Can the concept of Safety II be used to Improve diagnosis?

Woods, David woods.2 at OSU.EDU
Thu Jan 3 13:58:49 UTC 2019

You asked:
Does anyone has ideas or literature about this topic - Safety II - and suggestions how we can use this concept to improve diagnosis in medicine.

Yes there is a base relative to the Safety II view on diagnosis.
First, the metaphor goes back to the early days of the patient safety movement in the report “A Tale of Two Stories” Richard Cook and  I did  [].  All 3 of the “second” stories have diagnostic issues.  The three cases illustrate a couple of points:
1A.  The word diagnosis refers to forms of cognitive work done in systems.
1B. The label diagnosis as used in medical circles refers to a wide range of situations that differ substantially in terms of the cognitive work and processes involved as well as the larger systems issues.  As a result, widely different aspects of cognitive work get confounded in discussions.

2. Safety II says look at what makes cognitive work and joint activity hard and how people are able to carry out the cognitive work well, usually.  This is one of the original goals of Cognitive Systems Engineering when Erik and I started it (with others) 38 years ago.  (see attached on some aspects of its origins).

3A. Since then we have looked at aspects of diagnostic activity in cognitive work systems but much more in other areas than health care.  I got involved in anesthesia in 1989 when Dave Gaba from Stanford said the results from studying cognitive work in control rooms as faults occurred also applied to the operating room.  Thus there is a body of work one could use, but that work is not used much in looking at diagnostic aspects of cognitive work in health care.
3B.  A lot of the recent work on diagnostic aspects of cognitive work is happening in software engineering if you wanted to look at some detailed studies.

4.  The basic model to start from goes back to the early 80’s studies Gaba was referring to — on anomaly response as a set of cognitive work functions that includes and frames diagnostic activities [but even this doesn’t cover a lot of what gets called diagnosis in medicine, e.g., medical imaging brings a rather different set of cognitive/perceptual issues].  The best review of anomaly response is chapter 8 of our 2006 book using space operations examples (attached). Note the chapter never uses the word abduction, but anomaly response is the real world version of the fundamental form of reasoning called abduction which that differs substantially from induction (machine learning is mostly inductive), and deduction (abduction starts with Peirce in the 1930s).  A brief intro to abduction is attached too (others have offered a bayesian reasoning approach as well; there are differences with anomaly response from a global perspective, but bits of Bayes can be seen as part of anomaly response).

5.  The difficulty is that the anomaly response studies say that, fundamentally, diagnostic work is much much more than classification. Yet health care remains more than a little stuck in looking at diagnosis as a classification task that occurs in single episodes with a patient.  Starting with anomaly response would change the approach quite a bit, diversify the set of issues, and widen the potential for helping cope with the complexity of diagnostic work.  For example, just the issue of hypothesis generation would be a very large topic and debate for different health care settings, and that is fairly compact bit of the much larger set (and even here you will then find the approach in health care is a glass half full with basic findings at odds with some widespread beliefs).

6.  The Safety II and anomaly response/abduction root does not find counting diagnostic “errors” as informative or actionable. Ouch. So when you are ready, it is very different way to think about the problems in diagnosis in health care.  As Safety II and cognitive work perspective starts with what makes diagnostic work in context hard, the first step is to look at specific contexts, pull out the cognitive work functions, understand the larger systems factors, develop and test a variety of potential interventions to improve the cognitive work functions (hypothesis generation is one; another is recognizing what is anomalous or unexpected in context).

7.  In health care, a Safety II approach would begin with thinking about the implications of some basics about many health care settings, e.g.:
~ the patient comes with problems — physiology is disrupted and disease processes may be going on.
~ patient physiology and diseases are highly variable,
~ disease processes develop and change over time,
~ uncertainty is always high (what is a “correct” answer may not be knowable or only appears much later in the patient’s course of care),
~ our understanding of disease processes is limited,
~ simplifications due to hindsight block seeing the reality of cognitive work as done.
~ ‘blunt’ end organizational factors play large roles that affect coordination across roles, cross-checks, continuity (or lack thereof), resources, workload bottlenecks and more.

