England-wide prescribing patterns for heart failure with reduced ejection fraction under scrutiny
Heart failure with reduced ejection fraction remains one of the most consequential chronic cardiovascular syndromes in routine care, and this nationwide OpenSAFELY analysis was built to examine how treatment has been delivered across England from 2018 through 2024. The study sits squarely in the gap between guideline ambition and day-to-day prescribing behavior. By using a large primary care dataset linked to national hospital admissions and mortality records, it was designed to capture treatment patterns at a population scale rather than through the narrower lens of single-center audits or trial populations.
The question matters because HFrEF accounts for an estimated 56% to 70% of heart failure overall, and the condition carries a steep burden of morbidity, hospitalization, and premature death. Even when therapies are well established in guidelines, uptake can lag in routine practice, particularly when patients move between primary care, hospital care, and specialist follow-up. A national prescribing study can show where the care pathway breaks down: at diagnosis, at initiation, during titration, or in long-term persistence.
One sentence captures the central ambition of the project: prescribing behavior across England was tracked longitudinally against a shifting clinical backdrop.
A national dataset built to follow treatment over time
The OpenSAFELY platform gave the investigators access to a large, population-based primary care record system, with linkage to national hospital admissions and mortality data. That structure is important for HFrEF because no single data source fully describes the patient journey. Primary care records capture the everyday prescriptions issued in the community, hospital data add information on acute admissions and serious events, and mortality records help define downstream outcomes and censoring in a way that a simple practice database cannot.
The six-year span from 2018 to 2024 also allows the analysis to reflect more than a single prescribing snapshot. Heart failure treatment has evolved over that period, and clinical practice has been shaped by changing evidence, changing expectations, and increasingly explicit guideline pathways. A time-series approach can therefore show whether routine care kept pace with those shifts or remained stuck in older prescribing habits.
The core clinical value of the study lies in its scale. National datasets reduce the risk that observed prescribing patterns are driven by local practice style, a single specialist service, or a particularly motivated patient cohort. They also make it possible to examine treatment at the point where most heart failure care actually happens: in the primary care record, where repeat prescriptions are issued and medication lists are maintained over time.
Clinical coding, prescription records, and where misclassification can creep in
The same data breadth that makes the study powerful also creates familiar epidemiologic challenges. Reliance on clinical coding for HFrEF can misclassify phenotype, especially when left ventricular ejection fraction is not consistently recorded or when patients carry heart failure diagnoses without a precise subtype attached. Some people with reduced ejection fraction may be missed entirely if the phenotype is uncoded. Others may be included even when the heart failure category is incomplete or outdated.
This matters because HFrEF is not the same as heart failure with preserved ejection fraction, and treatments are not interchangeable across those groups. If subtype coding is inaccurate, prescribing estimates may be blurred by inclusion of patients who do not truly belong in the HFrEF cohort. In a research setting, that can dilute apparent treatment rates or distort comparisons over time.
A second limitation is even more practical at the bedside. Prescription records show what was issued in primary care, not what was actually dispensed, taken, or continued. A medication appearing in the record does not guarantee adherence, and it does not account for drugs supplied directly through secondary care. In heart failure, where therapy is often started in hospital and then handed over to general practice, that boundary can be especially difficult to measure cleanly.
Why prescribing studies still matter in heart failure care
Despite those caveats, a study of this kind is valuable because it examines the machinery of care delivery rather than just its theoretical ideal. Heart failure pharmacotherapy depends on repeated decisions: initiation, up-titration, substitution when side effects occur, and long-term continuation when symptoms improve. A population study can reveal whether the system is delivering evidence-based treatment at scale or leaving too much variation to local habit.
The long observation window is especially relevant in a condition like HFrEF, where treatment standards have been changing steadily. When guidelines evolve, one immediate clinical question is whether new recommendations translate into everyday prescribing across an entire health system. National data can provide that answer in a way that small retrospective cohorts cannot. It can also identify whether practice is changing uniformly or only in pockets of care.
Residual confounding remains a final and unavoidable issue. Treatment comparisons were not fully adjusted for every clinical factor that influences prescribing, and in heart failure those factors are numerous: symptom burden, renal function, blood pressure, frailty, prior intolerance, comorbidity load, and clinician judgment at the point of review. Those variables shape real-world prescribing in ways that administrative data cannot always fully capture.
The study therefore offers a strong framework for understanding how HFrEF treatment is delivered in England, while also showing how hard it is to measure true therapeutic exposure in routine care. The most durable contribution may be methodological as much as clinical: a reminder that population-level prescribing research can map broad practice patterns, but it cannot fully substitute for the granularity of a well-phenotyped clinical cohort. The unresolved issue is not whether national data are useful; it is how far they can go before coding limits and incomplete exposure capture begin to obscure the therapeutic picture.