Epidemiology and Statistics: Exploring Research Design and Prediction Modelling

The Epidemiology and Statistics group at The Health Research Unit Zimbabwe, fondly known as the Epi & Stats group, is a learning forum that brings researchers together to learn, discuss and engage with epidemiological and research methodologies.

Launched in June 2026, the group provides a collaborative space for researchers at different stages of their careers to share knowledge, explore methodological concepts and work through real-world research challenges.

In the second hybrid session, Dr Paul Nesara, a researcher of neonatal health, and Dr Edson Marambire, a researcher of tuberculosis and infectious diseases, explored two aspects of health research: How researchers can design studies to answer important questions, and how data can be used to predict health outcomes.

The Richness of Research Design

Dr Paul Nesara began by taking participants back to the foundations and philosophy of research, asking the question: How do we know what we know?

He introduced the concepts of ontology, epistemology and methodology before turning to a more practical question for researchers: how do you choose the right research design for your project?

The session revisited the definitions of quantitative, qualitative and mixed method approaches to research, highlighting how each one can be used to answer different research questions. It also explored the breadth of research methodology within these categories - from cohort studies to ethnography and convergent parallel mixed method designs, and considered how their applicability and use vary depending on the context. 

The richness of these approaches was central to Paul’s presentation and the key takeaway is that if you find yourself in a room full of researchers - as many of us at THRU-ZIM often do - and someone tells you they're using a quantitative, qualitative, or mixed methods research approach, be sure to ask: 

Which one?

Can we predict who is infected with Mycobacterium tuberculosis without a laboratory test?

This is the question with which Dr Edson Marambire opened his presentation on the use of predictive models. 

Drawing on work from his PHD and ERASE-TB, Edson discusses the gaps in TB preventing therapy targeting. He states that better targeting is needed as the World Health Organization mandates that all household contacts of a TB index case in high incidence countries receive TPT - which is costly, logistically demanding and only modestly predicts future disease. 

One proposed solution is to use questionnaire based prediction models to risk-stratify household contacts, potentially reducing the need for laboratory visits and allowing preventative therapy to be targeted towards those most likely to benefit. 

Edson walked the participants through the steps of developing a predictive model, from defining the outcome and population, selecting candidate predictors and preparing the data, through to fitting the model, selecting variables and assessing the model performance.

An important part of the discussion was the need to ask whether a model can be generalised to other populations and settings, and whether it can reliably be applied to data other than that from which it was trained and developed.

While his explanation clearly laid out how a prediction model can be created to explore alternative strategies for treatment targeting in the context of TB infection, Edson also ensured that the realities of how feasible such a predictive model is in practice were not lost on participants. He demonstrated that, although variables with strong associations can be combined to develop a predictive model, this does not necessarily mean that the resulting approach will improve upon current practice.

From choosing the right design to making better predictions.

Both presentations highlighted the importance of good data, appropriate methods, and careful interpretation, while providing engaging and thought-provoking insights into the topics discussed.

Thank you Dr Paul Nesara and Dr Edson Marambire for sharing your expertise with the group!