Deficits in Self-Reported Initiation are Associated with Subsequent Disability in ICU Survivors

Abstract

Purpose:

To determine whether deficits in a key aspect of executive functioning, namely, initiation, were associated with current and future functional disabilities in ICU survivors.

Methods:

A nested sub-study within a two-center prospective observational cohort. We used 3 tests of initiation at 3 and 12 months: the Ruff-Total Unique Design (Ruff-TUD), Controlled Oral Word Association (COWA), and Behavior Rating Inventory of Executive Function initiation (BRIEF-I). Disability in Instrumental Activities of Daily Living (IADL) was measured with the Functional Activities Questionnaire (FAQ). We used a proportional odds logistic regression model to evaluate the association between initiation and disability. Covariates in the model included age, education, baseline FAQ, pre-existing cognitive impairment, comorbidities, admission severity of illness, episodes of hypoxia, and days of severe sepsis.

Results:

In 195 patients, after adjusting for covariates, only the BRIEF-I was associated with disability at any time point. Comparing the 25^th versus the 75^th percentile scores (95% CI) of the BRIEF-I at 3 months, patients with worse initiation scores had 5.062 times the odds (95% CI: 2.539, 10.092) of disability according to the FAQ at 3 months, with similar odds at 12 months (OR: 3.476, 95% CI (1.943, 6.216). Worse BRIEF-I scores at 3 months were associated with future disability at 12 months (OR 95% CI): 5.079 (2.579, 10.000).

Conclusions:

Executive function deficits acquired after a critical illness in the domain of initiation are common in ICU survivors, and when they are identified via self-report tools,, they are associated with current and future disability in IADLs.

Introduction

Despite surviving a life-threatening ordeal, many survivors of a critical illness often endure significant challenges in its aftermath, including poor health status and a reduced quality of life. ICU survivors, who often face delirium in the context of their critical illness are frequently burdened with unemployment, impaired physical ability, new psychosocial states, as well as long-term cognitive impairment of a severity that impacts higher order functioning (i.e., living independently, managing finances) after an episode of critical illness (1–8). The long-term cognitive impairment occurring after critical illness is in need of further, more detailed characterization, as questions exist regarding its course and mechanisms. Cognitive impairment following critical illness typically affect a range of domains including executive functioning, a domain of unique importance as it is strongly implicated in key aspects of daily functioning. Executive functioning has diverse sub-domains including attentional and inhibitory control, working memory, reasoning, problem solving, planning and cognitive flexibility, as well as initiation. Much of the existing work on post-ICU cognitive impairment has focused on abilities such as memory, processing speed, and visual-spatial construction. Despite this, deficits in executive functioning subdomains, like initiation (the ability to start or engage in a task), may be particularly important to assess as they could independently explain some of the severe disabilities seen in survivors of critical illness.

The significant impairments and disabilities experienced by ICU survivors are well known, entrenched, and broad in scope (7, 9–13). For example, our group has recently published a pilot investigation showing that nearly one-half of patients surviving a critical illness who were previous employed were newly unemployed at 12 months (14). The dynamics that underlie and drive the development of disability after critical illness – leading to unemployment, problems with medication management, poor financial planning, and eventually an inability to live independently, etc. – are unclear but could be due to impairments in the ability to act and thoughtfully initiate. In this investigation, we sought to test whether characteristics of initiation are associated with disability in instrumental activities of daily living (IADLs) at 3 and 12 months in survivors of critical illness.

Methods

Study Population and Setting

The parent cohort, The Bringing to light the Risk Factors and Incidence of Neuropsychological Dysfunction in ICU survivors (BRAIN-ICU) (NCT00392795), was a multicenter prospective observational cohort study conducted at both the Vanderbilt University Medical Center and the Saint Thomas Hospital medical and surgical intensive care units (ICUs) in Nashville, Tennessee. Details regarding the inclusion and exclusion criteria have been described in detail (7). Briefly, we included adults with respiratory failure, cardiogenic shock, or septic shock. Individuals with a recent ICU exposure (defined as: as mechanical ventilation within the past 2 months prior to study enrollment, >5 ICU days in the month before study enrollment, or those with greater than 72 hours of organ dysfunction within the current ICU admission) were excluded. In addition, we excluded patients who we anticipated would have significant difficulty participating in the in-hospital or postdischarge phases of the study secondary to: blindness, deafness, an inability to speak English, residence ≥ 200 miles from the study site, current homelessness, current active substance use or psychotic illness, those for whom informed consent could not be obtained, patients who were unlikely to survive the next 24 hours, as well as those with suspected neurological or neurocognitive disorder. Informed consent was obtained at enrollment from either the patient or a suitable surrogate if the patient lacked medical decision-making capacity. If a surrogate provided initial consent, and if the patient eventually regained capacity, the patient was either re-consented or allowed to discontinue from the study. At 3 and 12 months (+/− 2 months) after hospital discharge, a master’s level professional or doctoral level neuropsychologist assessed patients’ ability to perform IADLs and executive function. Patients included in this analysis consented to participate in a nested cohort study of BRAIN-ICU participants who survived their initial hospitalization, participated in both 3 and 12-month follow-up, and agreed to participate in a detailed assessment of executive function. The BRAIN-ICU study and this sub-study were approved by IRBs at Vanderbilt University Medical Center and the Saint Thomas Hospital.

