Budget impacts associated with improving diagnosis and treatment of atrial fibrillation in high-risk stroke patients

The AHSN AF Business Case Model evaluates the budget impact of addressing identified and modeled gaps in diagnosis (the detection gap), identifying patients at risk of stroke and starting OAC treatment (protection gaps 1 and 2) and optimizing OAC treatment to minimize stroke risk (the perfection gap; Panel 1).

The GRASP-AF suite of audit tools [19] provides AF data that underpin key input assumptions made in NICE technology appraisals of OACs [12,13] and the NICE costing template for warfarin [7]. These data provide the default clinical (risks of stroke and bleeding) and cost inputs in the AHSN AF Business Case Model. However, it is possible for users to substitute their own data or those from other sources.

The AHSN AF Business Case Model provides a picture of the AF detection and care pathway, breaking down diagnosis and AF management into four categories to identify potential gaps in care. Protection gaps 1 and 2, which address OAC prescribing in diagnosed patients with AF, are based on robust real-world national (England) datasets and are, therefore, ‘visible’. By contrast, the detection (represented in this study by screening) and perfection (achieving appropriate and adequate treatment in high-risk patients with AF) gaps required greater modeling and projection. The breakdown enables health providers to identify and assess areas in which changes will achieve the greatest benefits for patients alongside the most manageable investments in money and manpower. We found that making changes from 2016/17 and for the following 3 years to completely close the care gaps in England would have had a substantial positive budget impact.

Protection gaps 1 and 2 together contributed to around a third of the projected costs required to close these gaps completely. With the recording of AF data in QOF, risk assessment and starting treatment are now the most visible parts of the AF care pathway, and yet a considerable number of patients – over 190,000 – appear to have been receiving suboptimum care. ‘Closing’ these two gaps in care was projected to help prevent 7058 strokes. This number alone exceeds the 2016 PHE target of 5000 strokes avoided [18] and almost reaches the target of 8000 set by NHS England [17]. It was projected that closing the perfection gap by ensuring that all patients with AF at high risk of stroke are receiving optimized OAC treatment could potentially add another 8000 avoided strokes to this number.

In terms of cost savings, we found that 97.7% of savings are accounted for by addressing two key gaps – protection gap 2 (starting anticoagulation in line with guidelines) and the perfection gap (optimization of therapy). For this analysis, in the model we assumed that 100% of AF patients at high risk of stroke would have OAC treatment initiated. We also assumed that patients receiving inadequate anticoagulation would have treatment optimized by switching 90% to a DOAC and the remaining 10% continuing to receive warfarin due to some patients being ineligible to swap to DOACs. At the time of developing the model, little information on the actual split was available and, therefore, this proportion was decided as a conservative estimate. Most of the data on groups of patients under-represented in trials, such as the very elderly, are derived from subgroup analyses, although they suggest that more patients might be eligible for treatment with DOACs [28]. The model estimated that 60% of people currently treated with warfarin will not achieve a TTR greater than 65%. This assumed value was based on anecdotal evidence but reflects the findings of published systematic reviews [29,30] and is thought to be considerably higher than the proportion of patients who achieve such a TTR in the real world [31]. Consequently, the opportunity to improve AF care and reduce the risk of stroke may actually be larger than demonstrated by the model.

Factors such as the location of the international normalized ratio monitoring clinic (e.g. hospital versus community clinic) and frequency of testing have been found to affect TTR, with patients assessed in community clinics having lower TTR than those assessed in specialized anticoagulation clinics [29–31]. Individual factors, such as diet, weight and other medications, can also affect effectiveness. In contrast to VKAs, DOACs have faster onset of action, more predictable pharmacokinetics and a wider therapeutic window, meaning that they are given at a fixed dose without monitoring. All four DOACs have been found to be at least noninferior to warfarin [32–35] and could be considered with a view to promoting the greatest AF control, most consistent anticoagulation and greatest adherence.

