Linking behavioral variability to daily function in stroke survivors: the mediating role of cognitive and motor mechanisms

Abstract

Background

Stroke significantly impacts an individual’s ability to perform instrumental activities of daily living (IADLs). Intraindividual variability (IIV) is characterized by heightened fluctuations across multiple attempts at a task. IIV is a behavioral indicator of central nervous system instability that may contribute directly to compromised capacity to perform IADLs following stroke. This study aimed to investigate the relationship between IIV and IADL performance in stroke survivors and determine whether executive control and/or motor capacity serve as mediating pathways.

Methods

In this cross-sectional observational study, 84 stroke survivors and 35 healthy older adults participated. We derived latent factor scores for IIV from reaction time and goal directed tasks, IADL performance from the Observed Task of Daily Living and Functional Activities Questionnaire, executive control from tests of divided attention, selective attention, cognitive flexibility, and processing speed, and motor capacity from grip and ankle strength. We examined the association between IIV and IADL performance and conducted a parallel mediation analysis to determine whether executive control and motor capacity mediated this relationship.

Results

Higher IIV was associated with compromised IADL performance. Mediation analyses showed that the relationship between IIV and IADL performance was significantly mediated by executive control but not motor capacity while controlling for the covariate of age.

Conclusions

Behavioral variability after stroke impacts everyday function primarily through a cognitive pathway. These findings highlight executive control as a key therapeutic target in stroke rehabilitation to mitigate the functional consequences of neural instability on instrumental activities of daily living.

Introduction

The capacity to perform instrumental activities of daily living (IADLs) is imperative for independent living, well-being, and quality of life. These IADLs comprise of various tasks that people do every day such as shopping, paying bills, and making phone calls [1]. Following stroke, an individual’s ability to perform IADLs is often compromised leading to poor life satisfaction and difficulty in returning to a life of independence [2]. Although challenges in performing IADLs are highly prevalent in the first year following stroke, ~ 40% of stroke survivors continue to experience persistent disability even 6 years after stroke [3]. Therefore, it becomes critical to identify potential mechanisms linked to declines in IADLs following stroke.

Intraindividual variability (IIV) in performance is an inherent element of human behavior. IIV refers to unintentional variations in a person’s behavior across successive attempts and represents within-person inconsistency. This behavioral variability is commonly assessed in the cognitive and motor domain. In the cognitive domain, trial-to-trial inconsistency in reaction time is a robust indicator of neural integrity [4, 5] and is believed to capture fluctuations in attentional or executive control mechanisms [6, 7]. In the motor domain, trial-to-trial inconsistency in endpoint during a goal-directed task is linked to difficulty navigating environments [8] and higher fall risk [9]. Interestingly, recent literature suggests that this motor inconsistency may reflect fluctuations in cognitive-motor processes responsible for the planning and execution of the required movement [10]. Moreover, motor inconsistency is apparent beyond changes in muscle strength in young and older adults [11]. Despite the well-documented implications of IIV in both the cognitive and motor domain, systematic investigations on the relationship between IIV and IADL performance in stroke survivors are notably absent.

Individuals with stroke often experience concurrent declines in cognitive and motor function which contribute to compromised IADL performance. Cognitive deficits in stroke survivors can include multiple domains with the most prominent being compromised executive control [12]. In addition, muscle strength is considered an important clinical indicator of progress toward stroke motor recovery that captures motor capacity, therefore its reduction represents the most clinically relevant motor dysfunction post-stroke [13]. These cognitive and motor disturbances are linked to reduced functional status, capacity to perform IADLs, and quality of life in stroke survivors [14].

Theoretically, the above evidence converges on a central principle that behavioral variability (i.e., IIV) serves as a behavioral indicator for the stability of the underlying neurobiological system. Within this framework, we propose that elevated IIV may not only indicate neural instability but also an upstream contributor to deficits in cognitive and motor control. Specifically, elevated behavioral variability potentially cascades into system-wide cognitive inefficiency and unpredictable movements, culminating in poorer performance during IADLs or avoidance of such tasks altogether. Alternatively, systematic neurobiological instability may directly interfere with IADL performance without affecting cognitive-motor performance [10]. For example, this instability could manifest as inconsistent or unreliable behaviors during complex, self-directed activities such as IADL performance. Although prior research has linked IIV to functional outcomes in aging and other neurological conditions, its relationship to IADL performance in stroke survivors and how executive control or motor capacity may mediate this relationship remains largely unexplored.