Measuring Disability in Instrumental Activities of Daily Living

We assessed disability in IADLs (i.e, the ability to shop alone, pay bills, prepare a balanced meal, etc.) at baseline, 3 months and 12 months using the Pfeffer Functional Activities Questionnaire (FAQ) (15). The FAQ is a 10 question assessment of an individual’s ability to participate in instrumental activities of daily living with each question scored from 0 (independent) to 3 points (complete dependence). Scores range from 0 to 30, with higher scores indicating greater disability. In general, the FAQ was completed by the patient. In the rare situation where the patient was unable to fill out the FAQ themselves, a first-degree relative (i.e., spouse or adult child) were asked to complete the questionnaire. For the purpose of this analysis, disability according to the FAQ was defined as a score of greater than or equal to 9 (indicating complete dependence on at least 3 IADLs.

Measuring Executive Function

Initiation, an executive functioning subtype, was assessed using objective cognitive tests and one brief self-report measure. Relevant cut-off points which indicated impairment on each respective psychological assessment were laid out in each assessment’s instructions, however in general, impairment on any battery was defined as performance at least 1.5 standard deviations (SD) below the population mean.

We assessed an aspect of executive functioning, initiation, utilizing a battery of instruments which included: the Ruff total unique design (Ruff-TUD), Controlled Oral Word Association (COWA) test, and the BRIEF-I (15–18). The Ruff TUD assesses the individual’s ability to think flexibly using visual cues (16). The Ruff-TUD specifically tests nonverbal fluency or the ability to plan strategically and coordinate in a fluid manner the tasks specified by the examiner by drawing as many unique designs as possible within 60 seconds, connecting the dots in different patterns (16). A score of greater than or equal to 36 on the Ruff was considered unimpaired, with lower scores indicating impairment. The Controlled Oral Word Association test (COWA) is a measure of verbal fluency or the ability to spontaneously generate words corresponding to a specific command during a defined period of time, i.e., unique words that being with the letter “F” in 60 seconds (18). A score of greater than or equal to 35 on the COWA was considered unimpaired, with lower scores indicating impairment. The Behavior Rating Inventory of Executive Function® initiation (BRIEF-I) subscale is a self-administered screening tool used to assess the ability to start an action or activity and generate new or novel strategies (17). A score of greater than 65 on the BRIEF-I was considered unimpaired, with lower scores indicating impairment.

Management of Missing Data

To avoid the bias associated with missing covariates or independent variable data, we used multiple imputation with 100 sets. We did not impute missing outcome measures. Those individuals with a missing outcome value at either 3 or 12 months were excluded from analysis.

Statistical Analysis

We collected baseline patient characteristics at the time of enrollment (from medical record abstraction and patient or surrogate interview): age, sex, race, years of education, and preexisting cognitive impairment (Informant Questionnaire of Cognitive Decline in the Elderly; IQCODE)(19). Additionally, an illness severity measure and a comorbidity index were calculated at enrollment using the modified Sequential Organ Failure Assessment Calculator (SOFA) and Charlson Comorbidity Index (20, 21). Data on severe sepsis and desaturations to <90% were collected daily by study team personnel through medical record abstraction through ICU discharge. Medians and interquartile ranges were provided for continuous variables and frequencies and percentages were provided for categorical data. To determine the association between measures of initiation at 3 months and 12 months with disability in IADLs at the same point in time (i.e., initiation measure versus Functional Activities Questionnaire or FAQ at 3 months) and at a future time point (i.e., initiation measure at 3 months versus FAQ at 12 months), we used proportional odds logistic regression, adjusting for covariates. Based on review of the literature and substantive knowledge, we a priori selected age at enrollment, years of education, FAQ score at enrollment, IQCODE at enrollment, number 15-minute interval oxygen saturations <90% during the study period, number of days of severe sepsis (using Sepsis-2 definition) during the study period, Charlson score at enrollment and modified SOFA score at enrollment. All covariates were modeled using restricted cubic splines to allow for a nonlinear relationship with the outcome. All variables (independent and dependent variables, covariates) were modeled continuously. The proportional odds assumption was tested and met. For all inferential statistics, the significance threshold was set at a p-value of ≤ 0.05. Descriptive statistics and logistic regression modeling and figure were produced using R studio (v. 1.1.423).