The greatest gains in stroke prevention seem numerically to be attainable from closing the detection gap, with a potential for avoiding 12,376 strokes. However, this was the only gap in care to incur costs across the 3 years. No country has yet established a cost-effective national screening program for AF. However, the 2016 European Society for Cardiology Guidelines [9] encourage further evaluation of systematic AF screening programs in at-risk populations. The AHSN AF Business Case Model calculated costs of screening with single-lead or limb-lead devices, but other methods might be more cost-effective. Given the opportunities in terms of significantly reducing stroke incidence and costs in this undetected population, further exploration of national screening approaches seems warranted. Welton and colleagues [36] suggest that national-level systematic opportunistic screening in primary care of patients aged 65–80 years by nurse pulse palpation or use of modified blood pressure monitors followed by ECG interpreted by trained GPs for confirmation of suspected AF and referral to a specialist for unclear cases is likely to be the most effective approach and to be a cost-effective use of resources. Screening by emerging technologies might reduce costs further once detection thresholds and reliability are resolved [37]. In the meantime, in localities where the detection gap is large, implementation of specifically targeted screening programs might be a beneficial use of resources.

This study has some limitations. The data that underlie the model were drawn from the most up-to-date available published sources at the time, but these data will not necessarily reflect current practice. Prevalence of diagnosed AF in QOF data continues to rise year on year (e.g. 1.63% in 2014/15 rising to 1.91% in 2017/18). Similarly, the prevalence study used to estimate true background prevalence of AF was several years old and was based on a Scandinavian rather than a UK population. There is, therefore, potential for our detection gap estimates to be inaccurate. Many DOAC studies for treatment of AF treatment exclude patients with severe clinically relevant mitral stenosis and artificial valve replacement, but other patients with valvular AF might be included. Owing to differences in methods between studies and in definitions of valvular AF, it was not possible to exclude these patients completely for the model, and we cannot guarantee that the data underlying the study refer only to patients with nonvalvular AF. However, patients with valvular AF make up a small proportion of the overall patient pool assessed, and we do not believe that recalculating values after exclusion of these patients would substantially alter the outputs.

Usage of DOACs within UK general practice is also changing rapidly. Data from 2015/16 were used in the model, at which time 28.4% of anticoagulant prescriptions in England were DOACs. Data from April 2019 show that this figure has now risen to 56.4% [38]. This change will affect the applicability of the estimates associated with the perfection gap.

Estimates for the impact of treatment on stroke and bleed risk are derived from the clinical trial literature. There are significant uncertainties as to the extent to which these data are translatable to normal clinical practice, particularly with reference to warfarin, where maintenance of anticoagulation within a narrow therapeutic window is critical to achieving the expected results. This factor might impact on our estimates for the protection gap.

Finally, the costs used for interventions and outcomes might be inaccurate. Although we followed the approach used by NICE to ensure maximum comparability, some parameters, such as the cost of a major bleed, warfarin monitoring and AF screening, could only be approximate at best.

Despite a decade of improvements, substantial AF detection, protection and perfection gaps were still found to be present in England using national datasets from 2016/17 and projected standards of care from the literature. The AHSN AF Business Impact Model indicates the size of the gap in care, the number of patients projected to benefit from addressing these gaps and the associated budget impact of doing so. The results can help to define investment priorities for money and resources in existing management pathways that would lead to long-term cost savings, while improving clinical outcomes for patients with nonvalvular AF, at national, regional (STP and CCG) and local (GP practice) levels. Improving the management of high-risk patients with AF through initiating and/or optimizing anticoagulation treatment, in line with guidelines, could yield immediate benefits with little planning or cost impact. The number of strokes that could be avoided would nearly meet or surpass national goals for NHS England and PHE. Closing the detection gaps – although likely to be high impact in terms of strokes avoided – will require further local evaluation, planning and investment, taking into account local and regional resources and services. At the local level, care gaps and numbers of strokes prevented might differ substantially from the national findings due to local factors such as socioeconomic status [39] and, therefore, might require different approaches, changes to practice and investments of time, resources and money from place to place. Ensuring input of appropriate, accurate and up-to-date information into the model will provide the most relevant indicators for potential strategies and investments to take forward.

All data underlying this analysis will be available on request.

A Orlowski has received speaker honoraria from Pfizer and provided consultancy to Daiichi Sankyo. J Belsey has received consultancy payments from Imperial College Health Partners to create the model on which this paper is based. He has also received consultancy payments from pharmaceutical companies with an interest in the field of atrial fibrillation (Bristol-Myers Squibb, Pfizer, Daiichi-Sankyo). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

This work is licensed under the Attribution-NonCommercial-NoDerivatives 4.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/