Validation of the Malay Version of the Cognitive Fusion Questionnaire-7 (CFQ-7-M)

1.1. Background

Cognitive fusion plays a key role in the development of psychological inflexibility. As such, it is defined as a person’s likelihood of becoming trapped or entangled in their own thoughts. A person with strong cognitive fusion is more vulnerable to mental health issues such as depression, anxiety, and even stress. Although the degree of outcome varies, in some severe cases, cognitive fusion can affect a person’s behavior and perception negatively [1-3]. People with high levels of cognitive fusion not only face increased vulnerability to emotional distress [4], but they also often struggle with maladaptive coping strategies, which further complicate the prognosis [5]. This is because individuals with a high level of cognitive fusion firmly believe and treat their internal experiences as the literal truth. However, cognitive fusion is not a fixed trait; rather, it is a dynamic and modifiable psychological process within the ACT framework. The framework involves six main processes, including acceptance, cognitive defusion, present-moment awareness, self-as-context, values, and committed action. ACT is a psychological intervention that aims to increase psychological flexibility, which refers to the ability of a person to stay present even in the moment of difficult experiences and continue to take meaningful actions based on their values in life. Within this framework, cognitive fusion is important because individuals who are highly fused with their thoughts may respond to those thoughts as if they are absolute truths, which can restrict flexible and values-based behaviour. Given that cognitive fusion is the key element of psychological inflexibility, the main therapeutic target within the ACT framework is to reduce it. Reducing cognitive fusion may help individuals relate to their thoughts more flexibly, allowing them to choose behaviors that are guided by personal values rather than being controlled by distressing or self-limiting thoughts [6, 7]. The CFQ-7 questionnaire is therefore useful in ACT-related research and practice because it's used to measure cognitive fusion, allowing researchers and clinicians to assess this specific process and monitor its changes following psychological interventions.

To enable the measurement of this construct, Gillanders and colleagues (2014) developed the Cognitive Fusion Questionnaire (CFQ-7) with seven items as a brief, unidimensional, psychometrically validated self-report tool [8]. Since then, the CFQ-7 has been adapted into different languages (e.g., Chinese, Spanish, Hungarian) across the world, showing excellent internal consistency and construct validity [9]. Among the studies, the CFQ-7 also revealed strong correlations with other related constructs such as experiential avoidance, depression, anxiety, and stress [10, 11].

Despite the growing importance and global utilization of CFQ-7 to measure cognitive fusion, there is no validated version of the CFQ-7 available in Malay or Bahasa Malaysia currently. There is a great need, considering the increasing utilization of ACT-based interventions across Malaysia in both clinical and subclinical studies [12, 13]. A culturally adapted and linguistically validated cognitive fusion questionnaire, such as CFQ-7-M, would enhance both clinicians' and researchers’ measurement of cognitive processes by accurately identifying the underlying problem and changes, especially during the intervention. Moreover, strong collectivist values and cultural norms in Malaysia may shape emotional expression and thought processes unique to the local context. As a result, the evaluation of the theoretical integrity and psychometric performance of the CFQ-7 is crucial to ensure its applicability in the Malaysian setting. Without such validation, the use of CFQ-7 in research and clinical assessments risks conceptual misunderstanding and measurement bias in the local context.

To close the gap, this study’s main objectivewas to translate, adapt, and validate the Malay version of the CFQ-7 (CFQ-7-M). The use of university students was considered appropriate for this initial translation and psychometric validation because the primary objective was to evaluate linguistic clarity, cultural comprehensibility, internal consistency, factor structure, and convergent validity of the CFQ-7-M. Additionally, cognitive fusion is relevant not only to clinical populations but also to non-clinical individuals who may experience varying levels of psychological inflexibility and distress [14]. Furthermore, this study was not designed to examine clinical severity, diagnostic accuracy, or cut-off scores. Therefore, university students provide a suitable starting point for examining whether the Malay version retains the expected conceptual and psychometric properties.

