Article Text

Download PDFPDF

Lean thinking in healthcare: a realist review of the literature
  1. Pamela Mazzocato1,
  2. Carl Savage1,
  3. Mats Brommels1,2,
  4. Håkan Aronsson3,
  5. Johan Thor1
  1. 1Medical Management Centre, Berzelius väg 3, Karolinska Institutet, Stockholm, Sweden
  2. 2Department of Public Health, University of Helsinki, Helsinki, Finland
  3. 3Department of Management and Engineering, Linköping University, Linköping, Sweden
  1. Correspondence to Pamela Mazzocato Medical Management Centre, Berzelius väg 3, Karolinska Institutet, SE-17177 Stockholm, Sweden; pamela.mazzocato{at}


Objective To understand how lean thinking has been put into practice in healthcare and how it has worked.

Design A realist literature review.

Data sources The authors systematically searched for articles in PubMed, Web of Science and Business Source Premier (January 1998 to February 2008) and then added articles through a snowball approach.

Review methods The authors included empirical studies of lean thinking applications in healthcare and excluded those articles that did not influence patient care, or reported hybrid approaches. The authors conducted a thematic analysis based on data collected using an original abstraction form. Based on this, they articulated interactions between context, lean interventions, mechanisms and outcomes.

Results The authors reviewed 33 articles and found a wide range of lean applications. The articles describe initial implementation stages and emphasise technical aspects. All articles report positive results. The authors found common contextual aspects which interact with different components of the lean interventions and trigger four different change mechanisms: understand processes to generate shared understanding; organise and design for effectiveness and efficiency; improve error detection to increase awareness and process reliability; and collaborate to systematically solve problems to enhance continual improvement.

Conclusions Lean thinking has been applied successfully in a wide variety of healthcare settings. While lean theory emphasises a holistic view, most cases report narrower technical applications with limited organisational reach. To better realise the potential benefits, healthcare organisations need to directly involve senior management, work across functional divides, pursue value creation for patients and other customers, and nurture a long-term view of continual improvement.

  • Lean thinking
  • Toyota production system
  • literature review
  • realist review
  • health services research
  • continuous quality improvement
  • healthcare quality
  • management
  • qualitative research
View Full Text

Statistics from


Lean is one of the latest management imports to the healthcare sector. Despite growing popularity,1–3 its applicability and utility remain unclear. The term ‘lean thinking’ is based on the production philosophy which evolved at Toyota (Toyota Production System (TPS)).4 5 TPS is based on a set of (technical) practices that focus on maximising value for the customer. Toyota itself highlights ‘just-in-time’ (‘what is needed, when it is needed, and in the amount needed’) and ‘jidoka’ (‘automation with a human touch’) as two important aspects ( Just-in-time is an approach to eliminating waste by making parts and goods available when they are demanded by a customer and in the quantity needed. To achieve just-in-time, parts are moved stepwise in small batches (with an optimal size of one, ie, one-piece-flow) with no work-in-process. When one-piece-flow is not achievable, pull replenishment systems that allow customer demand to initiate production are used. The consumption of an article then creates a demand for replenishment which is signalled by what are referred to as kanban cards. Jidoka is achieved through equipment that automatically stops when quality problems are detected. This allows one operator to visually monitor (through visual signals such as andon display boards) and control several machines and processes simultaneously. These practices furnish the ability to detect and signal problems in real time and lay the foundation for learning and improvement. Fujimoto6 describes TPS as comprising three distinct types of routines:

  • Routines for problem identification make it easy to detect problems or mistakes and indicate a need for improvement.

  • Routines for problem solving help identify feasible solutions AND manage the subsequent change process.

  • Routines for solution retention provide the stability necessary for individuals to internalise solutions through the formalisation and institutionalisation of new standard operating procedures.

While companies trying to adopt the ‘Toyota Way’ often focus on technical solutions to improve processes,7 it is the ability of managers to actively facilitate learning through questions and problem solving rather then telling workers what to do and how to do it that determines success.6 8

Given the challenges regarding safety, effectiveness and value shared by most healthcare systems,9 lean thinking has garnered considerable interest.10 We therefore sought to understand what about lean thinking works (or does not) in healthcare11 as well as why, how and when. To develop this knowledge, we undertook a realist review of the literature on lean applications in healthcare. With a view to assisting clinicians and decision-makers as well as researchers, we set out to clarify:

  1. Which components of lean thinking have been put into practice and in which settings?

  2. What outcomes have been attributed to lean applications in healthcare?

  3. Which components, in which contexts, produce which results? What are the mechanisms that make this possible?

Materials and methods

We found a realist review design, developed from realistic evaluation,12 to suit our research questions. Realistic evaluation is grounded on the belief that social interventions, such as quality improvement initiatives, are complex, and that the way they bring about change is influenced by, dependent on and in turn influences their contexts.12 That is, an intervention (I) in a context (C) triggers a mechanism (M) which generates an outcome (O). Mechanisms thus explain what happens when a particular context plays host to an intervention. A complex social intervention can therefore yield different outcomes in different contexts due to contextual factors at the level of the individual, interpersonal relationships, institutional setting and the wider infrastructural setting.12 13

A realist review synthesises research about how complex interventions work.14 Interventions are attended by explicit and/or implicit hypotheses and assumptions which can be articulated as an initial programme theory, followed, evaluated and then refined, by looking at the inter-relationships between context, interventions, mechanisms and outcomes (CO).13 14

This realist review started with a systematic search in PubMed, Web of Science and Business Source Premier (January 1998 to February 2008). Key terms were: lean healthcare, Toyota way, lean thinking, lean manufacturing, Toyota production system, lean service*, lean process*, lean enterprise, Toyota DNA, lean production, lean healthcare, lean method* and lean principle*. We excluded articles that did not concern patient care, or reported hybrid approaches (such as ‘Lean Six Sigma’). A further selection was made based on the presence (or absence) of empirical data about lean applications in healthcare. We included additional relevant articles cited in the previously selected articles. We then developed, piloted and refined a data abstraction form (see appendix available from the authors) to systematically extract relevant data. Data were collected in an Excel spreadsheet, organised into tables and queried as a database (Microsoft Access; Microsoft, Seattle, Washington).

Using thematic content analysis,15 we organised the data around context, outcomes of lean use in healthcare and components of lean interventions used in healthcare. We then coded and categorised the associated data within each topic using terms taken directly from the articles when available. In the absence of these, we created new descriptive terms. Three of the authors piloted the categorisation together. The first author completed the analysis and conferred with the other authors regarding any uncertainties. On multiple occasions, we revisited the full text articles. The categories of the different topics were compared (data available upon request) with each other and interpreted to identify candidate mechanisms. We then delved deeper into a case with rich accounts of lean applications to further articulate the CIMO configurations and formulate a cohesive theory of lean interventions in healthcare.


The systematic search yielded over 1000 article references, of which we identified 112 potentially relevant articles:

  • 27 empirical studies published in academic journals;16–42

  • eight tutorial articles with some empirical data or examples;

  • 21 tutorial articles and other publications without empirical data or examples;

  • 56 other kinds of publications (letters, commentaries, meeting abstracts, articles published in trade journals, cover stories and news stories).

A search through the reference lists of the identified articles yielded six additional empirical studies (from academic journals or conference proceedings)43–48 and some reports. In the end, 33 (27+6) empirical articles met our inclusion criteria.

Only 10 out of the 33 articles17 25 31–34 39 42 46 47 had an explicitly stated and transparent research methodology. The research methodology of the others was unclear, although some18 24 26 29 38 41 48 did include a description of the intervention and a discussion about the implementation process. While many of the articles did report aspects of the context, very few presented contextual factors at all four levels of the individual, interpersonal relationships, institutional setting and wider infrastructural setting. We did not identify descriptions of contextual changes over time.

Context (C)

We found lean applications in a wide range of organisational settings, clinical specialities and healthcare fields (see tables 1 and 2).

Table 1

Healthcare settings in which lean has been applied

Table 2

Clinical specialities and healthcare fields in which lean has been applied

Most of the articles report applications limited to specific processes within one unit or department. While parallel implementation efforts occurred simultaneously in some organisations, in only a few were organisational boundaries crossed.

Despite the great variety of settings, we identified some common contextual characteristics (independent of clinical specialty or field) of relevance to lean implementation efforts: a need/willingness to improve the organisational performance, unclear procedures and staff unaware of problems, workarounds, multiprofessional and hierarchical organisations, status differences, physician autonomy, inconsistent team communication, a culture of blame.

Components of lean interventions used in healthcare (I)

Many different lean interventions were presented in the articles. From these, we discerned four general components of lean thinking in use (table 3):

  • methods to understand processes in order to identify and analyse problems;

  • methods to organise more effective and/or efficient processes;

  • methods to improve error detection, relay information to problem solvers, and prevent errors from causing harm;

  • methods to manage change and solve problems with a scientific approach.

Table 3

Examples of lean tools and methods used in healthcare

Each component includes a number of tools and methods used in combination. Each component was used in a variety of different settings.

