Improvements in Performance Based Measures of Functional Capacity and Cognition After Computerized Functional Skills Training in older People with Mild Cognitive Impairment and Healthy Comparators

Abstract

With no pharmacological treatments for Mild Cognitive impairment (MCI), computerized training strategies have been attempted. A computerized skills training intervention, FUNSAT, previously produced training-related gains in cognition in MCI and in comparators with normal cognition (NC). A new remotely delivered version of FUNSAT was administered to a new sample of participants with NC and MCI. Outcomes measures included cognition (BAC) and functional capacity (VRFCAT) to examine training transfer. Participants with MCI (n=92) and NC (n=72) trained for up to 12 weeks on FUNSAT. Half the MCI participants started with 3 weeks of computerized cognitive training (CCT). Baseline, post-training, and 30-day follow-up scores on cognition and functional capacity were compared. Participants improved on both cognition (d=.80) and functional capacity (d=.64), with no differences in training gains across MCI and NC, although treatment with CCT in MCI was associated with similar gains with fewer FUNSAT training sessions. This is the first treatment study in MCI to demonstrate transfer to untrained measures of functional capacity. NC improved in cognition and functional capacity with skills training alone. These findings have implications for other conditions, such as schizophrenia, where functional capacity is a treatment target.

1. INTRODUCTION

On the spectrum between healthy cognitive function and dementia, mild cognitive impairment (MCI), especially its amnestic version (a-MCI), is a critical transitional condition. It is commonly reported that people with MCI have an increased risk for the development of Alzheimer’s disease (AD; Duara et al., 2011). Importantly, although initial conceptions of MCI suggested that functional impairment was absent (Petersen et al., 2001), research has shown that people with mild MCI (Goldberg et al., 2010) and related conditions, such as subjective cognitive complaints (Atkins et al., 2018), have impaired performance on structured performance-based assessments of everyday tasks. These impairments have been shown to be greater than those seen in participants with normal cognition, but less severe than seen in AD (Gomar et al., 2011).

Pharmacological therapies have historically been developed for the treatment of cognitive impairment in the AD spectrum, including MCI. These efforts have had limited success in MCI (Feldman et al., 2007) and acetylcholinesterase inhibitors have generally time-limited effects in AD (Rountree et al., 2013). Non-pharmacological therapies, such as Computerized Cognitive Training (CCT), have gained popularity in recent years due to their potential to impact cognitive performance in older people with MCI and normal cognition (NC). In these older populations with both normal cognition (NC; Lampit et al.,2014) and MCI (Zhang et al., 2019), meta-analyses of studies delivering CCT have concluded that these strategies show promise in enhancing cognitive performance. The largest limitation of these studies is that since they were commonly delivered without additional skills-training interventions, it has been difficult to find evidence of a significant transfer of training benefits to other domains such as the development of new real-world functional abilities (Harvey et al.,2018).

While functional capacity, which includes the capability to carry out everyday tasks required for independent living, has been studied recently in MCI, there are two main gaps in the literature. The first is that rehabilitation-focused treatment of changes in everyday functioning skills has been less well-developed in the AD spectrum, including MCI, compared to other conditions such as serious mental illness (SMI). Training programs targeting social competence (Patterson et al., 2003), everyday living skills (Bowie et al., 2012), and vocational outcomes (McGurk et al., 2007) are common in SMI. Secondly, there have been essentially no studies of performance-based functional capacity measures in MCI treatment studies, with most studies either measuring activities of daily living with self-reports or rating scales (Harvey et al., 2017).

A prior randomized clinical trial used computerized functional skills training (FUNSAT) alone or combined with CCT with the Brain HQ system (Harvey et al., 2020) in MCI and NC participants. The results suggested that CCT and FUNSAT training targeting 6 technology-based activities of daily living had benefits on the ability to perform the simulations of everyday tasks in trained by the FUNSAT, including improvements in both time to completion and errors (Czaja et al., 2020) across both populations. Further, synergistic benefits of combined training on cognitive performance with a well-normed assessment were seen in both NC and MCI with the combined training program, but participants with MCI showed a greater dropout rate than NC, particularly with combined training (Harvey et al., 2021). The FUNSAT program had previously been found to be sensitive to performance deficits in MCI (Czaja et al., 2017) and the factor structure and convergence with cognitive performance was quite robust and interpretable (Harvey et al., 2021).

