Chrisphine Onyango

Chrisphine Onyango

Auto-generated profile, based on publicly available information

Chrisphine Onyango

Department of Public Health, Anderson Centre for Autism, New York, USA

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 August 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
LNLispah Nyambura
Lispah Nyambura
COCalvince Oketch
Calvince Oketch
COChrisphine Onyango
Chrisphine Onyango
Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental condition for which early identification can facilitate timely assessment and intervention. Machine learning (ML) and deep learning (DL) have been investigated across behavioral questionnaires, home video, and neuroimaging data as potential tools to support earlier and more scalable screening. This systematic review synthesized evidence on ML/DL approaches for early ASD detection in children and additionally assessed their engineering readiness for clinical translation. Following PRISMA 2020 guidance, PubMed and IEEE Xplore were searched from inception to 29 July 2026, supplemented by reference-list screening and Google Scholar. Studies were eligible when they evaluated ML/DL methods for ASD-related classification in pediatric populations or populations meeting the prespecified pediatric eligibility rule. Study characteristics, model architectures, validation strategies, diagnostic-performance measures, and engineering factors were extracted. Risk of bias was assessed using PROBAST, with QUADAS-2 considered for diagnostic-accuracy studies. Fifteen reports met the final eligibility criteria, comprising 13 primary diagnostic studies and two supporting methodological/engineering studies. Behavioral questionnaires, home video, and resting state fMRI were the principal modalities. Retrospective behavioral studies reported very high accuracy, whereas prospective and cross-cultural validation generally produced lower performance. AUC was reported by only a small subset of studies and the estimates represented different diagnostic tasks and validation settings; therefore, a conventional pooled AUC was not calculated. Study-level AUCs included 0.965 for ASD versus ADHD using SRS data, 0.86 for an eight-feature video model in an independent validation cohort, and 0.76 and 0.85 for the two classification layers reported in the Bangladeshi video study. The evidence indicates that ML/DL methods are promising as adjunctive screening technologies, but current evidence is limited by heterogeneous populations and modalities, variable validation practices, potential data leakage, limited external validation, and incomplete reporting of explainability and reproducibility. Engineering priorities for translation include prospective multicenter validation, robust handling of class imbalance and dataset shift, explainable AI, privacy preserving learning, standardized preprocessing, and deployment-aware model optimization.

29 April 2026 in medRxiv
Verified
Preprint
Open access
PDF file
JOHN KYALO MUTHUKAJOHN KYALO MUTHUKA
JOHN KYALO MUTHUKA
RZRuvimbo Zimunya
Ruvimbo Zimunya
ASAndrina Simengwa
Andrina Simengwa
COChrisphine Onyango
Chrisphine Onyango
KOKelly Oluoch
Kelly Oluoch
DNDesire A. Nshimirimana
Desire A. Nshimirimana
This systematic review and meta-analysis aimed to estimate the overall effectiveness of ASD interventions and identify sources of heterogeneity using frequentist and Bayesian approaches. A systematic search of PubMed/MEDLINE, Embase, Web of Science, and Scopus was conducted for studies published between January 1, 2004, and April 30, 2025. Mainly, randomized controlled trial studies with extractable intervention outcomes were included. A total of 41 studies (n=3,008) were synthesized using random-effects models (REML), Bayesian hierarchical modeling, meta-regression, and sensitivity analyses following PRISMA guidelines. The pooled random-effects estimate showed a significant positive effect of ASD interventions (effect size = 0.506, 95% CI: 0.392–0.619; z = 8.72, p < .001), corresponding to an estimated success proportion of 62% (95% CI: 59%–65%). Heterogeneity was substantial (Q□(40) = 238.78, p < .001; I² = 82.45%; τ² = 0.069, 95% CI: 0.028–0.137; τ = 0.262), with H² = 5.70 and a wide prediction interval (−0.020 to 1.031), indicating strong between-study variability. Bayesian meta-analysis confirmed a comparable effect (posterior mean = 0.619, 95% CrI: 0.592–0.646), with τ = 0.273 and I² ≈ 82.5%, and MCMC diagnostics showed stable convergence (R-hat ≈ 1.00). Publication bias analyses indicated significant funnel plot asymmetry (Egger-type regression: z = 3.429, p < .001; weighted regression: t = 9.573, p < .001), while rank correlation was non-significant (τ = −0.178, p = .103). Trim-and-fill analysis imputed 10 studies, reducing the pooled effect to 0.374 (95% CI: 0.258–0.491; τ = 0.338), though the effect remained significant (p < .001). Sensitivity analyses excluding influential studies yielded a stable effect (0.505, 95% CI: 0.401–0.609), with persistent heterogeneity (I² = 75.49%; Q□(38) = 190.21, p < .001; τ² = 0.043). Subgroup analyses showed highest effects for digital/technology-based interventions (0.672; 67%; I² = 0%), followed by nutritional (0.635; 64%; I² = 73.81%), behavioral (0.630; 63%; I² = 74.78%), and pharmacological (0.627; 63%; I² = 0%) interventions, while physical/occupational therapies showed lower effects (0.523; 52%; I² = 63.35%) and combined interventions showed borderline effects (0.593; 59%; I² = 19.96%); subgroup differences were significant (Q□5) = 22.63, p < .001). Regional effects were similar and non-significant (Q□(2) = 0.73, p = .694): North America (0.619; I² = 84.21%), Europe (0.626; I² = 62.34%), and Asia (0.659; I² = 0%). Age at intervention onset did not significantly moderate effects (Q□(5) = 0.98, p = .964), although variability was observed across children, adolescents, adults, and toddlers. Meta-regression identified significant moderators including intervention context (Q□= 18.159, p = .020), outcome domain (Q□= 19.588, p = .003), age at start (Q□= 17.795, p = .003), and intervention category (Q□= 31.714, p < .001), while follow-up duration and intervention duration were not significant. Bayesian subgroup analyses confirmed robustness, with strongest evidence for pharmacological (BF → ∞), behavioral (BF ≈ 832.50), and digital interventions (BF ≈ 30.92). In conclusion, ASD interventions demonstrate a moderate and statistically significant overall effect (∼0.50–0.62), with substantial heterogeneity driven primarily by intervention type, context, and participant characteristics, and findings were consistent across frequentist, Bayesian, and sensitivity analyses, supporting robust but context-dependent effectiveness.