Trajectory of Cognitive Decline After Incident Heart Failure Hospitalization: Findings From the REGARDS Study

Abstract

Background

Cognitive impairment is common among adults with heart failure (HF) and associated with poor outcomes. However, less is known about the trajectory of cognitive decline after a first HF hospitalization. We examined the rate of cognitive decline among adults with incident HF hospitalization compared with those without HF hospitalization.

Methods and Results

The REGARDS (Reasons for Geographic and Racial Differences in Stroke) study is a prospective longitudinal study of 23 894 participants aged ≥45 years free of HF at baseline. HF hospitalization was expert adjudicated. Changes in global cognitive function (primary outcome) were assessed with the Six‐Item Screener (range, 0–6). Secondary outcomes included change in Word List Learning (range, 0–30), Word List Delayed Recall (WLD; range, 0–10), and Animal Fluency Test (range, 0+). Segmented linear mixed‐effects regression models were used. Over 5 years, mean scores across all 4 cognitive tests declined for all participants regardless of HF status. Those with incident HF hospitalization experienced faster declines in the Six‐Item Screener versus those who were HF free (difference, −0.031 [95% CI, −0.047 to −0.016]; P<0.001), a finding that persisted in fully adjusted models. Those with incident HF hospitalization did not experience faster declines in Word List Learning, Word List Delayed Recall, or Animal Fluency Test scores compared with those without HF hospitalization. Participants with hospitalization for HF with preserved, compared with reduced, ejection fraction had faster decline in Animal Fluency Test.

Conclusions

Global cognitive decline occurred faster among adults with incident HF hospitalization compared with those who remained free of HF hospitalization. This pattern was not seen for the other cognitive domains.

Nonstandard Abbreviations and Acronyms

AFT

Animal Fluency Test

HFpEF

heart failure with preserved ejection fraction

HFrEF

heart failure with reduced ejection fraction

REGARDS

Reasons for Geographic and Racial Differences in Stroke

SIS

Six‐Item Screener

WLD

Word List Delayed Recall

WLL

Word List Learning

Clinical Perspective.

What Is New?

• Global cognitive function declined at a faster rate among adults with an incident heart failure (HF) hospitalization compared with those who remained free of HF hospitalization.

• This pattern of decline was not seen for the cognitive domains of learning, memory, and executive function.

What Are the Clinical Implications?

• Our findings can inform clinicians' understanding of cognitive trajectories in HF and provide information for counseling patients and caregivers with HF about cognition following an incident HF hospitalization.

Cognitive impairment is highly prevalent among adults with heart failure (HF) ¹ , ² and is associated with impairments in self‐care and poor quality of life, as well as increased morbidity and death. ³ , ⁴ , ⁵ Although HF is associated with worse cognition, independent of age, less is known about the trajectory of cognitive decline among adults after their first hospitalization for HF. While cognitive impairment among adults with prevalent HF likely increases with duration of disease and disease severity, how rapidly cognition declines among patients after they experience a first hospitalization for HF is poorly understood. ⁶ Additionally, less is known about how specific cognitive domains decline after a first HF hospitalization, which is critical for understanding not only brain function, but patients' ability to navigate the health care system and carry out key aspects of HF self‐care.

Prior studies have aimed to elucidate the trajectory of decline but have been limited by small, nonrepresentative samples and a lack of comprehensive and repeated cognitive assessments over time. ⁷ , ⁸ One recent study addressed some of these methodologic shortcomings; using the Cardiovascular Health Study, Hammond et al found that global cognitive ability, but not processing speed, declined significantly faster after newly diagnosed HF, compared with those without HF. ⁹ However, this study did not have substantial racial diversity, and data were collected in the 1980s and 1990s. Notably, this study was the first to examine whether cognitive decline differed by HF with preserved ejection fraction (HFpEF) compared with HF with reduced ejection fraction (HFrEF), an important investigation given the underlying different pathophysiology; authors did not find significant differences.

An improved understanding of the trajectory of cognition among adults after their first hospitalization for HF has 2 important implications: first, it can help clinicians inform patients and caregivers of what they can expect after a first HF hospitalization. Second, it can provide the scientific and medical community with an understanding of how much of the cognitive decline experienced by patients is due to HF after adjusting for sociodemographics and other medical conditions. With its large, geographically and racially diverse sample, rigorously collected cognitive battery of validated instruments, and adjudicated HF hospitalizations, the REGARDS (Reasons for Geographic and Racial Differences in Stroke) cohort is well suited to fill these gaps in the existing literature.

Our primary objective was to describe the trajectory of cognitive function among adults after their first hospitalization with HF (incident HF hospitalization) compared with those without an incident HF hospitalization, across 4 cognitive domains: (1) global cognitive function; (2) learning; (3) memory; and (4) executive function. Additionally, we aimed to explore differences in the trajectory of cognitive function after incident HF by HF subtype (HFpEF versus HFrEF). We hypothesized that participants would have faster rates of cognitive decline after a first HF hospitalization compared with those without HF, and that the trajectory would differ by cognitive domain and HF subtype.

