Advertisement

Societal Costs of First-Incident Ischemic Stroke in Patients with Atrial Fibrillation—A Danish Nationwide Registry Study

Open ArchivePublished:March 24, 2016DOI:https://doi.org/10.1016/j.jval.2016.01.009

      Abstract

      Background

      Oral anticoagulation therapy reduces the risk of ischemic stroke in patients with atrial fibrillation (AF). However, more data on the costs of stroke in patients with AF are needed to assess how this therapy affects societal costs.

      Objectives

      The aim of the study was to estimate the average 3-year societal costs of first-incident ischemic stroke in Danish patients with AF, including costs of health care, social care services, and productivity loss.

      Methods

      The study was designed as an incidence-based cost-of-illness study covering the entire Danish population. All patients with a hospital diagnosis of AF were identified, and propensity score–matched analyses were used to estimate costs attributable to first-incident stroke among patients with AF in the period 2002 to 2012. All data were obtained from nationwide registries.

      Results

      A total of 21,673 patients with AF were identified with a first-incident stroke. The average 3-year costs attributable to stroke were US $30,925 per patient (present value) corresponding to US $19,989 in the incidence year and US $7,683 and US $5,176 1 and 2 years after the stroke, respectively. Health care accounted for 66% of the 3-year costs, with hospitalizations in the incidence year as the main cost driver. After the incidence year, costs of social care services exceeded health care costs. Sensitivity analyses showed that the cost estimates were relatively robust.

      Conclusions

      The societal costs of first-incident stroke in patients with AF are substantial. This new evidence can be valuable as an input for decision making regarding the treatment of AF and prevention of future strokes.

      Keywords

      Introduction

      Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia affecting 1% to 2% of the population [
      • Camm A.J.
      • et al.
      European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery
      Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
      ]. More than 6 million people in Europe and 5 million people in the United States currently suffer from AF [
      • Camm A.J.
      • et al.
      European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery
      Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
      ,
      • Colilla S.
      • Crow A.
      • Petkun W.
      • et al.
      Estimates of current and future incidence and prevalence of atrial fibrillation in the U.S. adult population.
      ], and projections suggest that the prevalence of AF will at least double by 2050 [
      • Camm A.J.
      • et al.
      European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery
      Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
      ,
      • Colilla S.
      • Crow A.
      • Petkun W.
      • et al.
      Estimates of current and future incidence and prevalence of atrial fibrillation in the U.S. adult population.
      ,
      • Savelieva I.
      • Camm J.
      Update on atrial fibrillation: part I.
      ].
      Cardioembolic strokes remain a major concern in relation to AF because AF is associated with a significantly elevated risk [
      • Ferro J.M.
      Cardioembolic stroke: an update.
      ,
      • Wolf P.A.
      • Abbott R.D.
      • Kannel W.B.
      Attrial fibrillation as an independent risk factor for stroke: the Framingham study.
      ]. Strokes due to cardioembolism account for about one-fifth of all ischemic strokes and are generally more severe than noncardioembolic strokes [
      • Ferro J.M.
      Cardioembolic stroke: an update.
      ,
      • Steinberg B.A.
      • Piccini J.P.
      Anticoagulation in atrial fibrillation.
      ,
      • Savelieva I.
      • Camm J.
      Update on atrial fibrillation: part II.
      ]. Continuous oral anticoagulation therapy has been shown to reduce stroke incidence [
      • Camm A.J.
      • et al.
      European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery
      Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
      ].
      Because the prevalence of AF is rising and budgets are under pressure, it is becoming increasingly critical to take costs into account when choosing among different treatment options, including oral anticoagulation therapy. Numerous studies on the costs of stroke exist, but few studies have examined the costs of stroke in patients with AF. We found only two large studies on the costs of stroke in patients with AF, and these studies included only hospital costs [
      • Ghatnekar O.
      • Glader E.L.
      The effect of atrial fibrillation on stroke-related inpatient costs in Sweden: a 3-year analysis of registry incidence data from 2001.
      ,
      • Wang G.
      • Joo H.
      • Tong X.
      • et al.
      Hospital costs associated with atrial fibrillation for patients with ischemic stroke aged 18-64 years in the United States.
      ]. Previous studies have shown that the costs of stroke are higher in patients with AF than in patients without AF [
      • Ghatnekar O.
      • Glader E.L.
      The effect of atrial fibrillation on stroke-related inpatient costs in Sweden: a 3-year analysis of registry incidence data from 2001.
      ,
      • Wang G.
      • Joo H.
      • Tong X.
      • et al.
      Hospital costs associated with atrial fibrillation for patients with ischemic stroke aged 18-64 years in the United States.
      ,
      • Brüggenjürgen B.
      • Rossnagel K.
      • Roll S.
      • et al.
      The impact of atrial fibrillation on the cost of stroke: the Berlin acute stroke study.
      ,
      • Hannon N.
      • Daly L.
      • Murphy S.
      • et al.
      Acute hospital, community, and indirect costs of stroke associated with atrial fibrillation: population-based study.
      ]. Additional studies are needed to ensure that valid cost estimates can be used to strengthen the foundation for evidence-based decision making regarding the treatment of AF and prevention of future strokes.
      Our aim was to estimate the 3-year societal costs attributable to first-incident ischemic stroke in patients with AF, including costs of health care, social care services, and productivity loss.

