Real-life cognitive functioning after acquired brain injury: An experience sampling study

Abstract

Background

Acquired brain injury (ABI) often leads to cognitive impairments, typically measured with cognitive tests in controlled environments. However, cognitive functioning in daily life is likely to fluctuate. This study explored the relationship between traditional subjective and objective cognitive measures and cognitive variability throughout the day and hypothesized that retrospective complaints are more strongly associated with momentary than baseline objective performance, and that greater cognitive variability would relate to higher subjective complaints. It also explored within-person associations between momentary cognitive performance and momentary affect, fatigue, social company and setting.

Methods

We conducted an experience sampling method study among 41 ABI patients. Baseline measures (Checklist for Cognitive and Emotional consequences following stroke, Montreal Cognitive Assessment and Digit Symbol Substitution Test) were administered during a briefing session. Participants responded to seven semi-random daily beeps over seven days using a mobile app. Momentary cognitive performance was assessed with a short digital version of the Digit Symbol Substitution Test.

Results

On average, participants completed 79% of beeps. Contrasting our hypothesis, cognitive variability was not associated with retrospective cognitive complaints or baseline cognitive performance. Multilevel analyses showed that momentary concentration complaints (β = 0.10, p = 0.02), fatigue (β = 0.10, p = 0.01) and being away from home (β = 0.23, p = 0.04) were associated with lower momentary cognitive performance. Affect and social company did not significantly impact performance. No between-person effects were found.

Conclusions

Momentary concentration complaints more accurately reflect real-time cognitive performance than retrospective questionnaires. Momentary fatigue impacts real-life cognition after ABI. Measuring cognitive variability can contribute to understanding the cognitive consequences of ABI.

Introduction

Cognitive impairments, such as memory or attention deficits, are among the most prevalent and impactful consequences of acquired brain injury (ABI), for instance, when patients want to re-engage in self-care, ¹ resume their social lives^2,3 or return to work. ⁴ Cognitive functioning is traditionally measured with a neuropsychological assessment (NPA), which includes a series of cognitive tests, each assessing specific cognitive domains in an optimal environment (e.g., quiet, free from noise, when the patient is well-rested). ⁵ These tests generally aim at providing an indicator of cognitive abilities (what you could do) and are indispensable for drawing up a profile of cognitive strengths and weaknesses. However, the results on these cognitive tests do not directly translate to performance in the complexity of daily life. Cognitive performance in daily life is likely to be affected by state-like factors such as fatigue or affect^6,7 and contextual factors such as noise, company and setting. ⁸ These factors impact cognitive capacity (what you are currently capable of), ⁹ causing variability in cognitive performance throughout the day and across days, which is not captured with cognitive tests in a controlled environment. As such, NPAs mainly give an indication about someone's cognitive abilities, while performance in daily life relies more on current cognitive capacity.

Patients’ perception of their own cognitive abilities is often assessed with retrospective self-report questionnaires. Although these measures are well-developed and useful, they have some limitations. When administered only once, self-report measures are unable to capture the large within-person variability caused by environmental and state-like factors, as they can only capture an aggregation of memories at a single moment, retrospectively. ¹⁰ People are susceptible to recall bias, which causes reported symptoms to be skewed towards their current state or extreme experiences, rather than reflecting all experiences comprehensively. ¹¹ Moreover, mood problems and limited awareness of deficits after ABI are known to affect appraisal of cognitive functioning on retrospective questionnaires. ¹² As a result, subjective self-reports of cognitive functioning after ABI often show weak to no associations with objective measures like cognitive tests.^13,14

One way to better understand the relationship between performance on cognitive tests and (self-reported) cognition in daily life is by testing the fluctuations in cognitive performance during the day using the experience sampling method (ESM). Experience sampling method is a digital self-monitoring method with high ecological validity, prompting repeated assessments of momentary experiences in daily life. ¹⁵ Whenever participants receive such a prompt (‘beep’), they are asked to reflect on their current status, environmental context and activities through Likert-scale questions on a mobile phone application. ¹⁶ The method has previously been used to examine the daily lives of patients with ABI and was found to be feasible.^17,18 Moreover, ESM measures of cognition have been proven valid and reliable in the general population, ¹⁹ and feasible for measuring within-person variability in cognitive performance. ²⁰ Previous research has found relationships between momentary affect, environment and cognitive performance in healthy adults ¹⁶ and older adults. ²¹ Furthermore, a study on cognitive performance in daily life after breast cancer found a link between within-person levels of fatigue (but not depressed mood) and momentary processing speed. ²² A recent study investigated the relationship between cognitive variability (with cognitive performance measured once a day for 14 days in a row) and daily functioning after ABI. According to the investigations carried out, it appears that greater cognitive variability was associated with lower daily and social functioning levels. ²³ To date, no study has yet examined the effects of affect, fatigue, social company and setting on momentary cognitive performance in patients with ABI.