Methods

The REGARDS study database includes identifying participant information and cannot be made publicly available because of ethical/legal restrictions. Deidentified data sets and statistical code specific to this article are available to researchers meeting criteria for access to confidential data. Data can be obtained upon request at regardsadmin@uab.edu.

Study Design, Participants, and Measurements

The REGARDS study is a prospective cohort study of 30 239 non‐Hispanic Black and White individuals initially designed to examine regional and racial disparities in stroke death. Details are described elsewhere. ¹⁰ Briefly, participants were enrolled between 2003 and 2007 using commercially available lists and a combination of mail and telephone contacts to recruit English‐speaking, community‐dwelling adults aged ≥45 years who were living in the continental United States. Race and sex were balanced by design, with oversampling of the southeastern United States. Baseline data collection included computer‐assisted telephone interviews gathering demographic information, medical history, and health status. In‐home examinations by trained staff followed standardized, quality‐controlled protocols to collect fasting blood and urine samples, ECGs, blood pressure, anthropometric measures, and data on medications via pill bottle review. Living participants or their next of kin were telephoned every 6 months with retrieval of medical records for reported hospitalizations. The REGARDS study procedures were approved by the institutional review boards at the participating centers and all participants provided written informed consent.

Study Population

For this study, we included REGARDS participants without suspected HF at baseline, which we ascertained with a previously developed medication algorithm that has a negative predictive value of >95% using Medicare claims data. ¹¹ , ¹² The medication‐based approach to determining suspected HF included use of digoxin in the absence of atrial fibrillation, angiotensin‐converting enzyme inhibitor/angiotensin receptor blocker plus β blocker in the absence of hypertension; carvedilol; spironolactone; loop diuretics including furosemide, bumetanide, or torsemide; or a combination of hydralazine and nitrates. We excluded participants with (1) missing data on self‐reported atrial fibrillation, (2) baseline medication use; and (3) participants with HF hospitalizations between the baseline computer‐assisted telephone interview and in‐home visit. Using this HF‐free cohort, we next excluded those with stroke at baseline since they represent a clinically distinct group with a different pathogenesis of cognitive decline; we also excluded those without a baseline measure of cognitive function.

To achieve the aforementioned primary study objectives, we divided our study population into 2 main groups: (1) participants who were free of HF at baseline and had no HF hospitalizations at follow‐up (no HF hospitalization) and (2) participants who were free of HF at baseline, had an incident HF hospitalization during follow‐up, and had at least 1 cognitive assessment after the HF hospitalization. Additionally, we identified a group of participants who were free of HF at baseline, had an incident HF hospitalization during follow‐up, and died before completing another cognitive assessment (HF decedents).

Main Exposure: Incident HF Hospitalization

Data from study baseline through December 31, 2017, were included. We defined incident HF as a first adjudicated HF hospitalization, which was clinician adjudicated. Adjudication of incident HF hospitalizations was performed throughout the study period. At 6‐month follow‐up telephone calls, interviewers asked participants if they were hospitalized during the prior 6 months and the reason for hospitalization. Medical records for potential cardiovascular‐related hospitalizations were subsequently retrieved and reviewed by 2 clinicians to adjudicate HF events. To identify HF hospitalizations, adjudicators considered symptoms, physical exam findings, laboratory values, imaging/echocardiography, and medical treatment. Differences were resolved by discussion. Next, HF was classified according to the lowest documented left ventricular ejection fraction during the hospitalization, determined by transthoracic echocardiogram or other imaging modalities. HF with left ventricular ejection fraction ≥50% was defined as HFpEF, and left ventricular ejection fraction <50% was defined as HFrEF, including HF with midrange left ventricular ejection fraction of 40% to 50%. ¹³ , ¹⁴

Main Outcome(s): Cognitive Function Across 4 Domains

REGARDS participants underwent cognitive assessments throughout the study period. These assessments were administered on the telephone by technicians who underwent formal training and certification.

Global cognitive function was assessed with the Six‐Item Screener (SIS), a validated screening instrument that can detect potential moderate or greater cognitive dysfunction in older patients and those experiencing acute illness. ¹⁵ Starting in 2003, the SIS was administered on an annual basis to participants. It consists of a 3‐item recall and a 3‐item temporal orientation; scores for the SIS range from 0 to 6. ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹

Learning and memory were assessed starting in 2006 and administered every 24 months with the Consortium to Establish a Registry for Alzheimer Disease's Word List Learning (WLL) test and Word List Delayed Recall (WLD). ²⁰ , ²¹ The WLL measures new learning and consists of 3 trials to learn a list of 10 unrelated words (scores range from 0 to 30). The WLD measures verbal memory, consisting of 3 trials to learn a list of 10 unrelated words and recall as many as possible after a 5‐minute delay (scores range from 0 to 10).

