Detecting and Predicting Cognitive Decline in Individuals with Traumatic Brain Injury: A Longitudinal Telephone-Based Study

Introduction

Traumatic brain injury (TBI) is a leading cause of disability in the United States, with an estimated 2.8 million individuals sustaining a TBI every year ¹ and 3.2 million people living with permanent TBI-related disability. ² Many individuals with a history of moderate-to-severe TBI experience enduring deficits in attention, processing speed, executive functioning, memory, and other cognitive domains. ^{3 –8} These long-term cognitive challenges can impact day-to-day functional independence, community integration, and life quality. ^{9 –14}

Evidence suggests that the most dramatic improvements occur in the first 6–18 months post-injury. ^5,15 However, changes in cognition after longer periods are not uncommon, ^5,16,17 and change may even be more common than stability. ^10,18 Improvements to cognitive performance have been detected up to 5 years following TBI, ^5,19 and retrospective data suggest that improvement can extend up to 10 years post-injury. ¹⁷ However, substantial evidence suggests that TBIs may contribute to cognitive decline, including risk for dementia and neurodegenerative disease. ^20,21 Several studies have demonstrated associations between TBI and later-life cognitive impairment and decline, including acceleration of cognitive decline and increased risk for developing Alzheimer’s disease (AD) and AD-related dementias. ^{22 –26} Given the concerns about long-term cognitive function and dementia risk expressed by TBI survivors and their loved ones, ^22,25,26 it is of particular clinical relevance to understand who is at risk for decline. Identifying early cognitive decline is of paramount importance for accurate diagnosis, informed treatment planning, and provision of long-term care for both patients and their caregivers. ^27,28

Although comprehensive in-person neuropsychological assessment is the gold standard for measuring cognitive performance, ²⁹ this resource-intensive and highly specialized evaluation is not always practical for the purpose of cognitive monitoring, especially for those with mobility and transportation limitations, compromised health, and socioeconomic challenges that impact access to care. ³⁰ Assessment tools that are capable of detecting subtle cognitive changes over time and that can be administered remotely (e.g., over the telephone) are needed for this purpose. The feasibility and validity of the Brief Test of Adult Cognition by Telephone (BTACT) have been established in individuals with TBI, including moderate-to-severe TBI, ^{31 –33} and the BTACT can identify cognitive change beyond the point of spontaneous recovery (i.e., 6 months to 2 years post-injury). ^34,35 It is not known, however, whether this brief telephone-administered test battery can detect more subtle changes in the chronic stages of TBI.

Detection of cognitive decline that is not just statistically significant but also clinically meaningful ³⁶ requires assessment tools that are sensitive to measuring fluctuations in cognition over time. Although statistical significance can be influenced by factors such as sample size, measurement sensitivity, and variability of data, clinical significance refers to a degree of change that impacts an individual’s life in a meaningful way. True change must be distinguishable from measurement error, practice effects, regression to the mean, or extraneous factors. ³⁷

Many approaches commonly adopted by investigators seeking to identify clinically meaningful change suffer from important drawbacks. ³⁸ Group differences based on means or difference scores may mask the direction of individual change, overlooking important subgroups that decline over time rather than improve or vice versa. Difference scores, while computationally appealing for their simplicity, assume a ratio scale and ignore variations in baseline level of functioning and measurement error. Purely distribution-based thresholds (e.g., standard deviation cut points) also overlook measurement error and test–retest reliability and do not take the individual’s perception of noticeable change into account. Minimum clinically important difference values also overlook issues such as reliability and measurement error, and they also rely on self-assessments of detectable changes, and some may argue that low self-awareness impacts accuracy of perceived change. ^{39 –41}

The reliable change index (RCI) ⁴² overcomes many of these drawbacks and has long been used to quantitatively define clinically meaningful change. The RCI distinguishes changes attributable to a process or intervention from measurement error by adjusting for measurement error and test–retest reliability in determining the threshold for change. RCI is widely regarded as the optimal approach to determining meaningful change. ^38,43,44

If changes in cognitive function are to be detected, it is of great interest to TBI survivors and their care partners to identify factors that may be associated with cognitive decline over time. Demographic factors (e.g., age at injury, socioeconomic status, gender), ^{16,45 –47} injury severity (e.g., loss of consciousness and/or post-traumatic amnesia duration), ^{4,14,45,47,48} age, health and lifestyle factors, and comorbid diseases (e.g., cardiovascular-related conditions, anxiety and depression, chronic sleep conditions, diabetes, cancer) ^12,49 are recognized contributors to cognitive decline in the general population; many risk factors for decline are more common among TBI survivors. Comorbid vascular risk factors, for example, have been associated with poorer cognitive prognosis following TBI. ^47,50,51 Other common co-occurring conditions with implications for cognitive function include sleep difficulties, ^12,52,53 substance use disorders, ¹² pain, ^54,55 and depression. ^12,56 To the extent that factors contributing to cognitive decline are modifiable, they may represent treatment targets to optimize long-term brain health following TBI.