The specific objectives of the study were to examine the item-level distribution of the CFQ-7-M and assess normality assumptions through descriptive statistics, skewness, and kurtosis. The study also aims to evaluate the internal consistency reliability by using Cronbach’s alpha, McDonald’s omega, and the Greatest Lower Bound (GLB) and to assess construct validity of the unidimensional model of CFQ-7-M using Confirmatory Factor Analysis (CFA) with established fit indices, besides establish convergent validity by examining the relationships between CFQ-7-M and theoretically related constructs, including psychological inflexibility (AAQ-II-M) and emotional distress (DASS-21 subscales).

By achieving these objectives, the study contributes to the growing body of ACT-related assessment tools in Malaysia, particularly in measuring cognitive fusion among Malay-speaking populations, supporting the culturally appropriate application of evidence-based psychological interventions in the local context.

2.1. Study Design

This is a cross-sectional instrument validation study conducted among UMS students. The process flow of the CFQ-7-M translation, pre-test, and psychometric validation process is described in the following sections and presented in Fig. (1). Permission to translate and use the original CFQ-7 was obtained from the original author of CFQ-7, Dr. David T. Gillanders, and ethical clearance was obtained from the UMS Medical Research Ethics Committee. Informed consent was also obtained from all participants following a full explanation of the study procedures.

2.2. Translation Phase

The CFQ-7-M was developed through a rigorous forward–backward translation process based on the US Census Bureau's guidelines [15]. The translation process was commenced on 14^th and 15^th April 2025 by six groups of UMS postgraduate students independently, each with five translators, all of whom demonstrated proficiency in English and Malay at the Sijil Pelajaran Malaysia (SPM) level (the ability to read and understand Bahasa Malaysia at the national secondary school examination standard). Then, the process was followed by a reconciliation meeting on 16^th April 2025 involving all translators and the research team to compare, discuss, and further refine each item until consensus was reached on the most acceptable items with conceptual and cultural similarity. The considerable similarities between the original CFQ-7 and the back-translated versions showed high translation quality and produced a harmonized CFQ-7-M.

Subsequently, face validity was assessed during the pilot-testing phase on 17^th April 2025, whereby 30 UMS students with a minimum SPM level of Bahasa Malaysia competency participated in a pilot test of the harmonized CFQ-7-M to assess its clarity, understanding, and cultural appropriateness. During the process, participants were asked to provide input on the items' understandability and identify any unclear phrases. Participants also reviewed each item for grammatical structure, subject-verb-object (SVO) order, and sentence flow to ensure clarity and preserve the original meaning of the questionnaire while removing any potential defects that might affect overall comprehension. As a result of this stage, the Malay version of CFQ-7 (CFQ-7-M) was finalized. The final Malay CFQ-7-M items, alongside the original English items, are provided in the Appendix.

2.3. Validation Study

This study employed a cross-sectional design to validate the CFQ-7-M among 220 UMS students from 19^th June to 11th July 2025. Convenience sampling was used to recruit participants. Participants were eligible for inclusion if they were aged 18 years or older, enrolled as students at UMS, able to read and understand Malay, and provided informed consent. The participants then completed a self-administered questionnaire through online Google Forms, comprising sections for sociodemographic information, the CFQ-7-M, the Malay version of the Acceptance and Action Questionnaire-II (AAQ-II) [16], and the Malay version of the Depression, Anxiety, and Stress Scale-21 (DASS-21) [17]. The data collected were noted to have minimal missing data.

There was only one missing observation (0.45%) for the length of work (years), while all other variables, including all CFQ-7-M items, had no missing observations. Therefore, CFA and convergent validity analyses were conducted on the full sample (N=220). The sample size was considered adequate for CFA-based validation of a 7-item scale [18].

