Assessing measurement invariance in the EORTC QLQ-C30

Purpose

We aimed to investigate measurement invariance (MI) in the European Organisation for research and treatment of cancer quality of life questionnaire core 30 (EORTC QLQ-C30) in a heterogeneous sample of patients with cancer.

Methods

Data from 12 studies within the PROFILES registry were used for secondary analyses (n?=?7007). We tested MI by successive restrictions on thresholds, loadings, and intercepts across subgroups based on primary cancer sites, age, sex, time since diagnosis, and life stage, using multigroup confirmatory factor analysis (MGCFA) for ordered categorical measures. We also evaluated the impact of potentially miss-specified parameter equality across groups on latent factor means by releasing threshold and loading equality constraints for each item at a time.

Results

Results showed that the highest level of MI (invariance of thresholds, loadings, and intercepts) was found across groups based on time since diagnosis and life stage and to a lesser extent across groups based on sex, age, and primary tumor site. On item level, however, changes in the item’s associated factor means were relatively small and in most cases canceled each other out to some extent.

Conclusions

Given only a few instances of non-invariance in our study, there is reason to be confident that valid conclusions can be drawn from between-group comparisons of QLQ-C30 latent means as operationalized in our study. Nonetheless, further research into MI between other subgroups for the QLQ-C30 (i.e., treatment effects and ethnicity) is warranted. We stress the importance of including MI evaluations in the development and validation of measurement instruments.

     

Psychometric properties of the Center for Epidemiologic Studies Depression Scale (CES-D) in French clinical and nonclinical adults

BackgroundPrevious research on the Center for Epidemiologic Studies Depression Scale (CES-D) has five main limitations. First, no study provided evidence of the factorial equivalence of this instrument across samples of depressive and community participants. Second, only one study included systematic tests of measurement invariance based on confirmatory factor analyses (CFA), and this study did not consider the higher-order factor structure of depression, although it is the CES-D global scale score that is most often used in the context of epidemiological studies. Third, few studies investigated the screening properties of the CES-D in non-English-language samples and their results were inconsistent. Fourth, although the French version of the CES-D has been used in several previous studies, it has never been systematically validated among community and/or depressed adults. Finally, very few studies have taken into account the ordered-categorical nature of the CES-D answer scale. The purpose of the study reported herein was therefore to examine the construct validity (i.e., factorial, reliability, measurement invariance, latent mean invariance, convergence, and screening properties) of the CES-D in a French sample of depressed patients and community adults.MethodsA total sample of 469 participants, comprising 163 clinically depressed patients and 306 community adults, was involved in this study. The factorial validity, and the measurement and latent mean invariance of the CES-D across gender and clinical status, were verified through CFAs based on ordered-categorical items. Correlation and receiver operator characteristic curves were also used to test the convergent validity and screening properties of the CES-D.ResultsThe present results: (i) provided support for the factor validity and reliability of a second-order measurement model of depression based on responses to the CES-D items; (ii) revealed the full measurement invariance of the first- and second-order measurement models across gender; (iii) showed the partial strict measurement invariance (four uniquenesses had to be freely estimated, but the factor variance–covariance matrix also proved fully invariant) of the first-order factor model and the complete measurement invariance of the second-order model across patients and community adults; (iv) revealed a lack of latent mean invariance across gender and across clinical and community subsamples (with women and patients reporting higher scores on all subscales and on the full scale); (v) confirmed the convergent validity of the CES-D with measures of depression, self-esteem, anxiety, and hopelessness; and (vi) demonstrated the efficacy of the screening properties of this instrument among clinical and nonclinical adults.ConclusionThis instrument may be useful for assessing depressive symptoms or for the screening of depressive disorders in the context of epidemiological studies targeting French patients and community men and women with a background similar to those from the present study.     

The perceived constraints subscale of the Sense of Mastery Scale: dimensionality and measurement invariance

Purpose

A number of studies have reported differences in sense of mastery and perceived control across different subgroups. Yet, few have examined measurement invariance, an important prerequisite for valid comparisons. This study examines the factorial structure and measurement invariance of the perceived constraints (PC) facet of Pearlin and Schooler’s (1981) Sense of Mastery Scale (SM) which is a commonly used short form of the widely used SM scale.

Methods

Confirmatory factor analyses using AMOS and Mplus were conducted to explore dimensionality and test for measurement invariance in factor structure, factor loadings, intercepts, and residual variances across gender, age, education, income, and employment status in a large (N = 19,858), nationally representative sample of Norwegian males and females aged 16–100.

