When the whole is the sum of its parts: validation of the Heart Failure Collaboratory medication composite score in Denmark

“The score will take care of itself when you take care of the effort that precedes the score.”

John Wooden

The treatment of heart failure with reduced ejection fraction (HFrEF) has benefitted from a growing number of evidence-based therapies that improve ventricular remodeling, treat symptoms, reduce hospitalization, and extend survival.¹ The effects of these treatments are generally additive, such that contemporary HFrEF management involves multiple medications used together, dose-maximized over time, with serial assessment of tolerance and side effects, ultimately to achieve significant improvements in health outcomes.² Given this relative complexity, which occurs in patients who frequently have issues like multimorbidity and polypharmacy, prescribing of guideline-directed medical therapies (GDMT) for HFrEF is rarely optimal. HFrEF registries and studies have consistently show significant underuse and underdosing in eligible patients.^3,4 This underuse deprives patients of important care, as the recent STRONG-HF trial showed that patients randomized to an intervention that aggressively intensified GDMT lead to an absolute reduction in HF hospitalization and all-cause death by 8% over 6 months.⁵

To quantify the overall quality of HFrEF prescribing within a patient and across populations, the Heart Failure Collaboratory created a composite score capturing the use and dosing of GDMT. Prescribing of each class of HFrEF medication earns points, with higher points for drugs with a survival benefit, and additional points given for dose optimization. The original score⁶ has since been updated⁷ to include the full complement of HFrEF drugs that reduce hospitalization (Figure, Panel A). A further modification of the HF Collaboratory score accounted for indications beyond the 4 foundational classes of medications, such as race for hydralazine-isosorbide dinitrate (H-ISDN) and heart rate for ivabradine, and normalized the score on a scale of 0 to 100 based on the total number of eligible points.⁸

Figure.

Proposed scores to quantify composite guideline-directed medical therapies for heart failure with reduced ejection fraction (Panel A, left) and a checklist designed for pre-clinic delivery to patients to encourage proactive discussions with clinicians regarding treatment intensification (Panel B, right).

While initially applied to clinical trial data to standardize measurement of background therapy and proposed as a surrogate endpoint for pragmatic studies, perhaps the greatest potential of the HF Collaboratory score is in application to broad HFrEF populations as a composite measure of quality of care. Since the strongest data for improved outcomes in HFrEF mostly come from the prescription of GDMT, assessment of overall use of these drugs should be a focus on quality measurement. However, the feasibility of data collection and the association of the score with patient-centered outcomes measures in routine care has not been tested.

In this issue, a group of investigators set out to calculate and validate the HF Collaboratory score in a large HFrEF population.⁹ They leveraged the Danish public healthcare system to cross-linked nationwide administrative registries containing information on inpatient and outpatient contacts, redeemed prescriptions, and laboratory values. They identified all Danish adults alive on July 1, 2018 with an existing International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) code of I50 (heart failure) who also had HFrEF based on a validated algorithm requiring at least one claimed prescription of a renin-angiotensin-system blocker and a blocker-blocker. Patients also needed a full year between diagnosis and prescribing assessment to allow adequate time for GDMT intensification. In total, nearly 27,000 individuals with HFrEF were identified. Assessment of medication prescriptions fills—notably without a detailed assessment of contraindications and intolerances—showed use of evidence-based beta blocker (EVBB) in 81%, angiotensin converting enzyme inhibitor / angiotensin receptor blocker (ACEi/ARB) in 77%, angiotensin receptor blocker + neprilysin inhibitor (ARNi) in 2%, mineralocorticoid receptor antagonist (MRA) in 30%, and ivabradine in 2%, with limited use of H-ISDN, vericiguat, and sodium glucose transporter inhibitor (SGLTi). The median HF Collaboratory score was 4 (out of a maximum of 12, Figure, Panel A) with an interquartile range of 2–4. After multivariable adjustment, higher HFC scores were independently associated with lower mortality in a graded fashion. The suboptimal uptake of foundational GDMT—in particular the low rate of MRA use—is largely consistent with other registries from around the world. However, the results are particularly sobering given the universal access to care, including free medications, in Denmark.

