A preliminary investigation into the factor structure of two psychological flexibility measures in a sample of community-dwelling older adults

Abstract

Objectives:

Acceptance and Commitment Therapy (ACT) targets psychological flexibility and the ability to identify behavioral function in context. Properly measuring these constructs is imperative to understanding whether these processes are mechanisms of change in treatment. The current study examined the factor structure of the Comprehensive Assessment of ACT processes (CompACT) and Tacting of Function scale (TOF) in community-dwelling older adults.

Methods:

Factor structure was examined with CFA. Eighty community-dwelling older adults completed questionnaires prior to an intervention.

Results:

While the original 23-item, 3-factor structure of the CompACT demonstrated inadequate fit, a modified 15-item, 3-factor structure demonstrated adequate fit. The original 1-factor structure of the TOF demonstrated inadequate fit. A modified 2-factor structure of the TOF fit significantly better than the original 1-factor structure; however, this model also demonstrated inadequate fit.

Conclusions:

Examining the factor structure of the CompACT and TOF in an older adult sample contributes to the field’s understanding of the constructs of psychological flexibility and tacting ability and to the clinical utility of these measures in an understudied sample.

Clinical Implications:

These findings provide preliminary support for the use of a short-form version of the CompACT to measure therapeutic processes of change in community-dwelling older adults.

Introduction

Acceptance and Commitment Therapy (ACT) is a transdiagnostic psychotherapy which facilitates behavioral change by increasing psychological flexibility (Hayes et al., 2011). Psychological flexibility, or the ability to engage in values-based actions with present moment awareness, despite the presence of internal struggles (e.g., thoughts or physical sensations) (Hayes et al., 2006), is comprised of three core processes that have been identified as key tenets to guide case conceptualization and promote overall psychological flexibility: openness, awareness, and engagement (Strosahl et al., 2012). Openness is willingness to experience one’s emotions, thoughts, physical sensations, memories, and urges (even when unpleasant), in the pursuit of personal values. Awareness is mindful attention to one’s experiences, behaviors, and functions of behavior. Engagement is connection with and the pursuit of personal values through consistent and flexible behaviors over time.

Psychological flexibility has been shown to be associated with well-being across samples and populations (Marshall & Brockman, 2016; Stenhoff et al., 2020; Wersebe et al., 2018) and has been found to be positively associated with desirable outcomes (e.g., health benefits) and negatively associated with adverse outcomes (e.g., psychopathology symptoms) (Kashdan & Rottenberg, 2010). Recently, there has been increasing interest in ACT as a treatment for improving health and quality of life outcomes in older adults (Pachana et al., 2014; Petkus et al., 2013). Increases in physical health problems and declines in physical functioning can lead to limitations in activities older adults can engage in. With ACT, clinicians can work with older adults to develop functional analyses of the presenting concerns, and help them workably commit to engagement in valued activities (Pachana et al., 2014). A functional analysis is used to examine the function of a person’s behavior in a given context. Topographically, an individual can engage in the same behavior, and the function of that behavior could be different under certain circumstances. For example, a person could drink alcohol to connect with friends, or they could drink to reduce or alter distressing emotions, thoughts, or memories. A functional analysis would help the clinician and client to determine the function of behaviors with the goal of building ongoing behavioral awareness. Functional analysis contributes to both case conceptualization and intervention planning through the identification of associations between environmental context and behaviors which may be causing distress (Assaz et al., 2018). Behavioral therapies, such as ACT, are guided by functional analysis, which allows clinicians to more effectively select therapeutic processes (e.g., psychological flexibility) to target in order to individualize treatment, rather than strict adherence to a protocol that may be less sensitive to contextual factors (e.g., age-related changes). As such, an ACT approach could be useful to improve successful aging, as it can be used to help older adults accept physical health declines that are not within their control, while helping them to continue identifying goals in adaptive ways (Petkus et al., 2013).

