Higher-Order Reasoning Training Years After Traumatic Brain Injury in Adults

Abstract

Objective: 

To conduct a feasibility study to compare the effects of top-down Strategic Memory and Reasoning Training (SMART) versus information-based Brain Health Workshop (BHW, control) on gist-reasoning (ie, abstracting novel meaning from complex information), memory, executive functions, and daily function in adults with traumatic brain injury.

Participants: 

Twenty-eight participants (of the 35 recruited), 16 men & 12 women, aged 20 to 65 years (M = 43, SD = 11.34) at chronic stages posttraumatic brain injury (2 years or longer) completed the training. Fourteen participants that received SMART and 14 participants that completed BHW were assessed both pre- and posttraining. Thirteen of the SMART trained and 11 from BHW participated in a 6-month testing.

Design: 

The study was a single blinded randomized control trial. Participants in both groups received a minimum of 15 hours of training over 8 weeks.

Results: 

The SMART group significantly improved gist-reasoning as compared to the BHW group. Benefits of the SMART extended to untrained measures of working memory and participation in functional activities. Exploratory analyses suggested potential transfer effects of SMART on memory and executive functions. The benefits of the SMART program as compared to BHW were evident at immediately posttraining and 6 months posttraining.

Conclusion: 

This study provides preliminary evidence that short-term intensive training in top-down modulation of information benefits gist-reasoning and generalizes to measures of executive function and real life function at chronic stages of post-TBI.

TRADITIONALLY, cognitive rehabilitation efforts following a traumatic brain injury (TBI) have focused on a structured “drill and practice” approach to remediate cognitive functions. Most treatment programs aim to improve (a) a specific cognitive process (eg, attention training), (b) functional tasks such as cooking or driving, or (c) use of external aids to compensate for memory impairments.^1–8 These training protocols have proven successful in improving general intellectual functioning, memory, and mastery of targeted skills in acute stages of recovery and continue to be beneficial posttraining for short periods of time (eg, 3–6 months).^9,10 However, a vast majority of survivors continue to present with impairments in daily functions that necessitate higher-order cognitive skills, especially in chronic stages of recovery (ie, 1 year or longer after injury).^11,12 In fact, the majority of existing cognitive treatments to remediate the sequelae of TBI occur during the acute phase of injury, despite the evidence that functional impairments produce lasting burden on the individual and family.¹³

The paucity of cognitive training programs that mitigate impairments of higher-order critical thinking appears to be driven by 2 primary factors. First, only a handful of cognitive training studies are grounded in a theoretical framework.¹³ Second, few controlled trials have been conducted to evaluate the treatment efficacy of cognitive training programs that specifically aim to improve higher-order skills.^9,13,14 To address these limitations, clinicians and theorists propose the application of theory-driven cognitive therapies that target frontal lobe-mediated top-down modulatory processes to improve higher-order cognitive skills in TBI.¹⁵ Top-down control processes are goal-oriented, internally driven, voluntary (not automatic) cognitive operations that both focus attention on task-relevant stimuli and ignore irrelevant distractions.^13,16,17 Neurally, top-down modulation involves bidirectional operations of both the enhancement and suppression of neural activity in cortical regions depending on the relevance of the information to our goals.^16–18 Increasingly, imaging data support the role of prefrontal cortices in top-down modulatory tasks.^18,19 Training top-down control processes in adults with TBI may be beneficial in restoring and improving cognitive function, as the frontal networks are typically the areas most affected by a TBI.^14,15

Researchers are beginning to adopt theoretical frameworks to guide cognitive training. For example, the goal-management training (GMT) and goal-oriented attention self-regulation training programs were developed based on the theory of goal neglect.²⁰ The GMT targets the disorganization of behavior which is commonly seen following TBI and aims to improve goal-directed behavior through training in discrete stages of goal completion, including assessing a situation and directing attention toward relevant goals, selecting appropriate goals and partitioning these into subgoals, and monitoring progress toward the goal. Levine and colleagues reported that post-GMT training (brief version, 1 session), adults with TBI showed improved performance on both trained skills and the untrained domains of simulated real-life tasks (eg, proofreading). In a single-case study of a postencephalitic participant who sought to improve her cooking ability, generalization benefits of the expanded version of GMT (8 sessions) were reported on other real life situations.²¹ A modified version of the GMT program (4 sessions) in healthy older adults also yielded generalized benefits to simulated real-life tasks and self-rated executive functions maintained at long-term follow-up.²² Similar to the GMT, the GbSM training protocol involved application of attention regulation strategies to participant-defined goals. Following a 5-week goal-oriented self-regulation program, researchers reported improved attentional functions (primary outcome measures) and generalized effects to executive functions examined on standardized neuropsychological measures of working memory, mental flexibility, and self-reported improved functional task performance.²³

Another cognitive training program that adopted a theoretical framework is the problem-solving training (PST) developed by Rath and colleagues. This training is based on the social problem-solving theory that refined the construct to emphasize problem-solving in everyday social contexts.^24,25 Specifically, the treatment program trains social problem-solving skills in the discrete steps of (1) problem definition and formulation (identifying the conditions and constraints of problematic situations and setting realistic goals), (2) generation of alternatives (brainstorming a range of possible solutions), (3) decision making (examining potential consequences of options and selecting an optimal one, given the conditions and constraints of the problem), and (4) solution implementation and verification (enacting solutions, monitoring their effectiveness, and making modifications as necessary). The 24-session PST uses worksheets, group discussions, and role-play through the 4 stages of the training. Improved performances on problem-solving measures (trained domain) were reported along with improved visual memory and decreased perseverative errors on the Wisconsin Card Sorting Test (a measure of executive function).

