Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0878
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-25T06:05:23.312689Z
status: published
lens: trace
sector: health
headline: Pathogenic Genetic Variants, Comorbid Autism and Adaptive Developmental Quotient as Independent Predictors of Intellectual Disability in Children With Global Developmental Delay: An Interpretable Machine Learning Model With Calibrated Risk Estimation
dek: Background Global developmental delay (GDD) frequently precedes intellectual disability (ID), but no validated multivariable prognostic tool exists to support individualised counselling during the initial diagnostic work-up. Existing risk indicators are typically considered in isolation, and their joint contribution within an interpretable predictive framework remains uncertain. Methods We retrospectively analysed 2453 children diagnosed with GDD between January 2014 and December 2023 at a provincial tertiary ch…
gain_title: Among 2453 tertiary-referred children with GDD, a Platt-calibrated L2 logistic regression integrating routine clinical, neurophysiological and genetic data stratified progression to ID with AUC 0.783 and near-ideal calibration, enabling high-PPV triage.
problem_title: In the same tertiary GDD cohort, the Youden-optimal model missed 33.9% of children who progressed to ID and had NPV 31.7%, limiting safe rule-out and indicating poor transportability beyond high-prevalence referral settings.
trace_subject: risk of progression from global developmental delay to intellectual disability predicted by interpretable machine learning using clinical, neurophysiological and genetic data
gain_reading: Among 2453 tertiary-referred children with GDD, a Platt-calibrated L2 logistic regression integrating routine clinical, neurophysiological and genetic data stratified progression to ID with AUC 0.783 and near-ideal calibration, enabling high-PPV triage.
gain_evidence: Platt-calibrated L2 logistic regression achieved an AUC of 0.783 (95% CI 0.735-0.828) with calibration slope 1.01 and intercept -0.005, and a Brier score of 0.113
problem_reading: In the same tertiary GDD cohort, the Youden-optimal model missed 33.9% of children who progressed to ID and had NPV 31.7%, limiting safe rule-out and indicating poor transportability beyond high-prevalence referral settings.
problem_evidence: 83.0% progressed to ID, reflecting the referral profile of a tertiary centre
quick_read: Researchers retrospectively analyzed 2453 children diagnosed with GDD between January 2014 and December 2023 at a provincial tertiary children's rehabilitation centre, followed to at least 60 months. Using 28 predictors across perinatal, developmental, neuroimaging, electrophysiological, genetic and comorbidity domains, they trained L2- and L1-regularized logistic regression, random forest, XGBoost and LightGBM, with Platt scaling for the L2 model, and evaluated on a held-out 30% test set.

By the August 2026 publication date, the calibrated L2 model showed moderate discrimination and good calibration for high-risk triage in this referral setting, with high positive predictive value but limited sensitivity and low negative predictive value. Because the cohort had 83% ID prevalence and came from a single tertiary centre, the authors note the tool is not ready for community use without refitting and external validation, and its impact on counselling decisions remains to be prospectively tested.
limitation: Single-centre retrospective design with high ID prevalence reflecting tertiary referral; model not validated for community populations and requires refitting and external validation before broader use.
tag: Dual reading
key_points: Retrospective cohort of 2453 children diagnosed with GDD 2014-2023 at provincial tertiary children's rehabilitation centre, followed to minimum age 60 months; 2037 (83.0%) progressed to ID. | 28 candidate predictors retained after multiple imputation and multicollinearity screening; five algorithms trained on 70% split with class-weight rebalancing, no synthetic oversampling; L2 logistic regression recalibrated by Platt scaling. | At Youden-optimal threshold 0.85, sensitivity 66.1%, specificity 76.8%, PPV 93.3%, NPV 31.7%; parsimonious five-feature model AUC 0.762; discrimination robust to exclusion of post-baseline predictors (AUC 0.769).
rundown: Training used stratified 70% partition with class-weight rebalancing and no synthetic minority over-sampling; evaluation on held-out 30% test set used 1000 bootstrap confidence intervals for discrimination, calibration slope and intercept after Platt scaling, Brier score and net benefit. Sensitivity analyses excluded post-baseline candidate predictors and benchmarked against one-, three- and five-feature regressions.

Strongest independent risk factors were pathogenic genetic variant pathogenicity, comorbid autism spectrum disorder and EEG epileptiform discharges; higher Gesell adaptive developmental quotient was strongest protective factor (OR 0.45 per standardised unit). Removal of early intensive intervention lowered AUC by only 0.012 (95% CI - 0.002 to 0.026), and complete-case analysis yielded AUC 0.839 and restriction to genetically tested children AUC 0.864.
sources:
- peer_reviewed | Journal of Intellectual Disability Research | https://doi.org/10.1111/jir.70164 | 2026-08-24
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

63dd7c6c34044e4e9bac2bb47672472b8bfa61f9308dc2cb53f2622362253421

previous

0000000000000000000000000000000000000000000000000000000000000000