Screening for atrial fibrillation to prevent strokes and systemic embolisms: modeling care potential and cost effectiveness

Screening for atrial fibrillation (AF) is recommended in the guidelines of the European Society of Cardiology (ESC) (Hindricks et al. 2020), but is currently not implemented in the form of systematic health programs in Germany. Given demographic changes and the age-dependent occurrence of AF, it is predicted that by 2060 the number of patients aged over 65 years with AF in Europe will increase to over 14 million (Krijthe et al. 2013). Since AF is a relevant risk factor for stroke, screening for AF will become increasingly important. In order to be able to assess the possible effects of introducing guideline-based screening for AF in the German health care context - in particular to avoid long-term health and health economic consequences - a systematic literature search was carried out in 2019. No evidence could be identified from completed, prospective studies in the German health care context that report long-term clinically relevant events after screening for AF. Against this background, a Markov model was developed with which the effects of a screening for AF can be simulated. This model is intended to contribute to the professional public discussion on the detection of AF and the prevention of thromboembolic events and to enable the derivation of adequate measures for better care. For this work, on the one hand, the long-term effects of consistent implementation of the screening measures provided for in the ESC guidelines with regard to the prevention of AF-associated strokes and systemic embolism were examined. On the other hand, the direct health economic implications were considered from the perspective of statutory health insurance (SHI).

>> A Markov model was developed consisting of a screening arm and a comparison arm (without screening), in which insured persons are observed over a period of 5 years (20 quarterly units).

Key points of modeling
The model population is derived from the GKV insured population using a prevalence-based approach. Based on the KM6 statistics (Federal Ministry of Health 2019), stratification according to age and gender (AGG) was carried out into 5-year age groups (up to 85+). The target group of insured people with previously unrecognized AF was determined by adjusting the insured population for the corresponding AGG-specific diagnosis prevalence (Ohlmeier et al. 2013). The proportion of insured people with undetected AF was determined based on literature information (Aronsson et al. 2015) on 32% of the diagnosis prevalence. In addition, the proportion of insured people with AF was characterized according to paroxysmal or persistent AF.
Based on the current ESC guidelines, in the screening arm, insured people aged 65 and over undergo an opportunistic screening and insured people aged 75 and over undergo a systematic screening approach. For the former, it was assumed that 65% of insured persons would be screened within 12 quarters. The systematic screening stipulates that 80% of the target group will be screened within 8 quarters. For both subpopulations, screening is assumed to be part of standard care; The screening services therefore take place as part of doctor's visits that are taking place anyway.
Both groups will undergo a one-time screening using 12-
Canal ECG undergone. For the diagnostic quality, a breakdown was made based on the overall sensitivity (Welton et al. 2017) and the patient proportions (Aronsson et al. 2015) for those insured by the German statutory health insurance (Federal Ministry of Health 2019). This results in a sensitivity of 100% for persistent or 44.15% for paroxysmal AF. The specificity was assumed to be 100% because the 12-lead ECG is the gold standard in AF diagnosis.
For patients with a diagnosis of AF, the model envisages the initiation of drug therapy with a vitamin K antagonist (VKA) or a direct oral anticoagulant (DOAC). According to the 2019 drug prescription report from the AOK Scientific Institute (Schwabe et al. 2019), prescription proportions were set at 32% VKA, 26% apixaban, 5% dabigatran, 28% rivaroxaban and 9% edoxaban.

Health effects
Clinically relevant events can occur at any time, regardless of whether screening is performed. The event rates depend on the presence of AF, the treatment (yes/no) and the anticoagulant used. The transition between the health stages is simulated using appropriate Markov iterations and takes into account, among other things, long-term therapy adherence (Collings et al. 2017).
Possible clinically relevant events include strokes and systemic embolisms, the prevention of which are central therapeutic goals of anticoagulation in AF. On the other hand, bleeding events can occur, for example, in the form of severe extra- and intracranial bleeding. Non-severe bleeding events were not considered due to the already complex model structure. The event rates used are based on German routine care (Hohnloser et al. 2018), with the calculation of missing event rates for edoxaban and placebo using rate ratios compared to warfarin from the meta-analysis by Tawfik et al. (2016) to the data on the VKA reference population in Hohnloser et al. (2018) took place (Table 1). Bleeding event rates were corrected for rates in the absence of OAC therapy. Mortality rates for events were derived based on literature data (Springer et al. 2018; Kolominsky-Rabas et al. 2006; Michalski et al. 2015).

Health economic effects
Screening costs, medical treatment costs, drug costs and direct costs of health outcomes were taken into account. For patients with an AF diagnosis, allocations from the morbidity-oriented risk structure adjustment (Morbi-RSA, as of 2020) (Federal Office for Social Security 2019a, 2019b) were taken into account. Due to the hierarchy of AF over hypertension, the proportion of approximately 87% of insured people with atrial fibrillation and corresponding comorbidities is relevant. For these patients was
only the allocation difference between hypertension and AF was taken into account.
If an ischemic stroke, a systemic embolism or an intracranial hemorrhage occurs, a permanently increased cost is assumed.
The cost of a stroke was calculated adjusted for inflation according to Kolominsky-Rabas et al. (2006). For intracranial hemorrhages and systemic embolisms, due to a lack of literature, the cost progression from ischemic stroke was transferred using a cost factor that was derived on the basis of the Diagnosis Related Groups (DRG) reimbursement for intracranial hemorrhages or systemic embolisms. For severe extracranial hemorrhages, purely inpatient treatment was assumed.
Table 2 lists the cost components considered in the model.
Results
Health effects of AF screening according to the guidelines
The determination of the model population to be screened is shown in Table 3.
In the modeling, 163,150 AF-associated strokes and 34,410 systemic embolisms occurred in the group without screening over a period of 5 years. With guideline-compliant screening for AF, the number of AF-associated strokes is reduced to 141,062 and the number of systemic embolisms to 29,873.
The results therefore show a potential of a total of 26,626 avoided events (including 22,088 strokes and 4,538 sys-
temic embolism) over 5 years. This compares to 3,841 additional intra- and extracranial hemorrhages. When looking at deaths after strokes, systemic embolisms and bleeding, there is a difference of 8,247 deaths avoided in the screening arm. Results of the model calculations are shown in Table 4.