In addition to theory-based intervention approaches that are either specific task-based or specific impairment-based (eg, problem-solving), researchers proposed learning-based approaches to remediate cognitive dysfunction.^26,27 For example, Gordon and colleagues' Executive Plus model proposes combining principles of top-down approaches to maximize learning so that treatment benefits generalize across a variety of life situations. Maximizing learning specifically refers to mastering the prerequisite skills required to remediate impairments. One example is to integrate attention training principles with PST to attend to necessary information to learn solve problems.^4,24,25 Another example is to combine PST with self-regulation training to facilitate an individual's ability to control impulses while solving a problem.^27,28 In essence, the Executive Plus model postulates a cumulative benefit of combining different intervention principles to maximize learning.

In addition to limited studies with theoretically motivated cognitive training programs, methodological issues have limited our understanding of the efficacy of many cognitive training programs. Large-scale reviews of cognitive training programs have identified that the majority of studies did not include a control group and/or employed small sample sizes or single-subject research designs.^29–31 Moreover, outcomes of training programs were limited to specific tasks trained or neuropsychological tests. Specifically, reviews of existing cognitive training programs conclude that treatment effectiveness often does not generalize to everyday life.^32,33 One recently developed strategy-based training program, labeled Strategic Memory and Reasoning Training (SMART), described later, was developed based on the theoretical construct of “gist-reasoning” that is relevant to everyday life tasks.^34–36

CONSTRUCT OF GIST-REASONING

One remarkable capacity of the human brain is its adeptness in processing large amounts of information.³⁷ To effectively process this vast amount of information, individuals show a preference for a top-down strategy of abstracting global meaning(s) or “gist” over a bottom-up strategy of encoding all the details, in a process of gist-reasoning.^38,39Gist-reasoning is defined as the ability to abstract gist meanings from information, which is conveyed in news articles, movies, and legal documents, to mention a few.^40,41 The ability to assimilate and interpret generalized meanings when faced with vast amounts of detailed information is not only an efficient cognitive function, but also has been associated with frontal lobe activation and performance on executive function measures as well as everyday life performance.^42,43 In other words, the propensity to engage in gist-reasoning versus remembering details not only minimizes cognitive overload of incoming stimuli, but also assists in constructing a form of meaning that is more robustly stored and retrieved than the composite details.^36,38

From a theoretical perspective, these 2 forms of memory, that is, gist-based versus detail-based, purportedly operate independently and are dissociable.^38,44 Whereas memory for gist involves assimilating and interpreting incoming information at an abstract level of meaning, memory at a detail or verbatim level is represented by the explicit facts. More recently, theorists adopted a constructivism view and proposed that although gist and detail are dissociable, gist representations can assist in memory for verbatim content (ie, details). That is, an individual with higher gist-reasoning skills may be more strategic in encoding details than an individual with lower gist-reasoning.⁴⁴

Empirically, distinctions between memory for gist and memory for details have proven to be clinically informative when elucidating cognitive impairments in healthy older adults and in clinical populations. For example, cross-sectional and longitudinal studies of cognitively normal older adults have found gist-reasoning to be relatively stable despite decline in detail memory.⁴⁵ The assumption is that older adults tend to rely on relatively intact gist-reasoning capabilities to compensate for declining episodic memory for details.⁴⁶ Whereas adults with aphasia present with relatively intact gist processing ability and decreased memory for details, individuals with right hemisphere brain damage demonstrate impaired gist-processing ability and relatively intact memory for details.^47–50 Adults with Alzheimer's disease experience both gist and detail deficits.⁵¹ A recent study with adolescents with moderate to severe TBI identified impaired gist-reasoning despite intact memory for explicitly stated details of texts.⁵² Preliminary evidence suggests lowered gist-reasoning performance in adults with TBI as compared to healthy adults.⁵³

Three cognitive control processes considered fundamental to gist-reasoning include (a) strategic attention (inhibiting less relevant information), (b) integrated reasoning (abstracting concepts by combining pre-existing knowledge with relevant facts), and (c) innovation (flexibly and fluently deriving multiple interpretations by interpreting the information from different perspectives). Previous research has shown that gist-reasoning and intelligence are related.^35,36 Moreover, gist-reasoning draws upon frontally mediated fluid intelligence domains of inhibition, working memory, concept abstraction, fluency, and cognitive flexibility, all of which engage frontally mediated top-down processes.^41,42,54 Nonetheless, the effects of TBI on gist metrics remain significant even after controlling for performance on intelligence quotient measures.⁴⁰ Thus, gist-reasoning may be related to intelligence, but may also represent a cognitive function that measures a unique ability beyond what is captured by standard intelligence quotient measures. With regard to the role of memory, evidence indicated significant contribution of working memory to gist-reasoning over and above what is explained by memory for explicit facts.^42,53

Additional support for the relevance of the construct of gist-reasoning comes from neural evidence of a significant role of frontal networks in processing gist. For example, recent evidence found brain regions to be differentially active during encoding of details versus gist.⁵⁵ Whereas encoding of details recruited a large area in the superior temporal gyrus in the left hemisphere and extended into the superior temporal sulcus and middle temporal gyrus, encoding of gist was associated with frontal lobe activation. Specifically, extensive frontal activation was reported in the region centered over the right precentral sulcus. Similarly, Robertson and colleagues found right inferior frontal activation when participants attended to gist-related information as compared to detail-related information, which activated predominantly the left anterior temporal region.⁵⁶ Right frontal activations have also been reported when eliciting gist-related constructs of inferring a moral of a story or constructing a title for a story.^57,58 In children with chronic TBI, researchers have reported a positive correlation between perfusion in right frontal regions and more proficient gist-reasoning abilities.⁵⁹

