“Developing high standards for learning from data will help build trust in the science”

Berkeley Public Health: Tell me about your research. What about it is most exciting to you right now?

Mark van der Laan: I am a biostatistician and my research is about methods that allow drawing conclusions about causal effects of treatment choices or exposures to environmental factors on health outcomes based on possibly adaptive clinical trials and observational studies.

My research team developed a general approach termed “Targeted Learning” following a principled roadmap for statistical estimation and causal inference. Its success in practice is reflected by UC Berkeley’s Center of Targeted Machine Learning and Causal Inference (CTML).

Our research is all about developing state-of-the-art tools based on super-learning. It’s an ensemble machine-learning approach that maps a large collection of candidate algorithms into a new algorithm that is more powerful. We also utilize targeted maximum likelihood estimation (TMLE) that tailors the estimate to the question of interest fully respecting the particular study design. Such methods are pre-specified to avoid human biases.

We are interested to not only develop formal theory and rationale, but also in bridging the gap from academic tools to tools that can be trusted to make important decisions such as drug approvals, based on real world data.

This requires working on real data collaborations; development of theory and methods to address the challenges of public health type studies; simulations to evaluate, refine, and optimize the practical performance of the methods; and building of pipelines for carrying out these sophisticated analyses that are accessible to the general practitioner.

For me, it has been a long and ongoing journey towards the building of these products and thereby moving towards a society where we have the tools to truly learn from the real world what works and what doesn’t work. This journey has been very exciting and inspiring, and I like to take people along in the challenges we have to address jointly.

Why is public health so important right now?

To achieve an overall healthy life for humans is clearly an enormous challenge throughout the world. Even though lots of medical progress and understanding of public health has been achieved, new challenges arise continuously. For example, the modern world lifestyle has resulted in an obesity crisis that is negatively affecting the health of so many.

Moreover, we cannot separate mental health and happiness from physical health. And it’s not enough to treat health symptoms through various medications or interventions, we need to understand and address the underlying situation.

What do you think the biggest challenge is in public health at the moment?

There are many, but from my professional perspective, dramatically improving the way we generate evidence from real world studies, clinical trials, and sequentially adaptive designs, is a challenge that keeps me fully occupied.

How do you think we can restore public faith in public health?

As a statistician, I believe that developing high standards and benchmarks for learning from data and its reporting will help to build trust in the science. Inconsistent and often unreliable data analyses results have harmed the public’s trust. This is what we are aiming to address with our research and CTML. Of course, this only represents my particular perspective as a statistician, but every professional public health scientist can contribute in important ways to restore public faith in public health.

How has the field changed during the course of your career?

The amount of data and the diversity of data has changed dramatically. These days we have measurements on complete genomic profiles, images of organs such as brains, doctor notes transformed into natural language features, and so on. This makes the traditional statistical methods obsolete and requires an integration of the state of the art in machine learning which has evolved immensely as well over the last decades. The Targeted Learning approach we developed can naturally adapt and integrate all these advances only enhancing the power of the method.

Who is your public health hero and why?

I would say that public health heroes come in all kinds of forms ranging from the people on the front line making a difference at a very personal level to innovative scientists who have the ability to recognize what is needed for true advancement of public health, often quite different from mainstream expert opinions, and subject to enormous resistance. For me, picking out one person would not do justice to the large diversity of public health heroes out there.