Methods
Explainable AI & Predictive Modeling
Developing interpretable prediction methods for rare outcomes, with a focus on SHAP-guided augmentation, class imbalance, and rigorous model evaluation.
SHAPrare eventssynthetic dataevaluation
Methods. Implementation. Impact.
Jonah Kwesi Amponsah, PhD
Statistician · Machine Learning Researcher · Implementation Scientist
Developing statistical and machine-learning methods to improve clinical prediction, implementation, and healthcare decision-making.
01 / Research Areas
Connecting methods
to better health.
Three connected areas: how we build predictive models, evaluate implementation strategies, and understand variation in health and care.
Translation
Implementation Science
Connecting implementation questions to experimental designs and quantitative evaluation, with applications to testing workflows and evidence-based cancer care.
cluster trialsSMARTMOSThybrid designs
Applications
Cancer & Population Health
Examining genomic testing, geographic disparities, and population health needs to inform care delivery and the prioritization of intervention resources.
oncologygenomicshealth equityspatial methods
02 / Featured Projects
Selected research projects.
Explainable AI
SHAP-Guided Hotspot Augmentation
A targeted augmentation framework for rare-event prediction that uses explainability and local model behavior to focus synthetic data generation where it is most useful.
Active researchImplementation Science · Oncology
Improving Genomic Testing in Cancer Care
Quantitative evaluation of genomic testing completion, workflow gaps, and implementation strategies across multiple cancer types.
Active researchPopulation Health · Spatial Methods
Geographic Prioritization for Depression Intervention
A spatial framework for identifying areas where intervention resources may have the greatest reach using demographic, rurality, and symptom-burden estimates.
Active research03 / Selected Publications
Contributing to
the evidence.
Selected peer-reviewed work across explainable AI, biostatistics, and population health.
See all publications →Explainable AI predicting Alzheimer’s disease with latent multimodal deep neural networks
Published · Journal of Biopharmaceutical StatisticsUnderstanding regional disparities in obesity: a multiscale geographically weighted analysis in the United States
Published · International Journal of Health Geographics04 / Current Work
Questions guiding
the methods.
Cross-cutting questions in evaluation, study design, and spatial analysis that inform my research agenda.
Model evaluation
Evaluating prediction for rare outcomes
How should repeated and nested evaluation be used to assess explainability-guided augmentation when outcomes are uncommon?
Study design
Matching trial designs to implementation questions
Which questions call for cluster trials, adaptive designs, or optimization strategies when evaluating implementation approaches?
Spatial methods
Accounting for place in population health
How can local variation, rurality, and symptom burden inform where intervention resources are prioritized?
05 / Talks & Presentations
Ideas in conversation.
December 2026
Accepted presentation
Choosing experimental designs for implementation strategy trials: A question-driven framework with practical tools and exemplars
19th Annual Conference on the Science of Dissemination and Implementation in Health
2026
Research portfolio
Explainable AI, predictive modeling, and applied biostatistics
Selected invited and scientific presentations
Collaborate