On the basis of the relevance of gist-reasoning as an informative metric, especially in clinical populations with frontal impairments, Chapman and colleagues³⁴ adopted the construct to formalize and test a training protocol, labeled SMART.³⁶ The SMART program utilizes a top-down strategy-based approach to train individuals to construct generalized meanings with no direct emphasis on remembering explicit facts. The strategy instruction is hierarchical, with each strategy dynamically building upon previous strategies to transform the explicitly encountered details into abstracted gist meanings through reasoning and inferencing (Table 1). In addition to the manualized content, the SMART program encompasses application of the learned strategies to activities relevant in daily life contexts.^36,41

TABLE 1: Strategic Memory And Reasoning Training (SMART) program

Recent studies show improved gist-reasoning performance following SMART in typically developing middle school children, in senior adults, and in adolescents with attention deficit hyperactivity disorder.^36,41,60,61 For example, Gamino and colleagues³⁶ found that middle school students who had cognitive strategy instruction focused on abstracting meaning significantly improved both their gist-reasoning and fact-learning abilities as well as improved academic performance on state-mandated tests (eg, Texas Assessment of Knowledge and Skills). In contrast, students who learned only rote memory strategies failed to show gains in gist-reasoning despite gains in fact-learning. Similarly, Anand and colleagues found significant improvement in gist-reasoning performance after just 8 hours of gist-reasoning training in older adults. Moreover, significant transfer effects were found on executive functions, including cognitive switching, concept abstraction, and verbal fluency.

Whereas limited research has demonstrated benefits of top-down training in adults with TBI, it is not known if the top-down cognitive control training of SMART would be beneficial in this population, especially in chronic stages of recovery. Therefore, a feasibility study was designed with specific aims as delineated below.

AIMS OF THE STUDY

The present feasibility study examined the ability to improve gist-reasoning in chronic stage TBI by comparing an experimental strategy-based SMART program versus an information-based Brain Health Workshop (BHW) control program (described in the Methods section) in an equivalent amount of training, immediately post- and at 6 months posttraining. The secondary aim was to explore whether the effects of the SMART program versus the control program generalized to untrained areas of memory, executive function, and self-ratings on daily function measures, immediately post- and at 6 months posttraining. It was postulated that a top-down SMART approach to learning would enhance gist-reasoning abilities. Moreover, we predicted that benefits of the SMART program would also generalize to untrained domains of memory, executive function, and daily function ratings. We also predicted that these gains would be sustained at 6-month follow-up testing.

METHODS

Participants

Thirty-five community dwelling individuals with TBI between the ages of 20 and 65 years at testing in chronic stages of recovery, at least 1-year postinjury participated in the study. The cause of TBI in 26 of the 35 participants was due to a motor vehicle accident (25 were involved in car accidents and 1 motorcycle accident), 4 were sports-related, 4 were work-related (eg, object hitting the head), and one was due to assault. Twenty-eight of the 35 individuals completed the training. As indicated in the Consolidated Standards of Reporting Trials (CONSORT) flow diagram-Figure 1, the reasons for dropout are considered random and were evenly divided between the 2 groups. Thus, the study results are based on the performance of those that completed the training program. Nine of the 28 participants were gainfully employed outside their homes, working between 15 and 30 hours a week. Ten of the rest of the 19 participants were engaged in other productive activities including volunteering at churches, food banks, and assisted living facilities on a regular basis. The remaining 9 participants were living alone in the community and were not gainfully employed or involved in volunteer activities. The majority (23 of 28) of the participants sustained their injuries over 15 years ago, and hence there was limited access to medical records that provided reports of the severity of the TBI. One participant had a documented GCS score of 7 (ie, severe), 5 participants had the severity of TBI documented as “severe” with no GCS scores, 2 participants had the severity documented as “mild” and the initial severity was unavailable for the rest of the participants. Therefore, the period of retrospective posttraumatic amnesia (PTA), a commonly used predictor of functional outcomes following a TBI, was used as an estimate of early injury severity. We acknowledge the limitation in using retrospective self-reported PTA, especially since some of the participants were injured at an early age and it was not possible to verify PTA with medical records. Although the accuracy of a retrospective self-reported PTA is equivocal, similar methods for examining PTA as a rough index of early injury severity have been utilized in previous chronic-stage studies when early medical information was not available.^62–64 The period of PTA was obtained during the initial testing by asking the participants to estimate how long it was between their injury and the time when he/she felt like they started remembering things continuously. Participants were told not to tell the time when they thought their memory was back, but rather when they started remembering day-to-day events in a consistent way.

Figure 1: CONSORT flow diagram. BHW indicates Brain Health Workshop; SMART, Strategic Memory and Reasoning Training.

In addition, since early injury severity has been questioned as a reliable predictor of severity of deficits years after sustaining a TBI, we adapted a methodology to establish current level of disability. Whereas a current level of disability cannot be directly or solely attributed to the earlier injury, it is relevant to documenting status at time of training. Specifically, we used 2 valid and reliable self-reported functional measures, the Glasgow Outcome Scale-Extended (GOS-E) and the Functional Status Examination (FSE), to characterize current severity of functional impairments. The GOS-E is a gross outcome measure that tracks degree of functional recovery with broad functional categories of cognition, mood, and behavior on an 8-point scale.⁶⁵ The possible score ranges from 2 to 8, with 8 indicating good recovery and 2 indicating a vegetative state. A score of 5 or 6 characterizes moderate functional recovery. The FSE offers some advantages over GOS-E as it provides a more detailed description of the functional deficits and the amount of assistance needed to accomplish the function. The self-reported FSE tests 10 functional categories including personal care, mobility/ambulation, travel, work and/or school, leisure and recreation, home management, social integration, cognitive and behavioral competency, standard of living, and financial independence.⁶⁶ A score of 10 signifies no impairment, 11 to 20 as mild impairment, 21 to 30 as moderate impairment, and a score above 30 is indicative of severe impairment. Participants with moderate functional impairments on the measures GOS-E (range = 5–6) and FSE (range = 21–30) were included in the study. These 2 functional measures were also included as pre-/posttraining comparison measures of daily function, as they are considered valid predictors of long-term functional outcome in adults with TBI.^67–69

Only native English speakers with at least a high school education who scored a minimum of ninth grade equivalency on vocabulary and comprehension on the Nelson-Denny reading test and had a minimum premorbid estimate of verbal intellectual functioning of 90 as measured by the North American Adult Reading Test were included in the study.^70,71 Participants were not provided with any transport to attend the sessions; hence, all participants were either independent drivers, used public transport, or had other means to attend the sessions.