Executive function (known as complex cognitive processing used in problem solving or complex action sequences) was assessed with the Animal Fluency Test (AFT) starting in 2006 at 18‐month intervals, and then changed to every 24 months in 2008. The AFT asks participants to name as many animals as they can in 1 minute; scores reflect the number of animals named, starting at 0. For scoring, repetitions and intrusions were removed. Prior studies have found that word list and verbal fluency can be measured reliably and precisely over the telephone in middle‐aged and older adults with scores virtually identical to those obtained in person. ²² , ²³

Because cognitive tests were administered on separate schedules, we established separate analytic samples to preserve cognitive data for all participants, as described below. For all cognitive tests, higher scores indicate better performance. Participants were censored if they experienced a stroke during follow‐up. According to the literature, a decline of ≥0.5 SDs from baseline has been defined as a clinically meaningful decline across these 4 cognitive domains. ²⁴

Covariates

We included baseline covariates that could influence incident HF hospitalization and cognition. These included sociodemographic factors: age, sex, race (self‐reported by participants), education, income, region (Stroke Belt, Stroke Buckle, non‐Belt, consistent with REGARDS sampling design), social isolation, and social network; clinical comorbidities: hypertension (defined as systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg [according to the guidelines at the time of REGARDS inception] or use of blood pressure–lowering medications), coronary heart disease (defined as self‐reported diagnosis or ECG evidence of previous myocardial infarction), hyperlipidemia (defined as self‐reported diagnosis or low‐density lipoprotein >130 mg/dL), diabetes (defined as intake of blood glucose–lowering medication or insulin, or fasting serum glucose ≥126 mg/dL or random glucose ≥200 mg/dL), body mass index (kg/m²), albumin to creatinine ratio (mg/g), atrial fibrillation (self‐reported diagnosis and/or ECG evidence), depressive symptoms (defined as a score ≥4 on the 4‐item Center for Epidemiologic Studies Depression Scale ²⁵ ), and self‐reported health (as assessed with the Physical Component Summary and the Mental Component Summary scores from the Short Form‐12) ²⁶ ; health behaviors: smoking (self‐reported cigarette smoking defined as current, former, or never), alcohol use (heavy, moderate, none using sex‐specific National Institute on Alcohol Abuse and Alcoholism guidelines), and engagement in physical activity per week sufficient to work up a sweat (none versus any).

Statistical Analysis

Within each cognitive domain, we first determined differences in baseline characteristics and baseline cognitive scores by HF status using a χ² test for categorical variables, an ANOVA for normally distributed continuous variables, and a Kruskal–Wallis test for nonnormally distributed continuous variables. Although cognitive function after first HF hospitalization could not be assessed for HF decedents, baseline characteristics were described for this group.

To compare the cognitive function outcomes for participants after incident HF hospitalization to participants of the same age with no HF hospitalizations, we followed the approach of prior studies using segmented linear mixed effects regression models with participant age as the time scale, random intercepts at the participant level, and random slopes for age at cognitive assessments. ²⁴ , ²⁷ We selected a normal distribution for the random effects and an independent correlation structure. We developed separate statistical models to examine the association between incident HF hospitalization and cognitive decline for each cognitive domain. Models included indicator variables for group (no HF hospitalizations and incident HF hospitalization), age at HF hospitalization (centered at age 65), slopes for annual rate of cognitive decline before and after the first HF hospitalization (for incident HF hospitalization group), and a slope for rate of cognitive decline for each additional year of age among the no HF hospitalization group. We also included a discontinuity term for change in cognitive function at the time of the first HF hospitalizations for those who experienced one. Models were conducted unadjusted and fully adjusted for covariates described above with continuous covariates centered at their mean. Models were used to graph predicted mean scores for the cognitive outcomes over time for participants with incident HF hospitalizations and participants without HF hospitalizations of the same age. Subgroup analyses were done to evaluate cognitive decline by HF subtype with differences in the slope by subtype tested using Q tests. Analyses were conducted using StataMP version 14 (StataCorp, College Station, TX), and P values <0.05 were considered statistically significant. Missing data were handled with a complete case approach.

Clinically meaningful differences in cognition (declines) were considered one half the SD of the baseline cognitive score (by domain) among the REGARDS study. We calculated this as 0.36 for the SIS, 2.55 for the WLL, 1.1 for the WLD, and 2.9 for the AFT.

Results

The main analytic sample was derived for the SIS outcome since this was assessed annually in REGARDS and contributed the most cognitive data. After excluding participants with suspected HF at baseline (n=4008), those with stroke at baseline (n=1486), those lost to follow‐up (n=359), a HF hospitalization before the first SIS (n=33), and those with missing information on the first SIS (n=356), we included 23 894 participants (Figure S1).