      Methods

      The study was designed as a nationwide registry-based cost-of-illness study from a societal perspective. We used an incidence-based approach and propensity score–matched analyses [
      • Rosenbaum P.R.
      • Rubin D.B.
      Constructing a control group using multivariate matched sampling methods that incorporate the propensity score.
      ] to estimate costs attributable to first-incident stroke among patients with AF in the period 2002 to 2012.

      Registry Data Sources

      The study was based on national Danish registries covering the entire population, which makes this a study at the population-level of a whole country. Every Danish resident has a permanent and personal registration number that enables linkage between registries at the individual level. Admissions and outpatient visits to hospitals are registered in the Danish National Patient Registry, with primary and secondary diagnoses coded according to the International Classification of Diseases, 10th revision (ICD-10) [
      • Lynge E.
      • Sandegaard J.L.
      • Rebolj M.
      The Danish National Patient Register.
      ]. Contacts to private practice health care professionals covered by national health insurance are registered in the National Health Insurance Service Registry [
      • Andersen J.S.
      • Olivarius Nde F.
      • Krasnik A.
      The Danish National Health Service Register.
      ]. Both the Danish National Patient Registry and the National Health Insurance Service Registry are used for payment purposes and the quality of data is considered to be high. All prescriptions dispensed from Danish pharmacies are registered in the Danish Registry of Medical Products Statistics (Prescription Registry) using the international therapeutic chemical classification system (Anatomical Therapeutic Chemical) [
      • Kildemoes H.W.
      • Sørensen H.T.
      • Hallas J.
      The Danish National Prescription Registry.
      ]. The Civil Registration System holds information on sex, age, residence, and vital status [
      • Pedersen C.B.
      The Danish Civil Registration System.
      ]. Furthermore, data on home help, education, labor market affiliation, and income are available from Statistics Denmark [
      • Baadsgaard M.
      • Quitzau J.
      Danish registers on personal income and transfer payments.
      ,
      • Jensen V.M.
      • Rasmussen A.W.
      Danish Education Registers.
      ,
      • Petersson F.
      • Baadsgaard M.
      • Thygesen L.C.
      Danish registers on personal labour market affiliation.
      ]. Data from different registries were linked together at the patient level using the personal registration number that all Danish residents have.

      Study Population

      Our basis population was identified from the National Patient Registry and consisted of all patients hospitalized in the period 1994 to 2012 with AF as the primary or secondary diagnosis (ICD-10 code I48) (see Fig. 1). We excluded patients who had been hospitalized with ischemic stroke (ICD-10 codes I63 and I64) in the period 1994 to 2001 to restrict the analysis to incident cases. For each of the following years (2002–2012), we identified the stroke group as patients who had been hospitalized with ischemic stroke as the primary or secondary diagnosis, conditional on them having been diagnosed with AF. For each year, we identified potential controls (the control reservoir) as the remaining patients with AF in the basis population. We censored patients at death. Furthermore, we excluded controls from the control group if they were hospitalized with ischemic stroke (ICD-10 codes I63 and I64) after the incidence year. From this year on, they were part of the stroke group. This was done to avoid contamination of the control group.
      Figure thumbnail gr1
      Fig. 1Flow chart. AF, atrial fibrillation. (Color version of figure available online).