Though limited, there is some empirical evidence suggesting ACT is an effective intervention in this population. ACT has shown to be a feasible intervention with older adults (Wetherell et al., 2011), and effective in reducing depressive and/or anxiety symptoms, and improving well-being and life satisfaction (Goetz & Hirschhorn, 2022; Jacobs et al., 2018; Karlin et al., 2013; Wetherell et al., 2011; Witlox et al., 2021). In a published case example , ACT was successfully paired with palliative care treatment for an older individual with a life-limiting illness experiencing suicidal ideations (Hinrichs et al., 2020). A 2016 meta-analysis (Kishita et al., 2017) of mindfulness-based cognitive behavioral therapies, including ACT studies, found that these interventions are effective for treating depressive and anxiety symptoms in older adults. Similarly, a newly published systematic review indicated that ACT may be a promising mindfulness intervention for improved mental health outcomes in older adults with chronic health conditions (Kayser et al., 2022). Further, ACT has shown to be effective in the management of chronic pain in older adults as well (McCracken & Jones, 2012). Thus, evidence exists to indicate that ACT is a feasible intervention that improves both psychological and physical health outcomes in older adults.

Despite the growing interest in the utility of ACT in older adults, more work needs to be done regarding the measurement of psychological flexibility in this population. Lifespan theories such as the Selective Optimization with Compensation model (which posits that as individuals age, they may adapt to focus on strengths-based goals and strategies that are available to them despite functional losses; (Baltes & Baltes, 1990) and Socioemotional Selectivity Theory (which suggests that as individuals age, they may be more likely to prioritize emotional meaning in relationships; (Carstensen et al., 1999), as well as research on the “positivity effect” in older adulthood (which suggests that older adults may tend to focus specifically on positive rather than negative information; (Carstensen & Mikels, 2005; Löckenhoff & Carstensen, 2004), give reason to postulate that processes such as psychological flexibility may not operate exactly the same in the context of older adulthood compared to other points in life. Appropriate measurement of therapeutic processes of change in different populations is critical in understanding how interventions may be working for specific groups.

Importantly, a recent systematic review of 46 articles examining psychological flexibility in older adults found that older adults showed greater awareness than younger adults (measured by the Mindful Attention Awareness Scale (Brown & Ryan, 2003), Kentucky Inventory of Mindfulness Skills (Baer et al., 2004), Philadelphia Mindfulness Scale (Cardaciotto et al., 2008), and Five Face Mindfulness Questionnaire (Baer et al., 2008)) and that the openness and awareness tenants in older adults had medium-to-large associations with depression and anxiety, in the expected directions (Plys et al., 2022). However, findings also indicated that there is a paucity of research examining the engagement process in older adults, and low internal consistency was found in many of the measures assessing the openness process (most notably in the Acceptance and Action Questionnaire (AAQ; first and second editions (Bond et al., 2011; Hayes et al., 2004)), a commonly used measure to examine lack of openness, termed psychological inflexibility by the authors). Thus, measurement of engagement and psychometrically stronger measures of openness are needed in older adult samples. Based on these findings, the authors recommended future research include rigorous testing of measures targeting psychological flexibility processes in older adult samples (Plys et al., 2022).

Given these current limitations, measures should be tested to examine the adequacy of using psychological flexibility process measures in older adult samples specifically. One measure that has potential utility in older adults is the Comprehensive assessment of Acceptance and Commitment Therapy processes (CompACT) (Francis et al., 2016), which assesses self-reported levels of the three core processes: openness, awareness, and engagement. The initial validation study of the CompACT proposed a correlated three-factor structure, demonstrated that the questionnaire possessed adequate internal consistency, and was positively associated with well-being and negatively associated with psychological distress (Francis et al., 2016). Given these properties, the CompACT may address the existing weaknesses (i.e., low internal consistency, lack of engagement measures) in the assessment of psychological flexibility in older adults. However, this measure was developed in a non-clinical sample of adults and has not yet been examined in an older adult population. Furthermore, the dimensionality of the CompACT has largely been assessed using exploratory factor analysis, rather than confirmatory factor analysis (CFA), with the exception of two recent studies (Trindade et al., 2022; Hsu et al., 2021; Hsu et al., 2022). Furthermore, these measurement refinement efforts for the CompACT have indicated that a shorter 18-item Portuguese-language version (Trindade et al., 2021; Trindade et al., 2022) and a 15-item English-language version (Hsu et al., 2021; Hsu et al., 2022) demonstrate better model fit across fit indices than the original 23-item version. Notably, these samples were not older adults, and generally were young to middle adult aged.

