The United States Preventive Services Task Force (USPSTF) recently highlighted the lack of studies examining universal autism screening in primary care and whether it leads to earlier diagnosis and treatment. This gap prevented them from being able to recommend universal autism screening until there is more evidence. CHOP has successfully implemented universal autism screening in primary care ? 83% screening compliance in 2016, and has both screening and follow-up primary care available in the electronic health record. This proposal is for a secondary analysis of this data to directly answer the question of whether universal screening results in earlier identification and earlier treatment entry for children with autism. Preliminary analyses on 13,503 children found the sensitivity of the M-CHAT was surprisingly low -- 40% (specificity 94%, positive predictive value 16%, and negative predictive value of 98%). These results show that, with only one screen, more children with autism screened negative (n=198) than screened positive (n=128). This project will include a sample double this size (as more children age into the 4-6 year age range), and include the wealth of data within the EHR to accomplish two aims.
Aim 1 will determine the age of diagnosis and entry into early intervention for children in a universal autism screening program. Many factors contribute to the age of diagnosis and intervention, but this dataset can answer three important questions: First, to determine the age of diagnosis and entry to intervention for children with autism who screened positive as toddlers at the first v. second screen (?true positives?). Second, to determine the age of diagnosis and early intervention for children who screened negative but went on to be diagnosed with autism (?false negatives?), since negative screening results can still lead to heightened awareness of milestones. Third, to explore the age of diagnosis and intervention for screened v. unscreened children, since a subset of children were not screened (approximately 17% of well-child visits, as well as some number of children who received sick care but did not present for well-child visits).
Aim 2 will identify factors that would improve screening accuracy, and then create and validate a clinical decision tree. Pediatricians use clinical judgment to interpret screening results?this project can provide much needed evidence. Variables such as autism risk factors (e.g., family history); child demographic, medical, and developmental factors; family variables including socio-economic status and parent concerns; and practice-level variables including patient population demographics, provider years of experience, provider compliance with screening, and provider rates of referrals, could inform clinical judgment. Upon successful completion, this project will contribute the exact type of evidence from universal screening in primary care requested by the USPSTF, as well as provide guidance for pediatricians on how to fine tune their interpretation of screening results. This can significantly impact the age of diagnosis and intervention, by revealing for whom screening is and is not working well.
We will analyze existing data from the highly successful universal autism screening program in our large primary care network. Our data can answer critical questions about the age of diagnosis, referral, and entry into intervention for children screened in primary care. We can also determine whether child, family, provider, and practice variables can be used to more accurately interpret early autism screening results.