The crisis paradigm: dismissing anomalies and pressing ahead

Thomas Kuhn famously argued that periods of relative stability in science, known as ‘Normal Science’, are characterised by such consistent successes.1 However, these periods of scientific successes are frequently disturbed by anomalies that cannot be explained using the very approaches that produced past successes.1

Indeed, in recent years, a series of comparatively large and high-quality trials have found no or small benefits from different kinds of HFDMPs.4 Recently, the Medicare Health Support Pilot Program trial of over 250 000 patients achieved only small improvements in a third of the outcomes measured, and identified that commercial disease management programmes for patients with HF, diabetes or both conditions, increased costs.5 Lack of benefits have also been found in recent reviews.6 How can these disparities be reconciled with the supportive consensus of current guidelines and past meta-analyses and trials?

Kuhn famously asserted that science does not progress through incremental fact-building, but is developed in ‘paradigm shifts’ in which progress is made only after stability is disrupted by crises in dominant paradigms.1 Even during this crisis phase, Kuhn argues that scientists—still encouraged by past successes—continue to dismiss inconsistent findings as ‘errors’ or small changes, and seek scientific progress using the same approaches that brought past successes.1

Accordingly, in response to the lack of benefits evident in recent trials of HFDMPs, negative results are still consistently attributed to various factors that are taken to ‘mask’ successful HFDMP interventions, including: small effects of the interventions,4 ,7 ,8 excessive information,8 low patient engagement,8 technical limitations,9 weaknesses in follow-up9 or atypically good usual care provided to comparison groups.4 ,7 These responses each draw on well-established methodological principles, and are understandable, justifiable and plausible. However, in the face of increasingly large volumes of inconsistent data, these attributions act to ‘explain away’ anomalies.

Absent in current approaches to HFDMPs is an openness to accept and learn from negative findings, to question the strength, sophistication or specificity of the underlying evidence from reviews and trials, or to pose different research questions in new ways.10 Basic and potentially pivotal weaknesses of past evidence, such as high levels of clinical, methodological and statistical heterogeneity,2 and very poor descriptions of usual care are almost totally ignored.2 In a recent analysis of past reviews,2 only three of 13 meta-analyses even acknowledged clinical or methodological heterogeneity. Scientists continue to pursue progress of HFDMPs via continued reassertion that they are effective or through the resynthesis of poorly described and disparate evidence driven by the same questions and approaches. The underlying premise consistently being to make the case that programmes ‘work’, to identify ‘what type(s) of programmes work best’10 and/or to identify the single or small set of characteristics that are common to universally effective programmes.6

This continuation of ‘normal science’ is understandable, because to do otherwise risks undermining past successes or conceding territory in a climate of fiscal insecurity and interdisciplinary competition, not only between advocates of different non-pharmacological approaches, but also between these approaches and pharmacological interventions.1 Conflicts of interests resulting from past advocacy of particular types of programmes risk prevailing over openness or willingness to question the consensus or assumptions underpinning past successes.

Revolutionary paradigms: complex and constructive

Yet, Kuhn's work reminds us that negative findings of recent trials are not problems to be dismissed, but are anomalies which should be an impetus for new approaches.1 These anomalies signal the need for an alternative paradigm to understand and research HFDMPs that asks new questions, and is informed by novel approaches to conceptualise what HFDMPs are, and what influences their outcomes.

Programmes should be theorised differently. As complex interventions, the determinants of programme effects should no longer be seen to reside solely in programmes themselves, or in a simple or small set of arbitrarily selected macro programme characteristics. Tools now exist to describe interventions in basic terms—such as, a taxonomy to systematically classify HFDMPs into six component characteristics.11 More sophisticated taxonomies should be developed to comprehensively describe interventions in terms of components, settings, mechanisms, context and theoretical basis. Research should then seek to understand the complex influence exerted by interactions between aspects of programmes, their settings, patients and providers on programme outcomes.

Future research should address fundamentally different questions. The search for a definitive kind of programme that is optimally effective across all types of settings and contexts should cease. While proponents of HFDMP interventions have historically tended to work around and advocate for single types of interventions, this approach more accurately reflects how researchers have become culturally accustomed to framing, debating and pitting interventions against each other, than actual differences in the interventions, or how programmes generate outcomes in the real world.12

Rather, evidence is needed of the likely effects of different types of programmes in different settings combined with an understanding of the mechanisms of these effects to better describe ‘what kinds of programmes work for whom, when and why’.12 This will produce evidence that is more responsive to variations in local settings and is, therefore, more likely to be used by clinicians and decision makers in practice.13 This approach is constructive: findings from trials can generate useful learning for intervention improvement, irrespective of whether programmes were effective or not.12

This new paradigm for HFDMPs is credible: in addition to reflecting Kuhn's historical analysis, the approach we suggest mirrors current reporting guidelines for non-pharmacological trials from CONSORT14 and the Medical Research Council framework for developing, implementing and reporting trials of complex interventions.15 Courage in adopting this new paradigm is now needed, not despite past results, but because of these results.