Relatedly, the Plys et al. (2022) review found that very few studies used measures to capture the engagement process of psychological flexibility. The engagement process includes understanding and clarifying one’s values and identifying values-based behaviors. A skill that includes aspects of engagement is the ability to tact (or label) the function (or purpose) of behavior in a given context. Behavioral function is typically labeled as avoidant- (e.g., drinking alcohol to numb pain) or values-based (e.g., volunteering in service of community). Despite this notable skill in ACT, until the recent development and validation of the Tacting of Function scale (TOF) (Pierce & Levin, 2019), measurements of this specific process have been limited to clinical observation rather than empirically based or validated tools. Pierce & Levin (2019) developed the TOF, a 10-item measure aimed at measuring one’s ability to tact, or label, functions of behavior. This skill integrates engagement with one’s values and noticing why one is engaging in a behavior, and as such, is important in promoting behavioral change. The initial validation study of the TOF indicated that the scale had a single factor that explained the majority of the variance, demonstrated strong internal consistency, and was negatively significantly associated with psychological inflexibility and obstruction in valued living (as measured by the AAQ-II and Valuing Questionnaire) and positively significantly associated with progress in valued living (also measured by the Valuing Questionnaire) (Pierce & Levin, 2019; Smout et al., 2014). Though these findings provide evidence for dimensionality and construct validity of the TOF, generalizability of these results may be limited because the sample consisted of a homogenous sample of college-aged students.

To summarize, ACT may be a useful intervention for older adults. However, there are limitations in the measurements currently used in older adults to measure psychological flexibility in that many of the measures have not been validated with older adult samples. Thus, the objective of the present study was to evaluate the factor structure of two promising measures of psychological flexibility processes, the CompACT and TOF, in a sample of non-clinical, community-dwelling older adults. The study examined whether the originally proposed factor solutions translated to the current sample. If the originally proposed models resulted in inadequate fit, alternative factor structures were explored.

Methods

Participants and Procedures

Participants included 80 community-dwelling older adults who were 65 years of age or older and residing in the Midwestern United States. The majority of the sample identified as Female (56.3%), with a mean age of 78.05 (SD = 5.08) years. Most of the sample identified their race as White (97.5%), with 2.5% identifying as “Other.” Most participants reported being college-educated, with 72.5% reporting a 4-year degree or higher.

All study procedures were approved by the University’s Institutional Review Board. The current analyses represent a sub-analysis of baseline data from a longitudinal intervention study (NCT [NCT03839329]). Of note, the parent study aimed to be powered to examine between group differences following a brief group intervention. Participants were recruited from an existing registry of older adults, where potential participants previously screened negatively for significant primary psychiatric disease, medications that affect cognitive functioning, neurological events, and major surgeries. After consenting, participants completed on-paper measures of demographic characteristics, psychological flexibility, and tacting ability as part of the baseline measures collected for the parent trial.

Measures

Psychological Flexibility.

The CompACT (Francis et al., 2016) was used to measure the three core processes of ACT: openness to experience (OE), behavioral awareness (BA), and valued action (VA). OE examines willingness to experience internal experiences such as thoughts or emotions without trying to control them (e.g., “Thoughts are just thoughts – they don’t control what I do.”). BA examines ability to direct mindful awareness to actions in the present moment (e.g., “I rush through meaningful activities without being really attentive to them.”). VA examines engagement in meaningful or values-based activities (e.g., “I can identify the things that really matter to me in life and pursue them.”). The 23 items are rated on a 7-point Likert scale (strongly disagree to strongly agree); 12 items are reverse scored, and items for each subscale are summed. Higher scores on each subscale indicate greater ability on each specific process.

Tacting Ability.

The TOF (Pierce & Levin, 2019) was used to measure the ability to label the function of behavior as either avoidant (e.g., “It is hard to tell whether or not my choices lead to personal growth”) or values-based (e.g., “I am aware when I feel a sense of meaning in my actions.”). The 10 items of the TOF are rated on a 7-point Likert scale (never true to always true) and assess tacting of function over the previous two weeks. Several items are reverse scored, and items are summed for a total score. Higher scores indicate greater proficiency in this skill.