Of the 23 894 participants, 584 (2.4%) experienced an incident HF hospitalization and completed a cognitive assessment after the HF event (incident HF hospitalization group); and 380 (1.6%) experienced an incident HF hospitalization and died before completing a cognitive assessment (HF decedent group) (Table 1). Median (interquartile range) follow‐up overall is 10 (5.0–12.0) years and mean (SD) follow‐up overall is 8.6 (4.1) years. Compared with participants without HF hospitalization, those with incident HF hospitalization and HF decedents were older, more likely to be men, had lower incomes, more cardiovascular comorbidities, and worse self‐rated physical health. They also smoked more and did less physical activity. A higher proportion of participants with incident HF hospitalization identified as Black compared with those who remained HF hospitalization free and compared with the HF decedents. HF decedents more commonly had HFrEF compared with those with incident HF hospitalization and subsequent cognitive assessment.

Table 1.

Baseline Characteristics Among REGARDS Participants by HF Status

Characteristics	Total	No HF*	Incident HF*	HF decedents*	P value
No.	23 894	22 930	584	380
Age, y, mean±SD	64±9.3	64±9.2	69±8.8	73±9	<0.001
Female sex, n (%)	13 273 (55.5)	12 843 (56.0)	271 (46.4)	159 (41.8)	<0.001
Black race, n (%)	9390 (39.3)	9024 (39.4)	241 (41.3)	125 (32.9)	0.023
Less than high school education, n (%)	2537 (10.6)	2362 (10.3)	96 (16.4)	79 (20.8)	<0.001
Income <$35 K, n (%)	9422 (44.7)	8923 (44.1)	301 (58.7)	198 (58.2)	<0.001
Region, n (%)
Stroke Belt	8225 (34.4)	7870 (34.3)	195 (33.4)	160 (42.1%)	0.018
Stroke Buckle	5010 (21.0)	4819 (21.0)	129 (22.1)	62 (16.3)
Non‐Belt	10 659 (44.6)	10 241 (44.7)	260 (44.5)	158 (41.6%)
Poor social network, n (%)	2824 (11.9)	2703 (11.9)	64 (11.1)	57 (15.1)	0.14
Social isolation, n (%)	3028 (13.5)	2900 (13.5)	75 (13.9)	53 (14.6)	0.80
Hypertension, n (%)	13 292 (55.6)	12 559 (54.8)	459 (78.6)	274 (72.1)	<0.001
Coronary heart disease, n (%)	3241 (13.7)	2903 (12.7)	202 (35.1)	136 (36.0)	<0.001
Dyslipidemia, n (%)	13 160 (57.1)	12 549 (56.8)	370 (64.7)	241 (65.5)	<0.001
Diabetes, n (%)	4149 (18.0)	3814 (17.2)	217 (37.9)	118 (32.0)	<0.001
Obese, n (%)	8614 (36.2)	8224 (36.0)	259 (44.7)	131 (34.9)	<0.001
Urinary albumin/creatinine ratio (mg/g), median (IQR)	7 (4.5–14)	6.9 (4.5–13)	13 (6.3–47)	16 (7.2–63)	<0.001
Atrial fibrillation, n (%)	2889 (12.6)	2585 (11.8)	179 (32.3)	125 (34.9)	<0.001
Depressive symptoms, n (%)	2297 (9.7)	2184 (9.6)	62 (10.7)	51 (13.5)	0.030
PCS score, mean±SD	48±9.8	48±9.7	44±10	44±11	<0.001
MCS score, mean±SD	54±8.1	54±8	54±8.6	54 8.7	0.81
Smoking status, n (%)
Current	3419 (14.4)	3262 (14.3)	92 (15.9)	65 (17.2)	0.007
Never	11 077 (46.5)	10 686 (46.8)	233 (40.2)	158 (41.8)
Past	9312 (39.1)	8903 (39.0)	254 (43.9)	155 (41.0)
Alcohol use, n (%)
Heavy	1021 (4.4)	993 (4.4)	17 (3.0)	11 (2.9)	<0.001
Moderate	8251 (35.2)	7973 (35.4)	164 (28.9)	114 (30.5)
None	14 174 (60.5)	13 539 (60.2)	386 (68.1)	249 (66.6)
No physical activity, n (%)	7452 (31.6)	7069 (31.3)	221 (38.2)	162 (43.3)	<0.001
LVEF, n (%)
≥50%			244 (46.7)	74 (35.2)	0.005
<50%			279 (53.3)	136 (64.8)

HF indicates heart failure; LVEF, left ventricular ejection fraction; MCS, Mental Component Summary; PCS, Physical Component Summary; and REGARDS, Reasons for Geographic and Racial Differences in Stroke.

^*Definitions for HF status: “No HF” participants were defined as those who were free of HF at baseline and at follow‐up; “incident HF” participants were defined as those who were free of HF at baseline, had an incident HF hospitalization, and subsequently completed a cognitive assessment; “HF decedents” were defined as those who were free of HF at baseline and had an incident HF event but subsequently died before the next cognitive assessment. Some participants with HF did not have LVEF documented in medical records.