This study differs from our earlier work in several significant ways. Specifically, an updated version of the FUNSAT^™ program was developed and tested in a new randomized clinical trial (RCT), in older adults with NC and MCI. In the new RCT, all training was delivered with fully remote cloud-based administration, with FUNSAT and Brain HQ accessed from home using Google Chrome on a study-provided Chrome Book device. Participants were assessed and trained in their commonly spoken language, either Spanish or English, with testing performed by bilingual testers and training and assessments performed in the same language. MCI status was assessed with a formal performance-based assessment with normed assessment tools. While the performance-based assessment of cognition was like the previous study, we added a well-validated performance-based functional capacity measure to examine far transfer to untrained functional skills.

This trial (NCT046779441) had three different pre-planned outcomes, presented separately. Improvements in performance on the training simulations, in errors and time to completion (Czaja et al., Under review) was the designated primary paper and real-world transfer of the technology-related skills assessed with Ecological momentary assessment (EMA; Dowell-Esquivel, et al., 2023) was the second. This paper presents the results of the third pre-planned analysis. The first two papers demonstrated substantial and wide-ranging training gains as well as real-world transfer of the trained skills everyday performance, examined with 6 months of Ecological Momentary Assessment (EMA) surveys.

In the study, NC received FUNSAT training only, to develop normative standards for improvement. MCI participants were randomized to either receive FUNSAT training only or to receive a three-week burst of CCT prior to the initiation of FUNSAT. Participants trained twice a week for 12 weeks or until they had mastered all 6 FUNSAT functional skills. On three different occasions throughout the study, we performed a cognitive and functional capacity assessment with two technology-based assessment tools: the i-Pad Brief Assessment of Cognition (BAC-app; Atkins et al., 2017) and Virtual Reality Functional Capacity Assessment Tool (VRFCAT; Keefe et al., 2016). Using the VRFCAT, we were able to evaluate improvements in a functional capacity assessment that examined performance of everyday functional skills that were not trained by FUNSAT, thus measuring far transfer to performance on simulations of untrained skills. Examining the cognitive benefits of skills training alone in the NC sample, we were also able to follow-up on previous studies reporting cognitive gains following novel skills training (Park et al., 2014; Leanos et al., 2020) in healthy older people.

Our hypotheses were that cognitive and functional capacity performance would improve with training and that combined training would lead to greater gains in MCI participants than skills training alone. We also hypothesized that training dose, which varied as a function of how rapidly participants mastered the tasks and “graduated” from training, would predict gains in cognition and functional capacity across both populations and across training conditions in the MCI participants. Finally, we performed more exploratory analyses to examine the correlations between subject characteristics, including scores on a performance-based assessment of cognitive performance (The Montreal Cognitive Assessment; MOCA; Nasreddine et al.,2005), academic skills indexed by reading performance, years of education, and training language, and the baseline scores and training gains on the BAC-App and the VRFCAT.

2.0. METHODS

2.1. Overall Study Design

This was a randomized trial that was carried out in community centers in New York City (n=10) and South Florida (n=4). Participants independently completed up to 12 weeks of computerized training at home after screening, informed consent, and an in-person baseline evaluation using one of three versions of a fixed-difficulty assessment of six technology-based functional tasks. There was a 30-day post-training follow-up assessment. The WCG Institutional Review Board approved the study, and each participant completed an informed consent form.

2.2. Participants

The sample consisted of English or Spanish-speaking adults who were at least 60 years old. They needed to have the ability to read computer screens, have at least 20/60 vision, and have sufficient physical dexterity to handle touch screens. There were no racial or ethnic background limitations on the recruitment of males or females. Exclusion criteria included a MOCA score of <18 and a reading score at less than a 6th-grade level in their commonly spoken language. A diagnosis of a significant psychiatric illness other than major depression, the inability to complete assessments in either English or Spanish, recent comparable interventions within the previous 12 months, or a medical history of brain diseases like CVA, seizures, tumors, or severe traumatic brain injuries with prolonged loss of consciousness were among the other exclusions. A final exclusion was participation in previous cognitive or functional skills training, including in our previous studies.