Exclusion criteria included pre-TBI histories of stroke, learning disability, communication disorder, substance abuse, or major psychiatric disorder. The current study did not include participants with current depression status, as determined by the Beck Depression Inventory (BDI-II), such that participants with a BDI score above 9 were not included in the study.⁷² In addition, participants who were receiving cognitive treatment(s) at the time of the assessment were excluded from the study. Three participants were from the University of Texas Southwestern Medical Center Traumatic Brain Injury Model System, 2 were referred by a local neuropsychology clinic, and the remaining participants were recruited from local brain injury support groups. Informed consent obtained from all participants was in approval and accordance with the guidelines of the institutional review boards of the University of Texas at Dallas and University of Texas Southwestern.

Procedure

The current study was a single-blinded randomized control trial, where participants were randomly assigned to one of the 2 protocols: (a) top-down SMART(experimental group) or (b) information-based BHW (control group).⁷³ Participants were informed that the goal of the study was to compare the benefits of 2 training programs that could be beneficial to adults with TBI. In addition, they were told that they would be randomly assigned to one of the 2 training programs after they agreed to participate. Measurement outcomes of both protocols included the same battery of experimental and standardized cognitive tests as well as functional measures. The examiners involved in testing, scoring, and data analyses were blinded to the group and time (ie, pre-, immediately post-, and at 6 month posttraining status). Training was initiated within 3 weeks of testing the participants. Posttesting took place within 3 weeks of completion of the training program. Six-month follow-up testing took place 6 to 7 months after the training sessions were completed.

Measures

Measures of gist-reasoning, memory, executive function, and day-to-day function were included in the study. The primary outcome measure of gist-reasoning was examined with the Test of Strategic Learning (TOSL) (S. Chapman J. Hart, H. Levin, L. Cook, J. Gamino, unpublished data, 2009). The TOSL measure has demonstrated sensitivity in examining gist-reasoning skills in clinical populations where higher-order cognitive functioning is compromised, including adults with stroke, adults with mild cognitive impairment or Alzheimer's disease, and children and adults with TBI.^47,51–53 The TOSL measure consists of 3 texts, designed to examine how one understands and constructs generalized/gist meanings from connected language. The 3 texts vary in length (from 291 to 575 words) and complexity. For each of the texts, the participant is asked to provide a summary or a shortened version of the original text that focuses on the general overview of what the text is about and does not include all of the details. Following these instructions, the examiner reads each text aloud, and participants are given a copy of the text to follow along. After the examiner completes reading the text, the participant's copy is taken away so that the participant does not have the option to refer to the original text while providing his or her summary. Participants' summary productions are audio-recorded and later transcribed for scoring.

The TOSL measure has a manualized objective scoring system wherein the summary productions conveying abstracted gist meanings received a higher score than those that focused on the stated details of the text. Each gist meaning conveyed by the individual receives 1 point. A total score (composite score) of 37 points is possible for the 3 summaries of the 3 texts. Two trained examiners, blinded to the participants' group status and time of testing (pre- vs immediately post- vs 6-month posttraining), independently scored the 3 summaries for inclusion of gist-based meanings. Interrater reliability of scores assessed on intraclass correlation coefficients in both groups combined for gist-reasoning performance was over 90% (Cronbach α range = 0.92–0.98, confidence interval = 0.78–0.98). The same measures were used at immediate post- and 6-months posttesting.

Two measures were administered to examine memory. Memory for details of the 3 TOSL texts (that were used to examine gist-reasoning) was examined by means of the “memory for details” recall measure (S. Chapman, J. Hart, H. Levin, L. Cook, J. Gamino, unpublished data, 2009). The “memory for details” measure includes probes that test memory for details from the texts considered important for understanding the key ideas of the texts. The probes followed summarization of each text. Each text has 8 probe questions, for a total of 24 probes for all the 3 texts combined. Responses to each memory for detail probe receives a score of 2, 1, or 0 points depending upon accuracy and completeness of the response, yielding a composite score of 48 points for all 3 texts. Memory was also examined by means of the Digit Span Forward subtest from the Wechsler Adult Intelligence Scale III.^74,75