We also included 2 other analytic samples, 1 for the WLL and WLD outcomes (Figure S2), which were assessed at the same time, and 1 for AFT (executive function) (Figure S3). Because WLL, WLD, and AFT were assessed less frequently than the SIS, the sample sizes were smaller; however, the demographic and clinical characteristics by HF status mirrored the SIS outcome (Tables S1 and S2).

The baseline cognitive status of participants across the 4 outcomes, by HF status, are shown in Table 2. In total, 1855 (7.8%) of the sample had global cognitive impairment as measured by an SIS score ≤4; by group, HF decedents had the highest rate of global cognitive impairment (13.7%) at baseline, followed by those with incident HF hospitalization (9.2%), and those without HF hospitalization (7.6%); these differences in global cognitive impairment by HF status were significant (P<0.001). The same pattern was seen for the WLL, WLD, and AFT tests (P<0.0001 for all).

Table 2.

Baseline Cognitive Status by HF Group in the REGARDS Study

Measure of cognition	Total (n=23 894)	No HF (n=22 930)	Incident HF (n=584)	HF decedents (n=380)	P value‡
SIS, impaired, n (%)	1855 (7.8)	1749 (7.6)	54 (9.2)	52 (13.7)	<0.001
SIS, mean±SD	5.6±0.7	5.6±0.7	5.5±0.8	5.4±0.8	<0.001
WLL, mean±SD†	17.5±5.1	17.6±5.1	15.8±4.8	14.5±5.0	<0.001
WLD, mean±SD†	6.5±2.2	6.5±2.2	5.7±2.1	5.4±2.3	<0.001
AFT, mean±SD*	17.0±5.8	17.1±5.8	15.1±5.2	14.0±5.2	<0.001

SIS scores range from 0 to 6, <5 impaired. WLL scores range from 0 to 30. WLD scores range from 0 to 10. AFT scores ranged from 0 and greater. AFT indicates Animal Fluency Test; HF, heart failure; REGARDS, Reasons for Geographic and Racial Differences in Stroke; SIS, Six‐Item Screener; WLD, Word List Delayed Recall; and WLL, Word List Learning.

^*Out of 16 168 sample.

^† Out of 16 924 sample.

^‡ The P value in Table 2 is comparing all HF groups to one another (except total column) using an ANOVA test.

Cognitive Trajectory by HF Status: Global Cognitive Function

The cognitive trajectories for participants after incident HF hospitalization compared with those without HF hospitalizations are shown by cognitive domain in Figures 1 and 2. These graphs depict predicted mean cognitive scores over a 5‐year period for participants of the same age with and without a first HF hospitalization; they are adjusted for sociodemographics, clinical comorbidities, and health behaviors. Minimally (age) adjusted models are shown in Figure 1. With respect to the SIS (Figure 1A), the mean SIS score declined over time for participants with and without incident HF hospitalization; however, the rate of decline in the SIS was significantly faster among those with incident HF hospitalization (difference in slopes, −0.02 [−0.03 to −0.08]; P<0.002). The fully adjusted models are shown in Figure 2. Here, the mean SIS score declined among both groups after adjustment (Figure 2A). Participants who remained HF hospitalization‐free had a 0.01‐point decline per year (−0.01 to −0.01; P<0.001) and those with incident HF hospitalization had a 0.04‐point decline per year (−0.06 to −0.03; P<0.001). The rate of decline in the SIS was significantly faster among those with incident HF hospitalization compared with those who remained HF hospitalization free (difference in slopes, −0.031 [−0.047 to −0.016]; P<0.001).

Figure 1. Minimally adjusted graphs comparing the cognitive trajectory of REGARDS participants after incident HF hospitalization event compared with those with no incident HF hospitalization event.

Minimally adjusted^# graphs comparing the cognitive trajectory for SIS (A), WLL (B), WLD (C), and AFT (D). Among REGARDS participants after incident HF hospitalization event (black line) compared with those with no incident HF hospitalization event (gray line). AFT indicates Animal Fluency Test; DIF, difference; HF, heart failure; REGARDS, Reasons for Geographic and Racial Differences in Strokes; SIS, Six‐Item Screener; WLD, Word List Delayed Recall; WLL, Word List Learning. * SIS scores range from 0 to 6; mean (SD) number of assessments over follow‐up for no HF group was 8 (4) and for HF group was 9 (3). ^†WLL scores range from 0 to 30; mean (SD) number of assessments over follow‐up for no HF group was 3 (1) and for HF group was 3 (1). ^‡WLD scores range from 0 to 10; mean (SD) number of assessments over follow‐up for no HF group was 3 (1) and for HF group was 3 (1). ^§ AFT) scores ranged from 0 to greater; mean (SD) number of assessments over follow‐up for no HF group was 3 (2) and for HF group was 4 (2). ^||Models used age as the time scale and include random intercepts for each participant and random slopes for age at the time of cognitive assessment. ^#Time 0 refers to the time of incident HF hospitalization event for those in the HF group. **Bayesian information criterion values for models: SIS, 414466.7; WLL, 282008.1; WLD, 198225.7; AFT, 308 231.

