Medical therapy for patients with recent-onset heart failure with reduced ejection fraction during the COVID-19 pandemic: Insights from the Veteran's affairs healthcare system.

This study aims to evaluate trends in guideline-directed medical therapy (GDMT) for patients with recent-onset heart failure with reduced ejection fraction (HFrEF) following the onset of the COVID-19 pandemic using an interrupted time series analysis in the Veteran's Affairs Healthcare System. Among 71,428 patients with recent-onset HFrEF between 1/1/2018 and 2/28/2021, we found the pandemic was not associated with differences in treatment rates for beta-blockers, renin-angiotensin-aldosterone system inhibitors, or mineralocorticoid receptor antagonists; there was a 2.6 % absolute decrease (95 % CI: 0.5 %-4.7 %) in ARNI rates in April 2020; which decreased over the pandemic. Despite the changes to healthcare delivery, the COVID-19 pandemic was associated with minimal changes in GDMT rates among patients with recent-onset HFrEF. Published by Elsevier Inc.

Introduction

Rapid optimization of heart failure (HF) medical therapy is critical given the condition's substantial morbidity [1], [2]. The coronavirus-19 (COVID-19) pandemic caused a major disruption in outpatient cardiovascular care, leading to both more telemedicine as a substitute for face-to-face visits and fewer overall encounters [3], [4], [5], [6]. Within the Veteran's Health Administration (VA), the largest integrated healthcare system in the United States, we evaluated whether this shift in outpatient care was associated with changes in the use of guideline-recommended medical therapy (GDMT) among patients with recent-onset HF with reduced ejection fraction (HFrEF) between 1/1/2018 and 1/28/2021.

Methods

Briefly, we used VA claims and electronic health record (EHR) data sources, including inpatient encounters, outpatient encounters, laboratory values, vital signs, and pharmacy data, to evaluate rates of GDMT eligibility and treatment among a cohort with recently-diagnosed HF. We previously described the methods of evaluating GDMT use among VA patients with HFrEF [7].

We included patients with a recent HF diagnosis at the VA between 1/1/2018 and 2/28/2021 and with left ventricular ejection fraction (LVEF) ≤40 % within 1 year of diagnosis. Index date was the later date between the diagnosis and LVEF documentation. We excluded patients with an earlier HF diagnosis, prior heart transplant or ventricular assist device, and lack of VA medication fills.

We captured patient medical characteristics (including diagnoses, vitals, and laboratory values), neighborhood social risk, drive time to VA specialty care, and VA facility COVID prevalence. We classified patients residing in the top quartile of the CDC/ATSDR Social Vulnerability Index (SVI), which ranks census tracts on 15 social risk factors, as high vulnerability [8]. We categorized patients in the top quartile of drive time to specialty care as having long drive times. We also classified the top quartile of facilities as high-COVID based on the number of COVID-19 infections through 11/30/2020, indexed to the number of 2019 patients.

We evaluated GDMT treatment rates based on prescription fills within 6 months of the index date. These included the following:1) guideline-recommended beta-blockers (BB) 2) any of angiotensin-converting enzyme inhibitors (ACEI), angiotensin receptor blockers (ARB), or angiotensin receptor-neprilysin inhibitors (ARNI) (ACEI/ARB/ARNI) 3) mineralocorticoid receptor antagonists (MRA); and 4) ARNI. In addition, we analyzed whether patients received ≥50 % of the target dose for (5) BB and (6) ACEI/ARB/ARNI therapies [9], [10]. We identified medications filled in the VA and non-VA medications entered into the VA EHR. We determined treatment rate based on the proportion of patients who filled a prescription, excluding patients who did not meet guideline-based indications or had contraindications to therapy based on allergies, comorbidities, vitals, or laboratory values. Details regarding the cohort design and estimating treatment rates have been published previously [7].

We used a “slope and level” interrupted time-series model with mixed-effects logistic regression to evaluate the association between GDMT rates and the pandemic. We included a continuous time variable for the date of recent-onset HFrEF to account for pre-pandemic temporal trends. Seasonal indicators and facility random effects were included. In the main analysis, we classified patients as post-pandemic if their 6-month follow-up extended to March 2020 or later. In a sensitivity analysis, we only classified patients as post-pandemic if their index date was in March 2020 or later. We assumed that the pandemic could be associated with both an acute (level) change and a gradual (slope) change. To compare the frequency of recent-onset HFrEF diagnoses over time, we used a level-change model with Poisson regression.