      Costs

      The costs attributable to stroke were estimated using the matched control/regression approach, which is considered the criterion standard method of cost-of-illness studies [
      • Akobundu E.
      • Ju J.
      • Blatt L.
      • et al.
      Cost-of-illness studies: a review of current methods.
      ]. We estimated costs for a 3-year period, counting from the year in which the stroke occurred (the incidence year). We calculated attributable costs as standardized costs incurred by patients in the stroke group minus standardized costs incurred by patients in the control group. Standardized costs were calculated as the costs incurred by patients in year t after the incidence year (t = 0, 1, 2) minus the costs incurred in the year before the stroke (the baseline year). Costs incurred in the baseline year were subtracted, because it was not possible to isolate diagnosis-specific costs when including costs outside the hospital. This was also the strategy used for the control group. A similar approach has been used in other cost-of-illness studies [
      • Luengo-Fernandez R.
      • Yiin G.S.
      • Gray A.M.
      • et al.
      Population-based study of acute- and long-term care costs after stroke in patients with AF.
      ,
      • Porsdal V.
      • Boysen G.
      Direct costs during the first year after intracerebral hemorrhage.
      ].
      Costs of health care included costs of hospitalizations, outpatient and emergency visits to hospitals, contacts to private practice health care professionals covered by national health insurance, and prescription medicine dispensed from Danish pharmacies. Hospital resource use and contacts to private practice health care professionals were priced according to tariffs effective in the year of delivery. Data on hospital tariffs were available only until 2011. Therefore, costs were estimated for the stroke groups from 2002 to 2011. Prescription medicine was priced by the pharmacy, including both the share covered by the national health insurance and patient co-payment.
      Costs of social care included costs of home help and nursing homes. Data were obtained from Statistics Denmark. Home help and nursing homes were priced by average tariffs effective in 2013.
      The productivity loss was estimated using the human capital approach, which is the most commonly used method for valuation of indirect costs [
      • Drummond M.F.
      • O’Brien B.
      • Stoddart G.L.
      • et al.
      Methods for the Economic Evaluation of Health Care Programmes.
      ,
      • Krol M.
      • Brouwer W.
      • Rutten F.
      Productivity costs in economic evaluations: past, present, future.
      ]. We used data from Statistics Denmark on income from employment to estimate the value of lost earnings of patients who were 18 to 65 years old. We assumed that patients older than 65 years were not engaged in active employment because 65 years is the normal retirement age in Denmark.
      The 3-year costs were calculated as the present value of the attributable costs in year t (t = 0, 1, 2) using a discount rate of 4% as currently recommended by the Danish Ministry of Finance [

      Danish Ministry of Finance. New and lower socio-economic discount rate. 2013. Available from: http://www.fm.dk/nyheder/pressemeddelelser/2013/05/ny-og-lavere-samfundsoekonomisk-diskonteringsrente. [Accessed October 30, 2015].

      ]. All costs were inflated to 2012 prices and converted to US dollars (US $) based on the average exchange rate in 2012 (US $ 100 = 579.72 DKK).

      Statistical Analysis

      We used nearest neighbor propensity score matching with replacement to select four controls for each patient in the stroke groups [
      • Leuven E.
      • Sianesi B.
      PSMATCH2: Stata Module to Perform Full Mahalanobis and Propensity Score Matching, Common Support Graphing, and Covariate Imbalance Testing (Statistical Software Components S 432001).
      ]. The propensity score was based on the following matching criteria: age, sex, comorbidity (measured by the Charlson index), and risk of stroke (measured by the CHA2DS2-VASc score) at baseline. The matching criteria were chosen to identify clinically comparable groups. The Charlson index was calculated using information from the National Patient Registry on diagnoses related to admissions and outpatient visits to hospitals 5 years before the incidence year [
      • Charlson M.E.
      • Pompei P.
      • Ales K.L.
      • et al.
      A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
      ]. The CHA2DS2-VASc score was calculated using information from the National Patient Registry and the Prescription Registry on patient’s age, sex, history of stroke and transient ischemic attack, hypertension, congestive heart failure, vascular disease, and diabetes in the baseline year [
      • Olesen J.B.
      • Lip G.Y.
      • Hansen M.L.
      • et al.
      Validation of risk stratification schemes for predicting stroke and thromboembolism in patients with atrial fibrillation: nationwide cohort study.
      ]. An absolute standardized difference of less than 10% and a variance ratio ranging between 0.8 and 1.25 were considered to support the assumption of balance between groups [
      • Austin P.C.
      An introduction to propensity score methods for reducing the effects of confounding in observational studies.
      ].
      Average attributable costs were estimated in a differences-in-differences regression model as β1 (see Table 1). T tests were used to investigate whether cost estimates were significantly different from zero. The statistical significance level was set at P < 0.05.
      Table 1Differences-in-differences regression model
      ΔCti=β0+β1Di+β2X1i+β3X2i++εi
      where:
      ΔCti=Cti-C-1i
      Cti=costs of patient i in year t
      C-1i=costs of patient i in the baseline year
      Di=dummy variable indicating whether the patient belongs to the stroke or control group
      Xi=covariates (see table 1)
      t=year 0,1 and 2 after the stroke
      Furthermore, sensitivity analyses were performed. First, we investigated how the cost estimates were affected if only patients in oral anticoagulation therapy were included. This group of patients was identified as patients who had redeemed at least two prescriptions in the baseline year with the Anatomical Therapeutic Chemical code B01AA (vitamin K antagonist, including warfarin) or B01AE07 (dabigatran), which were the only oral anticoagulation therapies on the Danish market until 2011. Second, we analyzed the robustness of the health care cost estimates if only patients who had been hospitalized with stroke as primary diagnosis were included. Third, we investigated the consequences of selecting another baseline year. Fourth, we analyzed changes in the health care cost estimates if only patients who survived the incidence year were included. Finally, we analyzed cost estimates beyond the 3-year period.
      Analyses were performed using the statistical package SAS 9.3 for Windows (SAS Institute, Cary, NC) and Stata 13.1 (StataCorp, College Station, TX).

