All BEMC Talks will be held online as Zoom webinars until further notice. Be sure to register early to secure your spot! Selected BEMC Talks will be recorded and posted on our YouTube channel.

BEMC Talks are generally held on the first Wednesday of each month at 4pm Berlin time (GMT+01:00) with some adjustments for holidays. They consist of an hour-long lecture followed by an interactive group discussion with the speaker.

2021 BEMC Talks

Date	Speaker	Talk title & description
January 13th, 2021	Matthew Sperrin, Manchester	When is causal thinking needed for prediction? When prediction models are implemented, they are often interpreted causally. For example, a model that predicts mortality in hospitalized covid-19 patients might be used to triage patients – with patients at low risk being considered for discharge. However, a patient may be predicted as low risk because of the care that similar patients received in hospital in the past – which would not be equivalent to the care that they would receive post-discharge. In this talk, I will describe some of the problems with misinterpretation of prediction models – and introduce some of the emerging methods to develop ‘counterfactual’ prediction models. > Watch webinar recording
February 3rd, 2021	Stefan Konigorski, Potsdam	Digital N-of-1 trials: connecting personalised medicine and population-level studies? Traditionally, guidelines and recommendations on health interventions have been obtained from studies of large groups of individuals. However, the derived average intervention effects do not allow meaningful insights on whether an intervention will help a given individual. N-of-1 trials are multi-crossover randomised controlled trials in a single participant which allow investigating individual-level intervention effects. Series of N-of-1 trials can then be aggregated to obtain population-level effect estimates. In this talk, I will present our open-source platform StudyU (arxiv.org/abs/2012.14201) which allows to digitally design, publish, and conduct N-of-1 trials. I will describe appropriate statistical methods for the analysis and discuss how this platform might be helpful to bridge individual-level and population-level studies. > Watch webinar recording
March 3rd, 2021	Anthony Matthews, Stockholm	Comparing effect estimates in randomised trials and observational studies The ability for observational data to deliver similar results as a trial that asks the same question about the risks or benefits of a clinical intervention can be restricted not only by lack of randomization, but also by study design and limitations of available data. This talk will give a brief introduction to the target trial framework, which is the application of design principles from randomized trials to the analysis of observational data. We will then use Swedish register data to apply this framework to a question on the effects of anticoagulant therapies under percutaneous coronary intervention, which will help us understand how and why differences may occur between results from randomized trials and observational analyses. > Watch webinar recording
April 7th, 2021	Sonia Boender, Berlin	Social media for public health #SoMe4epis Social media have become an integral part of our society, and therefore belong in the domain of public health. Sharing of content, the way we seek and judge information, and interaction on a personal and professional level is taking place both on- and offline. The flexibility and speed of social media provide new opportunities for professional networking, outreach to communities, and the way we conduct and disseminate research. This talk aims to share the unique advantages and functionalities of social and platforms that are not always well known to epidemiologists. Ultimately, this talk aims to address common fears, and encourages epidemiologists to use social media for networking with colleagues, as well as engagement in science communication with the general public. > Watch webinar recording
May 5th, 2021	Eleanor Murray, Boston	The circle of life: epidemiologic methods for dealing with treatment-confounder feedback In epidemiology, many (if not most) of the exposures of interest occur more than once during an individual’s life. These sustained or recurring exposures are more complicated to analyze because the reasons for exposure or non-exposure may change over time, and even worse, may be affected themselves by prior exposure patterns—i.e. treatment-confounder feedback. This feedback can create bias in our attempts to understand causal effects even in randomized clinical trials. In this talk, I will describe the problem that can arise when attempting to use traditional methods, how to think through whether this source of bias may be occurring in your study design using directed acyclic graphs (DAGs), and explain modern causal inference techniques for minimizing treatment-confounder feedback in epidemiologic research including inverse probability weighting and the parametric g-formula. > Watch webinar recording
June 2nd, 2021	Sabine Gabrysch, Berlin	A cluster-randomised controlled field trial in practice: the Food and Agricultural Approaches to Reducing Malnutrition (FAARM) study in Bangladesh The FAARM cluster-randomized trial tests the hypothesis that integrated agriculture, nutrition and hygiene interventions can reduce undernutrition when children benefit in their crucial first 1000 days. We included 2700 young women in 96 settlements in rural Habiganj District, Bangladesh in an impact evaluation of a Homestead Food Production program implemented by the NGO Helen Keller International. Furthermore, we assessed the program impact pathways to discern how any impact is achieved. The study design, field implementation and first results will be presented, as well as an overview of interdisciplinary add-on projects. > Watch webinar recording
July 7th, 2021	Sabine Oertelt-Prigione, Nijmegen	Current methodological challenges in the study of sex and gender in health research The study of sex and gender in biomedicine is a relatively new subject. Although data should be broken down “by sex and age”, a sex- and gender-sensitive analysis extends beyond this basic level and can add much value to our understanding of inequities in healthcare. We will discuss how sex and gender can be operationalized in more detail and how to address potential challenges. > Watch webinar recording
August 2021	No BEMC Talk	Summer break
September 1st, 2021	Ben Van Calster, Leuven	The ‘enemies’ of reliable predictive analytics In this talk, I will present an overview of the challenges of reliable predictive analytics. When in a cynical mood, these challenges may be labeled enemies. I divide the challenges into three categories: (1) general, (2) model development, (3) model validation and implementation. The issues I will discuss contribute to the reality that few models make it to clinical practice, and that those that do may not have optimal performance or impact on patient outcomes. Greater awareness of these challenges among all stakeholders (researchers, funding agencies, health professionals as end-users, and all of us as potential patients) may help to improve research and lead to better models that may have greater utility for society. > Watch webinar recording
October 6th, 2021	Tracey Weissgerber, Berlin	Taking shortcuts: great for travel, dangerous for writing reproducible methods sections This talk will focus on strategies for writing transparent and reproducible methods sections. We’ll also examine the following questions: 1. Why do authors use citations in the methods section of a paper? 2. How often do authors replace detailed descriptions of methods with “shortcut citations” to previous articles that used a similar method? What problems occur when readers consult shortcut citations to find methodological details? 3. How can we change journal policies to encourage transparent and reproducible reporting of methods?  Data shared during this talk are part of a study conducted by Berlin graduate students during a participant-guided, “learn by doing” meta-research course. > Watch webinar recording
November 3rd, 2021 Webinar registration	Sebastián Martínez, Glasgow	Causal inference in the presence of interference: generalised propensity score application in public health Traditional causal inference methodologies are based on the assumption that treatment does not spill over between units. There are several applications in which this assumption does not hold and the causal inference community has risen to the challenge, significantly expanding the literature that looks at this problem during the last 10-15 years. In this talk we are going to look at some of the different ideas and focus on disentangling main effects from spillover effects in observational settings using a generalised propensity score methodology developed by Airoldi et al. (2020). We use simulations to explore some of the boundaries of this particular method and apply it to a public health peer-driven intervention.
December 1st, 2021	Cecile Janssens, Atlanta	TBA