Health economic effects of screening for AF according to the guidelines
When looking at the difference in all total costs after 5 years, there is a cost advantage of around 274.6 million euros compared to the group without screening.
The screening cost per identified patient with AF is 349 euros, with the total cost of carrying out the screening over 5 years being approximately 99.1 million euros. If AF is diagnosed, additional costs are incurred for drug therapy, which amount to approximately 355.3 million euros within the observation period. These additional costs are offset by savings in the care of clinically relevant events amounting to 755.1 million euros. This leads to savings at the level of individual health insurance companies amounting to 966 euros per identified insured person with AF.
If the allocations from the Morbi-RSA are taken into account, there will be shifts of around 1.4 billion euros over the 5 years. At the level of the individual health insurance company, this results in savings of 4,854 euros per insured person with an AF diagnosis.
An overview of the results can be found in Table 5.
discussion
The results presented are based on
the most comprehensive modeling study to date on the implementation of guideline-based screening for AF for stroke prevention in Germany. In this context, a screening approach with opportunistic screening for AF from 65 to 74 years of age and systematic screening from 75 years of age was simulated based on the ESC guidelines. This revealed a relevant potential for identifying pa-
patients with previously unmet treatment needs and to prevent ischemic strokes and systemic embolisms.
To ensure the highest possible quality and plausibility of the modeling results, systematic literature searches were carried out for the individual model parameters and the necessary calculations were carried out on an evidence-based basis. Due to the lack of completed, prospective studies in the German health care context on the health and economic effects of screening for AF, there have so far been no reliable results on the expected effects of systematic implementation. This study has now created a basis for professional public discussion.
On the clinical side, it was shown that if screening was introduced over 5 years and a therapy mix was adopted, a total of over 26,000 strokes and systemic embolisms could be prevented. It is not easy to classify the magnitude of the effect for the German healthcare context. However, it can be assumed that the estimated number of strokes prevented by improved AF screening affects over 10% of all AF-associated strokes (assuming that approximately 20% of the 260,000 annual strokes are AF-associated).
An underestimation of the true effects is likely to result from conservative assumptions regarding AF detection rates. On the one hand, a several times higher proportion of paroxysmal AF is assumed than can be observed with diagnosed AF. On the other hand, the sensitivity of the 12-channel ECG for paroxysmal AF was conservatively calculated as approximately 44.2%.
The results of the Strokestop study (Svennberg et al. 2021) also showed that screening for atrial fibrillation in the population of 75 and 76 year olds is beneficial in preventing strokes and systemic embolism. In 14,387 patients in the intervention arm (screening), 160 fewer events occurred than in the control arm of the same size (no screening). An extrapolation of these results to the approximately 10 million screened insured persons as part of this modeling confirms the deliberately conservative assumptions of the model.
On the cost side, the model results show a savings potential of 274.6 million euros over 5 years. Indirect costs are not taken into account because these can be ignored due to the age of the model population (65+ years). Uncertainty arises from the limited data available on the current costs of strokes and hardly any data on the costs of systemic embolisms and major bleeding over time. However, the magnitude of the model assumptions made corresponds to the information in a recent review by Düvel et al. (2021).
Taking into account the allocations from the Morbi-RSA, there is a net effect of 4,849 euros over 5 years for each insured person with VHF. It should be noted that the estimated allocations from the RSA are based on the status of 2020. The full cost model will be implemented in 2021, which will make transferring the results more difficult. However, it can be assumed that the compensation effects of the RSA will increase from 2021. While the amount of allocations for AF has fallen slightly, from 2021 the dominant relationship with hypertension, which affects 87% of insured people with AF, will no longer apply. The interpretation of the results should take into account the uncertainty resulting from the limited data availability on individual model parameters. A large number of parameters were varied as part of sensitivity analyses. The core results, particularly with regard to the avoidance of effectiveness events and the economic advantage through screening, were robust (see supplementary material online at [email protected]). Given the increasing relevance of AF, which is associated with demographic change, it is desirable to improve the evidence base in order to be able to make informed decisions in the healthcare context in the future.
conclusion for practice
The results show a significant potential of guideline-based screening for AF to prevent AF-associated strokes and systemic embolisms while at the same time saving costs for statutory health insurance. The full potential for improving care could be achieved through consistent implementation of the guideline recommendations. For a comprehensive and comprehensive detection of insured people with undetected AF, it would be conceivable to establish programs for regular and systematic examination for AF for insured people with an increased risk of stroke due to age or other risk factors.

Notes
• By consistently implementing guideline-compliant AF screening, a significant number of ischemic strokes and systemic embolisms can be prevented each year.
• Prevention of ischemic strokes and systemic embolisms is cost-saving and economical from a statutory health insurance perspective. <<