      Ethical Issues

      The study was approved by the Danish Data Protection Agency (REF 2013–54–0410) and the National Health Surveillance and Research (REF 0036011). Approval from an ethics committee was not required by Danish law because the study was based on registry data only.

      Results

      We identified 21,673 patients with AF with first-incident ischemic stroke in the period 2002 to 2012, corresponding to an incidence rate of 1% to 2% per year (see Fig. 1). Of these, 26% were in oral anticoagulation therapy in the year before the stroke. The percentage of patients with stroke in prior anticoagulation therapy increased from 21% in 2002 to 28% in 2012. Among all patients with AF in the basis population, the percentage in anticoagulation therapy increased from 12% in 2002 to 43% in 2012.
      Table 2 presents background characteristics of the study population. There were no significant imbalances between patients in the stroke groups and the matched controls with regard to the covariates used in the propensity score matching (see Fig. 2). The imbalances for other covariates were also within the boundaries considered to support the assumption of balance between groups.
      Table 2Baseline characteristics of the study population
      CharacteristicStroke groupMatched controls
      Sex: female, %5454
      Average age (y)7878
      Age group (y), %
       <65 y1010
       65–74 y2121
       ≥75 y6969
      Charlton index, %
       05253
       1-23635
       ≥21212
      CHA2Ds2 VASc score, average3.013.00
      Education, %
       Primary and lower secondary school4039
       Higher secondary school and vocational training2324
       Higher education1011
       Missing2626
      Labor market affiliation, %
       Wage earner55
       Self-employed33
       Unemployed11
       Retired9089
       Other11
       Missing12
      Annual income, average (US $)26,85127,770
      Figure thumbnail gr2
      Fig. 2Quality of propensity score matching. (Color version of figure available online).
      The estimated average 3-year societal costs attributable to stroke were US $30,925 per patient (present value) corresponding to US $19,989 in the incidence year, US $7,683 in the year after, and US $5,176 2 years after the stroke (see Table 3). Health care costs accounted for 66% of the 3-year costs, whereas social care costs and productivity loss accounted for 26% and 8%, respectively. Most of the health care costs were due to hospitalizations in the incidence year. Social care costs and productivity loss exceeded health care costs 1 and 2 years after the stroke, respectively.
      Table 3Average attributable costs per patient, US $ (2012 prices)
      Types of costsTotal 3-y attributable costs (present value)Attributable cost estimates
      The significance of cost estimates related to hospital care, contacts to private practice health professionals, prescribed medicine, home help, nursing homes, and productivity loss was tested using t tests (H0: µ = 0 against the hypothesis H1: µ ≠ 0.). This is equivalent to testing whether standardized costs per patient in the stroke groups were significantly different from standardized costs per patient in the control group.
      Year 0 (incidence year)Year 1 after the strokeYear 2 after the stroke
      Direct costs
       Health care20,50717,7963354331
        Inpatient hospital care19,06617,100
      The null hypothesis was rejected (P < 0.001).
      2826
      The null hypothesis was rejected (P < 0.001).
      12
        Outpatient hospital care412532
      The null hypothesis was rejected (P < 0.001).
      63−177
      The null hypothesis was rejected (P < 0.05).
        Private practicing health professionals668124
      The null hypothesis was rejected (P < 0.001).
      297
      The null hypothesis was rejected (P < 0.001).
      308
      The null hypothesis was rejected (P < 0.001).
        Prescribed medicine36140
      The null hypothesis was rejected (P < 0.001).
      168
      The null hypothesis was rejected (P < 0.001).
      187
      The null hypothesis was rejected (P < 0.001).
       Social care services8,0801,9053,4853403
        Home help3,952699
      The null hypothesis was rejected (P < 0.001).
      1704
      The null hypothesis was rejected (P < 0.001).
      1917
      The null hypothesis was rejected (P < 0.001).
        Nursing home4,1271,206
      The null hypothesis was rejected (P < 0.001).
      1781
      The null hypothesis was rejected (P < 0.001).
      1486
      The null hypothesis was rejected (P < 0.001).
      Indirect costs
       Productivity loss2,338288
      The null hypothesis was rejected (P < 0.01).
      843
      The null hypothesis was rejected (P < 0.001).
      1442
      The null hypothesis was rejected (P < 0.001).
      Direct and indirect costs30,92519,98976835176
      low asterisk The significance of cost estimates related to hospital care, contacts to private practice health professionals, prescribed medicine, home help, nursing homes, and productivity loss was tested using t tests (H0: µ = 0 against the hypothesis H1: µ ≠ 0.). This is equivalent to testing whether standardized costs per patient in the stroke groups were significantly different from standardized costs per patient in the control group.
      The null hypothesis was rejected (P < 0.001).
      The null hypothesis was rejected (P < 0.05).
      § The null hypothesis was rejected (P < 0.01).
      Hospital cost estimates were statistically significant at a 5% level in the incidence year and the year after the stroke, whereas social care cost estimates and the estimated productivity loss were also statistically significant 2 years after.
      Sensitivity analyses showed that health care cost estimates were relatively robust (Table 4). The 3-year costs were essentially unchanged when estimated for the subgroup of patients with AF in oral anticoagulation therapy. When only the patients with AF who were diagnosed with stroke as the primary diagnosis were included, the cost estimates increased because these strokes are generally more severe. The cost estimates also increased when only the patients who were alive after the incidence year were included, and if the baseline year was 2 years before the incidence year. The latter was a consequence of increasing average costs incurred by patients in the stroke group in the years before the stroke, maybe due to warning symptoms. The cost estimates decreased when estimated over a 5- and 10-year period because standardized costs in the control group generally exceeded standardized costs in the stroke group from 3 years after the stroke and onward. However, the 5- and 10-year attributable cost estimates were still positive. Focus has been on the 3-year costs because the longer-term cost estimates were generally not statistically significant, and the comparability between the stroke and control groups decreased over time because of different mortality rates and so on.
      Table 4Results of sensitivity analysis
      ScenarioAttributable health care costs (US $)
      Base scenario20,507
      Only patients in oral anticoagulation therapy21,313
      Only patients with ischemic stroke as primary diagnosis26,221
      Only patients who survive the incidence year21,527
      Baseline year 2 y before the incidence year21,818
      5-y time horizon19,763
      10-y time horizon13,612

