Coffee is bad for your heart. Coffee lowers your risk of a heart attack. Both headlines may be based on real research-and still contradict each other. What does that mean for us? Which studies can we trust? And how can we tell whether an association really describes cause and effect?
These are the questions explored in the new episode of Einstein on Air - Große Fragen, Berliner Antworten. Our guest is Jess Rohmann, an epidemiologist at QUEST for Responsible Research at the Berlin Institute of Health at Charité. She studies the quality of research data and develops methods to help us draw more reliable insights from health data.
Epidemiology doesn't happen only in the lab. Jess Rohmann works with large datasets from healthcare settings, such as hospitals and emergency medical services. But one thing is always clear to her: behind every number are people and their health. Making sense of data therefore requires close collaboration with doctors, statisticians, public health researchers and other specialists. Only by bringing data analysis together with medical expertise can we ask the right questions-and find robust answers.
The conversation shows why observational studies can be so challenging. Take this example: patients discharged home after a stroke seem to recover faster than those who go into rehabilitation. Does that mean being discharged home causes better recovery? Not necessarily. Perhaps those patients had less severe strokes to begin with, were in better physical shape, or had more support from family members. These so-called confounding factors can influence both the discharge decision and the course of recovery.
To make these relationships visible, Jess Rohmann uses causal diagrams known as DAGs-"Directed Acyclic Graphs". She explains how easily correlation can be mistaken for causation with the seemingly quirky example of ice cream sales and sunburn: on hot, sunny days, more ice cream is sold-and more people get sunburned. Of course, the ice cream doesn't cause the sunburn. The common cause is the sun. This kind of reasoning matters greatly in health research. Many factors cannot simply be studied in a randomised trial: no one, for example, can be required to smoke or breathe in asbestos. Observational data are therefore indispensable-but they call for especially careful study design and statistical methods.
Coffee consumption is another example of how different studies don't always ask the same question. How much coffee do people drink? Under what living conditions? What role do stress, sleep, smoking and other health habits play? Jess Rohmann urges us not to judge studies by their headlines alone, but to look more closely: What did the researchers actually ask? How were concepts and data defined? Were the methods suited to the research question? And how transparently were uncertainties and limitations presented?
Another focus of the conversation is peer review-the evaluation of scientific work by other researchers. Through Peerspectives, a course she co-founded, Jess Rohmann teaches doctoral researchers how to assess studies critically and constructively. Her hope for the future is that methodological expertise will be brought into research projects from the outset-not only once the data have been collected or the findings written up.
If you'd like to read scientific headlines more carefully and assess them with greater confidence, listen to this episode. Available here on our website and on all major podcast platforms.
Please note: the episode is in German; English subtitles are available on YouTube.
Go to podcast on YouTube