These findings extend the work for the HF Collaboratory composite medical therapy score by showing that the score can be systematically calculated and is independently associated with survival. To our knowledge, this is the first published study to evaluate the pragmatic use of this score. This sets the stage for wider use of a composite score for assessment of GDMT use.

However, before advocating for widespread application of the score, several issues with this Danish application must be considered. They do not have clinical data on left ventricular ejection fraction (LVEF). Instead, they rely on previously validated algorithm that uses ICD codes and medication data to identify patients with HFrEF, which has a high positive predictive value but is also likely to exclude HFrEF not on treatment. This likely introduces bias, and the average HF Collaboratory scores for the total HFrEF population may be even lower. Furthermore, the patient population lacks diversity, which make its less applicable to diverse populations in countries like the US. For example, H-ISDN is very seldomly used in Denmark due to a predominantly White population. They are unable to distinguish between patients who have a low composite medical therapy score but are maximally tolerated therapy (either from physiological limitations or other contraindications) vs those due to low quality of care and clinical inertia.

Perhaps most limiting, the HF Collaboratory raises questions about attempts to simplify complex treatment into a single score. In the relatively short time since the score was published, the score has already changed to capture emerging data around SGLTi, ivabradine, and vericiguat. Although evolution of the score is appropriate, it complicates score comparisons over time. Perhaps more controversial is the somewhat subjective scoring that is applied: e.g., why should low-dose ARNI receive 3 points and high-dose EVBB receive only 2 points? Somewhat in conflict with the score, much of the HF community has worked to emphasize the use of those classes of medications with documented survival benefit, essentially setting up a prioritization for highest value “4-Pillar Therapy”, followed later by lower-value therapies once these 4 classes of medications have been maximized. A scoring system that focuses on only “4-Pillar Therapy” intensification may provide greater simplicity and prioritization and facilitate pairing with patient-activation tools, such as EPIC-HF (Figure, Panel B).⁴

For any version of GDMT score to be broadly adopted for quality improvement, several additional questions remain. Will stakeholders beyond the HF Collaboratory, such as the American Heart Association and the American College of Cardiology, adopt a version of the GDMT score as a performance or quality measure in their next update of the 2020 HF measure report?¹⁰ What should the total score be and how should contradictions and physiological limitations be factored into the GDMT score? For example, the presence of end-stage renal disease or type 1 diabetes would reduce the maximum score for a patient, and a lower score without integrating these co-morbidities may mispresent the level of quality of care. What changes are needed to electronic health record systems and workflows to identify patients more reliably with HFrEF and accurately capture the GDMT score? The availability of LVEF as structured data is heterogenous across care systems and using diagnosis codes or problems alone to subtype HF has been shown to lead to substantial misclassification.¹¹ Medication data, especially related to dosing, from the electronic health record can also be fraught with data quality issues.¹² Even if the score can be accurately calculated in EHR systems, well-tested patient and clinician clinical decision support tools that surface the more granular gap in care and facilitate GDMT intensification will be needed. Lastly, is a “prescription” GDMT score the correct measure, or, given the known multi-level and multifaceted barriers to accessing and maintaining GDMT, including cost, should we be focusing on a score that includes a measure of adherence?

The work presented here marks a significant step forward in efforts to improve the wide-scale use of GDMT for patients with HFrEF. This study has several strengths, including its large sample, pragmatic approach, high quality data (including use of pharmacy dispense data), and clear implications. Given the large amount of overall interest in improving GDMT uptake for patients with HFrEF and the number of pragmatic trials underway or being planning, this paper provides supporting evidence that a GDMT score may be a good surrogate endpoint for clinical trials and a reasonable quality measure for HFrEF processes of care, although additional research will be needed prior to wider adoption of a GDMT score as a quality or performance measure. Our patients can only win when we do the work of actually getting them the totality of evidence-based therapies.