Results

Study Population

We enrolled 195 patients who were a median of 59 years old (IQR: 49 – 67), with a median APACHE-II score of 17 (IQR: 12 – 20) and median score of 0 (IQR: 0 – 3) on the FAQ at enrollment. A FAQ was completed by 194 patients at baseline and 177 (91%) and 153 (79%) patients at 3 and 12 months, respectively. Additional study demographics are listed in Table 1 and flow of patients throughout the study is displayed in Figure 1.

Table 1.

Initiation versus FAQ^a at 3 and 12 Months in 195 Patients from the BRAIN-ICU Study^b

	N = 195
Age
Median (IQR)	59 (49 – 67)
Sex - no.(%)
Male	93 (48%)
Female	102 (52%)
Race - no.(%)
White	169 (87%)
African American	26 (13%)
Education (years)
Median (IQR)	12 (12 – 14)
Charlson comorbidity index
Median (IQR)	2 (1 – 4)
SOFA score at enrollmentc
Median (IQR)	6 (4 – 8)
APACHE-IId score at enrollment
Median (IQR)	17 (12 – 20)
Days of severe sepsis in the ICU
Median (IQR)	2 (0 – 6)
Number of interval oxygen saturations <90%
Median (IQR)	2 (0 – 8)
IQCODEe at enrollment
Median (IQR)	3 (3 – 3.1)
FAQ Median (IQR)
Enrollment (N=194)	0 (0 – 3)
3 Months (N=177)	2 (0, 7)
12 Months (N=153)	1 (0, 6)
Ruff-TUDf t-score (Median (IQR))
At 3 months (N = 151)	37 (34, 45)
At 12 months (N = 125)	41 (34, 48)
COWAg t-score (Median (IQR))
At 3 months (N = 164)	39 (30, 47)
At 12 months (N = 139)	40 (33, 50)
BRIEF-Ih t-score (Median (IQR))
At 3 months (N = 159)	58 (46, 64)
At 12 months (N = 134)	56 (47, 67)

^aFAQ = Functional Assessment Questionnaire;

^bBRAIN-ICU = Bringing to Light the Risk Factors and Incidence of Neuropsychological Dysfunction in ICU Survivors;

^cSOFA = modified Sequential Organ Failure Assessment;

^dAPACHE = Acute Physiology and Chronic Health Evaluation;

^eIQCODE = Informant questionnaire on cognitive decline in the elderly;

^fRuff-TUD = Ruff Total Unique Design;

^gCOWA = Controlled Oral Word Association;

^hBRIEF-I = Behavior Rating Inventory of Executive Function Initiation Subscale

Figure 1.

Flow of BRAIN-ICU participants into this executive function sub-study.

Cognitive Outcomes: Ruff-TUD (nonverbal fluency), COWA (verbal fluency), BRIEF-I (initiation subscale)

Median Ruff-TUD t-scores at 3 and 12 months were 37 (IQR 34 – 45) and 41 (IQR 34 – 48), respectively. Among patients assessed with the RUFF, 71 (47%) and 42 (34%) were impaired at 3 and 12 months, respectively. Median COWA t-scores at 3 and 12 months were 39 (IQR 30 – 47) and 40 (IQR 33 – 50), respectively. Of those assessed with the COWA at 3 months, 64 (39%) were impaired, compared to 45 (32%) at 12 months. The median BRIEF-I subscale t-scores at 3 and 12 months were 58 (IQR 46 – 64) and 56 (IQR 47 – 67), respectively. At 3 months, 73 patients (37%) were impaired on the BRIEF-I initiation measure versus 102 patients (52%) at 12 months.

