Why change can be the hardest word of all...

Dixon-Woods M, Bosk CL, Aveling EL, Goeschel CA, Pronovost PJ (2011).  Explaining Michigan: Developing an Ex Post Theory of a Quality Improvement Program.  Milbank Quarterly: 89(2); 167-205 http://www.kliniskbiokemi.net/pdfiler/DixonWoods2011.pdf

Why read this?

There is little good evidence on how or why service improvement initiatives work – or don’t work.   This paper highlights key findings from one of the most successful patient safety initiatives.  This was a campaign to reduce (and, in some cases, eliminate) harm in a hundred intensive care units in Michigan.  It resulted in remarkable reductions in central line infections and other avoidable harm.   The programme had a clear protocol and documented resources which others can use.   But as the authors note, programmes rarely keep to their original plans.  As other countries and regions try to replicate these results, greater understanding is needed of the real reasons for success in Michigan.

What distinguishes this review is the power of the analysis and elegance of the writing, combining rigorous theoretical knowledge with insights from the clinicians who led the initiative.  The paper combines the talents of two leading medical sociologists with an interest in patient safety, Mary Dixon-Woods from Leicester and Charles Bosk from the US (whose ethnographic account in the 1970s of why trainee surgeons make mistakes has become something of a classic and well worth a read), with the experience and insight from  the participating programme leaders, including Peter Pronovost.    This generates findings which have the texture of lived experience (and battle scars), as well as the benefits of a rich social science hinterland.  

The paper is also interesting for its methods.  Ideally, ambitious initiatives like Michigan should be accompanied by real-time evaluation, including observation of the programme in action.  However,  this was not possible.  Instead, the authors updated the programme’s initial theory of change with wider theoretical learning from social science and insights from key players – both those creating the programme and local participating intensive care units.   This provides a grounded reality to their findings – plus some useful methodology for others to adopt in retrospective evaluations (although general readers may want to skip the parts on ex post theorisation and Bayesian logic). 

What does this study say?

The key findings from this exercise combining theory, experience and more theory to understand how and why Michigan worked apply to many quality improvement initiatives.   Firstly, once enough organisations sign up to a programme of change, more will follow because of known patterns of peer influence (`isomorphic pressures’).   This may sound obvious, but has implications for new programmes – better to engage a large number of organisations at once, than have a few beacon or showcase sites at the start. 

Secondly, success at Michigan depended on strong horizontal links between units to generate the energy and momentum of a social movement.  These relationships were key in exerting peer pressure and supporting each other through change.  

An interesting third point (which probably could not have been articulated by the leaders themselves) was the need to construct blood-borne infections as a social problem that could be fixed.  At the time, infection was seen as inevitable and tricky to solve (`the problem of many hands’) – the programme re-framed this as an avoidable problem, using variation in infection rates mong participating units as evidence of the potential for improvement.   This was put into practice by the project leaders (or `vertical core’) who provided robust scientific expertise and credibility while also able to inspire and engage others – and respond to challenges from sceptics.  To win hearts and minds, the programme combined hard data and robust evidence-based practice with stories – the 18 month old toddler who died following a catheter infection.   This helped to disrupt existing beliefs about harm and potential for change. 

The fourth element identified by the authors was a few powerful interventions at the sharp end to change practice and culture, with much discussion of the checklist.   Popular accounts of the Michigan project reduce its success to the use of a five-point checklist of evidence-based practice (from hand hygiene to use of chlorhexidine before inserting a line).   This paper gives a careful account of the checklist’s social functions, making visible the discrepancies between actual and ideal practice and licensing challenge from nurse to doctor in a way which subverts usual practice.  The authors use theoretical knowledge to explain how a checklist institutionalises good practice by making it routine, or even ritual. 

A fifth key factor was the use of data collected systematically by the central team  (unlike many system improvements where it is generated locally in often haphazard ways) to measure and benchmark. 

Lastly, the programme provided hard as well as soft tactics to ensure engagement, and deter non-compliance, including sanctions by programme leaders.

(Aside – do the authors make enough of the effect of charismatic leaders?  From a few patient safety events, I can testify to the rockstar effect of clinical champion, Peter Pronovost – named as one of Time Magazine’s 100 most influential people in 2008 and holder of a `genius grant’ (McArthur fellowship).  Not many of them around..)

Bite-size take-home message for managers (with apologies to the authors)

Successful programmes need stories AND statistics.