Executive function measures included 2 measures of working memory, and single measures of inhibition, nonverbal reasoning, cognitive flexibility, and verbal fluency. The 2 measures of working memory included the modified version of the Daneman and Carpenter working memory Listening span task and the Letter Number-Sequence working memory span task from the Wechsler Adult Intelligence Scale III. In this listening span working memory test, participants listen to a set of unrelated sentences and recall the final word of each sentence in the set after listening to the whole sentence-set.⁷⁶ Participants are presented with increasingly longer sets of sentences, and working memory span was determined by the maximum number of sentences the participants could listen to while maintaining the recall of at least two-thirds of the sentence-final words. In the Letter Number-Sequence working memory task, the examinee is read a combination of numbers and letters and is asked to recall all the numbers first in ascending order and then the letters in alphabetical order.⁷⁴ Inhibition was examined with the Delis-Kaplan Executive Function System (D-KEFS) Color-Word Interference task, specifically the number of errors made on the inhibition (task 3) & inhibition/switching (task 4) conditions.⁷⁷ The D-KEFS Color-Word interference test measures the ability to inhibit an overlearned verbal response (ie, reading the printed words) to generate the conflicting response of naming the dissonant ink colors in which the words are printed. Nonverbal reasoning was examined on the Matrix-Reasoning task from the Wechsler Adult Intelligence Scale III.⁷⁴ This test is composed of 4 types of nonverbal reasoning tasks: pattern completion, classification, analogy, and serial reasoning. The participant chooses 1 of the 5 response options to complete a matrix. Cognitive flexibility was measured on the Trail Making Test-Part B for adults.⁷⁸ This test requires the participant to connect 25 circles of numbers and letters in an alternating pattern (1-A-2-B-3-C, etc) in as little time as possible. Verbal fluency was measured on the Controlled Oral Word Association Test that requires the participant to generate as many words as possible that begin with a specific letter in a specific amount of time.⁷⁹ These executive function measures were used at pre-, immediate post-, and 6-months posttraining.

Measures of GOS-E, FSE, and Community Integration Questionnaire (CIQ) were administered to measure daily function at pre-, immediate post-, and 6-month posttraining. Descriptions of GOS-E and FSE are provided in the previous section that characterized participant functional status at initial testing. The self-reported CIQ measure provides a general overview of an individual's functioning based on responses to 15 questions related to participation in activities at home, social, and education or vocation settings.⁸⁰ The score ranges from 0 to 29, with increasing scores indicating improved level of integration in the community. Several studies have established reliability and validity of the CIQ measures.^81–83

Training protocols

The training protocols included the experimental SMART program and the control BHW protocol. Both groups were controlled for the number of contact hours with the clinicians, group interactions within group members, and the remuneration for participation in the study. Both SMART and BHW programs offered a total of 18 hours of training during 12 group sessions (1.5 hours each session) conducted over 8 weeks. The first 15 hours of training over 10 sessions were conducted in the first 5 weeks (ie, 2 sessions per week). The final 3 hours of training, over 2 booster sessions, took place at spaced intervals over the next 3 weeks (ie, session 11 during week-6 and session 12 in the eighth-week). Two trained clinicians (a speech pathologist and an occupational therapist) who had experience in TBI rehabilitation led each group. Each group consisted of 4 to 5 participants.

Group sessions were chosen in favor of individual sessions, as we anticipated that group discussions in SMART and BHW could potentially reinforce participants practice in incorporating the strategies or relating brain health information to issues in their lives. Comparable number of group members in both SMART and BHW groups controlled for the amount of social bonding, which is considered therapeutic. To ensure adherence to the program, the facilitators stressed the importance of attending every session and diligence in completion of homework assignments during recruitment. Specifically, participants were also told that their contribution to this specific study and the science of cognitive rehabilitation in general greatly relied on these 2 factors (ie, attendance and homework assignments).

Sessions were both didactic and interactive in nature. Training materials, including power point handouts and preselected reading material, were given to the participants in both SMART and BHW in a binder to take home and to be brought back for the following session. Participants were allowed to keep the binders after completion of the training program. All of the participants received $10 per visit (including testing and treatment) during their participation in the study. It is interesting to note that individuals in both groups seemed to enjoy participating in the training. The SMART group participants discussed the benefits of the strategies in their everyday life and the BHW expressed benefits from learning about the brain and the effects of a TBI on the brain.

Strategic memory and reasoning training (experimental group)

The SMART strategies were introduced to the group in a power point format with a description of the strategy (Table 1). The description included explaining the role of frontal lobes in mediating these strategies and the relevance of these strategies to derive gist. Preselected materials of varying lengths, including newspaper articles and stories, were used to illustrate the strategies. More importantly, the application of these strategies in their daily life was emphasized. For example, to illustrate the strategy of “inhibition,” discussions of real-life examples included ignoring less relevant steps during event planning or ignoring distracting information when following a lecture and so on.

The strategies were introduced in a sequential manner (Table 1). Mastery of individual strategies was not necessary to move on to the next level, as strategies were continually reinforced at each stage of the program. Following each session, participants were given preselected homework assignments that predominantly utilized written material. The written material included articles from local newspapers, magazines, and lectures. Along with these preselected assignments, homework also included that participants identify or select specific life activities in which SMART strategies could be incorporated.

One specific real life example of writing a resume using the SMART approach was illustrated by one participant. The first step of strategic attention (ie, strategy of filter) involved deleting information that would not be relevant to include in the individual's resume. In addition, the individual had to selectively identify the most important information related to his/her own skills and strengths. The second step of integration (ie, strategies of focus and link) involved categorizing the individual's skills and strengths into academic accomplishments, leadership qualities, professional or personal characteristics, and work experience including volunteer work that related to the job requirements. These categories were then supported with relevant details to help the employer/interviewer capture the breadth and depth of that particular category. The third step of innovation (ie, strategies of zoom and generalize) involved summarizing the qualifications and abilities at a higher/broader level, into 2 or 3 succinct statements to provide the resume “objective” statement. Innovation also involved flexibility in preparing the resume in multiple formats and revising the objectives statements to adapt to the different employers needs and job requirements. Application of SMART strategies to other daily life tasks such as planning an event, learning from a lecture or discussion of a movie was also discussed.