Instrumental Activities of Daily Living

At 3-month follow-up nearly a quarter, 22% of patients (N =39 of 177) reported impairment on the FAQ, with median FAQ scores of 2 (IQR 0, 7). At 12 months, 17% of patients (N = 26 of 155) reported disability on the FAQ with median scores of 1 (IQR 0, 6). Of those considered disabled at 3 months, the majority (n=16, 62%) remained impaired at 12 months. Of non-disabled individuals at 3 months, 94% (N=105 of 112) remained non-disabled at 12 months.

Effect of Initiation on FAQ Scores

Neither the Ruff-TUD nor COWA were significantly associated with disability at either 3 or 12 months. In proportional odds logistic regression models adjusted for age at enrollment, years of education, baseline IQCODE and FAQ scores, SOFA (severity of illness) score, Charlson Comorbidity Index, days of severe sepsis in the ICU, and number of interval oxygen saturations <90% in the ICU, performance on the BRIEF-Initiation (BRIEF-I), reflected in t-scores, were significantly associated with a 1-unit change in FAQ score at all points in time (Table 2 and Figure 2). Odds ratios (and 95% CI) for the BRIEF-I subscale and FAQ were 0.060 (0.034, 0.086) for both assessments at 3 months and 0.060 (0.030, 0.089) for both assessments at 12 months. Odds ratios (95% CI) for the BRIEF-I at 3 months and FAQ at 12 months were 0.068 (0.035, 0.101).

Table 2.

Odds Ratios^a and 95% confidence intervals for 3 and 12-month initiation assessment measures versus FAQ^b at 3 and 12-month executive function follow up in the BRAIN-ICU^c cohort.

Modeldef	OR (95% CI)
3 m Ruff vs. 3 m FAQ	0.672 (0.377, 1.197)
12 m Ruff vs. 12 m FAQ	0.542 (0.253, 1.164)
3 m Ruff vs. 12 m FAQ	0.665 (0.361, 1.226)
3 m COWA vs. 3 m FAQ	0.539 (0.283, 1.028)
12 m COWA vs. 12 m FAQ	0.764 (0.416, 1.405)
3 m COWA vs. 12 m FAQ	0.934 (0.463, 1.883)
3 m BRIEF-I vs. 3 m FAQ	5.062 (2.539, 10.092)
12 m BRIEF-I vs. 12 m FAQ	3.476 (1.943, 6.216)
3 m BRIEF-I vs. 12 m FAQ	5.079 (2.579, 10.000)

^aResults reported as Odds Ratios (95% CI) using an proportional odds logistic regression model comparing the 25^th versus the 75^th percentile oa 1-unit change inf initiation measure (Ruff, COWA or BRIEF-I score versus a 1-unit change in the FAQ at different time points) bolded ORs and 95% CI indicate statistical significance;

^bFAQ = Functional Assessment Questionnaire;

^cBRAIN-ICU = Bringing to Light the Risk Factors and Incidence of Neuropsychological Dysfunction in ICU Survivors: ;

^dRuff-TUD = Ruff Total Unique Design;

^eCOWA = Controlled Oral Word Association;

^fBRIEF-I = Behavior Rating Inventory of Executive Function Initiation Subscale

^gBolded numbers represent statistically significant association between the initiation measure (Ruff-TUD, COWA or BRIEF-I) and FAQ and 3 different points in time.

Figure 2.

Initiation measures and odds of worse FAQ score at 3 and 12 months

In the figure above, the black dots represent Odds Ratio values (comparing the 25^th versus the 75^th percentile of initiation measure (Ruff, COWA or BRIEF-I) versus FAQ at different points above in time) which is fully described above in Table 2, the horizontal solid black lines represents the 95% Confidence Interval, with the central dot indicating the OR and the vertical dashed black line represents a referent Odds Ratio of 1.

Of note, only the BRIEF-I measure, a subjective measure of initiation, was significantly associated with FAQ score at any time point, suggesting that a patient’s individual appraisal of their ability to initiate tasks, etc., is more reflective of functional disability both now and in the future than objective measures alone.

For the logistic regression models, only the BRIEF-I models were significantly associated with higher FAQ scores (indicating greater degrees of functional dependence in IADLs) (refer to Table 2 and Figure 2 for full model results). Higher BRIEF scores (indicating greater impairment in initiation) at 3 months were significantly associated with a greater odds of a higher FAQ at both 3 (OR: 5.062; 95% CI: 2.539, 10.092) and 12 months OR (95% CI) 5.079 (2.579, 10.000). Additionally, a higher BRIEF score at 12 months was predictive of a greater odds of a higher FAQ at 12 months OR (95% CI) 3.476 (1.943, 6.216). In a sensitivity analysis, composite exposure variable (impairment on any 1 of the 3 initiation measures at 3 months was not significantly associated (OR = 1.185; 95% CI: 0.583, 2.406) with higher FAQ score at 12 months.

