Examination of Attention, Executive Function, and Memory as Predictors of Mortality Risk in Adults with Systolic Heart Failure

Abstract

Background:

The prevalence and impact of cognitive impairment in heart failure (HF) is increasingly recognized. Converging evidence points to global cognitive function as predictive of prognosis in adults with HF when assessed with screening tools. Additional work is needed to understand which domains of cognitive function are most relevant for prognosis.

Aims:

The present study sought to examine associations between domains of cognitive function and mortality risk in adults with HF.

Methods:

In the present prospective, observational cohort study, global cognitive function, attention, executive function, and memory were assessed by means of comprehensive neuropsychogical battery in adults with systolic HF. Mortality data were obtained from the National Death Index (median follow-up: 2.95 years). Relationships among each cognitive domain and mortality were assessed with Cox regression. Covariates included age, sex, HF severity, comorbidity and depressive symptoms.

Results:

Participants were 325 patients with systolic HF with a mean age of 68.6 years (59% male, 73% Caucasian). Following covariate adjustment, better global cognitive function, attention, and executive function were related to decreased mortality risk.

Conclusions:

Future research is needed to clarify the underlying mechanisms of the association between cognitive impairment and mortality.

Introduction

Heart failure (HF) remains a significant public health concern and a burden for healthcare providers seeking to improve morbidity and mortality.¹ Declining cardiovascular disease deaths over the past decade have offered promise of improved disease management.² However, increased survival has also coincided with higher rates of comorbidity and medication use for adults with HF.³ Thus, longer-term and more complex self-management needs exist.

Cognitive impairment is increasingly recognized as prevalent in HF, with estimates ranging from approximately 20% to over 70%.^{4, 5} Individuals with HF commonly display impairments in multiple domains of cognitive functioning that are impacted by vascular risk factors, including attention and executive function.^{6, 7} Impairments in memory have also been observed.⁸ Several reports have suggested a link between cognitive function and indicators of prognosis in patients with HF. For example, global cognitive function assessed through screening tools is associated with readmission and mortality in HF over varied follow-up periods.^9–12

Converging evidence points to the predictive prognostic value of global cognitive impairment. However, less work has investigated whether certain domains of cognitive function are differentially related to outcomes such as mortality in HF. Examination of domain-specific associations with prognosis may assist in understanding what component(s) of cognitive function drive the association with poorer outcomes. Pressler and colleagues examined the relationships among different domains of cognitive function (i.e., language, working memory, memory, visuospatial ability, psychomotor speed, and executive function) and 1-year mortality in 166 outpatients with HF.¹³ Global cognitive function, working memory, psychomotor speed, executive function, and visual and verbal memory were predictive of 1-year mortality, with visual memory impairment emerging as the strongest cognitive predictor of death at one year follow-up.¹³ As health status in HF worsens over time, it is important to understand whether specific domains of cognitive function is predictive of mortality over a longer follow-up period. This will assist in providing direction for future research investigating mechanisms of this association and in focusing intervention efforts to mitigate the effects of dysfunction in specific domains. Thus, the aim of the current study was to extend prior findings by examining domain-specific cognitive predictors of mortality risk beyond one year of follow-up in adults with HF. It was hypothesized that better global cognitive function, attention, executive function, and memory would each be related to decreased mortality risk.

Method

Participants

The current sample was drawn from a parent study called the Heart Adherence, Behavior, and Cognition (HeartABC) study (ClinicalTrials.gov Identifier: NCT01461629). HeartABC was a prospective observational cohort study of relationships between psychosocial factors, cognitive function, and self-management behaviors in adults 50 to 85 years with systolic HF recruited from inpatient and outpatient cardiology services. A detailed description of the HeartABC study and inclusion criteria have previously been published.^14–16 Of note, exclusion criteria included cardiac surgery within 3 months of enrollment, history of neurological disorder or injury (e.g., Alzheimer’s disease, dementia, stroke, seizures), moderate or severe head injury, past or current significant psychiatric disorders (e.g., psychotic disorders, bipolar disorder, learning disorder, developmental disability), renal failure requiring dialysis, or untreated sleep apnea, and current substance abuse or within the past 5 years.

Measures

Cognitive Function.