We reported the level and slope changes as odds ratios (ORs). In addition, we calculated the average marginal effect of the pandemic for patients with an index date in April 2020 (follow-up period of 4/2020–9/2020); April 2020 marked the maximal reduction in VA in-person cardiology encounters and was early in the pandemic before health systems could fully adapt [6]. The average marginal effect was the absolute difference between the adjusted therapy rate post-pandemic and the estimated rate based on pre-pandemic trends. By performing stratified analyses and comparing the average marginal effects, we evaluated whether the association differed based on the following: (1) neighborhood social risk, (2) specialty drive time, and (3) facility COVID-19 burden. As sensitivity analyses, we adjusted for medical characteristics and assumed that patients with missing vitals or laboratory values were ineligible for therapy. The study was approved by the Stanford University institutional review board.

Results

We identified 71,428 patients with recent-onset HFrEF diagnosed between 1/1/2018 and 2/28/2021 including 43,346 patients pre-pandemic, 9897 patients with partial pandemic overlap, and 18,185 patients with post-pandemic follow-up. The pandemic was associated with a 21.9 % (95 % CI: 19.7 %–24.0 %) decrease in the monthly number of recent-onset HFrEF diagnoses. Medical characteristics were relatively stable over time (Table 1 ).

Table 1

Cohort characteristics.