      Discussion

      Our study provides an estimate of the 3-year societal costs attributable to first-incident ischemic stroke in patients with AF, including costs of health care, social care, and productivity loss. We found that the cost estimates were essentially unchanged when estimated for the subgroup of patients with AF in oral anticoagulation therapy. However, use of oral anticoagulation therapies may still reduce net societal costs by reducing the number of strokes experienced by patients with AF.
      Few studies on the costs of stroke in patients with AF exist to which we can compare our findings. Ghatnekar and Glader [
      • Ghatnekar O.
      • Glader E.L.
      The effect of atrial fibrillation on stroke-related inpatient costs in Sweden: a 3-year analysis of registry incidence data from 2001.
      ] investigated the costs of stroke of 1619 patients with AF hospitalized with first-incident stroke in Sweden. They estimated the average 3-year inpatient hospital costs to be US $11,917 in 2001 prices. Wang et al. [
      • Wang G.
      • Joo H.
      • Tong X.
      • et al.
      Hospital costs associated with atrial fibrillation for patients with ischemic stroke aged 18-64 years in the United States.
      ] estimated the hospital costs associated with acute ischemic stroke among insured patients with and without AF aged 18 to 64 years in the United States [
      • Wang G.
      • Joo H.
      • Tong X.
      • et al.
      Hospital costs associated with atrial fibrillation for patients with ischemic stroke aged 18-64 years in the United States.
      ]. For nonrepeat stroke admissions, they estimated the average costs for patients with AF to be US $23,770 in 2012 prices, based on claims data on 2,407 hospital admissions. Luengo-Fernandez et al. [
      • Luengo-Fernandez R.
      • Yiin G.S.
      • Gray A.M.
      • et al.
      Population-based study of acute- and long-term care costs after stroke in patients with AF.
      ] investigated the acute and longer-term health care costs of stroke in 153 patients with AF in the United Kingdom [
      • Luengo-Fernandez R.
      • Yiin G.S.
      • Gray A.M.
      • et al.
      Population-based study of acute- and long-term care costs after stroke in patients with AF.
      ]. Costs were estimated for the first 90 days after stroke (acute period) and for the period following the acute period until 5 years after the event. Acute health care costs amounted to US $16,496, whereas annual postacute health care costs for the patients surviving past the acute period were US $1,274 in 2008/2009 prices. Finally, Brüggenjürgen et al. [
      • Brüggenjürgen B.
      • Rossnagel K.
      • Roll S.
      • et al.
      The impact of atrial fibrillation on the cost of stroke: the Berlin acute stroke study.
      ] estimated the average 1-year costs of stroke to be US $15,150 in 2002 prices based on 71 patients with AF admitted to four German hospitals in the period 2000 to 2001, including costs of health care and nursing homes. The cost estimates of our study are higher than the estimates of the studies described above.
      However, our cost estimates are slightly lower than the estimates made by Yiin et al. [
      • Yiin G.S.
      • Howard D.P.
      • Paul N.L.
      • et al.
      Age-specific incidence, outcome, cost, and projected future burden of atrial fibrillation-related embolic vascular events: a population-based study.
      ], who performed a prospective population-based study of strokes associated with AF in the United Kingdom [
      • Yiin G.S.
      • Howard D.P.
      • Paul N.L.
      • et al.
      Age-specific incidence, outcome, cost, and projected future burden of atrial fibrillation-related embolic vascular events: a population-based study.
      ]. They estimated the average 1-year hospital costs and 5-year residential care costs incurred by 190 patients with incident AF-related ischemic stroke to amount to a total of US $36,184 in 2008/2009 prices. The hospital and residential care costs amounted to US $20,037 and US $16,148, respectively. Moreover, our cost estimates are significantly lower than the estimates made by Hannon et al. [