In addition to a measure of global cognitive function, three additional domains of cognitive function commonly impacted by HF were assessed. First, to assess global cognitive function, the Modified Mini-Mental Status Examination (3MS) raw score was used.¹⁷ Possible scores range from 0 to 100, and higher scores reflect better function.

For each of the additional cognitive domains (i.e., attention, executive function, and memory) multiple neuropsychological tests with strong validity and reliability were administered. For each test, raw scores were converted to age-adjusted scaled scores using published normative data. Next, scaled scores were converted to T scores with a mean of 50 and standard deviation of 10. In order to create a score that represented each cognitive domain, a composite scores was created to reflect performance across tests measuring specific abilities within each domain. A composite score for each domain was created by averaging T scores of relevant tests. T scores <=35 (i.e., 1.5 SD below the mean) were considered suggestive of cognitive impairment. The following tests were used to assess each domain:

Attention:

Three attention tasks were administered. The Stroop Word and Color subtests, ¹⁸ Trail Making Test A (TMT-A),^{19, 20} and Letter–Number Sequencing,²¹ were used to assess an individual’s ability to attend to and process information, including psychomotor speed, visual sequencing, and working memory.

Executive Function:

Three tasks of executive function were included. The Stroop Color Word subtest,¹⁸ Trail Making Test B (TMT-B),^{19, 20} and the Frontal Assessment Battery²² were used to measure planning, problem-solving, inhibition, and reasoning skills.

Memory:

The Rey Auditory Verbal Learning Over Time, True Hits, Short Delay, and Long Delay scores were used to determine the ability to retain and recall verbal information.

Covariates.

Demographic, medical, and clinical covariates known to be associated with cognitive function and/or mortality risk were selected a priori. Covariates included age, sex, comorbidity, depressive symptoms, and disease severity. Sex was collected from self-report. Depressive symptoms were assessed using the Patient Health Questionnaire—9 (PHQ-9).²³ HF severity was determined using physician-documented New York Heart Association (NYHA) class,²⁴ which was verified at baseline through assessment of self-reported HF symptoms and limitations. The Charlson comorbidity index was used to summarize the presence of comorbidities.²⁵

All-cause Mortality.

The Centers for Disease Control and Prevention’s National Death Index (NDI) service was used to obtain survival status of study participants. The NDI compares study participants with death records from state vital statistics offices based on demographic variables obtained as part of the initial data collection and identifies vital status and coded cause of death information. Death records through December 2014 were searched.

Procedure

The HeartABC study was approved by the institutional review boards of the relevant sites (Kent State University, Summa Health System, and University Hospitals of Cleveland). The investigation conforms with the principles outlined in the Declaration of Helsinki. Recruitment occurred from August 2010 through October 2013. Detailed descriptions of study procedures have been described previously.^14–16 For the present analysis, data collected during the first two study visits were used. During the first visit, demographic, psychosocial, and medical history was collected through a combination of self-report and interviews. One to two weeks later, a trained research assistant administered a comprehensive neuropsychological battery, which lasted approximately 60 minutes. No additional contact with participants was required for the obtainment of NDI data.

Analytic Plan

The present study represents a secondary analysis from the HeartABC study. In the parent study, of 1140 screened patients, 425 individuals met inclusion criteria and 372 agreed to participate. Of the total 372 enrolled HeartABC individuals, 20 participants withdrew at Time 1, 15 withdrew at Time 2, and 5 withdrew at each Time 3 and Time 4. Listwise deletion was used to exclude individuals missing data on variables of interest (i.e., age, sex, PHQ-9 scores, NYHA class, education, cognitive domains). This exclusion yielded a final sample of 325. Descriptive statistics (means, standard deviations, frequencies) were used to summarize sample characteristics. To examine differences in baseline characteristics between individuals who died versus individuals who survived, t-tests were calculated for continuous variables and Pearson χ² tests were calculated for categorical variables.

Cox proportional hazards regression ²⁶ was used to determine the relationship between demographic, medical, and clinical baseline variable, cognitive function and all-cause mortality. First, univariate models were performed to examine the association between each variable entered alone and mortality risk. Second, for each domain of cognitive function (i.e., attention, executive function, and attention), a Cox proportional hazards regression model with two steps were conducted. Block 1 included only covariates (i.e., age, sex, comorbidity, NYHA class, PHQ-9 scores). Block 2 included the cognitive function composite domain score.