	Overall	Pre-pandemic	Post-pandemic	Standardized differences
	N = 71,428	N = 47,282	N = 24,146
Demographics
Age	71.8 (10.9)	71.8 (10.9)	72.0 (10.9)	0.02
Sex, women	2.6 % (1836)	2.5 % (1173)	2.8 % (663)	0.12
Race				0.01
Asian	0.6 %(399)	0.5 % (249)	0.6 % (150)
Black	19.5 % (13,932)	19.4 %(9157)	19.8 %(4775)
Native American/Alaska Native	1.0 %(709)	0.9 %(463)	0.9 %(246)
Pacific Islander	0.9 %(684)	0.9 %(432)	1.0 %(252)
White	71.8 %(51,316)	72.2 %(34,138)	71.1 %(17,178)
Missing	6.1 %(4388)	6.0 %(2843)	6.4 %(1545)
Ethnicity				0.01
Hispanic	4.6 %(3284)	4.6 %(2185)	4.6 %(1099)
Non-Hispanic	92.0 %(65,738)	92.1 %(43,554)	91.9 %(22,184)
Missing	3.4 %(2406)	3.3 %(1543)	3.6 %(863)
Vitals
Systolic blood pressure, mmHg	129.7 (20.0)	129.5 (20.0)	130.2 (19.9)	0.04
Systolic blood pressure available	99.5 %	99.7 %	99.0 %
Diastolic blood pressure, mmHg	75.1 (12.3)	74.9 (12.4)	75.5 (12.3)	0.05
Diastolic blood pressure available	99.5 %	99.7 %	99.0 %
Respiratory rate, breaths per minute	18.1 (2.4)	18.1 (2.4)	18.1 (2.5)	0.00
Respiratory rate available	99.0 %	99.2 %	98.4 %
Heart rate, beats per minute	79.8 (17.5)	79.5 (17.5)	80.3 (17.7)	0.05
Heart rate available	99.5 %	99.7 %	99.0 %
Oxygen saturation, %	96.2 (2.6)	96.2 (2.6)	96.3 (2.6)	0.04
Oxygen saturation available	97.3 %	97.4 %	96.9 %
Body mass index, kg/m2	29.7 (6.8)	29.7 (6.8)	29.7 (6.7)	0.00
Body mass index available	88.6 %	89.8 %	86.2 %
Left ventricular ejection fraction (%)	31.7 (7.7)	31.7 (7.7)	31.6 (7.7)	−0.01
Laboratory values
Glomerular filtration rate, mL/min/1.73 m2	66.7 (27.2)	66.8 (27.3)	66.4 (27.0)	−0.01
Glomerular filtration rate available	97.6 %(69,723)	97.8 %(46,244)	97.2 %(23,479)
Sodium, mEq/L	138.7 (3.4)	138.8 (3.4)	138.5 (3.4)	−0.09
Sodium available	97.8 %(69,826)	98.0 %(46,313)	97.4 %(23,513)
Hemoglobin, g/dL	13.0 (2.3)	13.0 (2.3)	13.0 (2.3)	0.00
Hemoglobin available	94.8 %(67,735)	95.1 %(44,968)	94.3 %(22,767)
Potassium, mEq/dL	4.2 (0.5)	4.2 (0.5)	4.2 (0.5)	0.00
Potassium available	97.1 %(69,321)	97.3 %(46,028)	96.5 %(23,293)
Hemoglobin A1c, %	6.6 (1.5)	6.6 (1.5)	6.6 (1.5)	0.00
Hemoglobin A1c available	85.5 %(61,047)	85.2 %(40,306)	85.9 %(20,741)
Comorbidities
Alcohol abuse	12.9 %(9218)	12.7 %(6002)	13.3 %(3216)	0.05
Atrial fibrillation	36.4 %(26,030)	35.5 %(16,796)	38.2 %(9234)	0.12
Cancer	16.9 %(12,079)	16.7 %(7903)	17.3 %(4176)	0.04
Cancer, metastatic	2.7 %(1893)	2.6 %(1223)	2.8 %(670)	0.08
Cerebrovascular disease	7.5 %(5340)	7.4 %(3493)	7.6 %(1847)	0.03
Chronic kidney disease	28.5 %(20,372)	28.3 %(13,367)	29.0 %(7005)	0.03
Chronic obstructive pulmonary disease	33.2 %(23,723)	33.8 %(15,974)	32.1 %(7749)	−0.08
Connective tissue disease	3.9 %(2781)	3.8 %(1795)	4.1 %(986)	0.03
Dementia	6.7 %(4779)	6.8 %(3221)	6.5 %(1558)	−0.05
Depression	26.4 %(18,834)	26.2 %(12,371)	26.8 %(6463)	0.03
Diabetes mellitus	50.1 %(35,784)	50.1 %(23,666)	50.2 %(12,118)	0.00
Frailty	18.0 %(12,832)	17.8 %(8411)	18.3 %(4421)	0.03
Hypertension	88.0 %(62,881)	88.0 %(41,589)	88.2 %(21,292)	0.02
Hypothyroidism	12.4 %(8862)	12.5 %(5895)	12.3 %(2967)	−0.02
Ischemic heart disease	63.6 %(45,433)	64.0 %(30,257)	62.9 %(15,176)	−0.05
Liver disease	10.7 %(7616)	10.5 %(4969)	11.0 %(2647)	0.05
Peptic ulcer disease	2.1 %(1510)	2.0 %(960)	2.3 %(550)	0.15
Peripheral arterial disease	25.6 %(18,304)	25.5 %(12,035)	26.0 %(6269)	0.03
Psychotic disorder	2.7 %(1929)	2.7 %(1299)	2.6 %(630)	−0.04
Substance abuse	7.6 %(5420)	7.3 %(3459)	8.1 %(1961)	0.11
Valvular heart disease	17.9 %(12,772)	17.6 %(8339)	18.4 %(4433)	0.05
Ventricular arrhythmia	5.5 %(3899)	5.5 %(2590)	5.4 %(1309)	−0.02

Continuous variables listed as mean (standard deviation). Categorical variables listed as percentage (frequency). The standardized differences for continuous and binary variables are Cohen's d; for categorical variables, Cramer's V was used. For both, an absolute value of the standardized difference of ≤0.20 is often considered small.

Fig. 1 displays the GDMT rates over time with the OR for the level change post-pandemic and the slope change. For BB, ACEI/ARB/ARNI, and MRA therapies, there was no immediate change in the level associated with pandemic onset, although a small increase in the slope was observed.

Fig. 1

Trends in GDMT over time

Abbreviations: ACEI: angiotensin-converting enzyme inhibitors; ARB: angiotensin receptor blocker; ARNI: angiotensin receptor neprilysin inhibitor; COVID: Coronarvirus-19; GDMT: guideline-directed medical therapy; MRA: mineralocorticoid receptor antagonist; OR: odds ratio. This figure displays therapy rates over time. The x-axis lists the 6-month follow-up date for patients with recent-onset HFrEF patients. The vertical dashed line (March 2020) represents the COVID-19 pandemic onset in the base case, in which all individuals with partial follow-up during the pandemic were designated post-pandemic. In a sensitivity analysis, only patients with complete post-pandemic follow-up were designated post-pandemic. The figure lists the OR for the slope and level change with 95 % confidence intervals in parentheses. An * indicates p < 0.05 for the OR.

