# Measuring functioning among youth using the Columbia impairment scale: investigating dimensionality and measurement invariance among 14–17 year olds using mental health services and their caregivers

## TL;DR

This study examines the structure and reliability of the Columbia Impairment Scale in measuring functioning among youth and their caregivers.

## Contribution

The study provides new evidence on the dimensional structure and measurement invariance of the CIS in youth and caregiver reports.

## Key findings

- A three-factor model (work/school, home/family, socializing) best fits the CIS with notable cross-loadings.
- Full metric invariance was found between youth and caregivers, but scalar invariance was not confirmed.
- The results suggest caution in interpreting global CIS scores due to cross-loadings and lack of scalar invariance.

## Abstract

Despite being a widely used and recommended measure of functioning, the Columbia Impairment Scale (CIS) lacks consensus on scale structure and whether child- and parent-report versions measure the same construct(s). This study aimed to better understand the structure and test for measurement invariance across groups of youth and their caregivers.

The sample included youth 14–17 years of age accessing mental health services, and their caregiver (most often mother), recruited from one of five mental health outpatient hospital sites in Toronto, Canada between September 2016 and March 2020. Exploratory Structural Equation Modeling (ESEM) was used to investigate dimensionality and test for measurement invariance using standard model fit statistics.

A total of 189 youth-caregiver dyads were included in the analysis. Youth were on average aged 15.7 (sd = 1.1); 64% were female. Caregivers had a mean age of 48.2 (sd = 7.4) and were 87% mothers. Using ESEM, evidence of a three-factor model was found (“work/school”, “home/family” and “socializing”), which included several, large conceptually relevant cross-loadings. Using this model, full metric invariance between youth and caregivers was established, but strong evidence of scalar invariance was not found.

While a multi-dimensional model provided the best fit for the CIS, the presence of several large cross-loadings calls into question whether and how the global scale can best be used in clinical and research settings. Lack of evidence of scalar invariance suggests that multi-informant data should be interpreted carefully. Next steps should include testing for essential unidimensionality.