Therefore, the current study was conducted to investigate cognitive fluctuations in daily lives of patients with ABI. The aim of the study was twofold. Our first research question was: What is the strength of the association between traditional (pen-and-paper) measures of objective cognition in a controlled environment, retrospective measures of subjective cognitive complaints, and momentary objective cognitive performance in patients with ABI? Given that momentary cognitive performance reflects cognitive capacity rather than ability, we hypothesized that the association between momentary cognitive performance and retrospective cognitive complaints would be stronger than the relationship between objective cognitive performance in a controlled environment and retrospective cognitive complaints. Additionally, we hypothesized that individuals with greater cognitive fluctuations may struggle to maintain their cognitive abilities throughout the day, leading to heightened subjective cognitive complaints. Our second research question was: are within-person momentary positive affect (PA) and negative affect (NA), fatigue, social company and setting associated with momentary objective cognitive performance?

Methods

Design

The current research is a prospective, observational ESM study into the relationship between several indices of cognitive functioning and the association between momentary cognitive performance and momentary measures of mood, fatigue, social company and setting in persons with ABI. This study was conducted in accordance with the Declaration of Helsinki ²⁴ and was approved by the Medical Ethical Committee of Maastricht University (reference number METC 2022-3570). All participants provided written informed consent.

Participants

Participants were selected from contact lists of individuals with ABI (aged 18 and older) who had previously participated in studies at the Limburg Brain Injury Centre (Maastricht, the Netherlands) and had given permission to be contacted for future research. Eligibility criteria included: 1) Medically confirmed ABI as a result of a stroke, traumatic brain injury (closed brain injury/contusion or penetrating head injury), brain infection, hypoxia (for instance after a cardiac arrest) or intoxication; 2) Good comprehension of the Dutch language; 3) Possessing and capable of handling a smartphone (based on self-report); 4) Be willing and able to give informed consent. Participants were excluded if they met one of the following criteria: 1) Being diagnosed with a neurodegenerative disease; 2) Visual limitations which cannot be compensated for by reading aids; 3) Being admitted to a hospital ward or rehabilitation centre during the time of the study. We initially based our target sample size (N = 40) on ESM studies with comparable designs,^16,21,22 as formal tools for power analysis in nested ESM data were not widely available at the time of study conception. Later, our power to detect within-person effects was confirmed using simulation-based analysis by Lafit et al., ²⁵ based on parameters from Verhagen et al. ¹⁶ Simulations showed high power (0.893) and good confidence interval coverage (94.4%) at N = 40.

Measurements

Baseline measures

At the start of the study, participants answered demographic (age, gender, living situation) and injury-related questions (type of injury, time since injury, previous ABI, hospital admission, length of hospital stay and discharge destination).

Subjective cognitive complaints: Cognitive complaints were measured using the self-reported cognition subscale of the Checklist for Cognitive and Emotional consequences following stroke (CLCE). ²⁶ The subscale contains 13 items on common cognitive complaints after ABI (e.g., ‘Since the brain injury, I have difficulty doing two things at the same time’). The items are scored as 0 (absence of problem) or 1 (presence of problem) and summed into a total score. The CLCE has good internal consistency. ²⁶