Discussion

In this nested observational cohort study of survivors of critical illness, we found that executive functioning deficits were extremely prevalent in a domain known as initiation, which refers to the ability to generate actions or responses. Moreover, we found that deficits in initiation as self-reported by the patient were strongly and uniquely associated with current and future of disability in IADLs, as reported by the patient’s family. Although initiation deficits have often been shown to be predictive of functional disability in previous investigations, objective neuropsychological tests of initiation were not related to functional disability. Whether subject reports are more reflective of cognitive difficulties than objective neuropsychological findings is the subject of vigorous ongoing debate. Our results suggest that, at a minimum, subjective patient reports should be carefully attended too and not necessarily dismissed if they differ from objective findings (e.g. such as when a patient reports marked cognitive symptoms but neuropsychological testing appears normal). In particular, the belief among patients and their families that their agency and their ability to actively engage stimuli in their respective environments (i.e. initiation ability) is reduced is of particular concerns, as this may be a harbinger of broader functional difficulties to come.

Interestingly, it was the self-report (BRIEF-I) measure of initiation, and not the objective measures of initiation (Ruff-TUD and COWA) that was associated with (family reported) functional deficits at 12 months. In this study, only patient reported deficits in initiation (as measured by the BRIEF-I) were significantly associated with functional disability, either at the same point in time or in the future. The two objective measures of initiation, the Ruff-TUD and COWA were not associated with the FAQ score at any point in time. These results suggest that patients themselves may be more aware of subtle and evolving deficits in executive functioning than can be assessed reliably with an objective measure. These subjective patient-led assessments may be more sensitive for patient reported outcomes and could potentially prove useful in identifying individuals vulnerable to comorbid or future functional deficits. One could consider, if patients with subtle executive functioning deficits could be identified early with targeted patient-reported screening, could interventions be put in place to help them achieve maximum independence and functionality in the future?

This investigation has several strengths. First, to our knowledge, this study represents the first attempt in a cohort of medically ill patients to examine a novel and to date, understudied cognitive domain, initiation as a potential independent variable associated with future disability in IADLs. Secondly, our study population employed multiple distinct measures of initiation, both objective and subjective, reflecting significant comprehensiveness. There are several important limitations to note as well. Our sample size was small and combined with the presence of relatively high attrition; the net effect may have been to limit generalizability. While we set out to study the executive dysfunction subdomain of initiation with the Ruff-TUD, COWA and BRIEF-I, we might actually have captured information about attention, memory and processing speed, which directly could affection functional outcomes. Additionally, cognitive tests, including tests of initiation, are typically measures of multiple domains, reflecting the fact that executive functioning is not a unitary construct but rather influenced by a diversity of abilities. In other words, deficits in initiation might just serve as a surrogate for broader deficits in general cognitive functioning. Additionally, while we utilized a well validated measure of functional disability – the FAQ – this measure relies on self or informant (typically a first degree relative, such as a spouse or adult child) report and in an ideal world, it would have been augmented by a more direct performance based assessment of functional disability (we didn’t add such a measure as we believed it would add undue respondent burden to our patients who were already often fatigued by the time they finished our 2.5 hour evaluation). Nonetheless, we believe the data obtained via the FAQ portrays a broadly accurate picture of the very common and often debilitating disability related challenges ICU survivors so often endure. Finally, initiation is closely related to self-efficacy, which in turn has been correlated with mood and anxiety. Depression and anxiety, which were not addressed, could be confounding factors in the subjects’ reports of initiation as well as contributing factors to the reported functional deficits.

In conclusion, we found deficits in initiation, to be associated with disability in IADLs. These data suggest a patient’s subjective experience with new-found executive dysfunction are associated with disability in IADLs at the present and in the future. Future studies, appropriately focused on early prediction of disability in IADLs, should incorporate measures of initiation into longitudinal assessments. ICU recovery clinics should consider querying patients about initiation deficits as they seek to anticipate the emergence of behavioral difficulties that could adversely impact vocational success, health management, and quality of life. Additionally, the question of whether these deficits in initiation can be rehabilitated once identified, leading to improved functional status, remains an intriguing question.