Medical Section meeting: Data and statistical methods in health

Thursday 20 June, the RSS Medical Section held a meeting titled 'Challenges and opportunities of data and statistical methods in health research and practice' at the Royal Statistical Society in London. The meeting was held in honour of the late, great Professor Doug Altman and was chaired by his long term colleague and good friend, Professor Martin Bland.

Professor Doug Altman was often the first in line to take on a new challenge or seize an opportunity relating to statistical methods or data in healthcare research and practice. The four invited speakers at this scientific meeting highlighted many challenges and opportunities in healthcare research and practice, both past and present, as well as a wide range of statistical methods to take on these challenges, many of which involved or were inspired by Doug.

Taking inspiration from the work of Matthews, Altman et al published in 1990, the first speaker, Professor Toby Prevost (Imperial College London), spoke about the 'Analysis of serial measurements in medical research and in clinical trials'. Toby described the application of the themes in 1990 article to the design of individual and cluster randomised trials and highlighted the parallels, yet very important differences, in the statistical approaches required for correlation due to clustering or due to serial measurement in a clinical trial setting.

The second speaker, Dr Camila Caiado (Durham University), spoke about statistical prediction model methodology meeting mobile app technology in a talk titled 'Developing planning tools for GP practices: Using population and healthcare records'. Camila provided a live demonstration of the mobile app which can show what the demographic and clinical characteristics of current population looks like and what the new population could look like. Such information is essential for GP practice managers, charged with allocating scarce resources.

The third speaker, a former GP and now a clinician with research interests in inner city general practice, Dr Mark Ashworth, spoke about 'Some problems with primary care data'. While primary care data is a rich statistical resource, there can be considerable problems for statistical analyses which may produce misleading data outputs. With many national databases, such as CPRD, THIN and QResearch as well as local databases, are these databases competing with or complimenting each other? Is primary care data consistent across all of these data bases? It is no wonder that, when GPs communicate risk to patients using numeric estimates such as relative risk, absolute risk and number needed to treat - which all give different perspectives of risks, as well as potentially misleading information from primary care resources - that patients may reject statistically based advice.

The final speaker of the evening, Professor Gary Collins (University of Oxford), a former student of Doug Altman, promised an interesting and potentially controversial talk on 'Issues in prognosis and prediction in medical research' to end the meeting. Gary highlighted that the clinical prediction model field is a very busy one, with hundreds of models published in disease areas such as cardiovascular disease, obstetrics, prostate cancer and traumatic brain injury. However, the quality of many of these models in terms of how they are reported, whether they have been validated and whether they are actually useful is questionable. Gary also spoke about the recent rise of machine learning algorithms in the clinical prediction field and raised concerns about transparency and reproducibility of these algorithms. The take home message for the audience: machine learning algorithms may choose great (or terrible) Netflix recommendations for us, but should we be risking patient outcomes with these methods...?

Throughout the meeting, chair Martin Bland entertained the audience with tales of his good friend Doug Altman, from the story that their first written work together was actually about real ale to the huge achievement of their Lancet paper in 1986 which has, to date, been cited over 40,000 times.

Martin closed the meeting thanking his dear friend for his vast contribution to improving medical research, with the backdrop of the words of Professor Doug Altman: 'To maximise the benefit to society, you need to not just do research, but do it well'.









Load more