Improving Healthcare in the Era of Precision Medicine
By Douglas A. Colquhoun, MB ChB, from the IARS, AUA and SOCCA 2019 Annual Meetings*
The AUA President’s Panel presented four perspectives on how to improve the delivery of healthcare in the emerging context of precision medicine. Drawn from the fields of statistics, informatics and anesthesia, each speaker explored the range of pressing challenges facing perioperative investigators.
The session was chaired by Michael Avidan, MBBCh FCA(SA), AUA President at Washington University School of Medicine at St. Louis, and featured Regina Nuzzo, PhD, Professor of Statistics at Gallaudet University, Philip Payne, PhD, FACMI, Professor of Medicine and Computer Science and Engineering at School of Medicine at St. Louis, C. David Mazer, MD, Professor of Anesthesiology at the University of Toronto and Bruce Biccard, MBChB, FCA(SA) FFARCI, MMedSci, PhD, Second Professor and Second Chair in Anaesthesia at University of Cape Town.
Dr. Nuzzo examined four problems facing the perioperative investigators and posed sample solutions, which she challenged the field to embrace. In proposing, among other solutions, the use of Bayesian methods, she demonstrated the lack of relevance of arbitrary p-value cut offs via an example in which the audience demonstrated their willingness to believe that a coin was biased and unwillingness to accept that an octopus could predict soccer results, despite similar probabilities! After highlighting examples of how the human mind is able to find patterns in random data, Dr. Nuzzo challenged perioperative investigators to hold themselves to higher statistical standards.
She proposed that blind data analysis, where investigators develop and lock statistical methods on without knowledge of the allocations or on simulated data, may prevent hopeful inferences being drawn. In an extension of this concept, Dr. Nuzzo identified the problem of investigators looking for data, which supports their hypothesis. She suggested the relevance of independent discussion, whereby a third party writes a discussion having read only the methods and results presented to ensure that the conclusions offered are truly supported by the data, an approach recently presented by Drs. Avidan, Ioannidis & Mashour. Finally, highlighting the real-world effects of citation and publication bias, Dr. Nuzzo used a stark example from the scientific literature in the treatment of depression where positive studies ultimately received attention disproportionate to their publication frequency. Dr. Nuzzo argued for the applicability of Registered Reports, whereby study designs and methodologies are subjected to peer review and decision to publication is made in advance of the results being available.
In recognizing that the era of precision medicine will herald new approaches to clinical trial design, Dr. Payne outlined the need for the development of true Learning Health Systems. In such organizations, each interaction with a patient is an opportunity to learn more about the patient, their disease processes and their response to treatment in real time which feeds forward to inform care delivered. Dr. Payne outlined the need for development of sub-phenotypes which encompass and describe the full range of available data about the patient. This would have the effect of moving clinical trial design to points which focus beyond the averages of a population included in traditional RCTs (which may be highly selected and not representative of the affected population at large). Dr. Payne proposed the use of adaptive designs which could define patients who would benefit from particular treatment strategies. He argued this was analogous to the ways in which physicians use their wealth of experience to make clinical decisions for the patients they care for. After carefully recognizing the potential for biased and faulty training data, Dr. Payne noted the role, which artificial intelligence may play in supporting this pursuit of a continuously evolving and adapting healthcare delivery environment.
Noting the temporal gap between evidence generation and the routine application in clinical practice, Dr. Mazer compared two examples of knowledge dissemination. He contrasted the 200-year gap between lemon juice being useful in the prevention of scurvy and its adoption by the British merchant navy and the far more rapid adoption of results from the SPRINT trial into clinical guidelines. From this, Dr. Mazer was able to explore the challenges of dissemination of trial findings into clinical practice via guideline development. He highlighted the reluctance of clinicians to consider single pivotal studies and instead seek for a preponderance of evidence. As a result, he noted that many clinical guidelines diffuse poorly into clinical practice and may rely heavily on expert opinion from non-diverse sources. He closed his talk with a pivot to a discussion on the challenges of trustworthiness in the dissemination of information, and noted the changing definition of a traditionally trustworthy resource in the age of social media-based information dissemination. He challenged the audience to understand and develop new means for public engagement to address this gap.
The panel closed with a challenging presentation from Dr. Biccard, presenting on the African Surgical Outcome Study he has led. Noting that 95% of the population in Africa have no access to affordable and safe surgical care, this initiative attempted to quantify the outcomes of surgical patients across 250 hospitals in 25 countries. He described complication rates, which were significantly higher than comparable rates in North America and significant mortality arising from these complication rates – illustrative of the concept of Failure to Rescue. He described forthcoming work, which will deploy via a cluster randomized design increased postoperative surveillance for complications. Dr. Biccard reflected that the ideal study in this context was one, which was short, simple, offered a practical intervention and required minimal data collection. In the subsequent question and answer session, Dr. Biccard reflected on the potential opportunities for use of mobile phones in the gathering of information and delivery of care given their substantial availability across the continent.
*Coverage from the AUA President’s Panel: The Reproducibility Crisis in the Era of Precision Medicine, presented by Michael Avidan, MBBCh, Regina Nuzzo, PhD, Philip R.O. Payne, PhD, David Mazer, MD, and Bruce Biccard, MBBCh, at the AUA 2019 Annual Meeting