AI-generated research brief — verify at source
Digital Developmental Screening Tool Shows Markedly Higher Sensitivity Than Routine GP Assessment
A cluster randomised controlled trial found that the WMG-I digital surveillance tool detected developmental concerns in children at age 2 with substantially greater sensitivity than routine general practice assessment — 93.9% versus 75.9% (95% CI 88.1%–99.7% vs 57.2%–94%; p=0.035). Specificity was similar between the two approaches and did not differ significantly, suggesting the improvement in case detection does not come at the cost of a meaningful increase in false-positive identifications.
What Was Studied
The study assessed the diagnostic accuracy of the Watch Me Grow Integrated (WMG-I) digital developmental screening and surveillance platform compared with routine general practitioner (GP) assessment in young children. The primary outcome was whether WMG-I could more accurately identify children with developmental conditions — as determined by validated reference standard assessments — than service-as-usual care delivered within primary care settings.
How It Was Studied
This was a prospective sub-study embedded within a longitudinal cluster randomised controlled trial involving 655 children aged 16 to 35 months attending 40 general practices across culturally diverse metropolitan areas of New South Wales and Queensland, Australia. Forty practices were randomised either to the WMG-I intervention or to service as usual (SaU). In the WMG-I arm, parents completed a digital screening weblink comprising a primary screener and, where concerns were flagged, an automated secondary screener; results were transmitted directly to both the parent and the GP to guide follow-up. The SaU arm received standard GP developmental checks without structured digital support. To anchor findings to a reliable diagnostic benchmark, all children in whom concerns were identified, plus a systematic 10% random sample of children without flagged concerns, underwent reference standard evaluation using the Mullen Scale of Early Learning, the Autism Diagnostic Observation Schedule, and the Vineland Adaptive Behaviour Scale; in total, 151 children received this comprehensive assessment.
What Was Observed
- WMG-I sensitivity was substantially higher than routine GP care. The digital tool correctly identified approximately 94 in every 100 children who had a developmental condition, compared with roughly 76 in every 100 under standard care — a statistically significant difference (WMG-I 93.9%, 95% CI 88.1%–99.7%; SaU 75.9%, 95% CI 57.2%–94%; p=0.035). This gap represents a clinically meaningful reduction in missed developmental concerns.
- Specificity was comparable between approaches and did not differ significantly. WMG-I correctly ruled out developmental concerns in approximately 71 of every 100 unaffected children, while SaU did so in approximately 67 of every 100 — a difference that was not statistically significant (WMG-I 71.4%, 95% CI 54.7%–88.1%; SaU 66.7%, 95% CI 44.9%–88.4%; p=0.743). The confidence intervals for both estimates are wide, indicating some imprecision around these specificity figures.
Why This Matters
Engagement with developmental surveillance programmes is reported to be low globally, creating a recognised gap between recommended policy and actual clinical practice. This study provides empirical diagnostic accuracy data demonstrating that a structured digital approach, integrated into routine primary care contacts such as immunisation visits, can meaningfully improve the early identification of developmental conditions in young children. The authors propose that the automated and scalable nature of WMG-I — including reminder systems tied to recommended surveillance ages — positions it as a practical population-level tool for closing this implementation gap.
How to Read This Result
While the sensitivity advantage of WMG-I is statistically significant and the effect size is notable, confidence should be tempered by the medium-quality study design and the risk of verification bias introduced by selective reference standard testing — a design feature that may artificially inflate sensitivity estimates in both arms.
Limitations
Only children with flagged concerns and a 10% random sample of those without concerns received the reference standard assessment; this selective approach introduces verification bias and may overestimate sensitivity for both groups. The trial was conducted exclusively in metropolitan regions of Australia, which limits how confidently findings can be extended to rural, remote, or internationally distinct healthcare settings. Additionally, because global engagement with developmental surveillance programmes is already low, the real-world effectiveness of WMG-I may depend heavily on initial enrolment rates, which the current study does not fully address.