Analytic Strategy

The factor structure and fit of the 23-items of the CompACT and the 10-items of the TOF were assessed with separate CFAs using robust maximum likelihood estimation (MLR) in Mplus v 8.4 (Muthén & Muthén, 1998-2017). In the context of scale validation, CFA is used to verify the number of underlying factors as specified by a theoretical model, and the specific patterns of item-factor relationships (Brown, 2015). In the CFA framework, an observed response y_is for item i and subject s is given as the sum of the intercept of item i (μ_i), the latent factor of subject s (F_s) weighted by the item specific loading (λ_i), and the error of item i and subject s (e_is) as y_is = μ_i + λ_iF_s + e_is. Models were identified by setting latent factor means to 0 and latent factor variances to 1, such that all items, intercepts, item factor loadings, and item residual variances were then estimated. Model fit was assessed with the obtained model χ², its scaling factor, degrees of freedom, and p-value. Other indices of fit included the comparative fit index (CFI), Tucker-Lewis Index (TLI), root mean square error of approximation (RMSEA), and standardized root mean squared residual (SRMR). Adequate fit was indicated by an obtained model χ² p-value of non-significance, CFI ≥ 0.95, TLI ≥ 0.95, RMSEA ≤ 0.06, and SRMR ≤ 0.08 (Hu & Bentler, 1999; Lei & Wu, 2007). Model comparisons were conducted using the rescaled −2ΔLL with degrees of freedom equal to the rescaled difference in the number of parameters between models. Finally, Omega model-based reliability was calculated for the sum scores of each factor (Brown, 2015). In recent years, omega model-based reliability has been recommended over Cronbach’s alpha, which has been shown to be based on assumptions which are unlikely to hold in empirical data (Dunn et al., 2014).

Results

Table 1 provides the descriptive statistics of all items in the final models, and Tables 2 and 3 provide the estimates and their standard errors for the item factor loadings and intercepts from both the unstandardized and standardized solutions.

Table 1.

Descriptive statistics of the 15-items from the Comprehensive Assessment of Acceptance and Commitment Therapy Processes (CompACT) short-form and 10-items from the Tacting of Function scale (TOF).

	Sample Size	Mean	SD	Minimum	Maximum
CompACT
Item 1	79	5.177	1.032	1	6
Item 2	74	2.865	2.901	0	6
Item 3	80	4.025	2.849	1	6
Item 4	76	3.829	2.852	0	6
Item 7	79	4.329	1.866	0	6
Item 8	79	3.797	4.111	0	6
Item 9	80	4.550	2.748	0	6
Item 10	78	5.436	0.861	1	6
Item 11	80	3.575	3.594	0	6
Item 12	77	4.234	2.802	0	6
Item 15	80	3.337	3.524	0	6
Item 16	80	3.975	3.199	0	6
Item 17	80	4.525	1.924	1	6
Item 19	80	4.625	2.759	1	6
Item 23	80	5.487	0.500	3	6
TOF
Item 1	80	5.438	1.596	1	7
Item 2	80	5.750	1.362	1	7
Item 3	80	4.612	2.187	1	7
Item 4	80	4.888	2.525	2	7
Item 5	80	5.237	2.006	1	7
Item 6	80	4.487	3.525	1	7
Item 7	80	5.388	2.387	1	7
Item 8	79	3.797	3.529	1	7
Item 9	79	4.304	3.022	1	7
Item 10	79	4.962	2.163	1	7

CompACT = Comprehensive assessment of Acceptance and Commitment Therapy Processes; TOF = Tacting of Function scale

Table 2.

Final model estimates for the Comprehensive assessment of Acceptance and Commitment Therapy Processes (CompACT).

Model Parameter	Unstandardized		Standardized
	Estimate	SE	Estimate	SE
OE Factor Loadings
Item 2	1.124	0.189	0.656	0.085
Item 4	1.277	0.156	0.753	0.071
Item 8	1.347	0.193	0.664	0.080
Item 11	1.412	0.186	0.745	0.092
Item 15	1.407	0.147	0.749	0.063
BA Factor Loadings
Item 3	0.895	0.182	0.530	0.105
Item 9	1.056	0.185	0.637	0.096
Item 12	1.251	0.166	0.744	0.094
Item 16	1.223	0.178	0.684	0.083
Item 19	1.264	0.136	0.761	0.059
VA Factor Loadings
Item 1	0.642	0.168	0.630	0.102
Item 7	0.650	0.187	0.476	0.145
Item 10	0.438	0.144	0.473	0.116
Item 17	0.982	0.169	0.708	0.094
Item 23	0.455	0.110	0.644	0.118
Factor Covariance
OE with BA	0.464	0.121	0.464	0.121
OE with VA	0.205	0.151	0.205	0.151
BA with VA	0.585	0.127	0.585	0.127
Residual Covariance
Item 11 with Item 4	−0.908	0.288	−0.642	0.252
Item Intercepts
Item 1	5.167	0.116
Item 2	2.829	0.195
Item 3	4.025	0.189
Item 4	3.776	0.192
Item 7	4.326	0.153
Item 8	3.779	0.228
Item 9	4.550	0.185
Item 10	5.429	0.106
Item 11	3.575	0.212
Item 12	4.192	0.190
Item 15	3.338	0.210
Item16	3.975	0.200
Item 17	4.525	0.155
Item 19	4.625	0.186
Item 23	5.487	0.079