Results

Sample Characteristics

Participants were 325 patients with systolic HF. Fifty-seven deaths (17.5%) occurred during the follow-up period, which averaged 1011.09 ± 333.54 days (median = 1072). Average scores on a measure of global cognitive function (3MS) was 92.10 (SD = 6.43), with 28.6% scoring below 90. Approximately 11.7% (n = 38) of the sample demonstrated impairment in attention. Executive function impairment was observed in 9.8% (n = 32). Memory deficits were noted in 5.8% (n = 19). See Table 1 for full demographic, medical, and clinical characteristics stratified by survival status.

Table 1.

Demographic and Clinical Characteristics of Participants at Baseline (n = 325).

	Total Sample	Mortality
		Yes	No	p value
Sample size (n)	325	57	268
Age	68.60 (9.62)	72.32 (9.02)	67.81 (9.58)	<.01
Male	193 (59.4)	42	151	<.05
Caucasian	238 (73.2)	45	193	ns
Highest Education Level
11th grade or less	37 (11.4)	7	30
High School	95 (29.2)	15	80
Some college/	121 (37.2)	21	100
Technical/trade school
Bachelor’s degree	42 (12.9)	8	34
Master’s degree	30 (9.2)	6	24
Months since HF diagnosis (n = 318)	104.59 (108.43)	109.36 (104.87)	103.59 (109.33)	ns
Married	185 (56.9)	29	156	ns
Charlson	3.31 (1.74)	3.60 (1.93)	3.24 (1.69)	ns
NYHA
Class I	32 (9.8)	3	29
Class II	75 (23.1)	12	63
Class III	202 (62.2)	39	163
Class IV	16 (4.9)	3	13
PHQ-9	4.26 (4.40)	5.44 (4.55)	4.01 (4.33)	< .05
Cognitive Function T-scores
Attention Composite	44.44 (7.49)	42.81 (6.34)	44.78 (7.68)	< .05
Stroop Word	42.97 (9.30)	42.01 (10.40)	43.16 (9.06)	ns
Stroop Color	45.18 (9.56)	43.37 (9.53)	45.60 (9.54)	ns
Trails A	42.54 (10.30)	40.44 (8.26)	42.99 (10.64)	< .05
Letter-Number Sequencing	47.07 (10.37)	45.35 (8.91)	47.43 (10.64)	ns
Executive Function Composite	46.04 (8.00)	43.20 (6.69)	46.64 (8.13)	< .01
Stroop Color Word	45.20 (10.02)	42.67 (8.37)	45.74 (10.28)	<.05
Trails B	41.84 (11.87)	37.93 (11.34)	42.67 (11.83)	<.01
Frontal Assessment Battery	51.08 (8.16)	49.01 (8.08)	51.52 (8.12)	<.05
Memory Composite	47.95 (7.75)	47.14 (7.72)	48.12 (7.77)	ns
Learning over Time	49.54 (10.74)	49.33 (10.78)	49.59 (10.75)	ns
True Hits	49.18 (9.04)	48.49 (9.44)	49.32 (8.96)	ns
Short Delay	45.72 (10.80)	44.56 (11.23)	45.96 (10.71)	ns
Long Delay	47.36 (9.39)	46.16 (9.94)	47.62 (9.26)	ns

Note. Continuous variables are described using means and standard deviations. p values for continuous variables calculated using t-tests. Categorical variables are described using frequencies and percentages. p values for categorical variables calculated using Pearson χ2 tests.

Abbreviations: Charlson = Charlson Comorbidity Index; NYHA = New York Heart Association; PHQ-9 = Patient Health Questionnaire—9.

Cognitive function and mortality risk

Cox proportional hazards regression analyses were performed to test the relationships between cognitive function domains (i.e., global cognitive function, executive function, attention, and memory) and all-cause mortality risk with and without controlling for covariates. Univariate analysis indicated better global cognitive function (3MS scores; HR = .95, 95% CI: .91 - .98), attention (HR = .96; 95% CI: .93 – .99) and executive function (HR = .95; 95% CI: .92 - .98), were associated with decreased mortality risk. No univariate relationship emerged between memory and all-cause mortality (HR = .99; 95% CI: .96 – 1.02). See Table 2.

