gms | German Medical Science

The notion of “big data”, a term roughly describing the collection and mathematical analysis of huge amounts of (unstructured) data, is currently used as public projection surface for numerous novel hopes and old anxieties. Proponents in the private sector realise that data analysis methods can partially predict and influence consumer behaviour in their favour. The public sector hopes to be able to control citizens in a more efficient manner. Opponents fear that both privacy and individual freedom are endangered by public and private data collection and analysis efforts and speak about “1984 on steroids”.

But what is the utility of the procedures summarised by the buzz word “big data” for empirical science? In general, many scientists hope that huge data amounts can help to uncover relationships of which no one thought before and fuel huge simulations, such as those needed in climate prediction. In medicine, our topic today, it is expected that new insights in biomedical, translational and clinical research can be obtained, and that the planning and delivery of medical care can be improved.

Using an example from the health insurance, it is shown that the real-time analysis of properly structured medical every-day data can yield great benefits for the planning and safeguarding of medical care supply by reducing health care costs and moral hazard in care delivery and consumption.

We then explain that for the generation of novel medical insights in basic research and clinical applications, the pure “big data” approach neglecting data structure and hypotheses-driven thinking brings little benefits. This is illustrated using examples from the areas of systems biology and clinical trials: in systems biology, screening experiments have yielded massive data amounts, but the gain has been very limited. Successful research has been based on comparatively well-understood, simple models and the evaluation of carefully planned, hypothesis-driven experiments.

In clinical medicine, “big data” has been associated with the concept of “real world trials”, in which pharmacologic interventions are delivered in a setting that is much closer to the clinical usage of drugs than a classically controlled clinical trial. Evidence is now emerging that although these trials are interesting for health economic outcome research, their usability to formally prove the efficacy and safety of therapeutic regimes is limited. Other approaches to use “big data” for medical purposes, e.g. the mining of social media data for drug safety purposes, are still in their infancy and their potential yield is still unclear.