Objective cognitive functioning: In order to measure cognitive functioning in a controlled environment, participants completed the 90-s paper-and-pen version of the Digit Symbol Substitution Test (DSST), a subtest of the Wechsler Adult Intelligence Scale – IV. ²⁷ The DSST assesses information processing speed and working memory and is a sensitive measure for detecting cognitive impairment. ²⁸ The scores on the 90-s version were extrapolated for calculating z-scores with the norms of the 120-s version of the DSST (Wechsler, 2012). A z-score lower than −1.5 was used as a cut-off score for cognitive impairment on the DSST. Additionally, as a global measure of cognitive functioning, patients completed the Montreal Cognitive Assessment (MoCA). ²⁹ The MoCA is a short screening assessment of six cognitive domains (memory, visuospatial abilities, executive functioning, attention, language and orientation). Lower scores indicate worse cognitive performance. Total scores range from 0 to 30; scores of ≥26 represent normal cognitive functioning. ²⁹

Experience sampling method measures

Momentary measurements were collected using the m-Health smartphone application m-Path (www.m-path.io). ³⁰ Data collection took place from July to September 2023. The application was programmed to deliver seven prompts (beeps) daily at semi-random moments between 9:00 and 22:00 (once within a timeframe of 111 min; beeps were separated by 15 min minimum). Momentary data were collected over the course of seven consecutive days, resulting in a maximum of 49 assessments per participant (7 beeps × 7 days). Each beep took 2–3 min to complete. Participants were instructed to respond as promptly as possible, and responses not provided within 15 min were marked as missing.

Each prompt comprised a 20-item self-report questionnaire consisting of questions on how participants were feeling at that moment. ¹⁶ For the current study, only a subset of selected items relevant to our research questions was selected and listed (Table 1). All administered items are displayed in Supplemental Table A. Momentary items for the current study consisted of eight 7-point Likert-scale questions on momentary PA and NA based on the Positive and Negative Affect Schedule ³¹ and previous ESM studies into mood^32,33; one Likert-scale question on momentary fatigue; and one Likert-scale question on momentary subjective concentration problems. All Likert-scale questions ranged from not at all (1) to very much (7). Moreover, participants completed two multiple-choice items on their location and social company. Afterwards, they completed the momentary Digit Symbol Substitution Test (mDSST) ¹⁶ as a measure of cognitive performance in daily life. The mDSST was based on the DSST ²⁷ and consists of item screens displaying the encoding key on top and response grid displaying symbols at the bottom. For each trial, a random number was displayed in the middle of the screen, for which participants had to select the matching symbol in the response grid by tapping it (Figure 1). The organization of the answering key changed with every trial. Each task lasted 30 s, and the outcome of the mDSST was the number of correctly completed trials. As a result of the shortened length of the task, the mDSST is believed to primarily measure processing speed (and not working memory), as participants lack the time to memorize the encoding key.

Table 1.

Momentary items in the current study.

Domain	Item	Response options
Positive affect (PA)	I feel energetic. I feel relaxed. I feel satisfied. I feel confident.	7-point scale (1 not at all – 7 very much)
Negative affect (NA)	I feel irritated. I feel lonely. I feel anxious. I feel sad.	7-point scale (1 not at all – 7 very much)
Fatigue	I feel tired.	7-point scale (1 not at all – 7 very much)
Subjective cognitive complaints	Since the last beep, I experienced concentration problems.	7-point scale (1 not at all – 7 very much)
Location	Where am I?	Multiple-choice: Home, With friends/ family, Work, Healthcare institution, Public space, Transportation, Other
Social company	Who am I with?	Multiple-choice: Partner, Family, Housemate, Friend, Colleague, Acquaintance, Stranger, Nobody
Cognitive functioning	30-s processing speed task	mDSST

mDSST: momentary Digit Symbol Substitution Test.

Questions and response options are in the same order as they were displayed in the m-Path questionnaire.

Figure 1.

Screenshot of a single trial of the momentary Digit Symbol Substitution Test (mDSST).

Note: Participants looked up the number in the encoding key at the top of the screen, after which they tapped the corresponding symbol in the response grid at the bottom of the screen.

Procedure

Participants were invited to take part in the study via email. Those who expressed interest were contacted by phone to receive further information and undergo an initial screening based on the inclusion and exclusion criteria. Eligible participants were sent an information letter, and a follow-up call was scheduled. After this second call, an appointment was made for the briefing session of ±1 h.