This study examines three questions. First, is IIV in behavior elevated in stroke survivors compared to adults who did not experience a stroke? We compared stroke survivors and healthy older adults on an IIV factor score derived from trial-to-trial variability in reaction time and goal-directed endpoint. We hypothesized that stroke survivors would demonstrate greater IIV than healthy older adults. Second, is IIV linked to detriments in IADL performance in stroke survivors? We assessed the magnitude and direction of the associations between IIV and IADL performance. We hypothesized that greater IIV would be associated with compromised IADL performance. Lastly, is the relationship between IIV and IADL performance in stroke survivors mediated by executive control, and/or motor capacity? We performed a mediation analysis with IIV as the independent variable, IADL performance as the dependent variable, and executive control and motor capacity as potential mediators. In this context, we conceptualize IIV not only as a behavioral marker of underlying neural instability, but also as a potential driver of deficits in both executive and motor functioning. We theorize that the detrimental effects of this instability on the performance of complex IADLs will manifest primarily through a failure of executive control, which is heavily taxed during the planning and execution of such tasks. Conversely, based on previous research showing that muscle strength does not predict functional mobility [15] and dexterity [16] especially in high-functioning stroke survivors, we did not expect motor capacity to impact IADL performance. Therefore, we hypothesized that executive control, but not motor capacity, would mediate the relationship between IIV and IADL performance in stroke survivors.

Methods

Participants

This cross-sectional study included 84 stroke survivors and 35 healthy older adults, recruited between May 2022 and November 2024 in Colorado and nearby areas. Potential participants were identified through neurology referrals, from a network of allied health professionals, and the Epic recruitment tool. Interested individuals were then contacted via an IRB-approved phone call or email to provide study details and assess initial eligibility before inviting them to the laboratory for study enrollment. All included stroke participants were community-dwelling and were not enrolled in any formal cognitive rehabilitation programs at the time of the study. Healthy older adult controls were recruited from the local community and were age- and sex-matched with the stroke survivors.

Study procedures took place at Colorado State University, with participants enrolled regardless of ethnic or sociodemographic background. Detailed participant demographics and characteristics are presented in Table 1. Inclusion criteria for the stroke participants included: (1) Being older than 21 years of age at the time of study participation; (2) diagnosis of stroke (ischemic or hemorrhagic; confirmed through medical records) at least 9 months prior to enrollment; (3) a minimum of 10 degrees of active movement in the paretic ankle (combined dorsiflexion and plantar flexion); and (4) ability to press down with the index finger. Exclusion criteria for all participants included being younger than 21 years of age, self-reported presence of any neurological disorder other than stroke (e.g., traumatic brain injury, Parkinson’s disease, epilepsy), substance abuse, aphasia, untreated sleep or anxiety disorders, or uncorrected vision and hearing impairments.

Table 1.

Descriptive sample characteristics (N = 119)

Variable	Stroke (n = 84)	Healthy older (n = 35)	p
Sociodemographic characteristics
Age, years, mean (SD)	67.39 (12.05)	68.47 (14.33)	0.624
Years of education, mean (SD)	15.77 (2.52)	16.89 (2.63)	0.076
Women, n	35	22	0.035
Ethnicity, n
Non-Hispanic White	78	29
Asian	1	4
African American	3	0
Hispanic	0	1
Other	2	1
Modified Rankin Score, mean (SD)	1.30 (0.86)	–
Type of stroke, n
Ischemic	64	–
Hemorrhagic	10	–
Both	1	–
Unknown	9	–
Years since stroke, mean (SD)	7.46 (19.93)	–
Side of the lesion, n
Right	23	–
Left	35	–
Both	8	–
Unknown	18	–
Lesion location, n
Infratentorial	11	–
Supratentorial	48	–
Both	7	–
Unknown	18	–

SD  standard deviation

Experimental protocol

The experimental protocol consisted of a single baseline session lasting approximately 90 min during which participants performed an upper limb reaction time task, a lower limb goal-directed task, assessments of cognitive and motor function, and assessments of IADL performance.

Intraindividual variability (IIV)

We quantified IIV using reaction time and goal-directed task. Participants performed both tasks with the left and right limbs.

Reaction time task. Participants were seated comfortably in front of a computer (Samsung DC32E, Samsung Electronics America, Resolution: 1920 × 1080, Refresh rate: 50 Hz) screen with their index finger rested on a red button within easy reach (Fig. 1A). At the beginning of each trial, the word RELAX appeared in the center of the screen for a random interval between 0.5 and 1 s. Immediately following, a red circle (stimulus) appeared on the screen. We instructed participants to press the button with their index finger as quickly as possible after seeing the stimulus. Each participant completed a series of 32 trials. We evaluated performance on this task using reaction time defined as the latency between stimulus appearance and button press.

Fig. 1.

