Validity and reliability of the Neurodevelopmental Screening Tool (NDST) in screening for neurodevelopmental disorders in children living in rural Kenyan coast [version 1; peer review: 2 approved]

Background: There are no data on the precise burden of neurodevelopmental disorders (NDD) in Africa, despite high incidence of risk factors. Ten Questions Questionnaire (TQQ) has been used extensively in Africa to screen neurological impairments but not autism spectrum disorders (ASD) and attention-deficit hyperactivity disorders (ADHD). The Neurodevelopmental Screening Tool (NDST) has reliably assessed NDD in Asia; its validity in Africa is unknown. Methods: Using NDST and TQQ, we screened 11,223 children aged 6-9 years in Kilifi, Kenya. We invited all screen-positives and a proportion of screen-negative children for confirmatory diagnosis of NDD using clinical history, neuropsychological assessments and interviews. Results: In total, 2,245 (20%) children screened positive for NDD. Confirmatory testing was completed for 1,564 (69.7%) screen-positive and 598 (6.7%) screen-negative children. NDST’s sensitivity was 87.8% (95%CI: 88.3-88.5%) for any NDD, 96.5% (95%CI:96.1-96.8%) ASD and 89.2% (95%CI: 88.7-89.8%) for ADHD. Moderate/severe neurological impairments’ sensitivities ranged from 85.7% (95%CI: 85.1-86.3%) for hearing impairments to 100.00% (100.0-100.0%) for motor impairments. NDST had higher sensitivities than TQQ for epilepsy (88.8 vs 86.7), motor impairments (100.0 vs 93.7) and cognitive impairment (88.2 vs 84.3). Sensitivities for visual and hearing impairments were comparable in both tools. NDST specificity was 82.8% (95%CI: 82.1-83.5%) for any NDD, 94.5% (95%CI: 94.0-94.9%) for ASD and 81.7% (95%CI: 81.0-82.4%) for ADHD. The specificities range for neurological impairments was 80.0% (95%CI: 79.3-80.7%) for visual impairments to 93.8% (95%CI: 93.4-94.3%) for epilepsy. Negative predictive values were generally very high (≤100%), but most positive predictive values (PPV) were low (≤17.8%). Domain specific internal consistency ranged from 0.72 (95%CI: 0.70-0.74) for ADHD to 0.89 (95%CI: 0.87-0.90) for epilepsy. Conclusions: NDST possesses high sensitivity and specificity for detecting different domains of NDD in Kilifi. Low PPV suggest that positive diagnoses should be confirmed when samples are drawn from a population with low disease prevalence