Cognitive status was determined based on whether participants met the Jak-Bondi criteria (Jak et al., 2009) for any of the three primary subtypes of MCI. If they did not, and their MOCA score was 26 or more, they were designated as normal cognition.

2.3. Cognitive Assessments

The participants’ preferred language—English or Spanish—was used for all assessments.

2.3.1. Baseline/Screening Measures

2.3.1.1. Montreal Cognitive Assessment (MOCA)

This test evaluates cognitive performance with scores ranging from 0-30. Certified bilingual raters performed assessments.

2.3.1.2. Reading Performance.

The literacy level of English speakers was examined with the Wide Range Achievement Test (WRAT;Jastak , 1993), 3^rd edition. Spanish speakers were assessed with the Woodcock-Munoz Language Survey, 3^rd edition (Woodcock et al., 2017).

2.3.2. Neuropsychological Assessment for MCI Determinations

2.3.2.1. Wechsler Memory Scale- revised, Logical Memory I and II (Anna Thompson Story).

Participants were read the story and asked for immediate recall, followed by a 20-minute delayed recall filled with other non-verbal assessments.

2.3.2.2. Brief Assessment of Cognition (BAC): App version.

The original paper version of the BAC assesses cognitive areas that are linked to daily functioning (Keefe et al., 2004). For simplicity of administration and uniformity, the same evaluations are delivered via a tablet linked to the cloud with the BAC App (Atkins et al., 2017).

The cognitive domains assessed include:

Verbal Memory:

5-trial 15-item word list learning test. Dependent variable was the total number of words learned.

Digit Sequencing:

Verbal working memory task. Dependent variable was the total number of digits correctly recalled in sequence.

Token Motor Task:

Measures motor speed and manual dexterity. Dependent variable was the number of tokens that can be moved into a cup in 60 seconds using both hands at the same time.

Verbal Fluency:

Measures category fluency (animals) and letter fluency (F and S). Dependent variable was the total score for 3 60-second blocks

Symbol Coding:

Processing speed task, measuring coding performance. Total score for correctly coded items in 60 seconds was the dependent variable.

Tower of London:

Executive functioning task, measures problem-solving. Participants view before and after pictures of balls on pegs. They respond with a decision as to how many moves were required to move from the first to the second. The dependent variable was the number of correct decisions.

The Jak-Bondi criteria (Jak et al., 2009) were applied to the results of the Anna Thompson story (immediate and delayed) and the 6 BAC-App performance tests to ascertain MCI subtype: non-amnestic (impairments on at least two non-memory domains, with no more than one memory domain); amnestic (impairments on at least two memory tests but no other domains); or multi-domain (at least two impairments on both memory and other domains). We used the age-corrected normative standards developed during the validation of expanded version of the BAC-app (Atkins et al., 2018) to identify whether performance on each of the individual tasks was worse than −1.0 SD, defining this score as reflecting impairment on that test. List learning was considered a memory domain, but digit sequencing was designated as working memory and not contributing to amnestic MCI determinations. We used the ADNI cut-offs for immediate and delayed recall on the Anna Thompson Story.

2.3.3. Development of Composite BAC-app scores.

Because we had a sample of participants that was older and composed of a heterogenous group of minority and non-minority participants, we created a composite score for the performance of the NC sample. We created standard scores for baseline performance on each of the 6 subtests, calculating the mean and standard deviation (SD) for the NC participants. We then adjusted each participant’s score on each subtest by calculating the difference of that score from the mean of the NC sample, dividing by the NC SD. This procedure yielded standard scores for each test, with the NC having a mean of 0.0. We then averaged all six scores to calculate a composite standard (z) score. Follow-up standard scores were calculated by taking the raw scores for each participant on each subtest at the follow-up assessments and subtracting that score from the baseline subtest mean of NC sample, dividing by the baseline SD. These scores were averaged into a composite for analysis. We used three different alternative forms of the BAC app; previous studies have shown retest effects across forms that were less than 0.1 SD for composite scores.