Incorporating the level and slope changes, the pandemic was associated with a 1.7 % absolute increase (95 % CI: 0.2 %–3.3 %) in the ACE/ARB/ARNI therapy rate for patients with a 6-month window from 4/2020–9/2020 (Table 2 ). With ARNI, we found an immediate decrease with pandemic onset (post-pandemic OR 0.84; 95 % CI: 0.75–0.94). However, there was a small, non-significant increase in the slope of ARNI uptake (OR 1.01 per month post-pandemic; 95 % CI: 1.00–1.02). The pandemic was associated with a 2.6 % absolute decrease (95 % CI: 0.5 %–4.7 %) in ARNI rates. For the other treatment metrics, including the proportion of ≥50 % of the target doses, estimated treatment rates were not significantly different from pre-pandemic trends.

Table 2

Percent change in GDMT rates for patients with recent-onset HFrEF in April 2020 compared with pre-pandemic trends, stratified by key characteristics.

	BB	ACEI/ARB/ARNI	MRA	ARNI	≥50 % BBTarget Dose	≥50 % ACEI/ARB/ARNITarget Dose
Overall	0.9 (−0.6 to 2.5)	1.7 (0.2 to 3.3)	0.7 (−0.9 to 2.2)	−2.6 (−4.7 to −0.5)	−0.0 (−1.9 to 1.8)	0.1 (−1.8 to 2.1)
Neighborhood social risk
Low risk	1.7(−0.1 to 3.4)	1.5(−0.3 to 3.2)	−0.1(−1.8 to 1.7)	−2.6(−5.0 to −0.2)	−0.2(−2.4 to 1.9)	−0.7(−2.9 to 1.6)
High risk	−1.4(−4.2 to 1.4)	2.4(−0.5 to 5.3)	2.7(−0.2 to 5.7)	−2.4(−6.3 to 1.6)	0.4(−3.1 to 4.0)	2.3(−1.4 to 6.1)
p-value for difference between subgroups	0.07	0.57	0.10	0.90	0.75	0.17
Drive distance from VA specialty care
Short drive time	1.0(−0.7 to 2.8)	1.8(0.1 to 3.6)	0.0(−1.8 to1.8)	−3.3(−5.8 to −0.9)	−0.7(−2.9 to 1.6)	0.0(−2.3 to 2.3)
Long drive time	0.6(−2.2 to 3.5)	1.6(−1.2 to 4.5)	2.3(−0.6 to 5.2)	−1.0(−4.7 to 2.6)	1.8(−1.5 to 5.2)	0.5(−3.0 to 4.0)
p-value for difference between subgroups	0.80	0.91	0.18	0.29	0.21	0.83
VA facilities with high COVID prevalence
Low COVID prevalence facilities	0.8(−1.0 to 2.5)	1.6(−0.1 to 3.3)	0.8(−1.0 to 2.5)	−1.4(−3.7 to 1.0)	−0.3(−2.4 to 1.9)	0.1(−2.1 to 2.3)
High COVID prevalence facilities	1.5(−1.5 to 4.6)	2.3(−0.7 to 5.4)	0.3(−2.8 to 3.4)	−6.7(−11.0 to −2.4)	0.6(−3.2 to 4.4)	0.2(−3.6 to 4.1)
p-value for difference between subgroups	0.67	0.66	0.79	0.03	0.69	0.95

Abbreviations: ACEI: angiotensin-converting enzyme inhibitors; ARB: angiotensin receptor blocker; ARNI: angiotensin receptor neprilysin inhibitor; BB: Beta-blocker; COVID: Coronarvirus-19; MRA: mineralocorticoid receptor antagonist; OR: odds ratio. For each of the three characteristics (neighborhood social risk, drive distance from VA specialty care, and VA COVID facility prevalence), the binary stratification was based on determining the upper quartile for each characteristic. These are treatment rates for patients with recent-onset HFrEF in April 2020 with follow-up through September 2020 compared with treatment rates pre-pandemic.

Treatment rates were similar with and without the Medicare Part D data. For ACEI/ARB/ARNI, the difference was 1.8 %; for BB, MRA, and ARNI, the differences were <0.8 %.