The briefing session was conducted either at the participant's home or at Maastricht University, depending on their preference, in a quiet environment. During this meeting, informed consent was obtained, and baseline assessments (demographic and injury-related questions, CLCE, DSST and MoCA) were conducted. Participants then installed the m-Path app on their smartphones, received detailed instructions on how to respond to the beeps and completed a test trial. To support participants throughout the study, they were provided with an information leaflet on completing the momentary items (see Supplemental Materials). Additionally, the research team was available via mobile phone to provide technical assistance when needed.

The seven-day ESM data collection period started the next day. Afterwards, participants were contacted for a debriefing session. During this session, those interested were offered a personalized summary of their cognitive performance and its association with relevant predictors as a non-monetary incentive. Moreover, a short evaluation questionnaire containing Likert scale questions (1 not at all – 7 very much) was administered to assess the acceptability of the paradigm and to enable data quality checks.

Analyses

Descriptive analyses were used to describe the sample, and compliance rates (# completed beeps/# total beeps) were calculated. Data of participants with a response rate of <33% were excluded, consistent with standard procedure in ESM research. ³⁴

To address our first research aim, Pearson's correlations were calculated between the pen-and-paper DSST, CLCE cognition subscale, MoCA and averages of the mDSST of each participant. Individual variability on the mDSST was assessed with two measures of variance; variability, defined as a measure of dispersion of scores from a central tendency, and instability, a measure of the extent to which consecutive scores are related, or show temporal dependency. ³⁵ The cognitive variability (CV) was used as a measure of variability and is calculated using the within-person standard deviation of mDSST scores divided by within-person mean mDSST score. The mean squared successive difference (MSSD) was used as a measure of instability and contains the squared distance from each measurement point to the next measurement point and is considered a measure of stability. ³⁵ MSSD scores were transformed using square root transformation to ensure the assumption of bivariate normality.

Our second research aim was addressed using multilevel regression analyses to determine the relationship between momentary predictors (PA, NA, fatigue, concentration problems, location and social company) as independent variables and mDSST scores as the dependent variable. Experience sampling method research typically results in a multilevel data structure, with beep-prompted observations (level 1) nested within individuals (level 2). Categorical data for setting and social company were dichotomized into ‘home/away’ and ‘alone/with others’, respectively. The Likert-scale items energetic, relaxed, satisfied and confident were aggregated to represent PA, and the items irritated, lonely, anxious and sad were aggregated to represent NA. Covariates included in the model were age (level 2) and the log transformation of consecutive beep number (level 1), with the latter included to control for potential practice effects. Covariates were selected based on their observed associations with the outcome measure; gender was also examined as a potential covariate but was found to be unrelated to mDSST scores. For all independent variables, two variables were created: the grand mean-centred person mean to assess between-person effects, and the person mean-centred value for assessing within-person effects. Missing data were not imputed, as multilevel models can handle missingness under the assumption that data are missing at random.

Identification of possible influential outliers was based on guidelines for standardized residuals (values greater than 3 or less than −3) and Cook's distance (values greater than [4/number of observations]). Due to the presence of autocorrelation in mDSST scores, an AR(1) autocorrelation model with continuous time structure was used. Models were fitted using maximum likelihood to allow model comparison with different fixed effect combinations using likelihood ratio tests. Model fit was assessed using the Akaike Information Criterion. The final model included PA, NA, fatigue, concentration, social company, location and age as fixed effects. Beep number was modelled with random slopes, and participant-level random intercepts accounted for individual differences. All analyses were performed in R (version 4.2) ³⁶ using the lme4 package. ³⁷ Alpha was set at .05 throughout.

Results

Sample characteristics

In total, 43 individuals participated in the study. One participant was excluded from the analyses due to not meeting the compliance threshold (compliance rate of 32%). One other participant was excluded due to technical issues (i.e., erroneous questionnaires), leading to a final sample of 41 participants. The overall compliance rate of the included participants was 79% (range 43%–96%). On average, participants were 56 years old (median; range 23–71) and had sustained their brain injury 6.3 years ago (median). Twenty-one participants (51%) scored below cut-off on the MoCA, and 15 participants (37%) had a z-score below −1.5 on the DSST. Participants reported an average of 6.8 cognitive complaints. Participant characteristics are summarized in Table 2.

Table 2.

Participant characteristics and scores on baseline measures (n = 41).