Descriptive statistics were used to determine the item distributions and skewness. Besides that, the Kurtosis values were checked to determine the normality of the data. Due to the non-normal distribution of the data, non-parametric analyses were eventually performed. Cronbach’s alpha, McDonald’s omega, and the Greatest Lower Bound (GLB) were then used to examine the internal consistency reliability. Since the CFQ-7 has a well-established unidimensional structure, this study chose confirmatory testing of the hypothesized one-factor model over the structure re-exploration with exploratory factor analysis (EFA). The confirmatory factor analysis (CFA) was conducted by treating CFQ-7-M items as ordered categorical indicators using the weighted least squares mean and variance adjusted (WLSMV) estimator (theta parameterisation). Model fit was further assessed using the comparative fit index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and standardised root mean square residual (SRMR). Convergent validity was assessed using both Spearman and Pearson correlation coefficients. Spearman correlations were used because the CFQ-7-M and validating measures were derived from Likert-type responses, and some variables showed non-normal distributions. Pearson correlations were also reported to support comparability with previous validation studies and to ensure the findings were consistent across correlation methods.

2.4. Measurement Invariance

Multi-group CFA (MG-CFA) with ordered categorical indicators and WLSMV estimation (theta parameterisation) was used to investigate measurement invariance across gender. This step is critical to ensure that the CFQ-7-M measures the same construct for men and women. Configural, metric (equal loadings), and scalar (equal loadings and thresholds) models were fitted, and changes in approximate fit indices were used to evaluate invariance. As a result, ΔCFI ≤ 0.010 and ΔRMSEA ≤ 0.015 were found to support invariance across groups. Furthermore, the two highest response categories in CFQ-7-M were collapsed (7 to 6) in the invariance analyses to account for sparse response categories in the smaller male sample.

2.5. Selection of Respondents

The sampling frame for this initial validation study included undergraduate and postgraduate students at Universiti Malaysia Sabah who were aged 18 years with the ability to read and understand Bahasa Malaysia at the SPM level. Participants were recruited using convenience sampling. Before data collection, all participants were provided with information about the study objectives and methods, as well as an informed consent form. Participation was solely voluntary, and the respondents’ confidentiality was maintained throughout the process.

2.6. Inclusion and Exclusion Criteria

Participants were eligible for inclusion if they were UMS students aged between 18 and 60 years, possessing SPM-level competency in both English and Bahasa Malaysia, and provided informed consent to participate in the study. Participants were excluded if they did not provide informed consent, had cognitive impairment, or had communication disabilities that could affect their ability to understand or complete the questionnaire.

3.1. Descriptive Analysis

Table 1 shows the sociodemographic characteristics of the overall participants. A total of 220 respondents with a mean age of 23 years, took part in the study . Amongst the participants, 58.6% were aged between 21 and 25 years, followed by 29.5% aged 20 years or younger, and only 11.8% were 26 years or above. In terms of gender, most participants were female (79.1%), and a large proportion of them were still single (92.3%), with only a small number married (7.3%) or divorced (0.5%). As for their educational background, most of the participants held a Bachelor's degree (75.0%), followed by those with a Master's degree (8.6%), a Diploma (10.0%), a Doctorate (4.5%), and a High School qualification (1.8%). Meanwhile, for work experience, the majority of participants had less than one year of working experience or were still students (78.2%), followed by 13.2% of them having more than four years of working experience, while another 4.5% reported they only had one to two years, and 4.1% had three to four years of working experience.

In Table 2, the descriptive statistics of the seven items in CFQ-7-M demonstrated moderate variability in responses as the mean scores ranged from 3.03 (Item 2) to 3.48 (Item 1), with standard deviations between 1.57 and 1.84. Additionally, all items showed slight positive skewness (ranging from 0.24 to 0.54), which suggested that more respondents had a tendency to pick the lower scores. This tendency is expected in a generally healthy or non-clinical university student population, as the majority of individuals are likely to report lower levels of cognitive fusion. Meanwhile, kurtosis values ranging from -0.42 to -0.95 suggested slightly flatter distributions than normal. However, the response range for all items was noted to be consistent, given the minimum and maximum values of 1 and 7, respectively, indicating the utilization of the full-scale range by participants. In summary, these results indicated an acceptable symmetric distribution with reasonable variability across all items.

