Causal interpretation of clinical prediction models: When, why and how.

Lijing Lin, Turing RA in Predictive Healthcare at the University of Manchester and Consultant Scientist at Unilever

Causal interpretation of clinical prediction models: When, why and how.

When developing models for prediction, neither the parameters of the model, nor the output predictions, have any causal interpretations. For pure prediction this is perfectly acceptable. However, prediction models are commonly interpreted in a causal manner - for example by altering inputs to the model to demonstrate hypothetical impact of an intervention. This can lead to biased causal effects being inferred, and thus misinformed decision making. We aimed to collect examples of use of prediction models in a causal manner in practice, and to identify and interpret literature that provides methods for enriching prediction models with causal interpretations.

We systematically reviewed literature to identify methods for prediction models with causal interpretations, by adapting a scoping review framework, and considering the interaction of prediction modelling keywords, and causal inference keywords. There were two broad categories of approach identified: 1) enriching prediction models with externally estimated causal effects, such as from meta-analyses of clinical trials; and 2) estimating both a prediction model, and causal effects, from observational data. In this talk I will present the main results from this review. We conclude that, there is a need for prediction models that allow for 'counterfactual prediction'. Methods exist but require development, particularly when triangulating data from different sources (e.g. observational data and randomised controlled trials). Techniques are also required to validate such models.


Currently, I am a postdoc at the University of Manchester, working on the project Advancing methodology for predictive healthcare, which is part of Turing Health Programme of the Alan Turing Institute. My research interests are in statistical machine learning and predictive modelling in healthcare and medicine. Before current post, I was a reserch associate and my work has been focused on disease subtype discovery, clinical phenotyping, and biomarker pattern recognition.