Brain health workshop (control group)

The control group was conducted using the BHW protocol that has been previously used as a control-training program for experimental cognitive training at the Rotman Institute, Toronto, and University of California, Berkeley.⁷³ Similar to SMART, the BHW is a manualized program. However, rather than being strategy-based, the BHW is information-based, covering predetermined topics on brain anatomy, brain functioning, general brain health, effects of lifestyles on brain health such as different diets and exercises, and cognitive changes following a TBI (Table 2). During each session, clinicians presented the assigned topics in a power point format. Participants were given the power point handouts in their binders to follow along. Group discussions provided clarifications and further explanations of the topics. At the end of each session, participants were also given take-home reading assignments on related topics that were discussed in the following sessions.

TABLE 2: Brain Health Workshop

DATA ANALYSES

Three analyses were conducted in the current study. First, comparability between SMART and BHW groups at pretraining was examined using 2-sample Student t tests. Specifically, comparability was examined on participant characterization measures, gist-reasoning performance, memory, executive function, and functional rating scales. Second, a repeated measures analysis of variance (ANOVA) examined the time by group interaction for gist-reasoning, memory, executive function, and functional rating scales. In the presence of an interaction, pairwise comparisons were examined with Bonferroni correction, where P-values .02 or less for any pairwise comparison were considered significant. Third, since this is a feasibility study, when measures failed to show a time by group interaction, we compared the main effects (means) across the 2 groups using the main effects F-tests to determine whether observed group differences might indicate promising hypotheses for future (or expanded) research. Analyses were performed using SPSS version 17.0 (SPSS Inc, Chicago, Illinois).⁸⁴

RESULTS

All participants received a minimum of 15 hours of training (ie, minimum of 10 sessions). At baseline, the t test analysis revealed comparability between the SMART and BHW groups on participant characterization measures of age at testing, age at injury, years since injury, educational level, estimate of premorbid verbal intellectual function (measured on the North American Adult Reading Test), and level of functional impairment as measured on GOS-E and FSE (Table 3). Furthermore, both groups were comparable in processing speed, as measured on the D-KEFS Color-Word Interference task conditions 1 (color naming) and 2 (word reading). Hence, these variables were not included for further inferential analysis.

TABLE 3: Demographic and clinical features of SMART (n = 14) and BHW (n = 14)

Interaction effects

Results indicated significant time by group interaction effects on measures of gist-reasoning, one measure of working memory (listening span task), and the CIQ. Gist-reasoning performance at pretraining was comparable between SMART and BHW groups, (t₂₆ = 0.33, P = .74 ns). Qualitatively, summaries on the TOSL (measure of gist-reasoning) at pretraining consisted predominantly of statements from the original text that were directly tied to the explicit content. At posttraining, the majority (12 of 14 at immediate posttesting and 11 of 13 at 6-months posttesting) of the self-generated summaries in the SMART group conveyed abstracted meanings that were not stated in the original text. Repeated measures ANOVA indicated significant group by time interaction for gist-reasoning, F_2,48 = 3.66 P = .03 (Table 4). Pairwise comparisons indicated that the SMART group made significant gains immediately posttraining (t₂₆ = 2.91, P = .007) and at 6 months posttraining (t₂₄ = 3.17, P = .004) when compared to pretraining level. However, the BHW did not show significant changes in gist-reasoning either at posttesting (t₂₆ = 0.775, P = .44), or at 6-month follow-up (t₂₄ = 0.65, P = .52) when compared to the pretraining level (Table 4).

TABLE 4: Significant time × group interactions with pair wise comparisons (Bonferroni adjusted)

At the pretraining level, both SMART and BHW groups were comparable on the working memory measure of listening span (t₂₆ = 1.22, P = .23 ns). At posttraining, a repeated measures ANOVA indicated significant group by time interaction for the working memory measure of listening span, F_2,44 = 14.57, P < . 001. Listening span capacity in the SMART group, both immediately posttraining and at 6 months posttraining was significantly higher than pretraining score (Table 4). The BHW did not show any significant changes in the listening span task at either immediate posttraining or at 6 months posttraining.

The t test analyses found that both SMART and BHW groups were comparable at pretraining on the functional rating scales of CIQ (t₂₆ = 0.99, P = .33 ns). At posttraining, repeated measures ANOVA indicated significant group by time interaction on the CIQ (composite) rating scale, F_2,46 = 4.90, P = .02. The CIQ (composite) score in the SMART group at 6 months posttraining was significantly higher than the pretraining score (Table 4). Furthermore, a significant time by group interaction was indicated for the CIQ subscale of social-integration, F_2,46 = 4.23, P = .02. The social-integration rating scale at 6 months posttraining was significantly higher than pretraining score (Table 3). No significant changes were noticed for the home-integration or productivity subscales of the CIQ. The BHW group did not show any significant changes over time on any of the functional rating scales.

Exploratory analyses

Exploratory between-subjects analyses examined the main effects of the groups. In terms of memory, at the pretraining level, both SMART and BHW groups were comparable on the measures of memory for text details (t₂₆ = 0.45, P = .65 ns) and digits forward (t₂₆ = 0.62, P = .53 ns). Significant main effects in the SMART group were found on memory for text details, F_1,22 = 6.10, P = .02 (M [SD]: pretraining = 33.5 [5.6], immediately posttraining = 41.2 [5.6], 6-month posttraining = 40.08 [3.2]). No significant changes were noticed in either group at immediate post- and 6-month posttraining period on the memory measure of digits forward (WAIS-III).

With regard to executive functions, both SMART and BHW groups were comparable at pretraining on the measures of inhibition measured on color-word interference D-KEFS tasks 3 and 4 (t₂₅ = 0.39, P = .69 ns), working memory on the Letter-Number Sequencing task (t₂₆ = 0.05, P = .95 ns), nonverbal reasoning on the Matrix Reasoning measure (t₂₆ = 1.23, P = .22 ns), verbal fluency measured on Controlled Oral Word Association Test (t₂₆ = 0.98, P = .33 ns), and cognitive flexibility measured on Trails B (t₂₆ = 1.16, P = .25 ns).