Table 2.

Univariate relationships between demographic, clinical, and cognitive factors and mortality risk (n = 325).

	All-Cause Mortality
	HR (95% CI)	p value
Age	1.05 (1.02 – 1.08)	< .01
Sex	1.83 (1.02 – 3.31)	< .05
Charlson	1.09 (.95 – 1.25)	ns
NYHA	1.35 (.75 – 2.44)	ns
PHQ-9	1.05 (1.00 – 1.10)	<.05
Global Cognitive Function	.95 (.91 – .98)	< .01
Attention	.96 (.93 – .99)	< .05
Executive Function	.95 (.92 –.98)	< .01
Memory	.99 (.96 – 1.02)	ns

Abbreviations: NYHA = New York Heart Association

PHQ-9 = Patient Health Questionnaire—9.

Note. Sex: Male = 1, Female = 0.

When covariates were entered simultaneously into a multivariate model, older age was associated with higher mortality risk (HR = 1.05, 95% CI: 1.02 – 1.08). Also, male sex predicted higher mortality risk (HR = 1.92, 95% CI: 1.06 – 3.48). Depressive symptoms assessed with the PHQ-9 also were associated with increased risk (HR = 1.07; 95% CI: 1.02 – 1.12). Following adjustment for covariates, better global cognitive function remained associated with decreased mortality risk (HR = .95, 95% CI: .92 - .99). Similarly, after controlling for covariates, higher attention (HR = .96, 95% CI: .92 - .99) and executive function were related to decreased mortality risk (HR = .95; 95% CI: .92 - .98); see Table 3.

Table 3.

Multivariate Relationships Among Cognitive Domains and Mortality Risk (n = 325) after controlling for age, sex, comorbidity, NYHA class, and PHQ-9.

	HR (95% CI)
Global Cognitive Function	.95 (.92 – .99)
Attention	.96 (.92 – .99)
Executive Function	.95 (.92 – .98)
Memory	.99 (.96 – 1.03)

Abbreviations: NYHA = New York Heart Association; PHQ-9 = Patient Health Questionnaire—9.

Note: Each cognitive domain was examined in a separate model.

As a post-hoc analysis, we further examined differences in individual cognitive test between individuals who survived versus those who did not (see Table 1). Consistent with the Cox regression analyses, each executive function task differed between those who survived and those who did not (p’s < .05). Of the attention-related tasks, only TMT-A, a measure of psychomotor speed, differed according to follow-up survival status (t(323) = 2.01, p < .05). No memory sub-tests differed across survival status.

Discussion

The present study examined associations between global cognitive function, attention, executive function, memory, and mortality risk in a stable sample of systolic HF outpatients with nearly three years of follow-up. As hypothesized, better global cognitive function, attention, and executive function were related to decreased mortality risk even after controlling for covariates. For both global cognitive function and executive function, a 1-unit increase in cognitive function was associated with a 5% decrease is mortality risk. Each 1-unit increase in attention was associated with a 4% decrease in mortality risk. Contrary to our hypothesis, memory was not associated with mortality risk in unadjusted or adjusted models.

These findings extend prior research that has primarily focused on correlates of poor prognosis or mortality with a shorter duration of follow-up. Consistent with prior work,¹³ global cognitive function, executive function, and attention were predictive of mortality risk in this sample of adults with systolic HF. Reduced cognitive function likely serves as an indicator of more severe disease pathology. For example, observed rates of cognitive impairment tend to increase alongside impairment in left ventricular ejection fraction.^{27, 28} Furthermore, deficits in executive function and attention may identify individuals at greater risk for subsequent physical and cognitive decline. Indeed, lower baseline cognitive performance has been noted in older adults who subsequently experienced decline in gait speed ²⁹ and motor performance.³⁰ In addition, HF requires ongoing disease self-management that spans symptom monitoring, medication adherence, daily weighing, and sodium restriction. Individuals with global, attentional, or executive deficits are likely to struggle with self-care decision making in response to symptom exacerbation that may require action beyond one’s daily regimen.³¹ Inadequate or delayed responding to symptom exacerbation can lead to re-hospitalization or worsening prognosis. Alternatively, other behavioral factors, such as lower levels of physical activity, may be related to both cognitive deficits, as well as increased mortality risk at follow-up.³² Alosco and colleagues (2014) reported that lower levels of physical activity assessed via accelerometer predicted lower attention, executive function, and cerebral blood flow 12 months later, and suggested that physical inactivity may exacerbate risk for subsequent cognitive decline among adults with HF. ³² In the current study, individuals with HF who are physically inactive may have been more likely to display poorer cognitive functioning at baseline and risk for additional cognitive and physical decline. Given that the contributors to associations between cognitive function and mortality risk are likely multifactorial, future work is needed to test these possibilities.