2.3.4. Virtual Reality Functional Capacity Assessment Tool (VRFCAT; Keefe et al., 2016).

The VRFCAT assesses four distinct functional abilities and 12 total task objectives: using a bus for transportation, in store shopping, managing currency, and determining whether items needed to finish a dish are available. A brief tutorial was administered to all participants, which included sample items like those from the test and practice in using a tablet computer with a touchscreen. For each objective, the dependent variable was time to completion. For all objectives, participants who were unable to complete the objective within 300 seconds or after making 6 errors were given a time-to-completion score of 300 seconds for that objective and automatically progressed to the next objective. Total time to completion across all objectives was used as the dependent variable at each assessment. Six different forms of the VRFCAT had been previously developed in multiple languages and versions 1-3, English and Spanish, were used in this study, with forms administered in a constant order across assessments. Previous studies with HC populations have found retest performance chances across forms that were d=.10 or less (Atkins et al., 2015).

2.4. General Procedures

The baseline fixed-difficulty FUNSAT, BAC-App, and VRFCAT assessments were performed in person by each participant. Participants were reassessed remotely with Form B of the fixed-difficulty assessments at the conclusion of the 12-week training period or following mastery of all six training tasks. There was a 30-day post-training assessment performed with no interspersed training, with these results presented in the current paper as 30-day followup.

2.4.1. FUNSAT^™ Program

2.4.1.1. Fixed Difficulty Functional Capacity Assessment.

The third generation of the FUNSAT^™ program assesses and teaches the same skills as earlier versions of the software, such as managing medication (understanding prescription labels and keeping medication organized), shopping at an online pharmacy, using a telephone voice menu for prescription refills, operating an ATM and a ticket kiosk, and internet banking (see examples in Figure 1). The cloud-based training was delivered via a touch-screen device (Chrome Book). Participants used their personal Wi-Fi connection or a hotspot that was provided to access the internet. The six tasks comprised three to six subtasks with sequential requirements, and they were presented in a multi-media style with improved graphic representations, text, and audio. In the telephone refill task, for instance, participants selected a preferred delivery method, selected a certain time and day for pick-up, refilled several prescriptions (pill bottles showed on the screen), and contacted the pharmacy using a simulated mobile phone keypad. Real-time data was gathered on mistakes and completion times. Only the time the participant was actively working on the activities was included in the completion time, which is in line with previous technology-based functional capacity assessments, such as the VRFCAT. If an item in a subtask (such as selecting the incorrect account in the ATM task) was incorrect more than four times, the FUNSAT program automatically moved on to the next question. Error feedback appeared as a pop-up window with the original instructions repeated and no attempts at training.

Figure 1.

Samples Depictions of Functional Skills Tasks Trained in the FUNSAT training System

2.4.1.2. FUNSAT Training.

The adaptive protocol used in FUNSAT^™ training provides instant feedback, progressive teaching, and increasing detailed feedback information in response to successive errors. For instance, If the participant entered the wrong PIN in the ATM task on their first attempt, they would receive the following feedback, repeating the original instruction, “Try Again! Your ATM PIN is 1234.” If they repeated the error for a second time, the feedback was “Try Again! Remember, your PIN is 1234. Please enter 1234.” Feedback for a third error was “Try Again! Press 1, then press 2, then press 3, and then press 4. Then press ENTER.” If they made a fourth error, the four keys lit up in sequence, with the participant instructed to touch them. An item, such as entering a PIN, was described as successfully mastered when it was completed once or twice in a row without making more than one mistake. A subtask was considered mastered when all items were mastered, and the six tasks were deemed mastered when all their separate subtasks were completed. Only the subtasks that the participants had not mastered were retaught when they returned to training. Once all six tasks were mastered, and training was designated as complete then follow-up assessments were initiated.

2.4.2. Computerized Cognitive Training

The MCI participants were randomized to receive FUNSAT^™ training alone or in combination with a burst of CCT. The Brain HQ program that we prioritized was “Double Decision”. The task was selected because of the advantages of processing speed training on the same our highly similar tasks documented both in our earlier study and in the ACTIVE study and related trials (Edwards et al., 2002; Ball et al., 2006; Wolinsky et al., 2015).Double decision requires participants to perform two tasks with simultaneous stimulus presentation: identify one of two centrally displayed objects (Car vs. Truck) and locate a simultaneously presented peripheral stimulus that differed from seven other stimuli in a semi-circular array. Duration of presentation of the target stimuli defines the difficulty of the task and duration is adjusted bidirectionally in response to trial x trial performance. Participants were also permitted to train up to 20% of their sessions on the extra task, “Hawkeye,” to offer variation in training experiences, in keeping with other clinical trials using Brain HQ (Mahncke et al., 2021). Hawkeye is a speed task that does not involve multi-tasking, wherein one item in an array of 6 birds is different from the other 5. The participants’ task is to click on the outlier, again with duration of presentation defining task difficulty and being dynamically adjusted in response to performance.

