Agreement Across Multiple Modes of Language Assessment in Minimally Verbal Children with Autism

Background:

Approximately 30-40% of children with autism remain minimally verbal past age 5, despite intense early intervention (Anderson et al., 2007).  These children are underrepresented in research, thus there is little information about their abilities and development.  This is partially due to difficulties inherent in assessing their skills, such as behavioral challenges and spoken language limitations (Kasari et al., submitted). 

Language is a primary concern with this population and is measured through direct assessment, observation and parent report.  Current recommendations are that a combination of modes be used to assess language in minimally verbal children with autism (Kasari et al., submitted).  However, little is known regarding relative accuracy and agreement across assessment modalities within this population. 

Objectives:

This study evaluates agreement among multiple measures of language ability in minimally verbal children with autism.  Parent report, direct assessment and observational measures were compared. 

Methods:

Participants included 60 minimally verbal children with autism (M age = 6.32 years) from a multi-site social-communication intervention study.  Mean developmental age was 4.01 years (Leiter-R age-equivalent). Participants had fewer than 20 words on a standardized language sample (M= 11.11).   Data for this study was taken from the entry timepoint.

            Parent report measures included Macarthur Bates Communication Inventory (MCDI; words produced, words understood scores) and Vineland Adaptive Behavior Scales (VABS; communication subscales).  Direct assessments included Peabody Picture Vocabulary Test (PPVT; receptive vocabulary) and Test of Early Language Development (TELD; expressive, receptive subscales).  Observational measures included a 20 min standardized language sample and data from the first two intervention sessions.  Language sample and intervention session videos were transcribed and analyzed using the SALT protocol (Miller & Iglesias, 2008).   The Number of Different Word Roots (NDWR) in each transcript was automatically calculated by SALT.

Results:

            Partial correlations were calculated controlling for the effects of developmental age.  Due to the number of correlations, p-values less than .01 were considered significant. Parent measures (MICDI, VABS) expressive scores were significantly correlated with NDWR from the language sample (r=.56, p<.001; r=.34, p=.007) and intervention sessions (r=.45, p=.004; r=.45, p=.002), but not TELD expressive scores (r=.29, p=.07; r=.33, p=.02).   Observational data (NDWR from Language Sample, intervention sessions) were significantly correlated with TELD expressive scores (r=.40, p=.002; r=.47, p<.001). Receptive skills reported by parents (MCDI, VABS) were significantly correlated with PPVT scores (r=.42, p=.008; r=.40, p=.005) but not TELD scores (r=.31, p=.06; r=.21, p=.16).  Assessments yielding age-equivalents were compared directly to each other.  PPVT scores were significantly higher than VABS (t=-3.12, p=.002) and TELD (t=7.27, p=.000) receptive.  TELD and VABS scores did not significantly differ for expressive (t=.68, p=.50) or receptive scores (t=-1.13, p=.27).

Conclusions:

Overall there was high agreement across assessment modalities for both receptive and expressive language.  Parent report reflected expressive language observed in a naturalistic context and observations were significantly correlated with direct assessment. PPVT scores correlated with parent report, and yielded the highest estimate of participant abilities. TELD scores were not correlated with parent report of expressive or receptive language, indicating that this assessment may not reflect abilities observed by parents.