Figure 2. Fully adjusted graphs comparing the cognitive trajectory of REGARDS participants after incident HF hospitalization event compared with those with no incident HF hospitalization event.

Fully adjusted^# graphs comparing the cognitive trajectory for SIS (A), WLL (B), WLD (C), and AFT (D). Among REGARDS participants after incident HF hospitalization event (black line) compared with those with no incident HF hospitalization event (gray line). AFT indicates Animal Fluency Test; DIF, difference; HF, heart failure; REGARDS, Reasons for Geographic and Racial Differences in Strokes; SIS, Six‐Item Screener; WLD, Word List Delayed Recall; and WLL, Word List Learning. *SIS scores range from 0 to 6; mean (SD) number of assessments over follow‐up for no HF group was 8 (4) and for HF group was 9 (3). ^†WLL scores range from 0 to 30; mean (SD) number of assessments over follow‐up for no HF group was 3 (1) and for HF group was 3 (1). ^‡WLD scores range from 0 to 10; mean (SD) number of assessments over follow‐up for no HF group was 3 (1) and for HF group was 3 (1). ^§AFT scores ranged from 0 to greater; mean (SD) number of assessments over follow‐up for no HF group was 3 (2) and for HF group was 4 (2). ^||Time 0 refers to the time of incident HF hospitalization event for those in the HF group. ^#Adjusted for sociodemographics (sex, race, education, income, region of residence, social network, social isolation), comorbidities (hypertension, history of heart disease, dyslipidemia, diabetes, body mass index, albumin–creatinine ratio, atrial fibrillation, depression, physical component summary score, mental component summary score) and health behaviors (smoking, alcohol use, and exercise). **Bayesian information criterion values for models: SIS, 290853.4; WLL, 202568.3; WLD, 141749.4; AFT, 221682.4.

Cognitive Trajectory by HF Status: Learning and Memory

With respect to WLL (learning) and WLD (memory), mean scores declined over a 5‐year period for those both with and without incident HF hospitalization (Figures 1B and 1C; 2B and 2C). In the fully adjusted models, for WLL, participants who remained HF hospitalization free had a 0.092‐point decline per year (−0.10 to −0.09; P<0.001) whereas those with an incident HF hospitalization had a 0.13‐point decline per year (−0.35 to 0.09; P=0.239). The rate of decline in WLL was not significantly faster among those with incident HF hospitalization compared with those who remained HF hospitalization‐free (difference in slopes −0.040 [−0.259 to 0.180]; P=0.723) (Figure 2B). A similar pattern was seen for WLD. That is, participants who remained HF hospitalization free had a 0.04‐point decline per year (−0.05 to −0.04; P<0.001); those with an incident HF hospitalization also had a 0.04‐point decline per year, but it was not significant (−0.13 to 0.06; P=0.46). Like WLL, the rate of decline in WLD was not significantly faster among those with incident HF hospitalization compared with those who remained HF hospitalization free (difference in slopes, −0.005 [−0.091 to 0.101]; P=0.917) (Figure 2C).

Cognitive Trajectory by HF Status: Executive Function

With respect to AFT (executive function) (Figures 1D and 2D), mean scores declined for participants regardless of HF status. In the fully adjusted models, participants who remained HF hospitalization free had a 0.24‐point decline per year (−0.25 to −0.24; P<0.001) and participants who had an incident HF hospitalization had a 0.42‐point decline per year (−0.62 to −0.22; P<0.001) (Figure 2D). The rate of decline in AFT was not significantly faster among those with incident HF hospitalization compared with those who remained HF hospitalization free (difference in slopes, −0.176 [−0.376 to 0.025]; P=0.08).

Cognitive Trajectory by HF Subtype

The cognitive trajectories among adults with incident HF hospitalization by HF subtype (HFrEF versus HFpEF) are shown in Figure 3. Participants with HFpEF had lower mean SIS scores at the time of diagnosis compared with those with HFrEF; however, this difference was not clinically meaningful (<0.5 SD). Participants with both HFpEF and HFrEF experienced statistically significant declines in mean SIS; each had a 0.03‐point decline per year (P<0.028 for HFpEF and P<0.037 for HFrEF) (Figure 3A). While participants with both HFpEF and HFrEF experienced declines in mean WLL and WLD scores, the declines were not statistically significant (Figure 3B and 3C). With respect to AFT, however, statistically significant differences were seen. Participants with HFpEF started with higher scores (versus those with HFrEF) and had a 0.55‐point decline per year (−0.81 to −0.28; P<0.000) which was statistically significant; those with HFrEF had a 0.15‐point decline per year (−0.43 to 0.130; P=0.291), which was not statistically significant (Figure 3D). Notably, the difference in cognitive decline between HFpEF and HFrEF was found to be statistically significant for AFT, with participants with HFpEF declining more rapidly per year (P=0.044).