Intraindividual variability tasks. A Schematic representation of the reaction time task and representative reaction time performance. Participants completed 32 trials in which they were required to press a button with their index finger as quickly as possible upon presentation of a visual stimulus (red circle on a screen). Reaction time was calculated as the latency between stimulus onset and button press. Intraindividual variability (IIV) was quantified as variance in reaction time across trials. B Schematic representation of the goal-directed task and representative goal-directed performance. The left panel illustrates the position of the foot. Participants were required to match a target of 10 N in 180 ms. The relative distance from the origin to peak of the bell-shaped performance represents the endpoint. IIV was quantified as the variance in endpoint across trials

Goal-directed task. Participants sat upright facing a 32-inch monitor (Samsung DC32E, Samsung Electronics America, Resolution: 1920 × 1080, Refresh rate: 50 Hz), positioned 1.85 m away at eye level. The hip joint was flexed to 90°, the knee was flexed to 90°, and the ankle was in a neutral position. The foot rested on a customized foot device with an adjustable footplate that we secured by straps over the metatarsals to ensure an isolated dorsiflexion of the ankle (Fig. 1B). The force exerted during the goal-directed contractions was measured using a force transducer (Model 41, Honeywell, Columbus, OH, USA) located parallel with the force direction on the customized foot device. The ankle force signal was sampled at 1000 Hz using a NI-DAQ card [model USB-6343 (BNC), National Instruments] and stored on a personal computer for offline analysis. The monitor displayed the contraction produced by ankle dorsiflexion using a custom-written program in Matlab (Math Works™ Inc., Natick, MA, USA). All participants confirmed that they could see the screen clearly. The protocol for rapid goal-directed contractions was adapted from previously published studies [17, 18].

Participants performed 30 goal-directed contractions (Fig. 1B). To perform the task correctly, participants had to perform a quick ankle dorsiflexion producing a target force of 10 N in 180 ms, creating a bell-shaped curve with a clear peak. The task consisted of three phases: GET READY, MOVE, and FEEDBACK. In the GET READY phase, a red target appeared on the monitor for 2 s, signaling participants to prepare. During the MOVE phase, the target turned green for 3 s, prompting participants to perform the goal-directed contractions. Participants had 3 s to begin, with force–time recording starting at the onset of contraction. No online feedback was provided during the task to avoid self-corrections. In the FEEDBACK phase, participants received visual feedback on their performance relative to the target force and time for 5 s, allowing them to plan for the next trial.

We evaluated performance in the goal-directed task using endpoint. Endpoint was derived from the peak force and time-to-peak force. Specifically, after converting to comparable units (%; Eqs. 1 and 2), endpoint was calculated as the hypotenuse between peak force and time-to-peak force (Eq. 3). This endpoint represents the exerted performance accounting for both force and time components.

1

2

3

Data preparation We applied the established statistical approach developed by Hultsch et al. (2000) to compute IIV [19]. These steps were performed on the left and right side, to account for outliers and confounding systematic effects.

First, we applied data purification to reaction times and goal-directed endpoints by identifying and removing outliers. Reaction times below 150 ms or above the mean + 3*SDs were adjusted to 150 ms or the individual’s upper limit, respectively. Goal-directed endpoints outside the mean ± 3*SDs range were replaced by the individual’s fence values (i.e., mean ± 3*SDs). Data purification mitigates errors from extreme responses (e.g., accidental key presses, task interruptions) and minimizes undue variability, ensuring a conservative analysis [19].

Next, we compute IIV by controlling for the systematic effects of practice and group means. To eliminate these effects, we input trial and group means as predictors of the reaction times and endpoints in a linear regression analysis and computed the resultant standardized reaction time and endpoint residuals (z scores). We then converted the resulting z-scores for each trial into standardized T-scores and computed the standard deviation of the reaction time and endpoint T-scores across the 32 and 30 trials, respectively.

Lastly, to account for age effects on IIV, we conducted a linear regression with age as a predictor of reaction time and endpoint T-score standard deviations. The residuals were then converted into standardized T-scores, ensuring that reaction time and endpoint IIV were independent of trial, mean performance, and age.

IIV factor score After computing the four purified IIV outcomes (for each limb and task), we conducted a principal components factor analysis and extracted individuals’ single factor scores (i.e. latent IIV), without rotation, from the four IIVs using the Bartlett method. The Bartlett method is an established and appropriate approach that produces factor scores that are unbiased estimates of the true latent factor. This ensures that our predictor variable is as accurate an estimation of the underlying construct as possible. This approach is well-supported in the psychometric literature for producing valid factor scores for use in subsequent analyses [20].

Additionally, to facilitate interpretation of subsequent regression analyses, we reversed the signs such that a positive score indicated lower IIV. This factor score represents an overall measure of IIV across two classic experimental paradigms and both sides of the body. We chose to create a single factor score for two primary reasons. First, from a theoretical standpoint, we conceptualize IIV from both tasks as reflecting a single, underlying latent construct of 'global neural instability.' Using a single factor score allows us to test our hypothesis about this global construct directly. Second, from a statistical standpoint, combining the two task scores into a single factor reduces measurement error by leveraging their shared variance and provides a more reliable and parsimonious predictor for our mediation model, increasing its statistical power.