      • Hannon N.
      • Daly L.
      • Murphy S.
      • et al.
      Acute hospital, community, and indirect costs of stroke associated with atrial fibrillation: population-based study.
      ], who investigated the costs of stroke associated with AF among 177 patients in Ireland. They estimated the average 2-year costs associated with AF-strokes to be US $76,602 in 2007 prices, including acute hospital, nursing home, community health care, and indirect costs.
      Differences in study design make it difficult to compare cost estimates between studies. We believe that our study has a number of advantages compared with existing studies. Most importantly, we used a controlled design and propensity score matching to minimize the risk of confounding by age, sex, comorbidity, stroke risk, and socioeconomic characteristics. None of the existing studies used a controlled design and may therefore have overestimated the costs attributable to stroke. Another important strength of our study is the use of national registry data, which renders our study less prone to selection and information bias because the Danish registries cover the entire population, because data are prospectively registered, and because the quality of data is considered to be high. Finally, our study was based on a large data set covering a 10-year period, making the cost estimates more robust.
      The inclusion of social care costs and productivity loss—additional to health care costs—was also a key strength of our study. However, not all relevant costs were included, for example, costs of rehabilitation, assistive devices, and home modifications financed by the municipalities and informal care costs, because these data were not available from the national registries. Therefore, our cost estimates remain conservative.
      We acknowledge that our study also has limitations. First, we only included those patients with AF who had been hospitalized with stroke as primary or secondary diagnosis in our study. Thus, our study provides an estimate of the costs for this subgroup of strokes only. Stroke patients treated in the community who were not hospitalized probably incurred less costs. However, because stroke patients who were not hospitalized are rare in Denmark, this problem is not regarded as a major concern. A Danish population-based study shows that only 12% of all stroke events are managed at home or in institutions other than hospitals [
      • Thorvaldsen P.
      • Davidsen M.
      • Brønnum-Hansen H.
      • et al.
      Stable stroke occurrence despite incidence reduction in an aging population: stroke trends in the Danish monitoring trends and determinants in cardiovascular disease (MONICA) population.
      ].
      Second, even though the quality of the data in the Danish national registries is considered to be high, there is a risk of misclassification. Most importantly, this study relies on the ICD-10 coding of AF and strokes in the National Patient Registry, which may not be entirely accurate. However, validation studies have reported high predictive values, with less than 10% to 15% of the patients being misclassified [
      • Johnsen S.P.
      • Overvad K.
      • Sørensen H.T.
      • et al.
      Predictive value of stroke and transient ischemic attack discharge diagnoses in The Danish National Registry of Patients.
      ,
      • Rix T.A.
      • Riahi S.
      • Overvad K.
      • et al.
      Validity of the diagnoses atrial fibrillation and atrial flutter in a Danish patient registry.
      ]. Furthermore, the Building Regulation Registry—which was used to estimate costs of nursing homes—may contain errors, because it is not audited whether the information reported to the registry by owners of buildings (e.g., municipalities as owners of nursing homes) is correct and up-to-date. Because possible misclassification is most likely nondifferential, this may have biased the results of this study in a conservative direction.
      Third, we cannot rule out confounding due to the observational nature of the study even though the risk of confounding was minimized by a well-balanced propensity score–matched design.