Results

The data supported a modified unidimensional model specifying correlations between the error terms of items 4 and 5, or possibly two highly correlated dimensions (r = 0.90). Metric invariance of the scale was shown for age, education, and employment, whereas invariance at the strong and strict levels was shown for gender and income. Partial invariance at the strong level was shown for age.

Conclusions

This Norwegian study supported a modified unidimensional structure for the abbreviated SM scale. Invariance testing indicated that comparisons across genders and income levels are unproblematic, whilst comparing mean scores across education and employment status is not justified. Latent, but not sum score means are comparable across age. Future studies using all 7 items of SM scale should provide more information on dimensionality and measurement invariance.

     

Measurement equivalence of the CES-D 8 in the general population in Belgium: a gender perspective

International research consistently finds gender differences in depression, but do women genuinely experience more complaints or are the findings contaminated by group-specific elements unrelated to depression but affecting its measurement? The study of gender differences in depression depends on the measurement quality of the instrument used to evaluate depression. In the present study we test the measurement equivalence of a shorter version of a commonly used instrument in mental health research, the Center for Epidemiologic Studies - Depression Scale (CES-D), using data from the Belgian sample of the third round of the European Social Survey (N = 1794). Evidence for measurement invariance can be established within the multigroup confirmatory factor analysis framework. This method allows us to evaluate a nested hierarchy of hypotheses to test different levels of cross-group measurement invariance: configural, metric, scalar and residual invariance, and clarifies under what conditions meaningful comparisons between the male and female respondents can be made. The best fitting factor model is then used to estimate the ''true'' prevalence of depressive symptoms for both groups. In our study measurement equivalence is established at all levels, indicating that the current depression scale allows defensible quantitative gender comparisons. Our data also confirm the epidemiological finding that women report more complaints of depression than men.     

Reference Values and Psychometric Properties of the Quality of Life After Traumatic Brain Injury-Overall Scale in Italy,The Netherlands,and the United Kingdom

ObjectivesThe Quality of Life after Brain Injury-Overall Scale (QOLIBRI-OS) is a short screening instrument for assessing disease-specific health-related quality of life (HRQoL) after traumatic brain injury. To date, no reference values are available for the QOLIBRI-OS in general populations. Thus, this study aimed to establish reference values for the QOLIBRI-OS in general population samples from Italy, The Netherlands, and the United Kingdom.MethodsData were collected using an online survey. The total sample comprised 11759 participants, consisting of 3549 Italian, 3564 Dutch, and 4646 British subjects. In this sample, 49% of the total sample did not report any health complaints, whereas 51% had at least 1 chronic health condition. Reference values were deduced for the QOLIBRI-OS for health-condition–related samples and total general population samples per country. To ensure the comparability of these values, measurement invariance was assessed using a multigroup confirmatory factor analysis. Covariates characterizing the reference values were selected with the help of regression analyses.ResultsThe confirmatory factor analysis confirmed that the QOLIBRI-OS scores measured the same traumatic brain injury–specific HRQoL construct across the 3 countries. Healthy individuals reported significantly higher HRQoL than individuals with at least 1 chronic health condition. Older age and higher education levels were significantly associated with higher HRQoL.ConclusionsBecause the reference values displayed differences in terms of age and education level across the 3 countries, we recommend using country-specific reference values stratified by sociodemographic and health status in research and clinical practice.     

Testing gender invariance of the Hospital Anxiety and Depression Scale using the classical approach and Bayesian approach

Purpose

Measurement invariance is an important attribute for the Hospital Anxiety and Depression Scale (HADS). Most of the confirmatory factor analysis studies on the HADS adopt the classical maximum likelihood approach. The restrictive assumptions of exact-zero cross-loadings and residual correlations in the classical approach can lead to inadequate model fit and biased parameter estimates. The present study adopted both the classical approach and the alternative Bayesian approach to examine the measurement and structural invariance of the HADS across gender.

Methods

A Chinese sample of 326 males and 427 females was used to examine the two-factor model of the HADS across gender. Configural and scalar invariance of the HADS were evaluated using the classical approach with the robust-weighted least-square estimator and the Bayesian approach with zero-mean, small-variance informative priors to cross-loadings and residual correlations.

Results

Acceptable and excellent model fits were found for the two-factor model under the classical and Bayesian approaches, respectively. The two-factor model displayed scalar invariance across gender using both approaches. In terms of structural invariance, females showed a significantly higher mean in the anxiety factor than males under both approaches.

