Reducing stroke risk for patients with heart failure and atrial fibrillation in primary care; researching cost-effective approaches to service improvement

Background: There has been a long-term problem with suboptimal anticoagulation prescribing for people with atrial fibrillation (AF) resulting in many strokes that were potentially preventable. More than one-third of people with AF will develop heart failure (HF) and people with both conditions are thought to be at particularly high stroke-risk. Aim: To investigate all-cause mortality, stroke and major haemorrhage among people with AF and HF, compared primarily to people with AF only and examine the economic evidence for service interventions to improve anticoagulation prescribing for stroke prevention in these groups. Methods: Three studies are based upon a large English primary care cohort (2000-2018), linked to hospital, deprivation and civil mortality databases. These analyses incorporate both traditional survival analysis techniques and competing-risks methods. Next, in a systematic review I establish the economic evidence for service interventions that might improve anticoagulation prescribing. Finally, I use decision analytic modelling to estimate the potential cost and clinical impact of implementing different service interventions at scale in the National Health Service. Results: The prognosis for people with HF and AF is typically poor, with close to two-thirds of people dying within five years of diagnosis. The cumulative probability of all-cause mortality far exceeds the cumulative probability of stroke or major haemorrhage in people with HF. Accounting for this competing risk, people with HF and AF were at decreased relative risk of stroke, but at similar relative risk of major haemorrhage compared to those with AF only. Increasing age, deprivation and an absence of anticoagulation were associated with mortality and stroke. I found limited economic data pertaining to service interventions that might improve anticoagulation prescribing. This led to uncertainty in my decision modelling, but there is some evidence to support either complex interventions or specialist anticoagulation management clinics to improve anticoagulation prescribing. Conclusion: People with HF and AF typically face a poor prognosis that far exceeds their relative risk of stroke. Risk prediction scores that do not account for this competing risk may over-estimate the importance of HF in predicting stroke but under-estimate its importance in predicting major haemorrhage among people with AF. Service interventions to improve anticoagulation prescribing might be better targeted at other populations with AF who are at high stroke-risk. Economic evidence is lacking as to which anticoagulation improvement initiatives can improve prescribing, meaning further primary research is needed.