Cognitive and motor function

Compromised executive control and muscle strength are the most prevalent cognitive and motor disturbances in individuals following stroke. Therefore, we quantified cognitive and motor function using assessments of executive control and motor capacity.

Executive control We assessed executive control using the Useful Field of View (UFOV) [21] test, Part B of the Trail Making Test (TMT-B) [22] and the Digit Symbol Substitution Test (DSST) [23, 24]. The UFOV is a highly robust tool used to capture processing speed in divided and selective attention tasks. Participants sat upright approximately 60 cm from a 32-inch monitor (Sync Master™ 275t + , Samsung Electronics America, NJ, USA) positioned at eye level, used to display the UFOV test. The divided attention task required participants to identify a central stimulus (car or truck) while simultaneously locating a peripheral stimulus (car). The selective attention task was like the divided attention task, except the peripheral stimulus was embedded among distractors (triangles). Longer times on the divided and selective attention tasks demonstrate poorer divided and selective attention, respectively. For TMT-B, participants alternated between numbers and letters (e.g., 1-A-2-B) as quickly as possible, and the time taken to complete the task was recorded. Longer times on TMT-B demonstrate poorer executive function. For DSST, participants were required to match symbols to corresponding numbers as quickly as possible within a set time limit and the number of correct responses was recorded. Lower scores on DSST demonstrate poorer executive function.

Executive control factor score Given that attention, cognitive flexibility, and visuospatial processing speed are complementary indicators of executive control [25], we conducted a principal components factor analysis and extracted individuals’ factors scores (i.e. latent executive control) from the four executive control outcome variables using the Bartlett method. Before deriving the factor variable, we converted the raw scores into z-scores and reversed the signs where necessary such that a positive z-score indicated better performance. Thus, the positive values in the subsequent derived factor score reflect better executive control. Aligning with our theoretical framework and supported by the presence of only one eigenvalue > 1.0, a single factor was extracted without rotation that represents an overall measure of executive control.

Motor capacity We measured grip strength using a digital hand dynamometer (Jamar Plus, model number 120604) and measured ankle strength using an electronic manual muscle tester (Lafayette Manual Muscle tester, 01165, IEC 60601–1-2; 2007–03). For ankle strength, participants lay in supine position. We placed the muscle tester against their metatarsals and instructed them to maximally dorsiflex and plantarflex their feet without bending their knees. We recorded the maximum voluntary contraction (MVC) for both extremities for grip strength, ankle dorsiflexion strength, and ankle plantarflexion strength, and computed the percentage of the strength of the paretic extremity relative to the nonparetic extremity. Values below 100% indicate that the strength of the paretic limb is lower than the strength of the nonparetic limb.

Motor capacity factor score We conducted a principal components factor analysis and extracted a factor score of motor capacity from the strength outcomes using the Bartlett method. Before deriving the factor variable, we converted the raw scores into z-scores and reversed the signs where necessary such that a positive z-score indicated better performance. Thus, the positive values in the subsequent derived factor score reflect better motor capacity. We extracted a single factor (eigenvalue > 1.0) without rotation. The strength outcomes capture motor strength in the upper and lower limbs and represent an overall measure of motor capacity.

Instrumental activities of daily living

Observed Task of Daily Living (OTDL) The OTDL is a performance-based measure that simulates real-world IADLs [26, 27]. Participants were asked to complete a series of tasks that reflect everyday activities, such as managing medications and handling finances. Each task was designed to assess various cognitive skills required for daily functioning. Participants were provided with the necessary materials and instructions for each task and were observed by trained assessors who scored their performance based on accuracy, time taken, and adherence to instructions. Higher scores demonstrated greater capability to manage IADLs independently.

Functional Activities Questionnaire (FAQ) The FAQ is a self-report measure that evaluates participants’ abilities to perform IADLs [28]. Participants rate their ability to complete complex daily tasks, such as shopping, preparing meals and remembering appointments. Each task is rated on a scale from 0 (independent) to 3 (dependent), with lower total scores indicating compromised daily function.

IADL factor score We conducted a principal components factor analysis and extracted a single factor score (eigenvalue > 1.0) without rotation of IADL performance from OTDL and FAQ using the Bartlett method. Before deriving the factor variable, we converted the raw scores into z-scores and reversed the signs where necessary such that a positive z-score indicated better performance. Thus, the positive values of the subsequently derived factor score reflect better IADL performance. The OTDL and FAQ outcomes capture objective and subjective based performance to accomplish IADLs and represent an overall measure of IADL performance.

