based on administrative records are probably the best tools to make such evaluation in view of the comprehensive information that they may provide. The study by Baturova et al. was performed in a specific population in Sweden and some of the results might not be extrapolated to other registries. It is possible that similar analysis in other settings and countries would have a similar reliability. Conflict of interest: none declared.
References 1. Friberg L, Rosenqvist M, Lip GY. Net clinical benefit of warfarin in patients with atrial fibrillation: a report from the Swedish atrial fibrillation cohort study. Circulation 2012; 125:2298 –307. 2. Olesen JB, Lip GY, Kamper AL, Hommel K, Kober L, Lane DA et al. Stroke and bleeding in atrial fibrillation with chronic kidney disease. N Engl J Med 2012;367:625 –35. 3. Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull 1999;46:263 –8. 4. Go AS, Hylek EM, Borowsky LH, Phillips KA, Selby JV, Singer DE. Warfarin use among ambulatory patients with nonvalvular atrial fibrillation: the anticoagulation and risk factors in atrial fibrillation (ATRIA) study. Ann Intern Med 1999;131:927 –34. 5. Baturova MA, Lindgren A, Carlson J, Shubik YV, Olsson SB, Platonov PG. Atrial fibrillation in patients with ischemic stroke in the Swedish national patient registers: how much do we miss? Europace 2014;16:1714 –9. 6. Frost L, Andersen LV, Vestergaard P, Husted S, Mortensen LS. Trend in mortality after stroke with atrial fibrillation. Am J Med 2007;120:47 –53. 7. Rix TA, Riahi S, Overvad K, Lundbye-Christensen S, Schmidt EB, Joensen AM. Validity of the diagnoses atrial fibrillation and atrial flutter in a Danish patient registry. Scand Cardiovasc J 2012;46:149 – 53. 8. Smith JG, Platonov PG, Hedblad B, Engstrom G, Melander O. Atrial fibrillation in the Malmo Diet and Cancer study: a study of occurrence, risk factors and diagnostic validity. Eur J Epidemiol 2010;25:95 –102. 9. Lowres N, Neubeck L, Salkeld G, Krass I, McLachlan AJ, Redfern J et al. Feasibility and cost-effectiveness of stroke prevention through community screening for atrial fibrillation using iPhone ECG in pharmacies. The SEARCH-AF study. Thromb Haemost 2014;111:1167 –76. 10. Lowres N, Neubeck L, Redfern J, Freedman SB. Screening to identify unknown atrial fibrillation. A systematic review. Thromb Haemost 2013;110:213 –22. 11. Banerjee A, Taillandier S, Olesen JB, Lane DA, Lallemand B, Lip GY et al. Pattern of atrial fibrillation and risk of outcomes: the Loire Valley Atrial Fibrillation Project. Int J Cardiol 2013;167:2682 –7.
Downloaded from by guest on December 21, 2014
were positive predictive value 74%, negative predictive value 95%, sensitivity 76% and specificity 94%. The positive predictive value of 74% is much lower than the values previously reported from the Danish Patient Register, where (for example) in 2007 Frost et al. found a positive predictive value of 99% using information from electrocardiograms, printouts from telemetry, Holter recordings and event recordings,6 while in 2012 Rix et al. found a positive predictive value of 92.6% using similar information.7 Notably, one of the co-authors from the Baturova et al. study published a study in 2010 reporting a positive predictive value of the AF diagnosis of 95% in the Swedish Patient Register.8 However, the difference between the predictive values found by Baturova et al. in the Swedish Patient Register and the previous Swedish and Danish results may be due to different validation techniques, for example, Baturova et al. only used information from electrocardiograms and cases with paroxysmal AF may have been missed. Indeed, screening for AF often yields many patients with previously undiagnosed AF, with a higher pickup rate if more prolonged monitoring is performed.9,10 The same can be said for associated comorbidities, where accuracy and completeness of diagnosis may vary. Indeed, the authors do not provide information on AF pattern, possibly paroxysmal, persistent or permanent. We do not know which pattern was the most accurately diagnosed, but it is very likely that permanent AF was easier to confirm on ECG records. By contrast, paroxysmal AF is not so easy to retrieve in ECG records, and a physician who diagnosed arrhythmia in an outpatient visit may have had the right diagnosis, while the authors had a wrong negative finding in their analysis. This once again may have more consequences if one considers implications for rhythm management than for antithrombotic management, the latter decision being unaffected by AF pattern or arrhythmia burden.11 Indeed, what is low AF burden today is not necessarily ‘low’ in the future. However, it was not the intention of the authors to make many statements concerning their ‘wrong’ AF diagnosis. Instead, the main finding of the paper was the significant underestimation of AF diagnosis in the patient registry, i.e. the lack of diagnosis while a diagnostic ECG was available. In that sense, the conclusion is reasonable when it states that we may underestimate prevalent and incident AF cases if only register data are used for identification of subjects with AF in epidemiology studies. We commend Baturova et al. for taking the important effort to validate one of the diagnoses so often used in the register-based studies. The validity of the diagnosis of AF was surprisingly low, and lower than previously reported, and this may be caused by a conservative validation technique. Certainly, Swedish studies on AF using the Swedish Patient Register should not be disregarded. The authors
Editorial