2.5. Data analyses

Analyses were performed with SPSS, version 28 (IBM Corporation, 2023). We examined training related gains in the BAC-App and VRFCAT with repeated-measures Analysis of Variance, with time (Baseline, Post Training, Follow-up) as the repeated measure, group (NC, MCI Skills-only, and MCI-Combined) as a between subjects’ factor, and total training sessions as between subjects’ covariate. Our primary interest was the interaction of group by time, because we knew that NC participants would have better overall performance because of baseline differences. We also compared training gains from baseline to endpoint between MCI participants across the two training conditions on composite cognition and VRFCAT time to completion using t-tests. We also correlated baseline and change scores on the Composite cognition and the VRFCAT time to completion with baseline MOCA scores, years of education and training sessions completed using Pearson correlations. Finally, we used regression analyses to examine the differential prediction of changes in the VRFCAT and composite cognition from baseline to endpoint based on the baseline scores on the two tasks, years of education, MOCA scores, and the number of training sessions.

3.0. Results

Demographic information is presented in Table 1 and the CONSORT diagram is presented in Figure 2. MCI participants had significantly less education and lower MOCA scores than NC participants but did not differ in age. There were no site, race, or training language differences across MCI status. There were slightly more Latinx participants and slightly more male participants in the MCI group than in the NC group. Completion rate for training was 92% for MCI participants who initiated training and 96% for the NC. There were two MCI participants and 1 NC who did not complete the follow-up assessment.

Table 1.

Demographic and Descriptive Information on Participants

Participants with MCI (N = 92)				NC Participants (N =72)
		M	SD	M	SD		t	p
Age		71.68	6.40	71.17	6.36		−.50	.65
MOCA score		22.45	3.21	27.15	1.37		11.50	<.001
Years of Education		13.29	3.94	15.56	2.52		4.18	<.001
		Mild Cognitive Impairment		Normal Cognition
			N=92	N=72
Site			N(%)	N(%)		X2	p
	South Florida		38 (41%)	32 (43%)		1.21	.55
	NY		54 (59%)	40 (57%)
Sex
	Male		18 (20%)	7 (8%)		6.00	.05
	Female		74 (80%)	65 (92%)
Race	White		36 (36%)	34 (47%)		7.95	.44
	Black		28 (30%)	14 (18%)
	Other/		28 (30%)	24 (35%)
	More than 1/None
Ethnicity
	Latinx		52 (57%)	34 (46%)		8.20	.04
	Non Latinx		40 (43%)	38 (54%)
Training Language
	English	50 (55%)		48 (67%)		2.64		.27
	Spanish	42 (35%)		24 (33%)
MCI Classification
Amnestic			14 (15%)
Multi-Domain			38 (41%)
Non-Amnestic			40 (43%)

Figure 2.

CONSORT diagram for patient flow to the end of the post training follow-up assessment

Gains in Cognitive Performance and Reductions VRFCAT time to completion.

Table 2 presents the performance of the three subsamples of participants on the BAC-app cognitive composite scores and VRFCAT time to completion across the three assessments. Figure 3 presents the baseline scores on the VRFCAT across the three training groups. As can be seen in the figure, the distributions of the three samples overlapped, with more occurrences of longer completion times in the MCI samples. There was a statistically significant (p<.001) overall effect of group on number of training sessions per functional task. Tukey post-hoc tests found that the NC performed significantly fewer fewest training sessions per task than MCI receiving skills only training, with no other comparisons significant. There were significant time effects for Composite cognitive performance and VRFCAT time to completion, with effect sizes for changes from baseline to 30-day follow-up for the different groups all larger than d>.55 for the VRFCAT completion time and all larger than d=.72 for changes in composite cognition. The interaction effect of time x group was not significant for VRFCAT Completion time, F=1.21, p=.31, or BAC App composite cognitive performance, F=2.18, p=.07, suggesting that performance improvements across the groups did not differ. The interaction effect of time x training sessions was statistically significant for improvements in composite cognitive performance. Following up this interaction, we repeated the analysis, eliminating total training as a covariate. The results suggested a statistically significant interaction of training group x time, F=3.74, p=.03. Thus, the lack of differences between groups in training gains in cognition may be accounted for by the MCI participants having more training exposure than the NC participants.