OE = Openness to experience; BA = Behavioral awareness; VA = Valued action

Table 3.

Final model estimates for the Tacting of Function scale (TOF).

Model Parameter	Unstandardized		Standardized
	Estimate	SE	Estimate	SE
Values-based tacting Factor Loadings
Item 1	1.025	0.171	0.811	0.087
Item 2	0.930	0.197	0.797	0.104
Item 3	0.435	0.157	0.294	0.100
Item 4	0.397	0.169	0.250	0.109
Item 5	0.697	0.202	0.492	0.134
Avoidant tacting Factor Loadings
Item 6	0.980	0.228	0.522	0.111
Item 7	0.220	0.195	0.142	0.127
Item 8	1.525	0.173	0.809	0.067
Item 9	1.567	0.155	0.898	0.061
Item 10	0.901	0.195	0.611	0.100
Factor Covariance
Values-based tacting with Avoidant tacting	0.464	0.137	0.464	0.137
Residual Covariance
Item 4 with Item 3	1.321	0.232	0.607	0.080
Item Intercepts
Item 1	5.438	0.141
Item 2	5.750	0.131
Item 3	4.613	0.165
Item 4	4.888	0.178
Item 5	5.237	0.158
Item 6	4.488	0.210
Item 7	5.388	0.173
Item 8	3.777	0.211
Item 9	4.283	0.196
Item 10	4.950	0.167

^*Note: The final model of the TOF resulted in inadequate fit on all indices; thus, factor loadings should be interpreted with this in mind.

CompACT

A CFA of the 23-items of the CompACT demonstrated that fitting the data to the three-factor structure proposed by Francis and colleagues (2016) resulted in inadequate fit (χ² (df = 227) = 377.24, p < .001, CFI = .75, TLI = .72, RMSEA = .09, SRMR = .12). One item (item 13) did not have a significant factor loading (unstandardized loading = .39, p = .11). The development paper of the CompACT did not examine a three-factor model with a CFA (Francis et al., 2016) and recent research (Hsu et al., 2021; Hsu et al., 2022) utilizing CFA has found acceptable fit with a 15-item English-language short form with a three-factor structure (OE, BA, and VA). This three-factor model with the 15-item version reported in (Hsu et al., 2021; Hsu et al., 2022) was examined in the current sample. However, this still resulted in inadequate fit according to most fit indices, χ² (df = 87) = 114.03, p = .03, CFI = .91, TLI = .89) with the exception of RMSEA (.06) and SRMR (.08).

In order to identify sources of misfit, the normalized residual covariance matrix, available via the RESIDUAL output option in Mplus, was examined. Relatively large residual covariances were observed among several reverse coded items, including items 2, 3, 4, 8, and 11. Modification indices, available via the MODINDICES output option in Mplus, corroborated this pattern, further suggesting additional remaining relationships among the reverse scored items that load onto the same factor. Each potential change was considered both statistically and in terms of theoretical similarity between the items.

Accordingly, a three-factor model with the inclusion of a residual covariance between items 4 and 11 was examined. This residual covariance was selected, as items 4 and 11 are both reverse scored, load on to the same factor (OE), and the item content is similar (i.e., both items reference avoiding thoughts or feelings). The revised three-factor model resulted in adequate fit for most fit indices (χ² (df = 86) = 101.62, p =.12, CFI = .95, RMSEA = .05, SRMR = .08), except for TLI (.94). Further, this three-factor model fit significantly better than the previous model (−2ΔLL (1) = 12.90, p < .01). Further examination of local fit via normalized residual covariances and modification indices yielded no interpretable remaining relationships, and thus this three-factor model was retained (Table 2). No further changes were made to the model given adequate fit.