Variable	Range	Med/M(SD)/n (%)
Age (median)	23–71	59
Gender (male)		18 (44%)
Type of injury Stroke (CVA/TIA/SAH) TBI Other		30 (73%) 7 (17%) 4 (8%)
Time since injury in years (median)	1–41	6.3
Cognitive functioning (MoCA) % impaired (<26)	18–30	25.4 (2.7) 21 (51%)
Cognitive complaints (CLCE – cognition subscale)	0–13	6.8 (3.1)
Paper-and-pencil DSST (baseline) % impaired (Z < -1.5)	23–53	40.5 (12.2) 15 (36%)

CVA: cerebrovascular accident; TIA: transient ischaemic attack; SAH: subarachnoid bleeding; MoCA: Montreal Cognitive Assessment; CLCE: Checklist for Cognitive and Emotional consequences following stroke; DSST: Digit Symbol Substitution Test; Med: median; M: mean; SD: standard deviation.

Patterns of daily cognitive functioning

On average, participants completed 11 mDSST trials per beep (Range = 1–22, SD = 3.3). Overall, participants experienced high PA (M = 5.0, SD = 1.2), low NA (M = 1.7, SD = 0.9) and moderate levels of fatigue (M = 4.4, SD = 1.8). All mean scores of momentary measures are presented in Table 3.

Table 3.

Mean scores on momentary measures.

Variable	Range	M (SD)
Overall response rate (% of completed beeps)	43–98	78.6 (13)
Momentary DSST	1–22	10.8 (3.3)
Momentary positive affect	1–7	5.0 (1.2)
Momentary negative affect	1–7	1.7 (0.9)
Momentary fatigue	1–7	4.4 (1.8)
Momentary concentration	1–7	2.6 (1.8)

DSST: Digit Symbol Substitution Test; M: mean; SD: standard deviation.

Figure 2 presents individual plots of cognitive variability for each participant. The figure illustrates differences in average performance levels (represented by the orange line) across participants. For instance, participant 041 exhibits the highest average number of correctly completed trials per beep (17.1), while participant 013 demonstrates the lowest average score (5.6). A clear practice effect is evident for most participants; as illustrated in the individual plots (e.g., participants 018, 039, 040 and 041), the total number of correctly completed trials is lower on the first day compared to subsequent days. Variability in within-person performance is also apparent from these plots, with participant 031 exhibiting the greatest fluctuations relative to their average scores. In contrast, participants 038 and 001 are examples of individuals with low variability.

Figure 2.

Individual plots of cognitive variability over the course of seven days. Orange lines represent within-person means on the momentary Digit Symbol Substitution Test (mDSST). Dotted lines represent within-person standard deviations. Vertical graph lines represent days. Horizontal graph lines represent the number of correctly completed trials on the mDSST per beep. Connected dots represent consecutive measurements.

Correlation between measures of cognition

The number of self-reported cognitive complaints did not correlate with any other measure of cognition. This suggests that the total number of cognitive complaints reported by participants was not associated with their average performance on the momentary cognitive test, cognitive variability (indexed by CV) or cognitive instability (indexed by MSSD). However, a strong negative correlation was observed between the mean mDSST score and cognitive variability, indicating that participants with greater cognitive variability overall performed worse on the momentary cognitive test (r = 0.54, p < 0.001). Participants with greater cognitive variability in daily life also showed lower performance on the pen-and-paper DSST (r = 0.42, p < 0.001). All correlations are displayed in Table 4.

Table 4.

Pearson correlation coefficients between traditional measures of cognition and momentary cognitive performance.

	Pen-and-paper DSST	CLCE cognition subscale	MoCA total	Mean mDSST	CV	MSSD
Pen-and-paper DSST	1
CLCE cognition subscale	−0.044	1
MoCA total	0.375*	0.183	1
Mean mDSST	0.763**	−0.154	0.267	1
CV	0.415**	0.174	−0.195	0.543**	1
MSSD	−0.033	0.112	0.031	−0.021	.423**	1

*p < 0.05; **p < 0.01.

DSST: Digit Symbol Substitution Test; CLCE: Checklist for Cognitive and Emotional consequences following stroke; MoCA: Montreal Cognitive Assessment; mDSST: momentary Digit Symbol Substitution Test; CV: coefficient of variance; MSSD: mean squared successive differences.