      Conclusions

      Our study is the first large study to provide an estimate of the societal costs attributable to first-incident ischemic stroke in patients with AF, including health care costs, social care costs, and productivity loss. This new evidence may be valuable as an input for economic evaluation and decision making regarding the use of anticoagulation therapy and prevention of future strokes in patients with AF.

      Acknowledgments

      This study was sponsored by Pfizer Denmark Aps and Bristol-Myers Squibb Denmark. We acknowledge Katja Lundberg Rand for insightful and valuable comments and suggestions to the manuscript during its preparation.

      References

        • Camm A.J.
        • et al.
        • European Heart Rhythm Association, European Association for Cardio-Thoracic Surgery
        Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC).
        Eur Heart J. 2010; 31: 2369-2429
        • Colilla S.
        • Crow A.
        • Petkun W.
        • et al.
        Estimates of current and future incidence and prevalence of atrial fibrillation in the U.S. adult population.
        Am J Cardiol. 2013; 112: 1142-1147
        • Savelieva I.
        • Camm J.
        Update on atrial fibrillation: part I.
        Clin Cardiol. 2008; 31: 55-62
        • Ferro J.M.
        Cardioembolic stroke: an update.
        Lancet Neurol. 2003; 2: 177-188
        • Wolf P.A.
        • Abbott R.D.
        • Kannel W.B.
        Attrial fibrillation as an independent risk factor for stroke: the Framingham study.
        Stroke. 1991; 22: 983-988
        • Steinberg B.A.
        • Piccini J.P.
        Anticoagulation in atrial fibrillation.
        BMJ. 2014; 348: g2116
        • Savelieva I.
        • Camm J.
        Update on atrial fibrillation: part II.
        Clin Cardiol. 2008; 31: 102-108
        • Ghatnekar O.
        • Glader E.L.
        The effect of atrial fibrillation on stroke-related inpatient costs in Sweden: a 3-year analysis of registry incidence data from 2001.
        Value Health. 2008; 11: 862-868
        • Wang G.
        • Joo H.
        • Tong X.
        • et al.
        Hospital costs associated with atrial fibrillation for patients with ischemic stroke aged 18-64 years in the United States.
        Stroke. 2015; 46: 1314-1320
        • Brüggenjürgen B.
        • Rossnagel K.
        • Roll S.
        • et al.
        The impact of atrial fibrillation on the cost of stroke: the Berlin acute stroke study.
        Value Health. 2007; 10: 137-143
        • Hannon N.
        • Daly L.
        • Murphy S.
        • et al.
        Acute hospital, community, and indirect costs of stroke associated with atrial fibrillation: population-based study.
        Stroke. 2014; 45: 3670-3674
        • Rosenbaum P.R.
        • Rubin D.B.
        Constructing a control group using multivariate matched sampling methods that incorporate the propensity score.
        Am Stat. 1985; 39: 33-38
        • Lynge E.
        • Sandegaard J.L.
        • Rebolj M.
        The Danish National Patient Register.
        Scand J Public Health. 2011; 39: 30-33
        • Andersen J.S.
        • Olivarius Nde F.
        • Krasnik A.
        The Danish National Health Service Register.
        Scand J Public Health. 2011; 39: 34-37
        • Kildemoes H.W.
        • Sørensen H.T.
        • Hallas J.
        The Danish National Prescription Registry.
        Scand J Public Health. 2011; 39: 38-41
        • Pedersen C.B.
        The Danish Civil Registration System.
        Scand J Public Health. 2011; 39: 22-25
        • Baadsgaard M.
        • Quitzau J.