Table 2:

Performance on Composite Cognitive performance and The Virtual Reality Functional Capacity Assessment Task at Baseline and After Treatment: Separated by Training Condition : Analysis of cases who competed all three assessments

Mean Training Sessions Per Task
	M	SD
Normal Cognition	5.38	4.37		ANOVA for Training Sessions Per task by Group F=9.06, p<.001
MCI Skills only	10.75	8.16		Follow-up tests: NC> Skills only, p<.001; NC=Combined, p=.14; Skills only=Combined, p=.10
MCI Combined	6.72	5.37
Outcomes Across Training Sessions
			Composite Cognition							VRFCAT Completion Time
	BL		Post		30-Day			BL		Post		30-Day
	M	SD	M	SD	M	SD	d	M	SD	M	SD	M	SD	d
Normal Cognition	0.00	.54	.11	.53	.27	.55	.82	1023.57	330.70	899.62	266.27	794.99	221.41	.81
MCI Skills only	−1.07	.63	−.68	.69	−.70	.64	.91	1316.99.	468.87	1171.89	446.80	1117.82	419.60	.62
MCI Combined	−1.15	.61	−.94	.79	−.72	.81	.73	1333.08	569.11	1177.66.	528.00	1121.12	367.77	.56

Notes.

No Significant Effects of Group or Interactions involving group Effect Sizes are within group change from baseline to 30-Day Follow-up

Time effects for Cognition, F=14.44, p<.001;

Time x Training Sessions for Cognition, F=4.03, p=.02;

Time Effect for VRFCAT, F=7.45, p<.001

Time X training Sessions for VRFCAT: F=2.79, p=.07.

Figure 3.

Baseline Scores on VRFCAT Displayed by Training Group: NC, MCI Skills only, and MCI Combined Treatment

3.2. MCI Training Condition.

There were no training condition differences within the MCI group for composite cognition, t=.86, p=.39, or VRCAT time to completion, t=.19, p=.85.

3.3. Baseline Participant Characteristics and Scores on BAC-App Composite Performance and VRFCAT time to completion.

Table 3 presents correlations between MOCA scores, years of education, and baseline and change scores in composite cognition and VRFCAT time to completion. In the NC sample, MOCA scores were correlated with baseline performance on composite cognition but no other baseline variables. In the MCI participants, MOCA scores correlated with baseline and change scores on composite cognition and baseline VRFCAT scores. Education was correlated with baseline VRFCAT scores in the NC participants, but in the MCI participants, education was also correlated with baseline composite cognition as well as baseline VRFCAT scores. Years of education was not correlated with treatment related changes in either sample. Lower baseline scores on both composite cognition and the VRFCAT were correlated with requiring more training sessions to graduate in participants with NC, but not in those with MCI. More improvement in both the VRFCAT and composite cognition were associated with more training sessions completed by the MCI participants, but not in those with NC. Across the two samples, MOCA scores shared at most 6% of variance in cognitive changes with training, while baseline scores on the BAC-App composite shared 30% of the variance with MOCA scores in the MCI sample.

Table 3.

