Article Text

PDF

‘Cognitive biases plus’: covert subverters of healthcare evidence
  1. Shashi S Seshia1,
  2. Michael Makhinson2,3,
  3. G Bryan Young4,5
  1. 1Department of Pediatrics, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
  2. 2Department of Psychiatry and Biobehavioral Science, David Geffen School of Medicine at the University of California, Los Angeles, California, USA
  3. 3Department of Psychiatry, Harbor-UCLA Medical Center, Torrance, California, USA
  4. 4Department of Clinical Neurological Sciences, Western University, London, Ontario, Canada
  5. 5Grey Bruce Health Services, Owen Sound, Ontario, Canada
  1. Correspondence to Dr Shashi S Seshia
    , Department of Pediatrics, University of Saskatchewan, 107, Hospital Drive, Saskatoon, Saskatchewan, S7N 0W8, Canada; sseshia{at}yahoo.ca

Statistics from Altmetric.com

Sir Francis Bacon (1620): “The human understanding when it has once adopted an opinion (either as being the received opinion or as being agreeable to itself) draws all things else to support and agree with it…” (Novum Organum: XLVI)

Introduction

Evidence-based medicine (EBM) was introduced in 1992 as a ‘new approach to teaching the practice of Medicine’.1 EBM's tenets were rapidly incorporated into all aspects and disciplines of healthcare, and its benefits and influence have been widely recognised.2 In 2014, Greenhalgh et al3 referred to the numerous successes of EBM but also drew attention to ‘some unintended consequences,’ and proposed open debate to help improve EBM. Our ‘perspective’ contributes to the dialogue.

Hypothesis

The ‘unintended consequences’ may be explained by: (1) limitations in the tenets of EBM, and (2) flawed or unethical decisions and actions (‘misuses of EBM’2) by several distinct yet interconnected healthcare-related organisations.4 ,5 At the core of both is the covert cognitive syndrome we have termed ‘cognitive biases plus’.4 Qualitative information is summarised and up-dated from a more detailed review,4 ,5 to support the hypothesis.

Definitions and concepts

Cognitive biases plus

Decisions are considered to be cognitively driven by two interconnected neurophysiological systems. System 1 (‘thinking fast’) is primarily mediated by the phylogenetically older limbic system, and involves cognitive processes that are predominantly automatic and unconscious; system 2 (‘thinking slow’) is more deliberate and conscious, with networks in the phylogenetically newer parts of the brain, especially frontal, being principally implicated.6 Most decisions occur primarily at system 1 level.6 Cognition and decisions are inextricably linked with emotion, and may be compromised by factors such as uncertainty, stress, high work load and complexity, sleep deprivation and social or organisational pressures, all common in healthcare.

There is considerable literature on conflicts of interest (CoIs), biases, fallacies and ethical violations (table 1); generally, these have been discussed in a silo fashion.4 We have suggested grouping them syndromically as ‘cognitive biases plus’ for the following reasons: (1) the four entities frequently co-occur in various combinations, (2) they share neurobiological processes (especially related to system 1), (3) many have evolutionary roots and (4) all are potentially influenced by prevailing organisational and sociocultural values. Individually and collectively, they undermine rational thinking, decisions and discourses.

Table 1

‘Cognitive Biases Plus’ and healthcare evidence

Figure 1

The putative ‘cognitive biases plus’ cascade and evidence pathway. CoIs, conflicts of interest; EBM, evidence-based medicine. Double-headed arrows reflect bi-directional reinforcing influences. This figure has been substantially revised from figure 3 in Seshia et al.5 Publisher: John Wiley & Sons Ltd.

Organisations that influence healthcare

In the current context, ‘organisation’ is a functional concept rather than a structural one, and refers to a body of people with common goals, vision and/or mission, often with a hierarchy of authority. The term ‘organisations’ incorporates the individuals in them; these individuals may or may not be employees but typically share or are governed by a common culture including values, beliefs and behaviour.