All factor loadings and factor covariances were statistically significant. The relationship between OE and VA was not significant (unstandardized loading = .21, p = .18). Finally, Omega model-based reliability was examined. Omega was .84 for the OE factor, .81 for the BA factor, and .72 for the VA factor, suggesting adequate reliability for each scale. See Figure 1 for the path diagram of the final model.

Figure 1.

Path diagram for the confirmatory factor analysis of the Comprehensive assessment of Acceptance and Commitment Therapy processes (CompACT) with unstandardized regression weights.

TOF

First, a one-factor model was tested, as this is what was initially posited to account for the pattern of covariance across the ten items (Pierce & Levin, 2019). Notably, however, the original development and validation paper of the TOF did not utilize a CFA, but rather a parallel analysis. The originally proposed one-factor model resulted in inadequate fit among all indices (χ² (df = 35) = 132.43, p <.001, CFI = .52, TLI = .38, RMSEA = .19, SRMR = .12). Further, two of the ten items (items 4 and 7) did not have significant factor loadings (unstandardized loadings = .39, p = .14 and .28, p = .14, respectively).

Next, separate latent factors for the value- and avoidant-worded items were tested by specifying a two-factor model in which the value-worded items 1-5 indicated a values-based tacting factor, and in which avoidance-worded items 6-10 indicated an avoidant tacting factor, and in which the two factors were allowed to correlate. The correlated two-factor model fit significantly better than the one-factor model, −2ΔLL (1) = 45.98, p < .001. However, the correlated two-factor model still had inadequate fit for all indices (χ² (df = 34) = 85.23, p < .001, CFI = .75, TLI = .67, RMSEA = .14, SRMR = .11). Again, sources of local misfit were identified using the normalized residual covariance matrix. Relatively large positive residual covariances were observed among several items, including items 3, 4, and 7 (the reverse scored items). Modification indices, available via the MODINDICES output option in Mplus, corroborated this pattern, further suggesting additional remaining relationships among the reverse scored items.

Accordingly, the two-factor model with the inclusion of a residual covariances between items 3 and 4 was examined. This residual covariance was selected, as items 3 and 4 load onto the same factor (values-based tacting). This two-factor model fit significantly better than previous two-factor model (−2ΔLL (1) = 35.93, p < .001). However, the model did not meet the threshold for adequate fit on any of the indices (χ² (df = 33) = 51.76, p = .02, CFI = .91, TLI = .87, RMSEA = .08, SRMR = .09). Further examination of local fit via normalized residual covariances and modification indices yielded no interpretable remaining relationships, and thus, this two-factor model was retained. Note that the factor loadings should be interpreted with the inadequate fit in mind.

Table 3 provides the estimates and their standard errors for the item factor loadings and Figure 2 illustrates the final model’s path diagram (i.e., the two-factor model with the inclusion of a residual covariances between items 3 and 4). All factor loadings and the factor covariance were statistically significant, with the exception of item 7 (unstandardized loading = .22, p = .26). Regarding model-based reliability, Omega was .64 for the values-based tacting factor and .77 for the avoidant tacting factor, suggesting marginal to adequate reliability for each of the scales.

Figure 2.

Path diagram for the confirmatory factor analysis with the Tacting of Function scale (TOF) with unstandardized regression weights.

Discussion

There is growing interest in the utility of ACT with older adults to improve psychological health (Goetz & Hirschhorn, 2022; Jacobs et al., 2018). However, a recent systematic review found that psychometrically stronger measures of psychological flexibility are needed, particularly in the openness and engagement processes (Plys et al., 2022). Thus, the present study examined the factor structure of the Comprehensive assessment of Acceptance and Commitment Therapy processes (CompACT; which measures openness, awareness, and engagement) and the Tacting of Function scale (TOF; a skill that includes aspects of engagement) in a sample of community-dwelling older adults.

Although our findings did not support the use of the 23-item CompACT in community dwelling older adults, the 15-item, three-factor short form proposed by (Hsu et al., 2021; Hsu et al., 2022) demonstrated mostly adequate fit (with the exception of TLI) after residual covariances were included between negatively-worded items on the OE scale. Notably, our findings with the CompACT are similar to those in a published manuscript on the psychometric comparison of psychological inflexibility measures (Ong et al., 2020), though these analyses were not conducted among a sample of older adults. According to the examination by Ong et al. (2020), the CompACT produced a different factor structure than what was found in the development and validation analyses across three samples: undergraduate college students, community members, and treatment-seeking individuals. Thus, our findings are in line with previous literature demonstrating that the CompACT may be less stable in terms of structural validity.