Correlations Between MOCA scores and Education and Training Sessions Prior to Graduation with Baseline and Change scores for Composite Cognition and VRFCAT from Baseline to Follow-up Assessments

	MOCA	Education	Training Sessions
Normal Cognition (n=68)
Baseline Composite Cognition	.32**	−.08	−.30**
Change in Composite Cognition	.13	.06	.16
Baseline VRFCAT	−.18	−.29**	.51***
Change in VRFCAT	−.19	−.08	.12
Training Sessions	−.23	−.21
MCI (n=84)
Baseline Composite Cognition	.55**	.37**	−.13
Change in Composite Cognition	.24**	.03	.28**
Baseline VRFCAT	−.24**	−.22**	.16
Change in VRFCAT	−.13	−.04	.24**
Training Sessions	−.03	−.17

4.0. DISCUSSION

There are several findings in this study that are novel and potentially important. Most importantly, computerized training on technology-related functional skills led to improvements in both MCI and NC participants in untrained outcomes: composite cognitive performance and a performance-based measure of functional capacity None of the component skills trained in FUNSAT are required to perform the VRFCAT, suggesting that FUNSAT training led to far transfer to other functional skills. Such a finding has not been previously reported in either of these populations. Also, FUNSAT training, without add-on CCT, led to improvements in both composite cognition and functional capacity in NC and MCI participants, suggesting that computerized functional skills training, even without concurrent CCT, has benefits in NC and MCI individuals. We did not find a synergistic effect of combined training on cognition or functional capacity in MCI, in terms of group-mean changes. However, training gains for the two MCI groups were not different and the number of FUNSAT training session associated those gains were less, albeit non-significantly (p=.10), with combined training.

In one of the other outcomes papers from this study (Dowell-Esquivel, et al., 2023), we found that functional skills training alone in NC individuals led to real-world training transfer, indexed by increases in real-world performance over 6 months in both trained and untrained technology-related skills. These increases in skills performance were correlated significantly with training gains on the simulations, for both trained and untrained technology-related functional skills. Thus, increased exposure to technology use alone cannot be the explanation for increases in real-world performance of these skills.

In terms of the MCI participants, combined CCT and skills training led to equivalent gains with a shorter duration of skills training sessions because of the CCT burst at the outset in those participants receiving combined training. Our cognitive training gains were reduced compared to our previous study for two likely reasons. First, NC participants did not receive combined training, reducing our ability to replicate a synergistic effect in a population with likely greater potential for cognitive gains with training. Second, because of our re-definition of graduation from all 6 training tasks, participants who mastered all 6 tasks received considerably fewer total training sessions than in the previous study. Although mastery of the tasks is important, our finding that exposure to skills training alone led to cognitive gains in the NC participants, and that these gains were increased with more training sessions for composite cognition and VRFCAT, suggests that continued skills training may have had the potential for additional gains in cognition and functional capacity.

There are limitations of the study, largely based on study design decisions. As we had previously demonstrated synergistic training gains in NC participants, our current study administered only skills training to NC, to develop normative understanding of expected training gains in NC training on FUNSAT. Our modification of the definition of graduation was designed to apply a more realistic criterion for adequate task performance, noting that making a single error entering a PIN or selecting a choice in a functional task (selecting the wrong account when viewing a balance) does not constitute a “Fatal error” for successful completion of a real-world functional task. That change in criteria led to more rapid mastery across tasks and reduced the number of training sessions delivered, reducing exposure to training that may have provided more cognitive benefits. There was no untreated or inactive treatment group, although the level of change detected with training was considerably larger than previously identified retesting effects across the different forms of the VRFCAT and BAC-app. A benefit associated with retesting alone would be expected to be 0.2 SD or less even with the same forms of the assessment and the level of training gains was 4 times that large for the BAC composite score and 3 times that for changes in completion time on the VRFCAT based on previous studies using multiple forms of the tests.

While this is the first study of computerized training in MCI that demonstrates far transfer to untrained functional capacity tasks, this is the second study overall that found statistically significant gains on the VRFCAT with computerized training, with the other study in schizophrenia (Nahum et al.,2021). Combined with our previous reports of real-world transfer of training to technology-related functional performance, these data suggest that computerized skills training provides wide-ranging cognitive and functional benefits. NC participants manifested gains in cognition and functional capacity with skills training alone, also demonstrating for the first time that training on cognitively demanding technology-related functional skills can improve functional capacity performance as well as replicating the previously reported gains in cognitive abilities in older people without cognitive impairments.

Highlights.

• Computerized functional skills training improved performance-based measures of cognition Functional capacity

• Changes were 3-4 times as large as expected practice effects.

• Changes in these two outcomes were similar in magnitude in participants with MCI and Normal Cognition

• Computerized cognitive training added to skills training led to greater gains per training session