We are all seeing more children on the autism spectrum than we ever expected. With a Centers for Disease Control–estimated prevalence of 1 in 44, the average pediatrician will be caring for 45 children with autism. It may feel like even more as parents bring in their children with related concerns or fears. Early entry into services has been shown to improve functioning, making early identification important. However, screening at the youngest ages has important limitations.
Sharing a concern about possible autism with parents is a painful aspect of primary care practice. We want to get it right, not frighten parents unnecessarily, nor miss children and delay intervention.
Dr. Barbara J. Howard
Autism screening is recommended by the American Academy of Pediatrics at 18- and 24-month pediatric well-child visits. There are several reasons for screening repeatedly: Autism symptoms emerge gradually in the toddler period; about 32% of children later found to have autism were developing in a typical pattern and appeared normal at 18 months only to regress by age 24 months; children may miss the 18 month screen; and all screens have false negatives as well as false positives. But even screening at these two ages is not enough.
One criticism of current screening tests pointed out by the U.S. Preventive Services Task Force has been a problem with the sample used to develop or validate the tool. Many test development studies included only children at risk by being in early intervention, siblings of children with diagnosed autism, or children only failing the screening tests rather than a community sample that the screen in actually used for.
Another obstacle to prediction of autism diagnoses made years later is that some children may not have had any clinical manifestations at the younger age even as judged by the best gold standard testing and, thus, negative screens were ambiguous. Additionally, data from prospective studies of high-risk infant siblings reveal that only 18% of children diagnosed with autism at 36 months were given that diagnosis at 18 months of age despite use of comprehensive diagnostic assessments.
Prevalence is also reported as 30% higher at age 8-12 years as at 3-7 years on gold-standard tests. Children identified later with autism tend to have milder symptoms and higher cognitive functioning. Therefore, we need some humility in thinking we can identify children as early as 18 months; rather, we need to use the best available methods at all ages and remain vigilant to symptoms as they evolve as well as to new screening and testing measures.
The most commonly used parent report screen is the 20-item Modified Checklist for Autism in Toddlers–Revised (M-CHAT-R), a modification of the original CHAT screen. To have reasonable positive predictive value, the M-CHAT-R authors recommend a clinician or trained staff member conduct a structured follow-up interview with the parent when the M-CHAT-R has a score of 3-7. Scores of 8 or more reflect enough symptoms to more strongly predict an autism diagnosis and thus the interview may be skipped in those cases. The recommended two-step process is called M-CHAT-R/F. At 18 months without the R/F, a positive M-CHAT-R only is associated with an autism diagnosis 27% of the time (PPV, 0.27); which is unacceptable for primary care use.
Unfortunately, the M-CHAT-R/F appears to be less accurate for 18-month-olds than 24-month-olds, in part because its yes/no response options are harder for a caregiver to answer, especially for behaviors just developing, or because of lack of experience with toddlers.
An alternative modification of the original CHAT called the Quantitative CHAT or Q-CHAT-10 has a range of response options for the caregiver; for example, always/usually/sometimes/rarely/never or many times a day/a few times a day/a few times a week/less than once a week/never. The authors of the Q-CHAT-10, however, recommend a summary pass/fail result for ease of use rather than using the range of response option values in the score. We recently published a study testing accuracy using add-up scoring that utilized the entire range of response option values, called Q-CHAT-10-O (O for ordinal), for children 16-20 months old as well as cartoon depictions of the behaviors. Our study also included diagnostic testing of screen-negative as well as screen-positive children to accurately calculate sensitivity and specificity for this method. In our study, Q-CHAT-10-O with a cutoff score greater than 11 showed higher sensitivity (0.63) than either M-CHAT-R/F (0.34) or Q-CHAT-10 (0.31) for this age range although the PPV (0.35) and negative predictive value (0.92) were comparable with M-CHAT R/F. Although Q-CHAT-10-O sensitivity (0.63) is less than M-CHAT-R (without follow-up; 0.73) and specificity (0.79) is less than the two-stage R/F procedure (0.90), on balance, it is more accurate and more practical for a primary care population. After 20 months of age, the M-CHAT-R/F has adequate accuracy to rescreen, if indicated, and for the subsequent 24 month screening. Language items are often of highest value in predicting outcomes in several tools including in the screen we are now validating for 18 month olds.
The Q-CHAT-10-O with ordinal scoring and pictures can also be recommended because it shows advantages over M-CHAT-R/F with half the number of items (10 vs. 20), no requirement for a follow-up interview, and improved sensitivity. Unlike M-CHAT-R, it also contributes to equity in screening because results did not differ depending on race or socioeconomic background.
Is there an even better way to detect autism in primary care? In 2022 an article was published regarding an exciting method of early autism detection called the Social Attention and Communication Surveillance–Revised (SACS-R), an eight-item observation checklist completed at public health nurse check-ups in Australia. The observers had 4 years of nursing degree education and a 3.5-hour training session.
The SACS-R and the preschool version (for older children) had significant associations with diagnostic testing at 12, 18, 24, and 42 months. The SACS-R had excellent PPV (82.6%), NPV (98.7%), and specificity (99.6%) and moderate sensitivity (61.5%) when used between 12 and 24 months of age. Pointing, eye contact, waving “bye, bye,” social communication by showing, and pretend play were the key indicators for observations at 18 months, with absence of three or more indicating risk for autism. Different key indicators were used at the other ages, reflecting the evolution of autism symptoms. This hybrid (observation and scoring) surveillance method by professionals shows hopeful data for the critical ability to identify children at risk for autism in primary care very early but requires more than parent report, that is, new levels of autism-specific clinician training and direct observations at multiple visits over time.
The takeaway is to remember that we should all watch closely for early signs of autism, informed by research on the key findings that a professional might observe, as well as by using the best screens available. We should remember that both false positives and false negatives are inherent in screening, especially at the youngest ages. We need to combine our concern with the parent’s concern as well as screen results and be sure to follow-up closely as symptoms can change in even a few months. Many factors may prevent a family from returning to see us or following our advice to go for testing or intervention, so tracking the child and their service use is an important part of the good care we strive to provide children with autism.