causal inference from CAUSALab

As researchers in the medical world, we strive to understand not just what’s happening, but why it’s happening.

We want to identify the true causes of diseases and health outcomes so we can develop effective interventions. Most of my previous research has relied on observational studies using existing databases. I’ve often been limited to interpret associations rather than causality, due to the possibility of unmeasured factors. However, I’ve come to realize that providing causal inference, rather than just association, is essential for patients and clinicians. This allows us to identify interventions that can truly benefit patients, rather than simply pointing to potential associations that might not translate into real-world improvements.

However, it’s a constant challenge to establish causal relationships in the real world. While randomized controlled trials (RCTs) are the gold standard for proving causality, they are often impractical or even impossible to conduct. However, there’s a growing number of observational research to mimic RCTs and explore causal inference. One prominent theory is the counterfactual framework developed by CAUSALab at Harvard.

It’s important to note that I don’t view causal inference as the only valuable type of research in the medical world. Traditional association studies still provide valuable insights and can generate hypotheses for further investigation. In reality, some information, not those key factors, may be missing in the existing database, which is not a reason to stop conducting research. By thinking from a causal inference perspective from the very beginning, we can design studies that better isolate the true impact of specific factors. It helps us to frame our research questions more effectively and to interpret the results with a deeper understanding of the underlying mechanisms.

This chapter will summarize the key ideas and methods presented in the book “What If,” which explores this exciting new frontier in causal inference.