The above points considered as a set mean it is likely initial assessments will be off and incomplete. The key is not being right the first time or every time, but how well the person, team, service or institution can revise as new evidence comes in, as interventions do not produce results as expected, ’steering’ the assessment to better help the patient, … , —  all key aspects of anomaly response in a system.

8.  Reframing ‘diagnosis’ away from maximizing correct classification toward supporting a dynamic process of anomaly response is a very big shift (again abduction/anomaly response does not apply every where in health care). Reframing to anomaly response would be broad change and challenge aspects of physician training/mindset that developed over long time periods as the medical system adapted over decades to constraints such as the ones under #7. Nevertheless, the classification approach is and will get stuck and frustrate progress in many areas, and there are other avenues to make progress despite the constraints that make forms of diagnostic cognitive work hard.


David Woods
Releasing the Adaptive Power of Human Systems

follow @ddwoods2<>

Department of Integrated Systems Engineering
The Ohio State University

Resilience Engineering Association
Human Factors and Ergonomics Society

woods.2 at osu dot edu

SNAFU Catchers Consortium, see results at <>  or video intro at

keynote on autonomy and people see
part 1:   part 2:  part 3:

keynotes on resilience and complexity see

on the Strategic Agility Gap

On Dec 9, 2018, at 8:00 AM, Gerrit Jager <gerrit.jager at PLANET.NL<mailto:gerrit.jager at>> wrote:

Dear members,

I have a question.

Diagnostic error is the ‘new frontier on patient safety’. In thinking about patient safety there has been a shift from Safety I to safety II.  To summarize Erik Hollnagel
‘Safety has traditionally been defined as a condition where the number of adverse outcomes was as low as possible (Safety-I). This means that safety management must start from the manifestations of the absence of safety and that - paradoxically - safety is measured by counting the number of cases where it fails rather than by the number of cases where it succeeds.’

‘Focusing on what goes right, rather than on what goes wrong, changes the definition of safety from ’avoiding that something goes wrong’ to ’ensuring that everything goes right’. More precisely, Safety-II is the ability to succeed under varying conditions, so that the number of intended and acceptable outcomes is as high as possible.  This means that safety is managed by what it achieves (successes, things that go right), and that likewise it is measured by counting the number of cases where things go right. In order to do this, safety management cannot only be reactive, it must also be proactive.’
(White paper about Safety is attached)

I experience the same paradox in our society. It is about Improving Diagnosis (safety II) but our conferences are about Diagnostic errors (Safety I).
It’s like a society that studies “the secrets of a happy marriage” by interviewing divorced couples.

The literature concerning Diagnosis and safety II is scarce.
Does anyone has ideas or literature about this topic and suggestions how we can use this concept to improve diagnosis in medicine.


Gerrit Jager
Officer Patient Safety



To unsubscribe from IMPROVEDX: click the following link:


Visit the searchable archives or adjust your subscription at:

Moderator:David Meyers, Board Member, Society for Improving Diagnosis in Medicine

To learn more about SIDM visit:

<Hollnagel From Safety I to safety II White paper.pdf>

To unsubscribe from the IMPROVEDX:
or click the following link: IMPROVEDX at LIST.IMPROVEDIAGNOSIS.ORG


For additional information and subscription commands, visit:

http://LIST.IMPROVEDIAGNOSIS.ORG/ (with your password)

Visit the searchable archives or adjust your subscription at:

Moderator: David Meyers, Board Member, Society to Improve Diagnosis in Medicine

To unsubscribe from the IMPROVEDX list, click the following link:<br>
<a href="" target="_blank"></a>

HTML Version:
URL: <../attachments/20190103/590a1671/attachment.html> ATTACHMENT:
Name: chapter8_JCS-Patterns.pdf Type: application/pdf Size: 1033065 bytes Desc: chapter8_JCS-Patterns.pdf URL: <../attachments/20190103/590a1671/attachment.pdf> ATTACHMENT:
Name: Origins_C003-corrections.pdf Type: application/pdf Size: 628465 bytes Desc: Origins_C003-corrections.pdf URL: <../attachments/20190103/590a1671/attachment-0001.pdf> ATTACHMENT:
Name: abduction basics.pdf Type: application/pdf Size: 32942 bytes Desc: abduction basics.pdf URL: <../attachments/20190103/590a1671/attachment-0002.pdf>

More information about the Test mailing list