Conclusion

The HADS demonstrated measurement invariance across gender and appears to be a well-developed instrument for assessment of anxiety and depression. The Bayesian approach is an alternative and flexible tool that could be used in future invariance studies.     

Use of factor analysis models to evaluate measurement invariance property of the Asthma Control Questionnaire (ACQ)

Purposes

To demonstrate the assessment of measurement invariance property in a health status instrument and to increase the awareness of its importance, we evaluate the measurement invariance of the Asthma Control Questionnaire (ACQ) across age and gender subgroups.

Methods

Data are obtained from children 7–12 years of age at entry into a randomized trial, which evaluates the effect of a telephone coaching program on improving asthma outcome. Multi-group confirmatory factor analysis is used to assess the comparability of factor loadings and intercepts across age and gender subgroups. Since age is a continuous variable, two different categorizations (7–10 vs 11–12 and 7–9 vs 10–12) are analyzed.

Results

The factor loadings and intercepts of all six items in ACQ are comparable across gender subgroups. Although the factor loadings are comparable across age 7–10 and 11–12 subgroups, one intercept is statistically but not practically different. For age 7–9 versus 10–12 subgroup comparison, the factor loadings are not comparable.

Conclusion

In children, the ACQ can be used to compare asthma control construct between boys and girls and between age 7–10 and 11–12 subgroups. Measurement invariance is an important property that should be examined when the latent construct(s) are compared across different subgroups.     

Factorial invariance of pediatric patient self-reported fatigue across age and gender: a multigroup confirmatory factor analysis approach utilizing the PedsQL™ Multidimensional Fatigue Scale

Objectives

In order to compare multidimensional fatigue research findings across age and gender subpopulations, it is important to demonstrate measurement invariance, that is, that the items from an instrument have equivalent meaning across the groups studied. This study examined the factorial invariance of the 18-item PedsQL? Multidimensional Fatigue Scale items across age and gender and tested a bifactor model.

Methods

Multigroup confirmatory factor analysis (MG-CFA) was performed specifying a three-factor model across three age groups (5–7, 8–12, and 13–18 years) and gender. MG-CFA models were proposed in order to compare the factor structure, metric, scalar, and error variance across age groups and gender. The analyses were based on 837 children and adolescents recruited from general pediatric clinics, subspecialty clinics, and hospitals in which children were being seen for well-child checks, mild acute illness, or chronic illness care.

Results

A bifactor model of the items with one general factor influencing all the items and three domain-specific factors representing the General, Sleep/Rest, and Cognitive Fatigue domains fit the data better than oblique factor models. Based on the multiple measures of model fit, configural, metric, and scalar invariance were found for almost all items across the age and gender groups, as was invariance in the factor covariances. The PedsQL? Multidimensional Fatigue Scale demonstrated strict factorial invariance for child and adolescent self-report across gender and strong factorial invariance across age subpopulations.

Conclusions

The findings support an equivalent three-factor structure across the age and gender groups studied. Based on these data, it can be concluded that pediatric patients across the groups interpreted the items in a similar manner regardless of their age or gender, supporting the multidimensional factor structure interpretation of the PedsQL? Multidimensional Fatigue Scale.     

Commentary on strengthening the assessment of factorial invariance across population subgroups: a commentary on Varni et al. (2013)

In his commentary on Varni et al.’s (Qual Life Res. doi:10.1007/s11136-013-0370-4, 2013) article, McIntosh (Qual Life Res. doi:10.1007/s11136-013-0465-y, 2013) has two main arguments. First, we should have paid more attention to statistical tests (i.e., χ ² values) instead of approximate fit indexes for our analysis, especially with the baseline model. Second, Bayesian methods are better than the frequentist methods we used in determining the model’s invariance across age and gender groups. We believe that statistical tests do have a place in assessing model fit, but overemphasis on them, especially with larger sample sizes, can lead to errant decisions. Second, while we agree that Bayesian methods have the potential to contribute much to the field of assessing invariance, more development needs to be conducted before they can be widely utilized in assessing factorial invariance across groups.     