Relationship between momentary factors and cognitive performance

The multilevel regression model showed that momentary fatigue, concentration and setting were significant predictors of the number of correctly completed trials on the mDSST, while controlling for all other variables (see Table 5 for model estimates). Specifically, fatigue was significantly negatively associated with the number of correctly completed trials at the within-person level, suggesting that higher-than-usual fatigue was linked to worse cognitive performance (β = −0.10, p =0.009). However, the between-person effect was not significant (p =0.45). Similarly, concentration problems exhibited a significant negative within-person association with the number of correctly completed trials (β = −0.10, p =0.019), indicating that greater concentration problems experienced since the last beep were associated with lower cognitive performance. The between-person effect of concentration was not significant (p =0.10).

Table 5.

Linear mixed model parameters on the association between momentary factors and number of correctly completed trials on the mDSST.

Predictor	β	SE	t(df)	p-value
(Intercept)	14.67	1.98	7.40 (1443)	<0.001***
Positive Affect (within-person)	0.12	0.08	1.56 (1443)	0.12
Positive Affect (between-person)	−0.18	0.56	−0.32 (35)	0.75
Negative Affect (within-person)	−0.05	0.10	−0.49 (1443)	0.62
Negative Affect (between-person)	−0.72	0.62	−1.16 (35)	0.26
Fatigue (within-person)	−0.10	0.04	−2.61 (1443)	0.009**
Fatigue (between-person)	0.26	0.34	0.76 (35)	0.45
Concentration problems (within-person)	−0.10	0.04	−2.35 (1443)	0.019*
Concentration problems (between-person)	−0.56	0.33	−1.69 (35)	0.10
Company: with others	−0.09	0.11	−0.87 (1443)	0.38
Setting: away	−0.23	0.11	−2.03 (1443)	0.04*
Beep number (log)	1.00	0.11	9.08 (1443)	<0.001***
Age	−0.12	0.03	−3.36 (35)	0.002**

mDSST: momentary Digit Symbol Substitution Test.

Note: Estimates are unstandardized regression coefficients (β) with standard errors (SE), t-values with degrees of freedom (df) and p-values. Within-person predictors were centred around person means (level 1). Between-person predictors reflect the grand mean-centred averages (level 2). *** p < 0.001; **p < 0.01; *p < 0.05.

Setting also influenced cognitive performance: being away from home was significantly associated with fewer correctly completed trials (β = −0.23, p =0.04), whereas the presence of others did not significantly affect the number of correctly completed trials (p =0.38). Furthermore, the log-transformed beep number demonstrated a strong positive association with the number of correctly completed trials (β = 1.00, p < 0.001), indicating an overall improvement over time. Lastly, age was negatively associated with the number of correctly completed trials (β = −0.12, p =0.002), suggesting that older individuals tended to perform worse on the task.

Acceptability

During the debriefing session, participants generally indicated that they did not think it was difficult to use the app (M = 1.3), found the questions easy to read (M = 6.9), and the verbal and in-text briefing about the app to be clear (M = 6.8 and M = 6.2, respectively). They found the number of beeps slightly burdensome (M = 3.0), but not the duration of the questionnaires (M = 1.9). No technical issues were reported. Data on all acceptability questions is reported in Supplemental Table B.

Discussion

This study aimed to investigate cognitive fluctuations in the daily lives of patients with ABI. First, we examined the relationship between traditional cognitive tests, specifically the DSST and MoCA, as well as retrospective self-reports of cognitive complaints, and momentary measures of objective cognition, respectively. Additionally, we explored how momentary PA, NA, fatigue, social company and setting were associated with momentary cognitive performance. We hypothesized that the association between momentary cognitive performance and retrospective cognitive complaints would be stronger than the relationship between objective cognitive performance in a controlled environment and retrospective cognitive complaints. Additionally, we hypothesized that individuals with greater cognitive fluctuations reported more subjective cognitive complaints on a retrospective questionnaire. However, neither of these hypotheses was supported, as retrospective complaints did not correlate with cognitive performance measures or variability indices. Despite this, a significant within-person association was observed between momentary cognitive performance and concentration complaints, indicating that subjective and objective measures of cognition are related in real life. Exploratory analyses further revealed that momentary fatigue, relative to a person's usual fatigue level, and being away from home were both associated with lower momentary cognitive performance. No significant associations were found between momentary cognitive performance and PA, NA or social company.

