Abstract
Clinical evidence should guide our actions in everyday care of our patients. We are experiencing remarkable advances in stroke management leading to saved lives, improved functional outcomes, reductions in societal costs and positive incidence trends. However, many areas of stroke care still lack evidence. Clinical trials have shown the value of stroke unit care and/or applying a bundle of care. Still, lack of identification of the value of common single interventions like, for example, early mobilization or blood pressure control remains a stumbling block for continued improvement. In this issue of the journal, Churilov et al. 1 present a timely and comprehensive report on rigorous approaches and methods for stroke researchers on how to use observational data when trying to solve a clinical problem.
To change guidelines, we need to establish causality. A randomized trial remains the golden standard when we want to establish evidence for superiority of a treatment. Randomization, well performed, eliminates the problems of matching individuals with distinct properties between treatment and control groups. We do not need to know about for example, genetics or environmental factors for every single subject. However, performing clinical trials requires important resources which in many cases is a challenge for an overly stretched health care.
In Europe, we have seen successful development of national stroke registries. When extensive coverage is achieved, registries provide high quality evidence in areas like implementation of guidelines, follow-up of outcomes and epidemiological trends. Registries still require dedicated resources for mostly manual reporting into electronic systems. Digitization of patient records has long held the promise of providing automated data for development and research. Building data lakes with digitalized health data and using AI have great potential. In a recent move, the European Health Data Space (EHDS) regulation was passed aiming to harmonize electronic health data records across countries and to enable the secure and trustworthy reuse health data for research, innovation, policy-making, and regulatory activities (health.ec.europa.eu).
It should be expected that these developments will further strengthen the area of observational studies. However, observational data even when very sophisticated, must be handled with care, as pointed out by Churilov et al. Data quantity and sophisticated methodology should not be confused with strength of evidence.
While regression analysis has been around for a long time, other tools like propensity matching are now commonly used in the search of evidence using observational data. Another recent development is the increased use of Mendelian randomization 2 which originates from genetic research but can also be applied on any population that contains individual properties, environmental risk factors or data on treatment. A major issue is that this method is often used on open-source data prone to contain uncontrolled bias and leading to numerous studies with low quality evidence. Again, computing a multitude of data does not equate solid evidence of causality.
While we continue and perform randomized trials, we must embrace the possibilities opening with rapidly increasing access to health data, but keep in mind that well established observational methodology including critical appraisal of data quality are essential.