The lack of association between memory and mortality risk is surprising for multiple reasons. First, the current findings are inconsistent with those previously reported by Pressler,¹³ which suggested memory was an important cognitive predictor of mortality. Some methodological differences may partially explain the contrasting results, such as differences in length of follow-up, the specific selection of tests, covariate adjustment, or treatment of scores in analyses (i.e., raw versus scaled scores). Additionally, our sample was slightly older, excluded individuals with baseline cognitive disorders, included a lower proportion of males, who are known to display poorer memory performance relative to females, and included a smaller proportion of individuals with NYHA Class IV HF. Given the conflicting findings, replication of the current findings in varied samples would be prudent.

Second, previously reported findings also from the HeartABC study indicated that medication adherence was more closely associated with memory than executive function and attention ¹⁵ and was also related to increased mortality risk.³³ As such, it is surprising memory was unrelated to mortality in the present analysis. As memory represented the cognitive domain with the smallest percentage of impairment, a sample with a greater proportion of impaired memory performance may reveal an association between memory and a prognostic outcome beyond one year of follow-up.

Other demographic, psychosocial, and behavioral mechanisms may contribute to the present findings, including the lack of observed association between memory and mortality risk. Individuals displaying signs of cognitive difficulties that are apparent to family or spouses may receive additional compensatory support in self-management. Nonetheless, these findings highlight the complex associations between cognitive function and self-management behaviors in adults with HF. Future studies are needed to evaluate whether changes in memory over time may relate to mortality risk and whether any relationship between memory function and prognosis is impacted by factors such as compensatory support.

Although not the primary focus of the present analyses, as part of screening, individuals with NYHA class IV HF were excluded for safety reasons. During the study, NHYA class was re-assessed and some participants who did not meet criteria for class IV HF upon study entry did upon subsequent assessment. Neither NYHA class nor the Charlson comorbidity index were related to mortality in this study. This finding is consistent with a prior report that a Charlson score of ≥ 4 was associated with 12 year mortality in individuals without HF, but not those with HF.³⁴ Regarding NYHA class, future studies may benefit from inclusion of additional metrics of disease severity, including objective assessments. Nonetheless, our sample was primarily comprised of those with Class II – III HF, which may have contributed to lower than expected rates of cognitive impairment and mortality.

The present findings should be interpreted in light of the study’s limitations. First, although exclusion of certain comorbidities known to contribute to cognitive impairment was important to elucidate the role of cognitive function in HF outcomes, the results may not generalize to other populations (i.e., HF with preserved ejection fraction) or HF patients with additional comorbidities (i.e., untreated sleep apnea, end-stage renal disease). It is likely that individuals with additional comorbidities may be more adversely impacted by cognitive impairment. Second, this study did not incorporate methods such as neuroimaging to better understand the underlying physiological correlates of our outcome measures. Additionally, changes in cognitive functioning over multiple time points were not examined. Individuals may experience a slight improvement in cognitive function if improvements in aspects of health occur, or alternatively may experience events such as cerebrovascular accident that lead to a significant change in cognitive abilities. Future studies incorporating repeated assessments of cognitive function over time will further clarify the present findings.

In sum, the current investigation of cognitive function and mortality risk in HF indicated that higher global cognitive function, executive function and attention, but not memory, were associated with decreased mortality risk. Clinically, the present findings highlight the importance of screening for cognitive impairment and monitoring individuals with diminishing cognitive function. Future studies to define the mechanisms of this relationship may ultimately provide insight to mitigate the effects of cognitive impairment on poorer prognosis in HF. Qualitative inquiry of patients and caregivers of those with HF may also increase understanding of patient perspectives on the role of cognitive function in HF self-management and assist in development of acceptable interventions to support adults with HF and cognitive decline.