Interestingly, the current study found a significant negative correlation between cognitive variability and both baseline and averaged momentary DSST performance, which contrasts Schmitter-Edgecombe et al., who did not find any relationship between average (momentary) cognitive performance and measures of cognitive variability (i.e., the extent to which an individual's cognitive performance fluctuates over time). ²¹ This discrepancy may be explained by the fact that the aforementioned study included a healthy population; however, it may also be due to differences in how variability was measured. Schmitter-Edgecombe et al. used within-person standard deviations of n-back scores as a measure of variability. Since within-person standard deviations are absolute measures of variance, they are known to be affected by within-person mean scores. ³⁸ Using a standardized measure of variance controls for the impact of overall performance level ³⁹ and may therefore be a more reliable estimate of variability. Moreover, the results of our study are consistent with previous studies on cognitive variability among older adults.^40,41 In a recent intensive sampling study, Munsell et al. studied day-to-day variability in cognitive performance among stroke patients. ²³ In addition to a significant association between higher variability and reduced daily activities and social functioning, they observed a trend towards significance in the differences between the neuropsychological profiles of patients across different cognitive variability classes. While further research is needed, existing evidence suggests that higher cognitive variability may be associated with poorer outcomes following ABI.

An explanation for the lack of correlation between measures of variance and retrospective questionnaires on cognitive complaints in our study may be that the CLCE covers all cognitive domains, whereas we only tested momentary speed of information processing. However, since the CLCE was also not associated with the MoCA – a cognitive screening tool that assesses multiple cognitive domains – the CLCE results likely reflect additional factors, such as recall bias and individual differences.^11,12 Since momentary concentration problems were associated with performance on a within-person, but not a between-person level, our results suggest that participants’ momentary cognitive complaints may be more reflective of actual cognitive capacity than retrospective complaint questionnaires, or overall levels of concentration complaints.

The findings of a previous study into momentary cognitive functioning among breast cancer patients with potential cancer-associated cognitive decline ²² were similar to our study. In their study, Small et al. also found an effect of higher-than-usual fatigue on momentary cognition. This means that, compared to patients’ average level of fatigue, higher-than-average levels contribute to lower cognitive performance. Similar to the current findings, the authors also found no within- or between-person effects of depressed mood on cognitive performance. In contrast, Verhagen et al. found a main effect of PA as well as an interaction effect of positive by NA on cognitive performance in a sample of healthy older adults. ¹⁶ The results indicate that, in healthy individuals, NA does not influence momentary cognitive performance when PA is high, but has a detrimental effect when PA is low. Concluding from the current and previous studies, it could be the case that affective state, in particular anxiety, drives cognitive performance in cognitively healthy individuals.^16,42 In contrast, in patients with cognitive impairments, factors like momentary fatigue may be more predictive of cognitive performance.

Implications for future research and healthcare practice

Given the high acceptability rates of the app and questionnaires used in the current study, our findings are of relevance to healthcare for patients with ABI, particularly those in rehabilitation settings. Beyond recognizing (invisible) cognitive variability, which may help reduce stigma following ABI, ⁴³ the proposed paradigm offers a valuable tool for collecting individual data on real-life cognition. For instance, healthcare professionals can use ESM tools like m-path to collect momentary data on cognitive performance through patients’ smartphones to assess individual patterns of cognitive variability throughout the day. These patterns can then be integrated with (neuro)psychological treatment, particularly within the context of ecological momentary interventions (EMIs) for improving cognitive functioning. Previous EMI studies in middle-aged and older adults included monitoring of biopsychosocial health aspects, focused on various populations, such as people with depression, chronic pain or eating disorders, were overall feasible, and improved outcomes.^44,45 Specifically, by offering feedback based on own real-life cognitive, behavioural or emotional pattern, a behavioural change may be promoted ⁴⁶ which is line with the social cognitive theory. ⁴⁷ To our knowledge, few EMI studies targeted cognition specifically,^48,49 and EMIs for people with neurological conditions remain rare. Future research may therefore co-create EMIs with people with ABI and iteratively evaluate them, ensuring that such a treatment meets the needs of people affected.