EBM expert groups

EBM expert groups can be collectively considered to be an organisation; included are not only the founding group at McMaster University (Canada) and the Centre for Evidence-based medicine in Oxford (UK), the late Dr David Sackett being the pioneering influence for both, but also EBM centres established elsewhere by trainees. The achievements of EBM testify to the contributions made by EBM expert groups to evidence-informed healthcare. At the same time, the potential for scientific inbreeding and groupthink increases the risk of shared intellectual CoIs and replication biases.4 ,5 ,7

EBM should be viewed as a ‘continuously evolving heuristic structure for optimising clinical practice’.8 In essence, the tenets are improved as limitations are recognised; GRADE (Grading of Recommendations, Assessment, Development and Evaluation) replacing the original EBM hierarchy being one example.9 Several limitations, including those within the original hierarchy were foreseen from the outset by many outside the core EBM groups;10–14 for example, Feinstein and Horwitz12 predicted all the problems listed by Greenhalgh et al.3 Reservations about systematic reviews,10 ,12 including Cochrane, have also been justified.15–17 We suggest that intellectual CoI, scientific inbreeding and other cognitive biases (table 1) resulted in the inability of EBM experts to anticipate the limitations or accept the ‘outside view.’ Planning and sunk cost fallacies may help explain the delay in rectifying oversights.

Healthcare-related organisations

EBM influences healthcare evidence indirectly through several organisations; those we studied are listed in the bottom extreme right panel of the figure 1, and are likely representative of organisations in most countries.

CoIs and ethical violations have been well documented in industry and among researchers,4 as have cognitive biases at the practicing physician–patient interface.18 Industry is the epicentre of financial CoIs, and its influence pervasive.3 ,4 Decisions by regulatory agencies for drugs and devices can be flawed and oversights inadequate.19 ,20 Cognitive biases of health authorities, especially anchoring bias and groupthink can lead to the imposition of inflexible rules and guidelines.3 CoIs and cognitive biases of reviewers, editors and editorial boards risk publication biases that can consolidate flawed or even fraudulent data;4 ,5 ,7 non-conforming views may be rejected or overridden.15 In addition, patients and their families often introduce cognitive biases into shared decision-making.21

Thus, clinical evidence can be compromised by ‘cognitive biases plus’ at several inter-related organisational levels.

The ‘cognitive biases plus’ cascade

The cascade is a simplified visual summary of several putative inter-linked complex sets of reinforcing primarily unconscious cognitive processes involving several organisations (figure 1). These processes have the potential to create, compound and entrench flawed opinions and evidence that converge to inform healthcare.

The authority bias of EBM can have a herd effect in healthcare-related organisations, reinforcing and entrenching not only the strengths but also the limitations associated with the paradigm. In addition, elements of ‘cognitive biases plus’ within each organisation at individual and group levels, can also subvert evidence and decisions. Collectively, these risk inappropriate healthcare practices at the point of care, examples of which Greenhalgh et al3 have provided.

Discussion

CoIs, self-serving bias, tendency to cheat, authority bias (the α animal), groupthink, social proof and the herd/lemming effect, are among the most fundamental of our evolutionary traits. In humans, intellectual CoI is a strong driver of beliefs and argumentations, potentially introducing subjectivity and fallacies into all discourses.4 ,5 ,22 ‘Cognitive biases plus’ are ‘human universals:’ that is, attributes to which all humans (and therefore organisations), including those involved in healthcare-related decisions, are potentially susceptible. The ‘multi-system failure’ related to anti-influenza drugs19 and the problems in the current approaches to drug safety20 can be explained by the concepts underlying the cascade, several organisations and elements of ‘cognitive biases plus’ (figure 1 and table 1) likely being responsible. The ramifications are multidisciplinary and global.19 ,20 ,23

On this background, suggestions for debate are outlined:

  • ‘Cognitive biases plus’

Decision-making, especially under uncertainty, is rarely rational, and experts are as prone to error as laypersons; however, some individuals may be more rational than others (Stanovich).6 Hence, experts, professionals and organisations must be constantly mindful of being in a perpetual ‘cognitive minefield,’ and welcome the ‘outside view.’6 Cognitive debiasing may help to minimise some biases.18 ,21 The concept of ‘cognitive biases plus’ needs to be validated. The role and management of ‘cognitive biases plus’ in medical decision-making require well-designed prospective studies; these studies should involve all professionals, including biostatistical experts and administrators, who contribute to or influence healthcare evidence, decisions and care.