We stratified the analyses based on three patient characteristics: (1) neighborhood social risk, (2) drive time to specialty care, and (3) VA facility COVID burden. Overall, our results were similar across the characteristics (Table 2). Hospitals with a high COVID-19 burden had larger reductions in ARNI rates.

Sensitivity analyses

In the sensitivity analyses, we varied our assumptions regarding patients with partial pandemic overlap. The reduction in ARNI rates persisted; however, when reclassifying the group with partial overlap as pre-pandemic or excluding them, the pandemic was no longer associated with a significant change in ACEI/ARB/ARNI rates. The results were also similar for the other GDMT rates. Adjusting for medical characteristics and excluding patients with missing relevant vitals/labs did not affect our findings.

Discussion

Among patients in a large integrated healthcare system, we found minimal change in the rates of HF GDMT early in the COVID-19 pandemic when in-person visits were at their nadir. However, ARNI therapy was an exception. While ARNI rates have gradually increased in VA, this progress slowed during the pandemic, primarily at facilities with high COVID-19 burden. The rapid expansion of VA telehealth services may explain stable GDMT rates; by June 2020, 58 % of VA encounters were provided virtual compared with 14 % before the pandemic [11].

Our analysis focused on patients with a HFrEF diagnosis. The largest pandemic effect for patients with HFrEF may be the likelihood of timely diagnosis. We also observed a decrease in the number of new HFrEF diagnoses. This may suggest a delayed diagnosis of incident HFrEF, consistent with the reduction in acute HF hospitalizations [12].

Given that the patients were diagnosed with HFrEF, it is unsurprising that first-line HFrEF therapy rates – BB and ACE/ARB/ARNI – remained unchanged. Subsequent treatment steps, such as transition from ACE/ARB to ARNI or dose increases may be more susceptible to less active disease management early in the pandemic when the focus on management of chronic conditions likely decreased. While the expanding use of ARNI temporarily slowed down, we did not find a significant drop in other metrics, demonstrating that VA successfully maintained the status quo of HFrEF management. However, this may also represent a gap in pre-pandemic management. If therapy were rarely up-titrated before the pandemic, less active management may have little effect.

There are important limitations to this analysis. First, we determined therapy exclusions using routinely collected data, although we incorporated laboratory values, vitals, and documented intolerances. Second, other important dimensions of HF disease management, such as adequate decongestion, were not captured. We were unable to evaluate sodium/glucose cotransporter-2 inhibitors because the trials demonstrating benefit among patients with HFrEF were published during the peri-COVID period. This limited the ability to establish pre-COVID trends.

The pandemic rapidly ushered in a new era of telehealth, which may explain why the HFrEF GDMT rates remained stable. However, beyond solely maintaining the suboptimal status quo, telehealth uptake may represent an opportunity to improve HFrEF disease management and GDMT rates.

Funding

This work was funded by the American Heart Association COVID-19 and Its Cardiovascular Impact Rapid Response Grant. AS receives research support from the (1K23HL151672-01).

Disclosures

Views expressed are those of the authors and not necessarily those of the Department of Veterans Affairs or other affiliated institutions. AS consults for Acumen, LLC. SK reports receiving unrestricted grants from Novartis and personal fees from Bayer and Bristol-Myers Squibb outside the submitted work. MPT reports employment with iRhythm Technology, Inc., personal fees from Medtronic Inc., personal fees from Abbott, grants from Bristol Myers Squibb, grants from American Heart Association, personal fees from Biotronik, personal fees from Sanofi, personal fees from Pfizer, grants from Apple, grants and personal fees from Bayer, personal fees from Myokardia, personal fees from Johnson & Johnson, personal fees from Milestone Pharmaceuticals, outside the submitted work.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

AS consults for Acumen, LLC. SK reports receiving unrestricted grants from Novartis and personal fees from Bayer and Bristol-Myers Squibb outside the submitted work. MPT reports employment from iRhythm Technologies, Inc., personal fees from Medtronic Inc., personal fees from Abbott, grants from Bristol Myers Squibb, grants from American Heart Association, personal fees from Biotronik, personal fees from Sanofi, personal fees from Pfizer, grants from Apple, grants and personal fees from Bayer, personal fees from Myokardia, personal fees from Johnson & Johnson, personal fees from Milestone Pharmaceuticals, outside the submitted work.