Figure 3. Cognitive trajectory of REGARDS participants after incident HF hospitalization, by HF subtype.

Cognitive trajectory for SIS (A), WLL (B), WLD (C), and AFT (D). Among REGARDS participants after incident HF hospitalization, by HF subtype (HFpEF=black line; HFrEF=gray line). ^||AFT indicates Animal Fluency Test; DIF, difference; HF, heart failure; REGARDS, Reasons for Geographic and Racial Differences in Strokes; SIS, Six‐Item Screener; WLD, Word List Delayed Recall; and WLL, Word List Learning. *SIS scores range from 0 to 6; mean (SD) number of assessments over follow‐up for no HF group was 8 (4) and for HF group was 9 (3). ^†WLL scores range from 0 to 30; mean (SD) number of assessments over follow‐up for no HF group was 3 (1) and for HF group was 3 (1). ^‡WLD scores range from 0 to 10; mean (SD) number of assessments over follow‐up for no HF group was 3 (1) and for HF group was 3 (1). ^§AFT scores ranged from 0 to greater; mean (SD) number of assessments over follow‐up for no HF group was 3 (2) and for HF group was 4 (2). ^||Out of incident HF sample (n=584). ^#Difference in 1‐year change in AFT after HF with preserved ejection fraction versus HF with reduced ejection fraction is statistically significant. Cochrane's Q test used to determine significance. **P value remains significant after Benjamini–Hochberg correction for type 1 error with multiple tests. ^¥Time 0 refers to the time of incident HF hospitalization event for those in the HF group.

Discussion

In this longitudinal study of Black and White community‐dwelling adults, participants who developed incident HF hospitalization experienced a steeper decline in global cognitive impairment compared with those who remained free of HF hospitalization, a finding that persisted after adjustment for age, other sociodemographic factors, and clinical comorbidities. Although statistically significant, this difference did not meet the threshold for a clinically significant change in global cognition, based on the accepted threshold of 0.5 SD. For those with incident HF hospitalization (compared with those who remained HF hospitalization free), the rate of cognitive decline was not steeper for the cognitive domains of learning and memory; it was for executive function but fell short of statistical and clinical significance. We did see a statistically significant difference in the decline of executive function by HF subtype, with participants with HFpEF experiencing more rapid declines than those with HFrEF. Significant differences in other cognitive domains by HF subtype were not seen.

Our study builds upon a few studies that have examined the trajectory of cognitive decline after a diagnosis of HF. For example, in the Cardiovascular Health Study, Hammond et al found that declines in global cognitive function were faster among community‐dwelling older adults who developed incident HF compared with those who did not but did not vary by HF subtype. ⁹ HF was not associated with significant declines in processing speed in their study, which occurred between 1989 and 1999. In another study of octogenarians in Sweden from 1991 to 2002, participants with HF experienced more rapid rates of decline in episodic memory over time compared with those without HF; there were no differences across the domains of processing speed, visuospatial ability, and semantic memory. ²⁸ Our study, which includes more contemporary data, adds to this body of literature in that we also assessed global cognitive impairment (SIS), confirming that incident HF hospitalization is associated with declines in this domain, while also assessing the domains of memory, learning, and executive function (which includes semantic memory). Similar to these other studies, we did not find that incident HF hospitalization was associated with significant declines across these other cognitive domains. Reasons for this remain unclear and warrant additional study in other cohorts.

Although evidence has accumulated that impaired cardiac function may hasten cognitive decline, most studies on cognition and HF are cross‐sectional. Among those that are longitudinal, it remains unclear if cognitive decline precedes the HF event or if cognition declines more rapidly subsequent to the HF diagnosis. ² , ²⁹ , ³⁰ , ³¹ Previously, we found that 14.9% of REGARDS participants had global cognitive impairment at the time of their incident HF hospitalization, a prevalence that was similar to an age‐, sex‐, and race‐matched group of participants with cardiovascular comorbidities but without incident HF hospitalization. ⁶ Building on these findings, here we found that there are likely to be differences in cognition depending on whether participants survived the incident HF hospitalization. That is, we found that baseline cognitive scores were lower across all 4 domains among the HF decedents, compared with those with incident HF hospitalization who survived and compared with those never hospitalized with the disease. Future studies are needed to confirm these findings and to assess the trajectory of cognitive decline before incident HF hospitalization to determine if these patterns exist. Declines in cognitive function combined with the development of HF may be an important signal of approaching end of life. Additionally, as cognitive trajectories may be heterogeneous after incident HF, future research on characteristics related to rate of cognitive decline among individuals with HF would also be valuable.