The aim of the present study is twofold. First, the study seeks to evaluate whether the BTACT can detect clinically meaningful change in cognitive functioning in the chronic stages post-TBI. Second, we determine the prevalence of clinically meaningful cognitive decline among individuals with chronic TBI and explore whether demographic, injury, and health conditions are associated with cognitive change.

Methods

Results

Discussion

The current study examined the ability of a brief telephone-administered cognitive test to detect cognitive change in a sample of chronic TBI survivors who were, on average, nearly 7 years post-injury. Although results varied by cognitive domain, we found evidence that the BTACT detects clinically meaningful RCI-based cognitive change (e.g., decline or improvement greater than RCI threshold) over a relatively brief period of time (i.e., ∼1 year) in this sample. Approximately 40% of participants in the current sample evidenced RCI-based changes in overall cognition, with greater rates observed for several cognitive domains. Current findings of improved performance in a subset of the sample not only challenge long-standing beliefs that cognitive recovery plateaus 1–2 years after injury ⁷⁸ but also converge with previously documented findings that a subset of survivors experience post-recovery decline in the chronic stages. ^5,16

Although our study identified significant cognitive changes over a 1-year period, we acknowledge that the terms “decline” and “recovery” may not imply continuous trends but rather episodic changes. It is essential to consider that cognitive fluctuations observed within this time frame may not necessarily represent ongoing trends toward decline or improvement but rather meaningful variations that could be influenced by various factors, including environmental influences and fluctuations in health conditions (e.g., anxiety, depression, and sleep difficulties). Moreover, our study does not determine the long-term trajectory of cognitive change beyond the observed 1-year period, nor does it establish whether these changes represent temporary fluctuations or ongoing trends.

Results of this study provide evidence that the BTACT is capable of detecting RCI-based change in a TBI cohort, which takes into account practice effects and test–retest reliability to infer clinically meaningful effects. This will be welcome news, given the growing number of TBI studies using this test, ^{29,31,32,65,79,80} many of which will be positioned to investigate whether RCI-based change documented herein reflects an ongoing trend. The BTACT is a brief, inexpensive, and repeatable measure of cognitive functioning that does not require in-person visits and, thus, does not result in a sample selected for the ability to attend in-person study visits. Given the high incidence among TBI survivors of factors rendering in-person testing unfeasible (e.g., health status and mobility considerations), ^29,31,32 evaluation of cognition over the telephone broadens inclusion in research and enables more representative investigation of cognitive functioning in this population.

The selection of RCI-based definition of change, as noted earlier, has distinct advantages over other approaches such as examining group-based change or calculating individual difference scores. RCI methods were recently applied to comprehensive neuropsychological test data and were able to identify change that corresponds to magnetic resonance imaging-based indices of biological aging. ³⁸ The use of brief, remotely administered tests of cognition may overcome some barriers associated with traditional in-person neuropsychological testing, making surveillance efforts more feasible, allowing for greater accessibility and convenience, and reducing the burden on both patients and health care providers.

We found relationships between demographic and health factors and decreased cognitive performance across multiple domains. Our objective was to comprehensively explore potential associations between various demographic and health factors and RCI-based cognitive decline across multiple domains. Participants whose cognitive performance declined between Visits 1 and 2 tended to be male and have greater lifetime head trauma exposure, self-reported dyspnea and seizure disorders, increased depressive and anxiety symptoms, and problematic substance use (e.g., illicit drug use). These findings are generally consistent with other studies investigating risk factors for cognitive decline and dementia in the general population. ^81,82 Certainly, substance abuse, ⁸³ depression, ⁸⁴ anxiety, ⁸⁵ and seizure disorders ⁸⁶ have all been associated with the elevated risk for cognitive decline and dementia. Similarly, dyspnea is a highly disabling symptom impairing cognitive functioning (e.g., poor executive function, attention, and processing speed). ^{87 –90}

There were some unexpected findings, such as the association between RCI-based cognitive decline and being a nonsmoker and having fewer cardiovascular diseases overall (e.g., stroke, heart attack, hypertension, heart failure, and other cardiac conditions). Smoking has been a curious factor in the neurodegenerative disease literature; for example, it is consistently found to be a protective factor for Parkinson’s disease. ^91,92 With respect to smoking and cardiovascular disease in the current study, it is possible that individuals with a history of excessive tobacco use, as well as those with multiple cardiovascular diseases (which tend to cluster in individuals), ⁹³ may die or refuse research participation before the age of greatest dementia risk, contributing to inconsistent findings with those documented previously in the dementia literature. ^82,94 The current finding that those who experienced an RCI-based decline in cognitive performance reported that they spent more hours sleeping on non-workdays and rarely felt unrested on non-workdays (e.g., once a month) is difficult to interpret based on the data available. Although it is possible that this could be interpreted as adequate sleep conferring protection against cognitive decline, it is also possible that these findings implicate undiagnosed obstructive sleep apnea (which affects 23–49% of those with TBI) ^95,96 and hypersomnolence (affecting 28–32% of those with TBI), ^{97 –101} which have been associated with cognitive impairment in TBI samples. ^79,102,103

