The predictive value of the ages and stages questionnaire in late infancy for low average cognitive ability at age 5

Predicting low cognitive ability at age 5 years using perinatal data and machine learning

BACKGROUND

There are no early, accurate, scalable methods for identifying infants at high risk of poor cognitive outcomes in childhood. We aim to develop an explainable predictive model, using machine learning and population-based cohort data, for this purpose.

METHODS

Data were from 8858 participants in the Growing Up in Ireland cohort, a nationally representative study of infants and their primary caregivers (PCGs). Maternal, infant, and socioeconomic characteristics were collected at 9-months and cognitive ability measured at age 5 years. Data preprocessing, synthetic minority oversampling, and feature selection were performed prior to training a variety of machine learning models using ten-fold cross validated grid search to tune hyperparameters. Final models were tested on an unseen test set.

RESULTS

A random forest (RF) model containing 15 participant-reported features in the first year of infant life, achieved an area under the receiver operating characteristic curve (AUROC) of 0.77 for predicting low cognitive ability at age 5. This model could detect 72% of infants with low cognitive ability, with a specificity of 66%.

CONCLUSIONS

Model performance would need to be improved before consideration as a population-level screening tool. However, this is a first step towards early, individual, risk stratification to allow targeted childhood screening.

IMPACT

This study is among the first to investigate whether machine learning methods can be used at a population-level to predict which infants are at high risk of low cognitive ability in childhood. A random forest model using 15 features which could be easily collected in the perinatal period achieved an AUROC of 0.77 for predicting low cognitive ability. Improved predictive performance would be required to implement this model at a population level but this may be a first step towards early, individual, risk stratification.

The ASQ did not predict low IQ scores when children born VPT were six years

AIM

There is a need for methods that can provide valid assessment tools in a follow-up programme without great financial costs. This study assessed the accuracy of the 60-month Ages and Stages Questionnaire as a screening tool to predict a low intelligence quotient score at 6 years in children born very preterm.

METHODS

Totally, 54 children participated in a six-year follow-up study, which included an intelligence quotient test at 6 years of age and a 60-month Ages and Stages Questionnaire at four and a half or 5 years of age at respond. We used the receiver operating characteristic curve and evaluated the optimal cut-off score to predict a low intelligence quotient score.

RESULTS

At four and a half years, the optimal cut-off value for predicting a low intelligence quotient score was 242, with a sensitivity of 67% and a specificity of 59%. At 5 years, only one child had a low intelligence quotient score, and the analysis was not performed.

CONCLUSION

Our results did not support the use of the 60-month Ages and Stages Questionnaire as a valuable screening tool to predict a low intelligence quotient score in children born very preterm at 6 years of age.

ASQ-3 and BSID-III's concurrent validity and predictive ability of cognitive outcome at 5 years

BACKGROUND

Early detection of cognitive disability is challenging. We assessed the domain-specific, concurrent validity of the ages and stages questionnaire (ASQ-3) and the Bayley Scales of Infant and Toddler Development (BSID-III), and their ability to predict cognitive delay at school age.

METHODS

Within a longitudinal birth cohort study, a nested cohort of children was assessed using ASQ-3 and BSID-III at 24 months, and at 5 years using the Kaufmann brief IQ test (KBIT).

RESULTS

278 children were assessed using BSID-III and ASQ-3 at 24-months; mean(SD) BW = 3445(506) grams, M:F ratio=52:48. ASQ-3 had reasonable predictive ability (AUROC, p value, sensitivity:specificity) of same domain delay for motor (0.630, p = 0.008, 50%:76.1%) and language (0.623, p = 0.010, 25%:99.5%) at 2 years, but poor ability to detect cognitive delay compared to BSID-III (0.587, p = 0.124, 20.7%/96.8%;). 204/278 children were assessed at 5 years. BSID-III language and cognition domains showed better correlation with verbal and nonverbal IQ (R = 0.435, p < 0.001 and 0.388, p < 0.001 respectively). Both assessments showed high specificity and low sensitivity for predicting delay at 5 years.

CONCLUSIONS

The ASQ-3 cognitive domain showed poor concurrent validity with BSID-III cognitive score. Both ASQ-3 and BSID-III at 2 years poorly predict cognitive delay at 5 years.

IMPACT

The ASQ-3 does not adequately detect cognitive delay or predict cognitive delay at 5 years, particularly for children with mild to moderate delay. The ASQ-3 shows reasonable concurrent validity with the motor and language subscales of the BSID-III. Neither early screening nor formal developmental testing demonstrated significant predictive validity to screen for cognitive delay at school age. This article highlights the need to analyse our existing model of using the ASQ-3 to screen for cognitive delay in children aged 2 years.

Longitudinal cohort study investigating neurodevelopmental and socioemotional outcomes in school-entry aged children after open heart surgery in Australia and New Zealand: the NITRIC follow-up study protocol

INTRODUCTION

Despite growing awareness of neurodevelopmental impairments in children with congenital heart disease (CHD), there is a lack of large, longitudinal, population-based cohorts. Little is known about the contemporary neurodevelopmental profile and the emergence of specific impairments in children with CHD entering school. The performance of standardised screening tools to predict neurodevelopmental outcomes at school age in this high-risk population remains poorly understood. The NITric oxide during cardiopulmonary bypass to improve Recovery in Infants with Congenital heart defects (NITRIC) trial randomised 1371 children <2 years of age, investigating the effect of gaseous nitric oxide applied into the cardiopulmonary bypass oxygenator during heart surgery. The NITRIC follow-up study will follow this cohort annually until 5 years of age to assess outcomes related to cognition and socioemotional behaviour at school entry, identify risk factors for adverse outcomes and evaluate the performance of screening tools.

METHODS AND ANALYSIS

Approximately 1150 children from the NITRIC trial across five sites in Australia and New Zealand will be eligible. Follow-up assessments will occur in two stages: (1) annual online screening of global neurodevelopment, socioemotional and executive functioning, health-related quality of life and parenting stress at ages 2-5 years; and (2) face-to-face assessment at age 5 years assessing intellectual ability, attention, memory and processing speed; fine motor skills; language and communication; and socioemotional outcomes. Cognitive and socioemotional outcomes and trajectories of neurodevelopment will be described and demographic, clinical, genetic and environmental predictors of these outcomes will be explored.

ETHICS AND DISSEMINATION

Ethical approval has been obtained from the Children's Health Queensland (HREC/20/QCHQ/70626) and New Zealand Health and Disability (21/NTA/83) Research Ethics Committees. The findings will inform the development of clinical decision tools and improve preventative and intervention strategies in children with CHD. Dissemination of the outcomes of the study is expected via publications in peer-reviewed journals, presentation at conferences, via social media, podcast presentations and medical education resources, and through CHD family partners.

TRIAL REGISTRATION NUMBER

The trial was prospectively registered with the Australian New Zealand Clinical Trials Registry as 'Gene Expression to Predict Long-Term Neurodevelopmental Outcome in Infants from the NITric oxide during cardiopulmonary bypass to improve Recovery in Infants with Congenital heart defects (NITRIC) Study - A Multicentre Prospective Trial'.

TRIAL REGISTRATION

ACTRN12621000904875.