The practice effect observed in our study was mainly limited to the first day, consistent with earlier ESM findings. ¹⁶ This suggests participants adapted quickly to the task, after which performance fluctuations reflected natural variability rather than continued learning effects. Prior research has shown that the steepness of early learning curves in repeated cognitive assessments may reflect individual differences in cognitive functioning, ⁵⁰ underscoring the potential value of capturing such patterns in ESM-based cognitive measures.

The present study provides evidence of intra-individual differences in cognitive variability throughout the day. However, little is known about the normative range of these fluctuations or the threshold at which variability becomes pathological. Future research should therefore focus on examining cognitive variability across different populations and comparing variability levels between patient groups and with a healthy control group. Additionally, the prognostic significance of cognitive variability needs to be further investigated, particularly in the context of treatment planning during the (post)acute phase of brain injury. Similarly, future studies should explore daily patterns of functioning, as previous research suggests that time of day may influence cognitive performance following ABI. ⁵¹

While our study controlled for the influence of age, other person-level factors such as injury type and time since injury were not included due to sample size constraints. Previous research indicated that cognitive variability depends on the recovery phase. ⁵² Future studies with larger samples should therefore incorporate injury-related characteristics to allow subgroup analyses and explore how demographic and injury-related factors interact with momentary predictors of cognitive performance.

Finally, future studies could build on our results by further distinguishing between situational predictors, such as being at work, or managing caregiving responsibilities, as these contexts are associated with increased cognitive demands in non-ABI populations.^53,54 Examining how contextual factors interact and influence momentary cognition could clarify the mechanisms through which fatigue and setting affect daily functioning.

Strengths and limitations

This study was the first to assess cognitive variability in an ABI sample. Given the high response rates, observed variability and strong correlation with the pen-and-paper DSST, the findings suggest that administering the mDSST via an ESM platform is a potentially valid and reliable method for measuring real-life cognition. Importantly, the mDSST demonstrated sensitivity to within-day fluctuations, highlighting its potential for capturing dynamic cognitive changes. Furthermore, this study contributes to understanding the discrepancy between self-reported cognitive complaints and NPA results, as well as identifying factors that may influence cognitive performance in daily life. However, this study was not without limitations. The mDSST only tested speed of information processing, yet no other cognitive functions. As the mDSST has been shown to be sensitive to fluctuations, ³² this was an appropriate target for the present study; however, additional cognitive domains should be explored. Furthermore, although ESM measures of cognition show satisfactory psychometric properties, ¹⁹ the validity of the mDSST has not been tested in an ABI sample before. Therefore, the validity and reliability of the mDSST for this population need to be further evaluated. Still, the results of our study provide preliminary support for its internal validity in this context and should be replicated in future research. Regarding our analyses, there was a practice effect which was not fully accounted for, though this was partially mitigated by adding a log-transformed beep number. The correlational analyses and between-person effects were based on participant-level variables (level 2), which may limit statistical power. Lastly, our sampling strategy recruited participants from previous studies on the consequences of ABI and required smartphone use. This may have introduced a selection bias towards a higher-functioning subgroup of our target population, potentially evidenced by the relatively low proportion of participants showing impairment on the MoCA (36%) compared with rates reported in other studies on MoCA scores among inpatients with ABI in the (sub)acute phase (84–89%).^55–57 As a result, our findings may not fully generalize to all persons with ABI, particularly those with more severe cognitive impairments, or those who have limited access to digital technology. Future studies should aim to include more representative samples to confirm the applicability of these findings to the broader ABI population.

Conclusion

This study provides new insights into cognitive fluctuations in the daily lives of individuals with ABI. By using momentary assessments, we found that cognitive performance varies within individuals and is influenced by momentary fatigue and being away from home. Retrospective cognitive complaints did not correlate with objective performance or variability indices. Instead, momentary concentration complaints were linked to real-time cognitive performance, suggesting that subjective cognitive experiences may be more accurately captured in the moment than through retrospective self-reports. Given the high acceptability of the ESM approach to measuring cognition, this method holds promise for both research and clinical applications, particularly in rehabilitation settings. Future studies should further explore cognitive variability across populations, its clinical significance, and the potential of ecological momentary cognitive rehabilitation interventions. Additionally, expanding assessments to other cognitive domains will enhance our understanding of real-life cognitive functioning after ABI.