Moreover, the findings demonstrated that the TOF may possess different factor structures in different samples, and future research is warranted to further understand the structure of this construct. To our knowledge, this is the first examination of this scale with CFA, as the original development and validation data were examined with a parallel analysis (Pierce & Levin, 2019) . Although the two-factor model demonstrated better fit than the one-factor model, the final two-factor model did not demonstrate adequate fit on any of the fit indices. Two items in particular loaded slightly lower than the other 8 items: items 4 (“It is hard to say if my choices are connected with my deepest held values”) and 7 (“I don’t notice when strong emotions have taken the “driver’s seat” of my actions”). It is possible that the language used in these two items in particular is more abstract than other items on the scale. Whereas other items ask if one’s actions align with the person that they “want to be,” if their actions fall short of “intentions,” or if they are aware of whether their choices are based on reducing feelings of “sadness” or increasing feelings of “happiness or joy,” items 4 and 7 may be more difficult to interpret. It is also possible that it was unclear to participants what “deepest held values” are. In ACT, values are discussed as who and what are important, and the qualities one wants to embody, but respondents may not have interpreted the item this way, particularly given that these data were collected prior to intervention. Similarly, it is possible that there was some ambiguity in the interpretation of “strong emotions,” or that this non-clinical group of participants answered the question with the belief that in their personal experience, strong emotions rarely have the stated impact. Relatedly, it is worth noting that increasing age has been shown to be related to alexithymia in the general population (Mattila et al., 2006), potentially leading to differential responding to this item among older adults.

There are some limitations worth considering in the interpretation of this study. First, a major limitation of the current study is that the sample size was relatively small (N = 80), and the CFA was likely underpowered. The precision of CFA model parameter estimates, as well as select model fit indices used in model evaluation are dependent on sample size (Brown, 2015). Therefore, future replications of the current study in larger samples of older adults are necessary. Furthermore, the sample was homogenous. Nearly all participants identified as White and highly educated, which may limit generalizability of the findings from the study. Participants were also non-clinical older adults, without significant primary psychiatric disease, neurological events, or impaired cognitive functioning. More research with these scales in a larger and more diverse sample of older adults, as well as a clinical sample, is warranted. Future research should explore the development of assessments that may better measure these processes in older adults, particularly in regard to tacting ability, as the TOF did not demonstrate adequate fit in the current sample. This may be due in part to the lack of power resulting from our small sample size. Despite these limitations, the authors believe that the current work is an important start to examining the appropriateness of these measures of psychological flexibility and tacting of function in an older adult sample. Given that older adults are an understudied group in psychological intervention research broadly, and in ACT in particular, this work serves as an important first step in examining which scales are appropriate to use to measure these processes, especially with the growing interest in the utility of ACT in this population. However, the sample size of the current study remains an important limitation to consider when interpreting these results.

Understanding transdiagnostic processes of change is imperative in improving clinical intervention, and thus, validated measures of these processes are critical. In particular, understanding the psychometric properties of these measures among the samples being examined is important to the larger literature. Recently, there has been a gaining interest in using ACT as a treatment for improving physical and psychological outcomes in older adults (Goetz & Hirschhorn, 2022; Jacobs et al., 2018; Kayser et al., 2022; Kishita et al., 2017; Pachana et al., 2014; Petkus et al., 2013). However, a recent systematic review identified a paucity of psychometrically sound measures of psychological flexibility (Plys et al., 2022). The present study evaluated the factor structure of the CompACT and TOF scales in a sample of community-dwelling older adults, in the context of the pre-intervention data from an RCT. These findings provide preliminary support for the use of a short-form version of the CompACT to measure therapeutic processes of change in a community-dwelling older adult sample. Future work should seek to replicate these findings in a larger, more diverse sample of older adults.

Clinical Implications.

• Clinicians treating community-dwelling older adults should consider the reliability of Acceptance and Commitment Therapy process questionnaires in this population specifically.

• Based on these preliminary findings, clinicians treating community-dwelling older adults with Acceptance and Commitment Therapy may opt to use the short-form version of the CompACT to measure therapeutic process of change in openness, awareness, and engagement.

Data availability statement:

Data can be made available upon reasonable request.