Potential bias

We implemented blinding where appropriate to minimize potential biases. Participants were informed of the broad scope and general procedures of the study but not of the study’s exact hypothesis, reducing the likelihood of unwanted behavior adjustment. Additionally, IADL assessments were scored by one author (Tiwari), whereas data analysis was conducted by separate authors (Delmas, Tseng) to prevent assessment bias.

Statistical analysis

The Shapiro–Wilk test was used to assess the normality of the data distributions on all the tests. Given failed assumptions of normality and imbalance in sample size between the two groups, group comparisons for continuous variables were done using the Mann Whitney U non-parametric test and chi-square tests were used for categorical variables. To examine sex as a biological factor, we compared IIV between males and females. We use Cohen’s d to report the effect size of group differences.

To assess the association between IIV and IADL performance, we examined their bivariate correlation. Then, we performed a single parallel mediation analysis to assess whether executive control and motor capacity significantly mediated the association of IIV with IADL performance. A bias-corrected bootstrap method (i.e., 5,000 bootstrap samples) was used to determine the significance levels and confidence intervals (95% CI) for the significant mediation effects. Parameters for total, direct, and indirect effects are presented in the accompanying diagrams. Outliers were identified as values exceeding ± 3 standard deviations from the group mean and using boxplot rules (i.e. 3 × IQR beyond Q1 or Q3).

Missing values in the data set include two stroke participants and one healthy older adult participant for OTDL, two stroke participants for TMT-B and DSST, and three stroke participants for selective attention. These missing values are attributed to the participant’s impairment and attrition. As missing data per measure did not exceed 10%, we applied linear imputation for continuous and median imputation for count variables. All statistical significance tests were conducted with an alpha level of 0.05, and all analyses were performed using SPSS 27.0 (IBM, Armonk, NY, USA).

To achieve 80% power at an alpha level of 0.05 and a moderate effect size based on our prior work [29, 30], 30 participants per group were needed to address the first primary question. To ensure a robust n/k ratio (data points/number of predictors) ≥ 15:1 for the bivariate associations and mediation analysis, a sample size of 75 stroke participants was needed for second and third questions. Therefore, we oversampled and recruited 84 stroke and 35 healthy control participants in this study.

Results

Participant demographics

The stroke and healthy older adult groups did not differ significantly on age or years of education (Table 1). However, the stroke group had significantly more females than the healthy older adult group. The primary ethnicity in both groups was White/Caucasian (> 80%). The stroke group contained one participant with Asian ethnicity, three participants with African American ethnicity and two participants with another ethnicity. The healthy older adults group had four participants with Asian ethnicity, one participant with Hispanic ethnicity, and one participant with another ethnicity. Moreover, as a cohort, stroke participants exhibited a low degree of disability indicated by the modified Rankin Scale (< 2).

Group comparisons on cognitive, motor and IADLs

Comparisons of IIV factor scores, and assessments of executive control, motor capacity, and IADLs between stroke and healthy older adults are presented in Table 2. Stroke survivors exhibited greater IIV factor score compared with healthy older adults (p = 0.018; d = 0.463). Notably, the IIV factor score was not significantly different between males and females (Mann Whitney U = 845; p = 0.910). We also found that the individuals with stroke exhibited significantly higher values in FAQ and TMT-B (p ≤ 0.01; d ≥ 0.55). Additionally, the stroke group had significantly lower values in OTDL and DSST (p < 0.025, d > 0.50). Divided attention, selective attention, grip strength, and ankle strength were not significantly different between the two groups (p > 0.05; d < 0.5). According to Cohen [31] the observed effect sizes were in the small to moderate range.

Table 2.

Intraindividual variability (IIV), executive control and motor capacity, and IADL outcomes for stroke survivors and healthy older adults

	Stroke (n = 84)	Healthy older (n = 35)	MW U	p	Cohen’s d
Intraindividual variability
IIV factor score (z-score)	− 0.13 (1.10)	0.32 (0.59)	1877.0	0.018	0.463
Executive control and motor capacity
Divided attention (ms)	177.77 (161.50)	109.83 (122.11)	1800.5	0.053	0.450
Selective attention (ms)	255.95 (142.17)	218.23 (129.71)	1697.5	0.184	0.272
TMT-B (s)	115.79 (75.50)	78.28 (39.71)	1989.5	0.002	0.559
DSST (# correct)	50.67 (18.49)	63.34 (13.60)	843.0	< 0.001	0.737
Grip strength (%)	92.96 (21.47)	93.92 (11.53)	1578.5	0.527	0.050
Plantarflexion strength (%)	95.16 (15.09)	100.62 (16.54)	1231.0	0.163	0.351
Dorsiflexion strength (%)	94.32 (21.48)	101.21 (16.52)	1328.0	0.408	0.342
Instrumental activities of daily living
OTDL (score)	21.08 (5.30)	23.43 (2.58)	1077.0	0.021	0.502
FAQ (score)	4.75 (5.28)	0.34 (0.80)	2443.0	< 0.001	0.986

