The Multiple Measure Approach to Diagnostics: Examining the Complexity and Stability of Autism Symptomatology

Poster Presentation
Thursday, May 10, 2018: 11:30 AM-1:30 PM
Hall Grote Zaal (de Doelen ICC Rotterdam)
M. Roberts1, Y. S. Stern1 and L. H. Hampton2, (1)Communication Sciences and Disorders, Northwestern University, Evanston, IL, (2)Northwestern University, Evanston, IL
Background: Despite the US Preventive Services Task Force recommendation to avoid early autism screening if concerns have not been raised by a parent or clinician (Siu & USPSTF, 2016), there is sufficient evidence that early autism screenings are beneficial to not only ensure earlier access to diagnosis for children from diverse backgrounds (Veenstra-VanderWeele & McGuire, 2016), but to also facilitate access to early interventions that can provide long-term benefits (Dawson, 2016). Any one autism screening measure does not result in optimal sensitivity or specificity, but recent studies have demonstrated that by combining screening measures we may be able to add positive predictive value and reduce stress and unnecessary assessments for children who receive false positive screenings (Khowaja, Robins, & Adamson).

Objectives: What is the positive predictive value of two common autism screening measures (STAT, MCHAT) (1) alone? (2) when either results in a positive? And (3) when both result in a positive?

Methods: A total of 158 toddlers were screened for an autism diagnosis after parent referral to a developmental clinic. Participants were 23% female, 30.6 months (sd=4.4) old, and 63% were ultimately diagnosed with autism. The STAT (Stone et al., 2004) and MCHAT (Robins, Fein, & Barton,1999) were administered at referral, and the ADOS was given as part of a full diagnostic evaluation 3 weeks later. Both screening measures were used to assess positive and negative predictive value of adding a second screening measure for autism screenings.

Results: The STAT alone resulted in a positive predictive value of 0.81 (SE=0.04) and a negative predictive value of 0.87 (SE=0.07). The MCHAT alone resulted in a positive predictive value of 0.84 (SE=0.05) and a negative predictive value of 0.52 (SE=0.06). Combing the STAT and MCHAT during screening resulted in significantly better positive predictive value (0.91, SE=0.04) than considering either the STAT or MCHAT (PPV=0.78, SE=0.04; p=0.027). Additionally, considering either the STAT or the MCHAT resulted in a greater Negative predictive value (0.90, SE=0.05) than combining the two measures (0.54, SE=0.06; p=0.000).

Conclusions: Multiple screening measures should be considered to optimize autism diagnostics after referral based on concern. Additional measures may be necessary to increase the positive predictive value of multiple measures in autism screenings while also optimizing negative predictive value. Recommendations for practice and future research will be discussed. By the time these data are presented, we anticipate a sample of 200 participants.