  • EBM

Any attempt to improve the evidence that informs current healthcare should also include a critical reappraisal of the tenets of EBM.5 Space does not permit the discussion this topic deserves.

  • Organisations

    • The apparent increase of CoIs in the culture of healthcare-related organisations may mirror the erosion of ethics in society.24 An editorial in the BMJ in response to publications in the NEJM reflects both sides of the debate on how to address financial CoIs;23 non-financial and intellectual CoIs may be equally important catalysts of the ‘cognitive biases plus’ cascade, and also need discussion. Potential remedies to address and minimise CoIs and ethical violations should be brainstormed with experts in ethics, law and behaviour research; patients should have a say. Anecdotally, altruistic ethical leaders and role models can catalyse positive change in organisations, an example of the beneficial effects that authority bias and the consequent herd effect can have.

    • The publication industry (print, electronic and preappraised sources of evidence) is the gatekeeper of analysing and disseminating healthcare evidence; high-impact publications have a halo effect and exert authority bias on those responsible for funding, directing, managing and performing research, patient care and teaching.4 Greenhalgh et al's3 call for publishers to raise the bar is justified. For example, safety information in clinical trials is often under-reported in RCTs,9 ,20 even in those published in high-impact journals and formal warning of toxicity or withdrawal of drugs associated with death are often delayed.20 One hundred and two drug studies were retracted from the biomedical literature for the period 2000–2011, the median time to retraction being 31 months (range: 1–130 months):25 frequently, flawed evidence lingers. The responsibility of editors to lead the drive to improve the reliability of information has been acknowledged.23 The open data movement should improve the quality of published evidence.23 All preappraised sources of evidence are not equally trustworthy, and users are advised to always critically appraise the information.9 The use of preappraised sources of evidence through applications (apps) at point of care risks automation bias: consequently, ‘rules’ based, that is, cookbook medicine.3 ,12 ,14 Regulation and quality control are urgently needed for medical apps, as are well-designed studies to better assess risks and benefits.26

  • Healthcare regulatory agencies for drugs and devices across the world must be autonomous and standards global. Resources, including the services of unbiased methodological experts should be pooled. These experts must be capable of (1) setting standards for trials of efficacy and effectiveness with meaningful clinical end points, (2) establishing robust pharmacovigilance systems and (3) detecting the increasingly subtle and harder to detect statistical and methodological manipulations.4 ,20 Benefits and risks should be assessed consistently, comprehensively and continuously throughout the market life of a drug.27

Conclusion

Like all arguments, ours are likely coloured by intellectual CoI and cognitive biases. In addition, the best supportive evidence is qualitative, observational and expert opinion. Nevertheless, there is tenable evidence to support the occurrence of a complex cascade of ‘cognitive biases plus’ that has the potential to undermine evidence and decisions, which inform the EBM paradigm and person-centered healthcare.

The landscape of healthcare problems is ever-changing. Hence, the principles of evidence-informed practice will have to be improved continuously;8 an appreciation of ‘cognitive biases plus’ may help in this endeavour. Evidence alone should never drive clinical decisions;9 critical thinking and appraisal at individual and organisational levels must remain the core of EBM and evidence-informed practice.3 ,5 ,9 ,28

Acknowledgments

The authors are grateful to Dr Dawn F Phillips for her participation in the early brainstorming of the concepts, Dr Molly Seshia for helping to refine the figure and reviewing all the drafts, and to Drs Bill Bingham, Bill Albritton and Ira Lesser for their reviews. Anonymous reviewers drew attention to deficiencies and suggested improvements for revision, confirming the importance of the ‘outside view.’ Flaws are our own.

References

Footnotes

  • Contributors SSS was responsible for the concepts and drafts. All authors brainstormed the concepts, contributed to the contents, revisions and approved the final version of this draft.

  • Funding (1) Presented in part at EvidenceLive, Oxford, UK, April 13–14, 2015 & by invitation at the 2nd European Conference on Person-Centered Healthcare, Madrid, Spain, 18–19 June, 2015; (2) SSS is grateful for support from the department of Pediatrics and the College of Medicine, University of Saskatchewan, and the Children's Foundation of Saskatchewan.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.