Values are reported as mean (SD). MW U  Mann Whitney U, TMT-B  Trail Making Test Part B, DSST  Digit Symbol Test, OTDL  Observed Task of Daily Living, FAQ  Functional Activities Questionnaire

Association between IIV and IADLs in stroke

The strength and direction of the association between IIV and IADL performance among the stroke participants is presented in Fig. 2. Lower IIV significantly correlated with better IADL performance (r = 0.537; p < 0.01) for the stroke survivors. However, as can be seen in Fig. 2A, two stroke individuals exhibited extreme IIV values relative to the rest of the group as identified by our outlier method and largest Cook’s Distance (approaching > 1). For this reason, we repeated the correlational analyses after removing these outliers. The results showed that IIV continued to be significantly associated with IADL performance (r = 0.384; p < 0.01; Fig. 2B), albeit in a somewhat attenuated way.

Fig. 2.

Associations between IIV and IADL in stroke survivors. We computed a factor score for IIV using the reaction time and goal-directed task and a factor score for IADL using the Observed Tasks of Daily Living and Functional Activities Questionnaire. Compromised IIV (positive values) associated with compromised IADL (positive values) while retaining the entire sample (A) and when removing stroke participants with extreme (red circles) IIV values (B)

Mediating factors between IIV and OTDL in stroke survivors

Prior to analysis, we confirmed that the assumptions of linearity, homoscedasticity, and normality of residuals were met by visually inspecting residual plots. No violations of concern were observed. Moreover, given that mediation results can be sensitive to the influence of outliers, we performed the mediation model with (Fig. 3A) and without (Fig. 3B) the outliers in IIV. Lastly, we checked if the potential covariates of age, years of education, and time since stroke were correlated with the mediation parameters. Only age significantly correlated with our mediation parameters and, thus, was entered as a covariate for the subsequent mediation analysis including the latent variables of IIV, executive control, motor capacity, and IADL performance.

Fig. 3.

Analyses testing potential mediators between IIV and IADL. We tested the mediators of executive control and motor capacity in a parallel mediation model. The indirect pathway of executive control was significant whereas the indirect pathway of motor capacity was not significant (A & B). While retaining the entire sample (A), we observed that the direct pathway of IIV to IADL was attenuated but significant, suggesting partial mediation. However, when removing stroke participants with extreme IIV values (B), the direct pathway of IIV to IADL became nonsignificant, suggesting full mediation. Values are reported as standardized beta coefficients. * p < .05; ** p < .01; *** p < .001

The statistical parameters of the mediation analyses with IIV as the independent variability and IADL performance as the dependent variable are summarized using standardized coefficients in Table 3 and illustrated in Fig. 3. The indirect effect of executive control was significant whereas the indirect effect of motor capacity was not significant. The association between IIV and IADL performance was attenuated but remained significant when controlling for executive control, motor capacity, and age (i.e. direct effect; Fig. 3A). Excluding outliers yielded similar results in that the indirect effect of executive control remained significant whereas the indirect effect of motor capacity was not significant. However, the direct association between IIV and IADL performance was not significant after adjusting for all mediators and covariates (Fig. 3B). Therefore, these findings suggest that the association between IIV and IADL performance is partially mediated by executive control but not by motor capacity.

Table 3.

Parameter estimates derived from mediation analyses

M	Total (c)		Direct (c')		Path a		Path b		Indirect
	β	SE	β	SE	β	SE	β	SE	β	SE	95% CI
With outliers (N = 84)
Executive control	0.52***	0.09	0.18*	0.09	0.48***	0.08	0.74***	0.11	0.35*	0.07	[0.19, 0.46]
Motor capacity					0.11	0.11	− 0.10	0.07	− 0.01	0.02	[− 0.05, 0.02]
Without outliers (N = 82)
Executive control	0.32***	0.10	0.05	0.09	0.34***	0.09	0.80***	0.10	0.27*	0.07	[0.13, 0.40]
Motor capacity					0.15	011	− 0.05	0.08	− 0.01	0.02	[− 0.04, 0.02]

IIV is the independent variable and IADL is the dependent variable. CI  Bootstrapped Confidence Interval

^*p < .05; ^**p < .01; ^***p < .001

Discussion

This study addressed three questions. First, using two classic experimental paradigms, we identified that intraindividual variability is elevated in stroke survivors. Second, we examined if IIV is associated with poorer IADL performance in stroke survivors. We found that higher IIV relates to poorer IADL performance measured via objective and subjective assessments. Lastly, we tested potential mechanisms mediating the association between IIV and IADLs by examining executive control and motor capacity. We found that executive control emerged as a significant mediator, whereas motor capacity did not. This is the first evidence in the stroke literature showing that executive control processes mediate the impact of IIV on the ability to manage daily IADLs.

