Discussion
In this study, parents of 488 children completed the ITC at their child’s 18-month health supervision visit. On the basis of a 10th percentile cut-off, 48 children were expected to have a positive ITC screen. In our sample, 46 children (9.4%) had a positive ITC screen, including 5.7% with concern for speech delay, 6.2% with concern for other communication delays and 2.5% with concern for both. At follow-up, at a mean age of 4 years, approximately 5% had a developmental diagnosis, including developmental delay, ASD, ADHD and learning problems. Of four children with a diagnosis of ASD at 3–5 years, three had a positive ITC screen, all with concern for other communication delays, which is notable as this ITC component is thought to capture ASD.24–26 Children with a positive ITC at the 18-month visit had 4.48 higher odds of a developmental diagnosis at follow-up. The high specificity (92%–95%) and NPV (96%) suggest that most children with a negative ITC screen will not have a later developmental diagnosis. The low false positive rate (5%–8%) suggests that use of the ITC will result in few unintended harms related to overdiagnosis and over-referral.
We also identified low sensitivity of the ITC suggesting that a positive ITC at the 18-month visit cannot accurately identify those who will have a developmental diagnosis at 3–5 years. Low sensitivity of developmental screening tools to predict later outcomes has been recognised as a challenge, due to the dynamic nature of children’s development.27 30 Factors associated with lower sensitivity include younger age at screening and longer latency from screening to outcome.30 Therefore, it is not surprising that screening at 18 months resulted in a low sensitivity to predict developmental diagnosis at 3–5 years. This underscores the importance of continuous developmental surveillance at all health supervision visits, as recommended by professional organisations.2 5
There have been three recent systematic reviews of predictive validity of developmental screening in early childhood, highlighting the interest in understanding which tools best identify children who may benefit from early identification and intervention.28–30 Of included studies, few examined screening at ≤18 months, and none examined the ITC.
Sim et al examined the predictive validity of language screening tools at 2–6 years in six studies.28 Studies of children <18 months were not eligible, and only two studies had a time to follow-up of ≥12 months, with a mean sensitivity of 54%. Cairney et al examined the predictive value of developmental assessment at 1–5 years in 13 studies and found a positive association between poor early child development and later educational difficulties, high specificity and NPV, and low sensitivity.30 Three studies included children <18 months, which assessed association but not screening test properties.
Schonhaut et al examined the predictive validity of the Ages & Stages Questionnaire (ASQ), which is of great importance given its common use in many countries.29 Of five included studies, three screened children at 36 or 48 months; and while the remaining two studies screened some children at 18 months, they included children born preterm. Lamsal et al (not included in any systematic review) examined the predictive validity of the ASQ for parent report of a developmental diagnosis at 4–5 years.15 At 24 months (the youngest age examined), using the 1 SD cut-off for the ASQ, sensitivity was 84% and specificity was 69%; using the 2 SD cut-off, sensitivity was 32% and specificity was 91%. In summary, little is known about the predictive validity of the ASQ in healthy term infants screened at 18 months.
Low to moderate sensitivity has been found in concurrent validity studies of the ASQ.12 13 16 Warren et al found a sensitivity of 55% and specificity of 86% in a meta-analysis of four studies.16 Sheldrick et al found a sensitivity of 35% and specificity of 89% in children 9 months–5.5 years (mean 2.6 years).13 Wilson et al found a sensitivity of 55% and specificity of 95% on the ASQ Communication Scale in children 23–30 months (mean 25 months).12
Low sensitivity of ASD-specific tools has also been found. In a predictive validity study of the Modified Checklist for Autism in Toddlers with Follow-Up in children 16–26 months in primary care, Guthrie et al found a sensitivity of 39% and specificity of 95% for ASD; and a sensitivity of 12% and specificity of 97% for any delay.14
The ITC was developed by Wetherby and Prizant in US children.18 For identification of communication disorders, concurrent validity was assessed in children 12–17 and 18–24 months: sensitivity 86%–89% and specificity 74%–77%21; and predictive validity in children 12–24 months, with language assessed at 3 years: sensitivity 83% and specificity 70%.21 For identification of ASD, in children 9–24 months (n=5385), 60 who were ≥4 years received a diagnosis of ASD; of these, the ITC identified 56, for a sensitivity of 93%.24
Pierce et al selected the ITC in their screen–evaluate–treat model embedded in primary care, aiming for early ASD intervention before 2 years.25 26 A network of 203 paediatricians screened more than 44 000 children at 12, 18 and 24 months using the ITC. Approximately 39% of children with a positive ITC were referred for diagnostic evaluation. Of these, almost half received a diagnosis of ASD and about one-third received a diagnosis of developmental delay, language delay or other delays.26 Pierce et al found that a diagnosis of ASD becomes stable starting at 14 months, that the most common diagnostic transition was from language or developmental delay to ASD, and that almost 24% of children with an ASD diagnosis at 3–4 years were late identified.40 In our study, three of four children with an ASD diagnosis at 3–5 years had a positive ITC at 18 months, supporting the potential for early identification of ASD using the ITC.
Strengths of the ITC include its focus on communication, an important developmental domain at 18 months for which interventions are available. The ITC distinguishes concern for speech delay (monitor/refer) and other communication delays such as ASD (refer). Overall, the ITC screen positive rate is about 10% and false positives are low, minimising overdiagnosis and over-referral. Additional advantages include its one-page format, ease of completion and availability free of charge.
Limitations of this study include parent report of a developmental diagnosis rather than a standardised clinician assessment. However, parent report of physician diagnosis of developmental disorders including ASD has been used extensively in national surveys such as the US National Survey of Children’s Health and the National Health Interview Survey.33–36 Kogan et al have summarised the evidence supporting the validity of parent report compared with physician diagnosis.33 In our study, maternal education and family income were high, which may limit the generalisability of our findings. However, family income was included as a covariate in our analysis and the overall prevalence of a positive ITC screen in our sample was close to the expected 10th percentile cut-off score.19