Outcomes of lean use in healthcare (O)

All the articles report successful lean applications. The most common areas of improvement included time-savings and timeliness of service,16–20 23–32 34 35 40 41 43 44 46–48 cost reductions or productivity enhancements,16–20 24–26 28–32 34 35 39 47 48 and several quality aspects including reduction in errors or mistakes,19 24 29–35 43 45 46 48 improved staff16 18 24 25 29 35 43 45 and patient satisfaction,18 25 29 45 47 and reduced mortality.34 44 Interestingly, reports about time-savings do not include how the time saved is redistributed in terms of redeployment or reduction in staff or other assets. Easily quantifiable benefits such as time or cost reductions were reported most frequently and often together. Other dimensions, such as patient or staff satisfaction, were presented anecdotally and without systematic measurement.

A number of intermediate outputs were also reported. These included the reduction of steps in a process,24 47 48 reduction in staff walking distance,29 35 increased process understanding,18 21 23 41 42 staff engagement and willingness to collaborate,22 26 34 37 38 calmer and more focused working environments,24 45 reduced time to resolve error alerts,22 34 37 increased number of signalled errors22 37 and improved teamwork.38 44

Candidate mechanisms for how lean works in healthcare (M)

Based on our review of how the components were used in the various settings and the resultant outcomes, we can begin to assemble candidate mechanisms. We propose one programme mechanism for each lean triggering component.

Candidate Mechanism I (triggered by Component I): Using methods to understand processes and to identify and analyse problems generates a shared understanding

Lean offers methods to conceptualise clinical work as a process that creates value for its customers, be it patients or other healthcare providers. In a context where staff are unaware of problems, where there are variations in work processes or an absence of routines, these methods enable problem solvers to recognise and diagnose the problems and to determine how they cause variation and ambiguity.18 24 34 35 Developing a shared understanding of what is important helps members of different professions to communicate and see how their roles and their work relate to the bigger picture. Jimmerson et al report that:

…staff for the first time began to see the waste they lived with every day and had previously assumed was part of their job. They also put a dollar value on their wasted time and began to realise the fiscal significance of addressing small problems.24

Candidate Mechanism II (triggered by Component II): lean methods generate practical suggestions for how to organise and design work flows and the working environment in the local setting in order to improve effectiveness and efficiency

All the articles describe cases with ineffective and inefficient processes of varying degrees. We found that similar methods were used in different settings to address problems and that they yielded concrete and easily implementable suggestions.24 34 This indicates either the existence of similar needs (eg, reduce excessive inventory, delays and waiting times) irrespective of setting and/or a flexibility of the lean methods of component II. A kanban replenishment system, for example, has been used in diagnostic units16 30 43 and also in nursing and endoscopy units26 29 to ensure just-in-time replenishment of supplies which reduces inventory. For example,

…when two disposable biopsy forceps are used during a procedure, the two biopsy forcep wrappers are placed in the plastic chart holder, which alerts the technician to replace those products.26

Only two of the articles describe changes in how staff work in terms of team composition17 or multidisciplinary task training.34 This is perhaps due to legal constraints in healthcare which delineate task responsibility and make it harder to transfer this responsibility between professions.

Candidate Mechanism III (triggered by Component III): methods that improve error detection work by making the process more explicit and staff more aware of how things should be done, thus improving process reliability

Lean offers different ways to enhance stakeholders' ability to detect deviations from ideal processes by making explicit the way things should be done. Simple techniques like colour wristbands with text to visualise and signal special conditions in patients, such as allergies to certain drugs, can be used to reduce communication errors.22 Adherence to standard procedures was improved by introducing directives and worksheets outlining guidelines and working procedures,27 34 48 implementing self-monitoring tools such as checklists,34 39 40 45 and adopting standard terminology.33 Clarity about how things should be done makes workarounds or a lack of routines more noticeable and enables stakeholders to promptly address deviations and workarounds.

The standardised practices allowed variations to be easily identified, so their consequences could be contained before they propagated into an infection.34

Enabling staff to easily identify process deviations allows them to be transformed into opportunities for continual learning.34

Candidate Mechanism IV (triggered by Component IV): a team approach to problem solving can create a shared understanding of a problem, how it can be solved and an acceptance of the subsequent countermeasures. The presence of stable and systematic approaches to problem solving can nurture a culture of continual improvement and learning.