        Danish registers on personal income and transfer payments.
        Scand J Public Health. 2011; 39: 103-105
        • Jensen V.M.
        • Rasmussen A.W.
        Danish Education Registers.
        Scand J Public Health. 2011; 39: 91-94
        • Petersson F.
        • Baadsgaard M.
        • Thygesen L.C.
        Danish registers on personal labour market affiliation.
        Scand J Public Health. 2011; 39: 95-98
        • Akobundu E.
        • Ju J.
        • Blatt L.
        • et al.
        Cost-of-illness studies: a review of current methods.
        Pharmacoeconomics. 2006; 24: 869-890
        • Luengo-Fernandez R.
        • Yiin G.S.
        • Gray A.M.
        • et al.
        Population-based study of acute- and long-term care costs after stroke in patients with AF.
        Int J Stroke. 2013; 8: 308-314
        • Porsdal V.
        • Boysen G.
        Direct costs during the first year after intracerebral hemorrhage.
        Eur J Neurol. 1999; 6: 449-454
        • Drummond M.F.
        • O’Brien B.
        • Stoddart G.L.
        • et al.
        Methods for the Economic Evaluation of Health Care Programmes.
        2nd ed. Oxford University Press, New York1997
        • Krol M.
        • Brouwer W.
        • Rutten F.
        Productivity costs in economic evaluations: past, present, future.
        Pharmacoeconomics. 2013; 31: 537-549
      1. Danish Ministry of Finance. New and lower socio-economic discount rate. 2013. Available from: http://www.fm.dk/nyheder/pressemeddelelser/2013/05/ny-og-lavere-samfundsoekonomisk-diskonteringsrente. [Accessed October 30, 2015].

        • Leuven E.
        • Sianesi B.
        PSMATCH2: Stata Module to Perform Full Mahalanobis and Propensity Score Matching, Common Support Graphing, and Covariate Imbalance Testing (Statistical Software Components S 432001).
        Boston College Department of Economics, Boston2003
        • Charlson M.E.
        • Pompei P.
        • Ales K.L.
        • et al.
        A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
        J Chronic Dis. 1987; 40: 373-383
        • Olesen J.B.
        • Lip G.Y.
        • Hansen M.L.
        • et al.
        Validation of risk stratification schemes for predicting stroke and thromboembolism in patients with atrial fibrillation: nationwide cohort study.
        BMJ. 2011; 342: d124
        • Austin P.C.
        An introduction to propensity score methods for reducing the effects of confounding in observational studies.
        Multivariate Behav Res. 2011; 46: 399-424
        • Yiin G.S.
        • Howard D.P.
        • Paul N.L.
        • et al.
        Age-specific incidence, outcome, cost, and projected future burden of atrial fibrillation-related embolic vascular events: a population-based study.
        Circulation. 2014; 130: 1236-1244
        • Thorvaldsen P.
        • Davidsen M.
        • Brønnum-Hansen H.
        • et al.
        Stable stroke occurrence despite incidence reduction in an aging population: stroke trends in the Danish monitoring trends and determinants in cardiovascular disease (MONICA) population.
        Stroke. 1999; 30: 2529-2534
        • Johnsen S.P.
        • Overvad K.
        • Sørensen H.T.
        • et al.
        Predictive value of stroke and transient ischemic attack discharge diagnoses in The Danish National Registry of Patients.
        J Clin Epidemiol. 2002; 55: 602-607
        • Rix T.A.
        • Riahi S.
        • Overvad K.
        • et al.
        Validity of the diagnoses atrial fibrillation and atrial flutter in a Danish patient registry.
        Scand Cardiovasc J. 2012; 46: 149-153