Significant main effects were found in the SMART group on 3 executive function measures. The 3 measures included (a) inhibition F_1,21 = 7.63, P = .01, where lower mean indicates fewer errors on the color-word interference task (M [SD]: pretraining = 8 [5.01], immediately posttraining = 3.7 [3.09], 6-month posttraining 2.08 [1.7]), (b) nonverbal reasoning (on Matrix Reasoning measure) F_1,22 = 14.06, P = .001 (pretraining = 12.21 [2.9], immediately posttraining 14 [2.82], 6-month posttraining = 14.62 [2.69]), and (c) cognitive flexibility F_1,22 = 7.17, P = .01, where lower mean indicates less time (in seconds) taken to successfully complete the Trails-B task (pretraining = 70.57 [32.93], immediately posttraining = 68.36 [56.01], 6-month posttraining = 59.08 [24.18]). No significant main effects were evident on the functional scales of GOS-E and FSE.

DISCUSSION

The current study presented a theoretical construct and empirical evidence for a strategy-based approach to promote gist-reasoning in adults with TBI. The construct of gist-reasoning elucidates an ecologically valid function of abstracting meaning from information in light of (a) characterizing abstraction skills, and (b) as a methodological framework to improve abstraction skills. The 3-top-down cognitive control processes of strategic attention, integration, and innovation in gist-reasoning represents the core components around which the strategy-based SMART program was developed.

Empirically, 3 major findings emerged from the current study that examined the effects of strategy-based SMART versus information-based BHW training protocols in adults with TBI at chronic stages of recovery. First, our findings revealed that 15 to 18 hours of SMART enhanced gist-reasoning in adults with TBI. Second, the effects of SMART generalized to untrained domains such as on the working memory measure of listening span and ratings of increased participation in daily activities. Third, there appeared to be sustained benefit (6 months posttraining) of SMART as compared to the control group (BHW).

Similar to the results in the current study, improvement in gist-reasoning following the SMART program were also reported in typically developing middle school students and in cognitively normal senior adults.^36,41 The consistent findings across studies indicate both specific and generalized effects of a top-down oriented strategic approach to learning, especially for complex/lengthy information, in both cognitively impaired and healthy adults.

Another promising finding of this study was the generalized effects of SMART training to improved performance on measures of executive function, especially that of the working memory listening span task that elicited the person to recall the last word of an increasing number of sentences (2–7 sentences). One possible explanation for enhanced gist-reasoning affecting working memory listening span is that the participants adopted gist-based (higher-level) strategies to connect meaning between the words to be remembered (eg, forming associations, or forming a new sentence or visual imagery) that helped with effective recall. One might also argue that improved working memory, either with the help of strategies of “filter” or “chunk” may have led to improved gist-reasoning performance. The present findings do not inform whether both skills, that is, gist-reasoning and working memory span, improve independently or are interdependent. Future studies should help elucidate this distinction of the benefits of the SMART program. In either case, we speculate a supportive dynamic relation between working memory and gist-reasoning. A supportive relation between working memory and gist-reasoning (examined on summary production tasks) has also been reported in previous studies.⁴⁰ Moreover, previous evidence has suggested that the dynamic nature of working memory may be more relevant to gist reasoning than straightforward memory. Specifically, a significant contribution of working memory to gist-reasoning over and above what is explained by memory for explicit facts has been reported in adolescent TBI studies.⁴²

Generalization effects following the SMART program have previously been reported both in typically developing middle school adolescents and in cognitively normal senior adults.^36,41 Following SMART, normally developing middle schoolers trained in gist-reasoning showed significant improvements in their academic performance on state-mandated critical thinking tests post-SMART. Similar benefits of top-down approaches to teaching abstract principles versus use of concrete examples in content areas such as math have also been reported.⁸⁵ Similarly, following SMART, senior adults improved on untrained measures of cognitive switching, concept abstraction, and verbal fluency.

Generalization benefits have also been observed on cognitive measures following theory-based cognitive training programs such as the GMT, goal-oriented attention self-regulation, and PST that involved application of strategies to a preset goal or a predetermined problem scenario, respectively.^21,23,25 These training programs were at the forefront in adopting theoretical frameworks to cognitive training programs. The SMART program broadens the scope of such theoretical strategy-based approaches to remediate executive control deficits. The top-down SMART approach uniquely contributes to cognitive remediation paradigms by training adults with TBI to be strategic thinkers. Specifically, the program distinguishes from existing strategy-based approaches in 2 ways. (a) Whereas existing approaches apply strategies to solve or accomplish a preset goal or a predetermined problem, participants in SMART program were trained to apply strategies across domains during training sessions. For example, during the SMART program, participants applied the strategy of “filter” during conversations, where participants ignored less relevant details while conveying an idea, or selectively deleted less relevant steps to plan an event, and so on. (b) In addition, the intensive SMART program focuses on homework assignments to ensure carryover of strategies beyond training sessions. The discussions of homework assignments in the groups provided therapeutic milieu to motivate, problem-solve, and innovate ways to efficiently approach a real life task. In essence, SMART strategies were new learning tools to repair and strengthen impaired higher-order cognitive skills following a TBI. Furthermore, the present study is one of the first studies to examine long-term benefits of a theory-backed training program.

