Autism Screening Using the Parent's Observation of Social Interactions in a Large Integrated Healthcare System.

Objective: To determine the accuracy of the Parent's Observation of Social Interactions (POSI) when deployed for universal autism screening within a large healthcare network.

Study design: Retrospective analysis of electronic health record (EHR) data from children screened for autism spectrum disorder (ASD) using POSI at 18- and 24- month pediatric well child care (WCC) visits across Southern California Permanente Medical Group facilities throughout the 2022 calendar year. Data on ASD diagnoses placed in the EHR problem list were analyzed one year later (until end of 2023) to calculate sensitivity, specificity, and positive/negative likelihood ratio (LR+ and LR-) values.

Results: At 18-month WCC, 8014 (26.4%) out of 30,375 children had elevated POSI scores of >3 (positive screen), and the sensitivity and specificity were 77.9% and 76.1% respectively. At 24-month WCC, 5988 children (21.4%) out of 27,975 had positive POSI screens, and the sensitivity and specificity were 76.5% and 81.4% respectively. Likelihood ratio for positive screen (LR+) was 3.3 at 18 months and 4.2 at 24 months. Likelihood ratio for negative screen (LR-) was 0.29 at both 18 and 24-month time points.

Conclusion: The POSI screen for ASD had adequate sensitivity and specificity (both >76%) when systematically deployed in a primary care setting at 18- and 24-month WCC visits. Children diagnosed with ASD were 3.3 and 4.2 times more likely to have screened positive at 18 and 24 months, compared with children who were not diagnosed with ASD. Conversely, children not diagnosed with ASD were 3.4 times more likely to have screened negative than children with ASD at 18 and 24 months.