Microsoft Research Lab – Africa, Nairobi

August Seminar

Sub-Population Identification of Multimorbidity in Sub-Saharan African Populations

Join us on Thursday 14th at 3pm EAT for a talk by Dr. Skyler Speakman (opens in new tab), a Senior Research Scientist at IBM Research Africa where he uses Machine Learning and Data Science to impact peoples’ lives across the continent. The Kenya-based team works on scaling AI by improving the cost/performance of data pipelines and better understanding how data can be represented for multiple downstream tasks.

Abstract:

This work bridges medical literature on multimorbidity with data science approaches to exploratory health data analysis. We define multimorbidity as the co-occurrence of at least two disease diagnoses from a pre-determined list, contributing to evolving definitions in the field. Applying this definition to two sub-Saharan populations—Nairobi, Kenya and Agincourt, South Africa—using data from the Africa Wits-INDEPTH Partnership for Genomic Studies, we explore patterns of disease co-occurrence across diverse demographic groups.

We automatically stratify the data to identify sub-populations with unusually high or low multimorbidity rates, offering a scalable alternative to traditional confirmatory methods. Notably, high-risk groups in one location often mirror those in the other, and we uncover nuanced risk profiles beyond age and sex, challenging common stratification practices. This work demonstrates how data science methods can enhance public health research by revealing complex patterns in large-scale datasets.