In addition to improved performance on executive function measures, the SMART group participants made significant gains on measures of daily function. Improved self-rating on the CIQ measure (composite score), especially at the 6-month follow-up testing (and not at immediate posttraining) may indicate that a long-term sustained effort of implementing the strategies may lead to improved participation in day-to-day tasks. Significantly increased ratings on the social integration subscale of CIQ were not due solely to the social interactions with other group participants, since the BHW control group had similar social interactions and yet failed to show increased scores at either immediate posttesting or at 6-months posttesting. Improved social integration in the SMART group could be secondary to improved self-regulatory executive functions, resulting in effective communication abilities such as not dwelling on less relevant details, or improved life skills in terms of planning and orchestrating tasks involved in execution of a family event, or restarting or developing new leisure interests (eg, golf, reading novels, and church-related activities).

The exploratory analyses results elucidated the generalizable potential of the SMART program on memory function and the executive functions of inhibition (on Color-Word interference tasks), nonverbal reasoning (on the Matrix Reasoning task), and cognitive flexibility (measured on trails B). The main effect of SMART on memory for details provided preliminary support for the notion that improved gist-reasoning may enhance an individual's ability to recall facts from texts, which is congruent with another recent SMART study in adolescents. Gamino and colleagues³⁶ showed that typically developing middle school students who were trained to abstract meaning also showed significant gains in ability to recall specific information from texts. In contrast, middle schoolers trained in rote memorization skills did not improve on gist performance. Although the current study did not compare the effects of bottom-up straightforward rote memorization training versus top-down processing, we anticipate that top-down modulation of information has a positive impact on bottom-up processes such as recall of details. Specifically, we propose that clustering ideas into more abstract concepts serves to encode details at a deeper level for learning and memory retention. Clearly, the relation between top-down and bottom-up processes is not typically bidirectional. That is, improvement in top-down processes has the potential to strengthen bottom-up processes, yet strengthened bottom-up processes may not necessarily improve top-down functions.¹³ Theoretically, this unidirectional benefit of top-down processes reflects the view of Reyna and Brainerd's “fuzzy-trace” theory, that gist memory can in fact improve (shape) memory for verbatim details.³⁸

In addition, main effects on measures of executive functions of inhibition, nonverbal reasoning (on Matrix Reasoning task), and cognitive flexibility (on Trails B) suggest preliminary trends for the potential of SMART to transfer to other untrained domains. With regard to improving inhibition, the strategy of “filtering” ideas may have helped override the automatic responses on the Color-Word interference task (used as measure of inhibition), or the “zoom out and zoom-in” strategy may have assisted with the Matrix Reasoning task that requires the participant to identify a missing detail of a larger pattern. Whereas we cautiously interpret the main effects of SMART on these 3 executive function measures, the evidence of similar transfer effects in studies of older adults reinforces the possibility that the findings may be upheld in future studies of TBI in adults. Nonetheless, we acknowledge that the failure to find significant group by time interaction on these measures may be due to small sample size, individual variation, or need for longer training period in some individuals to achieve more widespread effects. Future studies with larger sample sizes would help address these concerns.

Limitations and future implications

The current findings, while promising, must be interpreted cautiously, as they require further validation to address at least 3 major limitations. First, we recognize that the participants in the present study were recruited primarily from the community at periods years after brain injury, and we were not successful in obtaining reliable documentation of acute severity of TBI. Documenting initial injury severity is critical to accurately establish the relation between initial injury severity, later recovery level, and response to cognitive treatment protocols.^29,30 Although the contribution of acute severity of injury to long-term functional outcomes is still not definitive, reliable documentation of associated medical conditions and preinjury function would assist with improved characterization of the sample.^67,68 Despite the shortcoming of limited data on acute severity level, we are encouraged by the improvements in the SMART group, in both cognitive function and self-rated daily function at posttraining periods (both immediately post- and at 6-months posttraining), as compared to the pretraining level and as compared to a control group. Although we do not anticipate that obtaining more definitive information regarding the initial injury documentation would have significantly altered the findings; we propose that future studies should more rigorously specify initial injury severity to validate the current findings.

Second, the current study included wide age ranges, including age at injury (3–40 years), and age at training (20–62 years) that may have affected the outcomes of the study. For example, the majority of the participants (19 of 28, comparable number in both groups) sustained their injury in their preteen, teen, or early adulthood years (before the age of 25 years). Research has demonstrated that frontal network myelination continues into the early third decade of life (ie, into the early twenties), and a TBI disrupts the maturation of frontal functions that affect functional outcomes.^85,86 In support of this evidence, Chapman and colleagues⁵⁴ postulate that disruption to frontal networks (as in a TBI) prior to full maturation may be associated with a neurodevelopmental stall, that is, a failure to develop higher-order cognitive skills, including gist-reasoning. Future studies could examine the effect of age at injury while examining higher-level cognitive function, either as an outcome measure or treatment factor. In addition to including a homogenous group based on age at injury, future studies should also consider age at training as a factor. It is possible that one age group may have benefited more than the other. However, this speculation needs further verification.

Third, the current study examined functional gains on self-rated questionnaires that may represent one's perception of gains made posttraining. Assessment of real-life task performance could provide a more accurate characterization of benefits to daily function posttraining. In addition to addressing these 3 limitations, future endeavors could examine the benefits of SMART (a) beyond 6 months (eg, in yearly intervals), (b) in an acute rehabilitation setting, and (c) in conjunction with imaging measures examining brain changes and potential repair.

CONCLUSION

The current pilot study proposes a theory-based top-down learning approach as an alternative to targeting specific impaired cognitive processes (ie, memory, attention). Specifically, we propose that strategy-based top-down modulation engaging strategic attention, integration, and innovation, has the potential to improve executive control deficits that not only are often times significantly impaired years post-TBI, but also are considered the most challenging to remediate. Moreover, the current randomized small sample control trial provides preliminary evidence that higher-level gist-reasoning skills training delivered at chronic stages postinjury (years or even decades postinjury) is beneficial to untrained cognitive domains and indices of real-life function both immediately and in the short-term, at least up to 6 months posttraining.