There are numerous mechanisms by which HF contributes to cognitive impairment. ³² Adults with HF, including those in this study, often have multiple chronic conditions, which include hypertension and atrial fibrillation, all of which impair hemodynamics. It is thought that cerebral hypoperfusion, disruption of the blood–brain barrier, oxidative damage, and brain‐derived cytokines result in impaired cognition among adults with HF, with the severity of the impairment increasing as the HF worsens. ² , ³ , ⁸ , ³³ , ³⁴ It is likely that these mechanisms are responsible for the myriad of cognitive deficits that adults with HF can have, including attention, learning ability, working memory, executive function, episodic memory, and processing speed. ² , ⁸ These deficits can impact patients' lives, including their ability to carry out HF self‐care activities such as recognizing worsening symptoms and adhering to lifestyle and medication regimens. ³⁵ , ³⁶ , ³⁷ For example, a prior study found that mild cognitive impairment, assessed with the Montreal Cognitive Assessment, was an independent predictor of the ability of a patient with HF to independently recognize symptom changes and make appropriate self‐care decisions. ³⁶ While our adjusted models found that incident HF hospitalization was associated with declines across the global cognitive function and executive function domains, the differences in cognition between individuals with and without incident HF hospitalization seen in this study population were modest and unlikely to adversely impact patients' HF self‐care.

There are a few potential reasons as to why we did not observe clinically meaningful declines in cognition. First, a ceiling effect could be occurring, whereby participants who are operating at a high level to begin with would need to experience a large drop in cognitive function for a screening tool to detect a deficit. Second, the cohort's overall health could be a factor. The REGARDS study recruited community‐dwelling adults, and there could have been differential loss to follow‐up among those with the greatest cognitive decline. Finally, while we assessed 4 domains of cognition, which is more comprehensive than most other studies, the literature suggests that other domains are impacted by HF, and we were unable to assess these here.

We had hypothesized that we would detect differences in cognitive decline by HF subtype since prior cross‐sectional studies have observed that lower left ventricular ejection fraction is associated with worse cognition. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ Longitudinal cohort studies of cognitive decline by HF subtype, however, have been lacking. ⁹ Notably, we found that declines in executive function differed by HF subtype, with participants with HFpEF experiencing more rapid declines than those with HFrEF. This finding supports recent evidence that different types of cardiac dysfunction have differential patterns of performance on cognitive and neuropsychological tests. ³⁹ , ⁴¹ It also supports a growing call for research that can elucidate the biologic mechanisms behind cognitive impairment in HFpEF, which have not been clearly established. ⁴¹ , ⁴² Finally, although it needs to be confirmed in other studies, it may signal that clinicians caring for adults with HFpEF should not be aware that patients may have difficulty carrying out key aspects of HF self‐care (many of which require executive function) but consider strategies to both mitigate decline and accommodate deficits through the support of caregivers, for example. ³⁷ , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶

Strengths and Limitations

Our study has several strengths. We had longitudinal cognitive assessments among a large cohort to estimate cognitive decline after incident HF hospitalization. Incident HF hospitalizations were expert adjudicated on the basis of medical record review. REGARDS systematically measured cognitive domains commonly impaired among individuals with HF including global cognition, learning, memory, and executive function. However, our study is not without limitations. First, results are generalizable only to community‐dwelling adults with HF hospitalization, and not those who were in institutional settings after their HF diagnosis, or those with a diagnosis of stroke. Next, we studied incident HF hospitalization, and some individuals may be diagnosed as outpatients. Our results should not be generalized to incident HF diagnoses. Beyond HF subtype, we were not able to account for some factors that impact the severity of the HF diagnosis or cognitive decline over time, including hospitalizations, adherence to medication and other treatments, and interactions with the health system. Finally, because participants with worse cognition may die, stop participating, or require a proxy, we may have underestimated the degree of cognitive decline in this population. Additionally, the cognitive assessments may not be sensitive to small to moderate changes among participants with good cognitive function at baseline (ie, ceiling effect).

Conclusions

Incident HF hospitalization was associated with a faster decline in global cognitive function compared with similarly aged participants without HF hospitalization in this longitudinal study. This pattern was not seen for the other 3 cognitive domains we assessed. Differences in the rate of cognitive decline after incident HF hospitalization by HF subtype were only seen for the domain of executive function. Our findings contribute to our understanding of cognitive trajectories in HF and provide information for counseling patients with HF and caregivers about cognition following an incident HF hospitalization.

Sources of Funding

The REGARDS study was supported by cooperative agreement U01 NS041588 cofunded by the National Institute of Neurological Disorders and Stroke and the National Institute on Aging, National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institute on Aging. Representatives of the National Institute of Neurological Disorders and Stroke were involved in the review of the manuscript but were not directly involved in the collection, management, analysis, or interpretation of the data. Additional funding for this study was provided by National Heart, Lung, and Blood Institute grants R01HL080477 and R01 HL165452 (Safford/Levitan). Dr Sterling is supported by the National Heart, Lung, and Blood Institute (K23HL150160 and R01HL169312), which pertain to the work; and an American Heart Association Award (23SCISA1142170), outside the submitted work.

Disclosures

Dr Levitan reports research funding from Amgen, Inc, unrelated to the current work. Dr Safford receives salary support from Amgen for investigator‐initiated research. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S2

Figures S1–S3