Factor structure for the frailty syndrome was consistent across Europe

ObjectivesWe explored the measurement properties of frailty with the goal of optimizing frailty assessment according to phenotype definition of Fried and comparing measurement properties across countries.Study Design and SettingData are from the Survey of Health, Ageing and Retirement in Europe (n = 27,938), a population-based study of community-dwelling adults aged ≥50 years. Frailty was specified as a unidimensional construct, and measurement invariance across the 12 countries was tested. To assess our measurement model, we used confirmatory factor analysis (CFA) and multigroup CFA to assess measurement invariance.ResultsThe unidimensional model fit the data well (adjusted χ²(48) = 82.74, P = 0.001), and the same structure was satisfactory for all countries. Inclusion of equality constraints led to significant model deterioration (adjusted χ²diff(88) = 995.05, P < 0.001), suggesting differences in parameters across countries. Spain was removed from further analyses, and equality constraints for Greece, Sweden, Israel, Italy, and France were not tenable. Accounting for these led to satisfactory model fit (adjusted χ²(113) = 414.33, P < 0.001). Significant mean frailty differences were identified.ConclusionThe relationships between the construct of frailty and indicators, although broadly constant, do vary across some countries. Furthermore, there was evidence of differing levels of frailty for the middle-aged and older populations across European countries.     

Exploring measurement invariance by gender in the profile of mood states depression subscale among cancer survivors

Purpose

The Profile of Mood States-Short Form (POMS-SF) is a well-validated tool commonly used in medical/clinical research. Less attention has been paid to the measurement invariance of the POMS—the degree to which the structure and items behave similarly for different groups (e.g., women and men). This study investigated the measurement invariance of the POMS Depression subscale across gender groups in a sample of cancer survivors.

Methods

The POMS Depression subscale has 8 items (Unhappy, Sad, Blue, Hopeless, Discouraged, Miserable, Helpless, and Worthless). Invariance was measured using multigroup confirmatory factor analysis. This study used data from American Cancer Society Studies of Cancer Survivors-II, a population-based survey of adult cancer survivors (n = 9170).

Results

We found factor structures and factor loadings were invariant for gender groups, but moderate differential item functioning (DIF) in the question containing the word blue.

Conclusion

With regard to cancer survivors’ gender, we found the Depression subscale of the POMS-SF had configural invariance, and partial metric and scalar invariance. This suggests that results should be interpreted with caution, especially when gender is considered important. More broadly, our finding suggests that questions with the word blue may introduce DIF into other measures of depressive mood. More research is needed to replicate these findings in other samples and with other instruments.

     

Confirmatory Factor Analysis and Sample Invariance of the Chinese Patient Satisfaction Questionnaire (ChPSQ-9) among Patients with Breast and Lung Cancer

Objectives:  Previous exploratory factor analysis of the 9-item Chinese Patient Satisfaction Questionnaire (ChPSQ-9) identified two dominant factors: doctor and nurse. The present study employed confirmatory factor analysis (CFA) to examine the factorial invariance of the ChPSQ-9 between and within samples of Chinese patients with breast or lung cancer.
Methods:  Longitudinal data were analyzed from Chinese breast and lung cancer patients who had completed the ChPSQ-9 during their first outpatient visit, at 3 months, and at 6 months after baseline. CFAs tested the fit of a one-factor model, a hierarchical model that comprised a general latent factor and two first-order factors, and a correlated model that comprised two correlated first-order factors to the data. The factorial invariance of the ChPSQ-9 between six independent samples across time was investigated using multigroup CFAs.
Results:  The CFA's results demonstrated a better fit of the correlated model over the one-factor model and the hierarchical model in the breast and lung cancer samples. The correlated model showed evidence of cross-sample and longitudinal factorial invariance. Patients were generally satisfied with services provided by doctors and nurses. Internal consistency of the scale was also good for both cancer samples across time.
Conclusions:  The ChPSQ-9 is a valid and reliable instrument to be employed among breast and lung cancer patients, in clinical settings or intervention research, to evaluate group differences in patient satisfaction and its association with intervention effectiveness.     

Factorial Invariance of Child Self-Report across Age Subgroups: A Confirmatory Factor Analysis of Ages 5 to 16 Years Utilizing the PedsQL 4.0 Generic Core Scales