Multidisciplinary process improvement teams were generally composed of staff and management representatives, but also included experts in the form of facilitators, lean experts, problem solvers or coaches. The approaches ranged from creating ad hoc teams to solve a specific problem to more stable and systematic ones. An example of the latter is a ‘learning line,’

…a model line created by taking a small representative horizontal slice of the organisation… In these lines, [a] few people can do work, encounter problems in doing that work with high frequency at low cost with rapid feedback, test in actual use changes in work (countermeasures) that will remove the root cause of the problem, and, hence, improve quality, safety, cost, or responsiveness.35

A team-based (collaborative) approach for systematic problem solving reinforces the understanding and values which can transform an error into a learning opportunity. When problem resolution required authority and information beyond that of the front-line staff, stable structures involving managers were effective.22 34 The Virginia Mason Medical Center's Patient Safety Alert (PSA) system requires staff to immediately report actual or potentially harmful situations using a 24 h hotline.22 37 The vice presidents and medical chiefs have all made a commitment to drop what they are doing and go with the patient safety specialist to the floor to see and observe the problem at its source and perform a root cause analysis within 24 h to 1 week.22 37 This strengthened the belief among staff that errors are preventable and changed a culture of blame into one of safety and continual improvement.22 34

Further articulating the CIMO configurations

We have identified four general mechanisms involved in the successful application of lean in healthcare. However, to deepen our understanding, we re-examined the articles for a specific case which would allow us to articulate more fully the inter-relationships between context, interventions, mechanisms and outcomes.

The Allegheny General Hospital (AGH) in Pennsylvania, USA, is part of the Pittsburgh Regional Health Initiative, which has developed its own version of TPS, ‘Perfecting Patient Care’ (PPC). AGH applied a top-down intervention to reduce central line-associated bloodstream infections (CLABs) in two ICUs which led to dramatic and sustainable reductions in mortalities.34 Table 4 illustrates how different components of the lean initiative interact with specific aspects of the context to activate certain mechanisms which in turn yield specific outputs.

Table 4

Allegheny General Hospital CIMO table (illustrating what about lean worked in a specific context and why)

This case allows us to tease out in more detail the CIMO interrelationships by illustrating that:

  • Context is more than just the institutional setting. The description of the context is different in each row because different aspects of the same setting interact with the intervention to activate and reinforce mechanisms.

  • Context is dynamic. The outputs of one intervention can alter the context and create the conditions necessary for activating the next mechanism. For example, team-based observations of line placement and maintenance (component in row 2) lead to awareness of variations in central line practice (outcome in row 2). This awareness becomes part of the context (rows 3 and 4) and is a necessary precondition for, among other things, interpreting each new CLAB occurrence as a learning opportunity (mechanism row 4).

The case also illustrates a number of ingredients for sustainable results. Lean goals were linked to the hospital's overarching goals. Management involvement in rapid problem investigation mitigated organisational hierarchies. Establishing uniform practices and educating new and rotating staff in both process improvement and new techniques helped institutionalise the collaboratively designed solutions and reinforced a culture of learning. This was further aided by other simple measures such as constructing vignettes which enabled nurses and physicians to overcome emotional hinders and more openly discuss complications and their consequences. Safety values were also emphasised in weekly working sessions.

Discussion and practice implications

Formulating a cohesive theory of how lean interventions work in healthcare

The success of lean thinking in healthcare rests on the ability to orchestrate a complex intervention process that incorporates and integrates multiple variations of four types of components. The development of a shared understanding among different professionals about healthcare as a process allows staff to collaboratively develop and design more effective, efficient and stable processes and makes deviations easier to detect and counteract. This impacts work practice and culture, enabling further improvement. As illustrated by the Allegheny case, these improvements can be reinforced through education and stable structures for continual improvement which resonate with the macro, meso and micro systems' vision,50 and through concrete and palpable demonstrations by management that improvement is everyone's responsibility.

That the four mechanisms impact and change the context suggests that becoming ‘lean’ is an evolutionary process over time and not just a matter of putting in place the four components, something which resonates with previous research on introduction of improvement strategies in healthcare.51 Toyota itself developed lean over decades in response to organisational, managerial, political, and competitive changes and challenges.6 Their competitive advantage appears to lie in this evolutionary learning capability. This

…dynamic capability encompasses making good decisions, learning from mistakes, and grasping the competitive benefits of unintended consequences. Manufacturing companies that survive for decades don't succeed just because they implement the right systems or routines at a certain point in time; they also have a long-term ability to generate effective routines even without prior knowledge of their competitive effects.6

Improving the application of lean in healthcare

Tools and methods are important at Toyota, but even more important is the development of a ‘routinised learning capability’ based on stable practices for problem identification, problem solving and solution retention.6 Echoing Radnor and Walley's analysis,52 we found that most organisations limited themselves to adopting specific lean techniques to solve a target problem within one specific unit or department. But Toyota has learnt that given a state of constant change and flux, the best one can hope to do is develop countermeasures that will only ‘serve until a better approach is found or conditions change.’8