Participants performed below the MIDUS II/MIDUS Refresher sample on most subtests, on average. There was, however, wide variability, suggesting good sensitivity to performance differences across individuals in this sample. Thus, proactive screening and treatment of conditions known to be associated with cognitive decline and dementia risk are warranted in groups with elevated rates of dementia such as TBI survivors. Although effective treatments for substance use, ^104,105 depression, ^106,107 anxiety disorders, ¹⁰⁸ and seizure disorders ^109,110 exist for the general population, there remains a need to tailor these interventions to accommodate TBI-related impairments to optimize accessibility, compliance, and efficacy in the TBI population. ^{111 –113} Relatedly, identifying high-risk subgroups with nonmodifiable risk factors (e.g., repeated TBIs across the lifespan) for surveillance is also important, as it can help to target resources and interventions to those in greatest need. Individuals in whom multiple risk factors converge may benefit from preventative interventions and more frequent cognitive monitoring.

There are several limitations to this study that should be noted. The BTACT provides a convenient and efficient cognitive assessment; however, domain-specific performance is based on a single subtest, and caution should be exercised when interpreting results, including observed changes in performance over time, in the context of complex cognitive functioning. The small sample size of the current study precludes more granular analyses of interactive effects of risk factors for cognitive decline. We focused on the state of health at the first visit and did not examine the effect of new diagnoses rendered between study visits, or whether diagnosed conditions were treated, and this study did not consider longevity of health conditions or whether they preceded or followed the TBI. These data would permit more detailed analyses of factors associated with cognitive improvement or decline in the chronic stages of TBI. The design of the parent study relied on MIDUS self-report of medical and behavioral conditions, which were not objectively corroborated. Calculations of RCI for each BTACT subtest relied on test–retest parameters generated on relatively small sample sizes from the original psychometric evaluation of the measure ³¹ ; updated results with larger samples would allow for more precision in RCI thresholds. Because we only had two assessments, we also cannot determine the shape of change over time; the trend could well be something other than linear and would require repeated assessments to ascertain. Furthermore, it cannot be known based on two assessments alone whether RCI-based change reflects a temporary fluctuation or an ongoing trend; this work requires extension with longitudinal data that include more than two follow-ups. Finally, the variability of RCI across subtests, which, in part, is the result of differences in test–retest reliability, reaffirms that approaches that take specific test characteristics into account are more psychometrically sound than those that rely on a simple, SD-based threshold.

Although the current study provides important insights into the factors associated with RCI-based cognitive decline, larger studies in diverse populations with more comprehensive neurobehavioral characterization and longer follow-up periods are necessary for a more comprehensive assessment of the clinical phenotype of post-traumatic neurodegeneration. Such studies could help identify specific cognitive domains that are particularly susceptible to decline following head trauma, as well as the temporal sequence of changes in these domains over time. Moreover, future research could examine interactions between various risk factors, which could help identify individuals who are particularly vulnerable to cognitive decline following head trauma. Furthermore, although not a focus of the current study, it should be noted that many individuals (rates ranging 6% to 22% across subtests) improved from Visit 1 to 2, underscoring the fact that cognitive improvement may continue many years post-injury. Understanding the factors contributing to continued improvement may similarly elucidate novel treatment targets.

Conclusion

Current findings indicate that a brief remotely deployable performance-based test of cognition can detect subtle changes in cognitive function, even in a chronic TBI cohort, over a relatively short period of time. Our findings support the notion that cognitive change is more likely than stability, consistent with evolving conceptualizations of TBI as a chronic disease process. ¹¹ The BTACT may prove useful in TBI surveillance efforts, particularly among high-risk subgroups, and in contexts in which in-person visits are unfeasible. Results described herein highlight the importance of considering a wide range of risk factors when examining cognitive decline following TBI and support the deployment of routine screening, prevention strategies, and targeted interventions aimed at addressing modifiable risk factors such as substance abuse, depression, anxiety, and sleep disturbances to reduce the risk of cognitive decline in TBI survivors.

Transparency, rigor, and reproducibility summary

The present study was not formally registered on ClinicalTrials.gov because it is not a clinical trial. The current research constitutes a secondary analysis of data from the TBI and Health Study, and therefore, no pre-specified analysis plan was outlined in the funded grants, nor was one registered before data collection. Out of the 126 participants initially enrolled, 90 met the inclusion criteria at the time of analysis; 36 were subsequently excluded. Recruitment occurred from February 2013 to December 2020, with follow-up data collected through May 2023. Various participant data, including demographics, head trauma history, medical conditions, and physical and psychosocial functioning, were collected via self-report questionnaires. The BTACT, a well-validated neuropsychological tool, was administered by trained data collectors under clinical neuropsychologist supervision. These examiners were unaware of the article’s specific aims and hypotheses. All data were securely stored and managed in REDCap. We encourage replication and external validation of our findings, with data accessible upon reasonable request to the corresponding author, facilitated by a subsequent data-use agreement.