Background:  The utilization of health-related quality of life (HRQOL) measurement in an effort to improve pediatric health and well-being and determine the value of health care services has grown dramatically over the past decade. The paradigm shift toward patient-reported outcomes (PROs) in clinical trials has provided the opportunity to emphasize the value and essential need for pediatric patient self-report. In order for HRQOL/PRO comparisons to be meaningful for subgroup analyses, it is essential to demonstrate factorial invariance. This study examined age subgroup factorial invariance of child self-report for ages 5 to 16 years on more than 8500 children utilizing the PedsQL™ 4.0 Generic Core Scales.
Method:  Multigroup Confirmatory Factor Analysis (MGCFA) was performed specifying a five-factor model. Two multigroup structural equation models, one with constrained parameters and the other with unconstrained parameters, were proposed to compare the factor loadings across the age subgroups.
Results:  Metric invariance (i.e., equal factor loadings) across the age subgroups was demonstrated based on stability of the Comparative Fit Index between the two models, and several additional indices of practical fit including the Root Mean Squared Error of Approximation, the Non-Normed Fit Index, and the Parsimony Normed Fit Index.
Conclusion:  The findings support an equivalent five-factor structure across the age subgroups. Based on these data, it can be concluded that children across the age subgroups in this study interpreted items on the PedsQL 4.0 Generic Core Scales in a similar manner regardless of their age.     

Assessment of the Dermatology Life Quality Index (DLQI) in a homogeneous population under lifetime arsenic exposure

Purpose

The psychometric property of the Dermatology Life Quality Index (DLQI) is underappreciated in public health settings. Our study aimed to assess the reliability, validity, and measurement invariance of DLQI in a homogeneous population with arsenic-related skin lesions and symptoms.

Methods

A cross-sectional study was conducted in communities under lifetime arsenic exposure. The DLQI was measured through a face-to-face interview. Skin examinations were performed by certificated dermatologists. The intensity of itching was measured by a numerical rating scale. Reliability, structural validity, and measurement invariance were determined using classical and modern test theories, including confirmatory factor analysis and item response models.

Results

465 participants with arsenic-related skin lesions and symptoms completed the DLQI assessment. The Cronbach’s alpha was 0.79, and the split-half reliability was 0.77. A two-factor model exhibited the best model fit among models evaluated, but local dependencies among items were identified. The model showed good root mean square error of approximation (0.031) and acceptable Tucker–Lewis index (0.92). Multi-group confirmatory factor analysis showed no measurement invariance across subgroups of age, gender, ethnicity, and intensity of itching.

Conclusions

The DLQI had acceptable psychometric properties, but measurement invariance was not observed across different groups of participants.

     

Construct validity of a revised version of the Occupational Balance Questionnaire

Abstract

Background: Occupational balance is fundamental to occupational therapy and occupational science. Therefore, the Occupational Balance Questionnaire (OBQ) was developed and has previously been found psychometrically valid according to classical test theory.

Aim: To investigate the internal construct validity of the OBQ using Rasch measurement theory.

Material and methods: Data from two general population samples were used to investigate the psychometric properties of the OBQ according to Rasch measurement theory.

Results: The analyses identified problems with the current response scale and multidimensionality of two items. As a result, a revised version, the OBQ11, was suggested and exhibited response categories that worked properly, good reliability (0.92), model fit and measurement invariance across age and gender groups. The hierarchical item ordering was in agreement with previous research.

Conclusion: The new OBQ11 satisfies the measurement criteria defined by the Rasch model. However, further studies of additional samples are needed to validate its generic properties.

Significance: The purpose of the OBQ11 is to measure occupational balance of individuals or groups, and to identify aspects of occupational balance in need of improvement. Our observations suggest that the OBQ11 is a valid and promising complement to other instruments.     

Factorial validity of the effort–reward imbalance scale: evidence from multi-sample and three-wave follow-up studies

Purpose

The present study investigated whether the factor structure of the effort–reward imbalance (ERI) scale (Siegrist et al. Soc Sci Med 58:1483–1499, 2004) remains same across two white-collar samples (i.e., factorial group invariance) and across three measurement times (i.e., factorial time invariance).

Methods

The factorial group invariance was tested using two different samples including 1,301 managers and 758 young white-collar professionals. The factorial time invariance was tested in the latter sample with a four-year three-wave follow-up design.

Results

The confirmatory factor analysis performed supported the theoretically based structure of the ERI scale, that is, the scale included two first-order factors of effort and overcommitment and one second-order factor of reward with first-order factors of esteem, career opportunities, and job security. The factorial group invariance of this structure was also supported. In addition, the factor loadings of all factors remained same across three measurements; thus, lending support for the factorial time invariance.

Conclusion

The ERI scale was found to be a valid tool to measure costs and gains of social exchanges at work as well as individual orientation toward work in these two occupational samples including a longitudinal study design.     