Inconsistency in performance is associated with compromised IADL performance following stroke

Many common activities, such as navigating stairs or taking one’s medications correctly, rely on consistency in performance. Substantial evidence from aging and neuropathological research identifies IIV as a key marker of neurobiological disturbance [32–34]. Our findings provide robust evidence that IIV is significantly amplified in stroke survivors. The stroke-related increase in IIV is likely attributed to neuronal deterioration responsible for stable and consistent performance [4, 5, 35]. Compromised neural pathways may contribute to inefficient signal transmission and increased noise resulting in inconsistency in behavioral performance. Notably, our findings suggest that exacerbated IIV in performance is related to compromised IADL performance in stroke participants. This finding extends previous work in neurological populations. For example, Munsell et al. (2024) showed that stroke survivors with higher levels of IIV in cognitive performance experienced greater challenges in performing everyday functional activities, such as using the phone and grooming, and participating in social activities [36]. Moreover, Yacoubi and Christou (2024) recently described how age and disease-related increase in IIV in motor performance leads to behavioral changes that can subsequently impact quality of life [10]. Together, current and previous work clearly demonstrates that behavioral variability is a critical factor influencing functional independence and quality of life. These findings emphasize the importance of developing IIV-focused interventions to enhance functional outcomes in stroke survivors.

The methodology used to assess the association between intraindividual variability in behavior and IADL performance has two major strengths. First, we assessed IADLs using both self-reported and performance-based measures. Daily functional activities require the complex interplay of physical, cognitive, and psychological factors [37]. For instance, managing finances requires the physical ability to write checks or use a banking app, the cognitive skills of tracking expenses, performing mental arithmetic, and understanding bills, and the confidence to make financial decisions without feeling overwhelmed or anxious. Self-report measures capture an individual’s awareness of their abilities or lack thereof. The subjective self-reports, however, reflect mostly psychological factors and do not capture behavioral skills that become evident in standardized performance-based assessments [38]. In contrast, objective measures of IADLs, such as the OTDL, provide performance-based assessments of concrete abilities under ecologically valid and controlled conditions [39]. Thus, objective assessments may detect deficits that an individual may not recognize or report. Both self-reported and objective assessments provide complementary information, offering critical insights to facilitate clinical decision-making. Therefore, this comprehensive approach strengthens the ecological and predictive validity of our findings on how behavioral IIV impacts daily functioning.

The second key strength of our methodology is that we leverage two classic experimental models previously used to examine IIV, reaction time [19, 40] and goal-directed movements [11, 41, 42]. Prior research findings quantifying IIV through these paradigms reveal its strong link to neuropathology and dysfunction in older adults. For example, higher reaction time IIV is associated with deterioration of neural network [43, 44], reduced white matter integrity [4, 45], and is sensitive to subtle cognitive changes [35]. On the other hand, higher goal-directed IIV represents a motor control disturbance that interferes with motor learning [46] and accuracy of movements [19, 47–49] which can potentially lead to maladaptive or compensatory behavioral changes [10]. By integrating these paradigms into a single factor IIV score, we captured a comprehensive measure of IIV that reflects both cognitive and motor attributes. Therefore, this composite metric may be more sensitive in detecting meaningful associations with IADL performance.

Executive control is a significant mediator of inconsistency following stroke

Importantly, our findings showed that executive control significantly mediated the association between IIV and IADL performance in stroke survivors. Executive control is a critical link between IIV and IADL performance for at least two reasons. First, executive function processes, including planning, problem-solving, and cognitive flexibility, enable individuals to manage and adapt to complex or novel situations [50, 51]. For example, everyday activities like preparing meals or locating a misplaced item require planning and problem-solving skills. Moreover, following stroke, previously familiar everyday activities may seem novel and may require re-learning. Second, the systematic neurobiological variability captured by IIV likely disrupts the central executive function processes needed to execute complex behavior. Instability within the executive control system may manifest as an unreliable and fluctuating capacity to plan, sustain attention, and flexibly switch between subtasks. This unreliability in critical executive processes could then translate directly to functional failures, where a momentary lapse in executive control may result in a missed medication dose or an abandoned financial transaction.

Notably, our analysis revealed that motor capacity as indicated by muscle strength was not a significant mediator of the association between IIV and IADL performance. This finding can be attributed to the fact that the stroke group included relatively high functioning individuals (mRS < 2) with strength outcomes comparable with healthy older adults. Consistent with prior work showing minimal contribution of muscle strength to functional mobility [15] and fine motor tasks [16], our results suggest that in high-functioning stroke survivors in our cohort with minimal physical limitations, the successful execution of complex daily tasks was more dependent on the stability of central cognitive processes than on motor capacity.