To improve the effectiveness of lean thinking in healthcare, organisations are well advised to:

  • Implement methods that help relay information to problem solvers and create stable structures for continual improvement. In this way, standards can be improved, learning enabled, and positive results sustained.6 In other words, avoid commissioning ad hoc ‘lean teams’ to address specific problems as these will hardly yield a culture where quality improvement becomes everyone's responsibility.53

  • Engage management in continual problem solving. In contrast to the large fraction of time Toyota managers spend on problem solving,8 54 only a few of the articles we reviewed mentioned management involvement in rapid problem investigations. Letting staff enthusiastically identify and solve problems on their own, as some advocate,55 is contrary to Toyota's approach to lean,8 and dissuaded in other healthcare studies.56 57 In fact, management often misunderstands the role they have in lean implementation.58 Part of the challenge lies in helping senior management reject quick-fixes7 55 59 in favour of addressing root causes with a long-term philosophy.10

  • Embrace a more holistic approach and connect to the larger context. Virginia Mason and the Pittsburgh Regional Healthcare Initiative renamed lean, as they translated it to their local contexts and integrated it with their organisational visions—the ‘Virginia Mason Production System’22 29 37 and ‘Perfecting Patient Care,’ respectively.16 21 32 34 35 43

Study limitations

We found the realist review approach useful and challenging. Disentangling patterns of CIMO relationships proved difficult because of unclear (and many times absent) study designs or outcome measures, a challenge we share with Vest and Gamm.60 The low number of studies specific to a certain setting, the limited data on contexts, and the variation in terminology, tools and methods used, made it hard to identify which aspects of lean worked best, in which settings, and in what way.

We suspect a publication bias, since all articles report positive results; surely there are (instructive) failed lean applications waiting to be studied. Furthermore, very few of the articles discuss any limitations to the application described, to the study design or to the generalisability of the study findings. Given this, we could not find conditions that enabled or constrained the success of lean in the same manner in which Greenhaalgh et al did in their realist review of the mechanisms behind the success or failure of school feeding programmes.61

All these aspects suggest that the field is perhaps not yet mature enough for a realist review. Inspired by Denyer et al,13 we tried to deal with this ‘immaturity’ by studying how lean interventions interact with contextual aspects common to healthcare instead of treating each setting as a unique context. An orthopaedic ward is different from an internal medicine ward, and medical humour is based on caricaturing differences between specialities, but what are interesting are the contextual factors that influence the interventions. Thus, we described the mechanisms at a level of abstraction which holds true to the components of the interventions and at the same time is true to the context. This is convincingly demonstrated by the AGH example. Indeed, as Greenhalgh et al argued in their realist evaluation of a ‘whole-scale transformation of a health service’ in London, mechanisms of change can only be found at a relatively high level of abstraction and through an interpretative process.62 However, we recognise that the same interventions will probably work differently within each case, depending on the tools and methods chosen, the implementation process, the context and the observed outcomes.

Narrowing the scope of the review to one lean tool or method or to one category of setting or clinical field could possibly generate more detailed CIMO configurations. However, given lean's inherent holistic nature, this approach might also lead to a rather limited understanding of lean.

Future studies of lean could focus on particular aspects, such as the role of management to improve implementation and sustainability. Moreover, learning could be enhanced by employing more rigorous approaches to research and reporting, for example the SQUIRE guidelines.63 Realistic evaluations12 could help practitioners and researchers alike better understand ‘how’ lean applications work (or do not) in interaction with each local context. Neglecting this interaction risks producing merely a collection of descriptive articles, a concern Walshe has raised regarding improvement research in general.64


Lean thinking has been applied successfully in a wide variety of healthcare settings. Benefits include improved quality, access, efficiency and reduced mortality. While lean theory emphasises a holistic view, most cases report narrower technical applications with limited organisational reach. To better realise the potential benefits of lean, healthcare organisations need to increase solution retention, involve senior management, work across functional divides, pursue value creation for patients and other customers, and nurture a long-term view of continual improvement.


The authors thank C Wannheden at Karolinska Institutet who developed an MS Access database for the data in this study. They also thank the anonymous reviewers who gave valuable feedback on an earlier version of this manuscript.


View Abstract


  • All authors had full access to all of the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

  • Funding This study was performed as part of a project on innovation implementation in healthcare, funded by the Swedish Vinnvård programme. PM and HA's salaries were covered by the project funding. The other authors were supported by their respective employers in conducting this research as part of their work.

  • Competing interests None.

  • 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.