Testing the Invariance of the National Health and Nutrition Examination Survey’s Sexual Behavior Questionnaire Across Gender,Ethnicity/Race,and Generation

Federal and state policies are based on data from surveys that examine sexual-related cognitions and behaviors through self-reports of attitudes and actions. No study has yet examined their factorial invariance—specifically, whether the relationship between items assessing sexual behavior and their underlying construct differ depending on gender, ethnicity/race, or age. This study examined the factor structure of four items from the sexual behavior questionnaire part of the National Health and Nutrition Examination Survey (NHANES). As NHANES provided different versions of the survey per gender, invariance was tested across gender to determine whether subsequent tests across ethnicity/race and generation could be done across gender. Items were not invariant across gender groups so data files for women and men were not collapsed. Across ethnicity/race for both genders, and across generation for women, items were configurally invariant, and exhibited metric invariance across Latino/Latina and Black participants for both genders. Across generation for men, the configural invariance model could not be identified so the baseline models were examined. The four item one factor model fit well for the Millennial and GenerationX groups but was a poor fit for the baby boomer and silent generation groups, suggesting that gender moderated the invariance across generation. Thus, comparisons between ethnic/racial and generational groups should not be made between the genders or even within gender. Findings highlight the need for programs and interventions that promote a more inclusive definition of “having had sex.”     

Hypersexuality,Gender, and Sexual Orientation: A Large-Scale Psychometric Survey Study

Criteria for hypersexual disorder (HD) were proposed for consideration in the DSM-5 but ultimately excluded for a variety of reasons. Regardless, research continues to investigate hypersexual behavior (HB). The Hypersexual Behavior Inventory (HBI) is one of the most robust scales assessing HB, but further examination is needed to explore its psychometric properties among different groups. Therefore, the aim of the present study was to examine the generalizability of the HBI in a large, diverse, non-clinical sample (N?=?18,034 participants; females?=?6132; 34.0%; M_age?=?33.6 years, SD_age?=?11.1) across both gender and sexual orientation. Measurement invariance testing was carried out to ensure gender- and sexual orientation-based comparisons were meaningful. Results demonstrated when both gender and sexual orientation were considered (i.e., heterosexual males vs. LGBTQ males vs. heterosexual females vs. LGBTQ females), LGBTQ males had significantly higher latent means on the HBI factors. Results also demonstrated LGBTQ males had the highest scores on other possible indicators of hypersexuality (e.g., frequency of masturbation, number of sexual partners, or frequency of pornography viewing). These findings suggest LGBTQ males may be a group most at risk of engaging in hypersexual behavior, and LGBTQ females are at a higher risk of engaging in hypersexual activities due to coping problems. Given the large-scale nature of the study, the findings contribute to the currently growing body of the literature on hypersexuality.     

Measurement invariance of the PROMIS pain interference item bank across community and clinical samples

Purpose

This study examined the measurement invariance of responses to the patient-reported outcomes measurement information system (PROMIS) pain interference (PI) item bank. The original PROMIS calibration sample (Wave I) was augmented with a sample of persons recruited from the American Chronic Pain Association (ACPA) to increase the number of participants reporting higher levels of pain. Establishing measurement invariance of an item bank is essential for the valid interpretation of group differences in the latent concept being measured.

Methods

Multi-group confirmatory factor analysis (MG-CFA) was used to evaluate successive levels of measurement invariance: configural, metric, and scalar invariance.

Results

Support was found for configural and metric invariance of the PROMIS-PI, but not for scalar invariance.

Conclusions and recommendations

Based on our results of MG-CFA, we recommend retaining the original parameter estimates obtained by combining the community sample of Wave I and ACPA participants. Future studies should extend this study by examining measurement equivalence in an item response theory framework such as differential item functioning analysis.     

Longitudinal measurement invariance of the Dutch Outcome Questionnaire-45 in a clinical sample

Purpose

In the absence of measurement invariance across measurement occasions, change scores based on pretest–posttest measurements may be inaccurate representations of real change on the latent variable. In this study, we examined whether measurement invariance held in the Dutch version of Outcome Questionnaire-45 (OQ-45).

Method

Using secondary data analysis of a sample of N?=?540 Dutch outpatients, we tested the stability of the factorial structure (gamma change) and the metric and scalar invariance (beta change) across pretest and posttest measurements using a combination of factor analysis and item response theory methodology.

Results

Results revealed a stable factorial structure from pretest to posttest and minor violations of metric invariance for two items in the Dutch OQ-45.

Conclusion

Even though for two items the assumption of invariance was violated, results suggest that the effects of these violations on practical change assessment using the OQ-45 were negligible.