3.2. Reliability Analysis

As shown in Table 3, the internal consistency of the CFQ-7 Malay version was excellent, as indicated by the results of the reliability analysis (McDonald’s ω = 0.937, Cronbach’s α = 0.936, and Greatest Lower Bound (GLB) = 0.956). Furthermore, Table 4 showed that individual item-rest correlations ranged from 0.724 to 0.856, which exceeded the recommended threshold of 0.30, indicating that all items had contributed meaningfully to the scale’s total score and were believed to be strongly related to the underlying construct. Overall, these results support the reliability and consistency of the CFQ-7-M in measuring cognitive fusion.

3.3. Construct Validity (Confirmatory Factor Analysis)

The items were treated as ordered categorical indicators and estimated using WLSMV (theta parameterisation) during the CFA. As shown in Table 5, the one-factor CFQ-7-M model had high incremental fit and minimal residual misfit (χ²(14) = 46.187, p < .001; CFI = 0.9977; TLI = 0.9966; SRMR = 0.0387). Although RMSEA was elevated at 0.102, it is interpreted cautiously because, for short, low-degree-of-freedom models such as CFQ-7-M, RMSEA can be inflated [19, 20]. Therefore, emphasis is placed on CFI, TLI, SRMR, and the magnitude of the factor loadings instead. Standardised factor loadings were uniformly strong, with a range between 0.796 and 0.901, supporting a robust unidimensional structure of the CFQ-7-M.

3.4. Convergent Validity

Table 6 shows the evidence of convergent validity by using Spearman’s rank correlation as the primary analysis due to its appropriateness for ordinal Likert-type responses. Meanwhile, Pearson’s correlation is reported for completeness and to ensure robustness across correlation methods. As a result, the CFQ-7-M’s total score was noted to be strongly associated with AAQ-II-M (Spearman ρ = 0.788; Pearson r = 0.788; all p < 0.001). Besides that, the CFQ-7-M also shows moderate-to-strong positive associations with psychological distress as measured by DASS-21 depression (ρ = 0.616; r = 0.592), anxiety (ρ = 0.569; r = 0.547), and stress (ρ = 0.694; r = 0.705), all p < 0.001. The comparable Spearman and Pearson coefficients suggest that convergent validity is robust across the correlation methods for this study.

3.5. Measurement Invariance across Gender

Measurement invariance across gender was supported as MG-CFA indicated configural invariance (the same one-factor structure across males and females). Metric invariance was supported (ΔCFI = −0.00084; ΔRMSEA = +0.00565), and scalar invariance was also supported (ΔCFI = +0.00164; ΔRMSEA = −0.05208), indicating that CFQ-7-M item thresholds and loadings function similarly across gender.

Table 7 shows configural, metric, and scalar invariance testing across gender (Female = 174; Male = 46) using MG-CFA with WLSMV (theta parameterisation) and ordered indicators. To account for sparse response categories in the smaller male subgroup (an unobserved category in one item), the two highest response options were collapsed for invariance testing (7 recoded to 6) to ensure stable threshold estimation. The configural model indicated the same one-factor structure across groups (CFI = 0.9976; TLI = 0.9965; RMSEA = 0.1053; SRMR = 0.0442). Constraining loadings supported metric invariance (CFI = 0.9967; TLI = 0.9965; RMSEA = 0.1109; SRMR = 0.0542), with negligible changes in fit (ΔCFI = −0.000842; ΔRMSEA = +0.005646). Adding equality constraints on thresholds supported scalar invariance (CFI = 0.9984; TLI = 0.9984; RMSEA = 0.0588; SRMR = 0.0493), again showing minimal change (ΔCFI = +0.001641; ΔRMSEA = −0.052077). Overall, our findings show that CFQ-7-M item functioning is consistent across gender at the level of factor loadings and thresholds.