Date Presented 03/27/20
The analysis in this paper tests the utility of the FYI 3.1 as an autism screening tool and how that utility varies over age. Initial screening was done on a community sample of 1,893 children between 8 and 16 months of age. Sensitivity and positive predictive value were tested over a range of specificity from 85 to 99%.
Primary Author and Speaker: John Sideris
Contributing Authors: Grace Baranek, Linda Watson, Elizabeth Crais, Yun-Ju Chen
PURPOSE: Current estimates of the prevalence of ASD in children in the United States indicate the rate is as high as 1.7% (Baio, et al., 2018) and research suggests that early screening and intervention (e.g., Johnson & Myers, 2007; Watson et al., 2018; Zwaigenbaum et al., 2015) may be helpful. In this study, we evaluated the predictive power of the latest revision of the FYI and recommend cut points over a range of sensitivity (Se) and specificity (Sp) in a community sample. Analyses of these kind often provide very restrictive information, with Sp and Se constrained to a single value. We present a range of potential cutoff scores on domains of the measure in order provide a deeper description of the trade-off between the two metrics and allow for greater flexibility in the application of this screening tool.
DESIGN: Responses to the FYI were collected online or by mail by the parents of children between the ages 8 and 16 months, recruited through North Carolina vital records in order to reach a broad community sample.Diagnostic assessments were done at 3 years of age for 1893 children, 67 (3.5%) of whom were classified with ASD. Subjects were grouped in to three age blocks: 8 to <11 months (N=591), 11 to <14 months (N=940), and 14 to <16 months (N=362) for analysis. Analyses in this presentation are restricted to only those either classified with ASD or with no other development delay.
The FYI is a 68-item measure with Items on the scale are grouped into two domains, Sensory-Regulatory (SR) and Social-Communication (SC). To reduce respondent burden, two short forms of the measure were created, each with 47 items and with 26 common across the forms. Each item asks parents about a specific behavior with five response categories ranging from “Never” to “Always.” For each item, we created weighted risk scores based on the prevalence of each response. Domain scores for SR and SC are created by taking simple means of the relevant items.
METHOD: For each of the age blocks, we conducted ROC curve analysis using logistic regression to determine targets for Se and Sp. The models included SR, SC, and their interaction. Examination of the ROC curve from each model indicated target Se when Sp was set between 85% and 99%. Scatterplots of SR and SC were examined to determine specific cut-points on each domain. Respondents were classified based on either having 1) high scores on both domains or 2) at least one extreme score (< mean + 2.5 SD) on one domain. Finally, frequency analyses were conducted to estimate empirical values for Se, Sp, PPV and Negative Predictive Value (NPV).
RESULTS: The logistic regression models indicated relatively strong predictive power for the combination of FYI domains with areas under the curve of .75, .82, and .67 for 8 to <11 month, 11 to <14 month, and 14 to <16 month groups, respectively. In all age groups, Se was very high, over 50%, when Sp was fixed at 85%, but when Sp was set at the most restrictive level, 99%, decreased to between 15% and 29% depending on age. Correspondingly, PPV increases over this same range of Sp.
CONCLUSION: The latest version of the FYI is efficacious in distinguishing between typically developing children and children with ASD in a community sample. Results of these analyses allow for selection of different cut-points on the domain scores at varying levels of Se and Sp. These results will provide clinicians with specific guidelines while also allow for flexibility in balancing Se and SP.
References
Baio, J., Wiggins, L., Christensen, D. L., Maenner, M. J., Daniels, J., Warren, Z., ... & Durkin, M. S. (2018). Prevalence of autism spectrum disorder among children aged 8 years—autism and developmental disabilities monitoring network, 11 sites, United States, 2014. MMWR Surveillance Summaries, 67(6), 1.
Johnson, C. P., & Myers, S. M. (2007). Identification and evaluation of children with autism spectrum disorders. Pediatrics, 120(5), 1183-1215.
Watson, L. R., Nowell, S. W., Crais, E. R., Baranek, G. T., Wakeford, L., & Turner-Brown, L. (2018). Supporting Families of Infants At-Risk for ASD Identified Through Community Screening and Surveillance. In Handbook of Parent-Implemented Interventions for Very Young Children with Autism (pp. 25-43). Springer, Cham.
Zwaigenbaum, L., Bauman, M. L., Choueiri, R., Fein, D., Kasari, C., Pierce, K., ... & McPartland, J. C. (2015). Early identification and interventions for autism spectrum disorder: executive summary. Pediatrics, 136(Supplement 1), S1-S9.