Clinical implications

Current stroke rehabilitation focuses on improving mean-level motor function [52], but our findings suggest that performance consistency and cognitive function also deserve attention. Our findings suggest that elevated behavioral variability contributes to poorer IADL performance, both directly and indirectly through reduced executive function. Training paradigms aimed at reducing inconsistency could enhance confidence in performing daily activities. Supporting evidence comes from Hammerbeck and colleagues (2017) who demonstrated that training-induced reduction in inconsistency improved movement accuracy in healthy older adults [53]. Furthermore, cognitive training targeting compromised executive control processes may benefit IADL performance. For example, Willis et al. (2006) showed that older adults who received cognitive training in inductive reasoning experienced less difficulty performing IADLs compared with a control group [54]. Although further research is needed, our results highlight the importance of addressing both behavioral variability and cognitive function in stroke rehabilitation to promote optimal daily function following stroke.

Limitations and further considerations

Despite the study’s strengths, we acknowledge several limitations. First, most stroke survivors in our sample (81%) had relatively high functional status (FAQ < 9), and low disability (mRS < 2), limiting the generalizability of the study’s findings to more heterogeneous stroke populations. Second, our cross-sectional design prevents causal conclusions. Nevertheless, these cross-sectional analyses offer the first evidence that specific cognitive functions may mediate the association between IIV and IADL performance in stroke survivors. Third, although using structural equation modeling (SEM) would be ideal to allow for simultaneous modeling of multiple predictors and mediators and reduce error variance, we opted for a mediation approach because sample size and model instability precluded robust SEM modeling. Fourth, several key domains of executive control, such as inhibition, complex planning, and working memory, were not assessed in our study. We selected specific tests to assess executive control based on their pragmatic clinical feasibility. Our use of a factor-analytic approach, including the Bartlett method, is appropriate for estimating latent constructs and aligns with the primary aims of this study. However, we acknowledge that this method produces a composite measure of ‘general’ executive control, which limits our ability to disentangle the unique contributions of individual executive functions. Although our findings support the mediating role of executive control broadly, they do not allow for precise inferences about which specific executive processes may be most relevant to the relationship between IIV and IADL performance in stroke survivors. Fifth, although the motor capacity is reflected by strength in the upper and lower limbs, other aspects of motor function such as coordination and dexterity may play a role in our findings. In future studies, motor assessments such as the Fugl-Meyer Assessment, Action Research Arm Test, and dexterity tasks could add important aspects of motor function to the mediation analysis. Sixth, other cognitive deficits, such as memory impairments, could also influence how behavioral IIV relates to IADL performance. Seventh, the influence of lesion characteristics on our findings remains unclear and may interact with the association between IIV and IADL performance in stroke survivors. We did not account for lesion location and our model may therefore underestimate the true strength of the IIV-IADL performance association that may exist within specific, more homogenous subgroups of stroke survivors. Eighth, this study did not account for the presence of comorbidities, physical activity, or rehabilitation history which may significantly impact IADL performance. Lastly, psychological factors like self-efficacy or optimism were not considered but warrant future investigation.

Conclusions

Here, we demonstrated that the inability to execute consistent performance following stroke is linked to reduced capacity to perform daily IADLs. Executive control, but not motor strength capacity, plays a key role in mediating this association. These findings highlight the need to consider behavioral variability (i.e., IIV) and executive control when developing rehabilitation strategies to facilitate functional recovery after stroke.

Abbreviations

CI

Confidence interval

DA

Divided attention

DSST

Digit Symbol Substitution Test

FAQ

Functional activities questionnaire

Hz

Hertz

IADL

Instrumental activities of daily living

IIV

Intraindividual variability

mRS

Modified Rankin Scale

ms

Milliseconds

MVC

Maximum voluntary contraction

N

Newtons

OTDL

Observed task of daily living

s

Seconds

SA

Selective attention

SD

Standard deviation

SE

Standard error

TMT-B

Trail Making Test Part B

UFOV

Useful Field of View

Author contributions

SD drafted and edited the manuscript. SD, AT, and HT performed data processing, data analysis, and reviewed the manuscript. SD prepared the figures. SD and AT performed data collection. SP assisted with participant recruitment. SP and MD reviewed the manuscript. NL provided study design, conceptual framework and approved the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by the National Institutes of Health (K01 AG070327 to NL).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on request.

Declarations

Ethics approval and consent to participate

The Institutional Review Board of Colorado State University approved the experimental procedures and all participants gave written informed consent prior to study procedures (20-9927H v. 11/18/2020).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.