Tony Cornford

Department of Information Systems 

London School of Economics and Political Science 

Houghton Street 

London WC2A 2AE 

Yesterday we heard from people who are deeply involved in this field. Today we hear from people from the outside, looking in. So I start with an information systems perspective, to be followed by perspectives from a sociologist, an economist and then a modeller. When we drafted the programme we assumed that each of the presenters would want to rewrite their title, but no one did. So did we get the programme topics right first time?

My intention is to explain the nature of information systems and perhaps through discussion, discover their contribution to understanding the causes of errors and possible means of risk reduction. I and my colleagues at LSE see information systems as a social science and the strapline for my department is 'the social study of information technology'. Therefore we have a stronger emphasis than you might expect on a mix of technology, business and management.

What are information systems? Well, technology is certainly there, but seen in the context  of organisations and the uses they find for it. However at the moment organisations are trouble enough! The other key part of the academic study of information systems (that we almost take for granted) is processes of organisational change. So you very rarely study technology in the context of organisational stasis. This is worth saying and although it may sound airy-fairy, we can easily start talking in a lot of detail whilst not acknowledging that it is almost inevitable that new technologies bring change. Change comes in different ways some strategic, some planned, some of it is technical design which we assume is based on good design but quite a lot is of a different character that I classify as opportunistic and emergent. The first is exemplified by someone saying 'Oh, if I can do that then I can do something else' and yesterday we had examples of technology enabling managerial change to occur e.g. monitoring what nurses are doing on night duty. Emergent is a rather vague, all encompassing term but being neither planned nor opportunistic without any body being responsible. So we are interested in planned and emergent change that accompanies information systems technologies and Anne's case study showed both aspects. Several of yesterday's talks showed that sometimes radical change was talked about - we are going to completely change the way we do this and deliver much better healthcare. Then almost in the same sentence: how are we going to do this? We are going to reinforce the routine - not quite a contraction in terms because by reinforcing the routine you could alter healthcare but it tells you something about the balance of changes in medicine. The levers for change are the routines of doctors and nurses. Change also comes from above and from below so emergent and opportunistic changes using technology coming from below and then there is the top-down change and yesterday we had examples of both. Finally, there are changes from the centres and from the periphery - the global centres of technical expertise. The periphery is where there are people a long way away from the coalface but who somehow manage to innovate and change and that feeds back. The most obvious example is software that now comes from North America and software developed elsewhere is noticeably different because the product comes from another source of ideas. This is also true of medical practices more generally, so the centres where agendas are created, ideas are generated and being pumped out from are interesting.
These centres are a very powerful source but there is the periphery too which sometimes pushes back with its own ideas and has an influence locally.

This is meant to be a slide about information systems as a general discipline that my colleagues and I practice. The other thing the information systems field tries to do is to give some account to technology and sometimes we criticise one another for not doing this sufficiently. It is quite easy to study organisations and how they change with a nominal account of technology and the contemporary view in my field is that we need to get back to technology. How are going to do that? Well one way is to say that technologies are the products of social antecedents involved in their production and which sustain them. American software reflects the American healthcare and social organisation. We should link technologies and their 'construction' to diverse ambitions, interests and social forces. Why we get the technologies we get? Microsoft and open-source software are more than just technologies. You probably need an alliance between multiple interests and forces.

The third bullet in this slide moves it on a bit. Information technology used to be talked of as new technology so the newness is quite interesting because the technologies herald something new and different. Thus accounting for technology is in part accounting for its newness and you can do this by responding differently. Resistance is a classic reaction about which there is a huge literature. Hostility and hospitality - two words with the same roots. Hospitality is a way of handling hostility or the reverse and you have to find ways of accommodating the new 'technology' guests. This is an interesting way of examining how organisations deal with new technology. They have some formal rules of dealing with this stranger or alien and when a new piece of technology is plonked down in a ward the attitude of nurses is that they have an alien present. This is quite a good way of describing their reaction. Domestication is another kind of metaphor because sometimes you can embrace the technology to become an everyday activity and exploring the role assigned and/or adopted - automate, 'informate', ''transformate'. Technologies automate - pretty straightforward - informate is a nasty word but the suggestion is that technologies are there to produce information for people to use. A good decision-support software informates, it puts back to people a picture of the situation and useful levers to help its resolution. Transformate is the really ugly word but it is there because technology has often been given an assigned role, or assumes one, to change the organisation. This is part of the radical agenda for change so in the context of this Meta-network what can we do? I think what we have been talking about can serve the activities of design, implementation and evaluation.

Whilst I am interested in evaluation I am also interested in why it is an unsatisfactory process. I think we can also and almost uniquely have a way of dealing with hybrids of technical and human factors. Being able to cross difficult boundaries is useful. I think also we have an opportunity (coming back to the evaluation issue) to support the search for cause (the traditional evaluation agenda) but also accounts of technology that represent consequence - what the technology becomes - not that the computer has caused something but that computers have certain broad consequences. The last point is to give some credence to the things we say in the margins. The things that do not get in the papers or in the studies and people have said them here and finding a way to pull these comments in may be part of our function here.

Questions

DB
I warmed to your comment about why we get the technologies we get and I can think of several examples in this field that do not meet our needs at all. The companies that make the drug information e.g. First Data Bank have an overriding interest that they do not get sued so they set very high warning levels. Likewise the vendors of hospital I.T. whose main motivation is to maintain their installation base make it very difficult to move in and out very easily of proprietary software and I am sure there are many other examples.

TC
The example I like is the Internet itself, which in a sense is the Cold War made technical and reflects the need for a network that would be immune from nuclear attack on and destruction of several centres. Building a network with this purpose made no financial sense but it made sense in a world were you worried about nuclear war and you wanted a robust network but you reap the whirlwind when you design a network that is immune from nuclear attack.

EK
I am taking what you have said and taking it further and looking not just at the interests that are represented in technology but two other things. How technology changes practices sometimes (implemented specifically to do so and occasionally in a most subtle way) but also technology may frame our understanding of things. After a while our universe closes around and we perceive things in a way determined by embedded understandings and practices in the system. Looking at medical error - it is how medical error is understood in those systems - that might in some way close our understanding. When we think first the ideas are fluid but suddenly become codified in information systems and so our universe has changed our thinking. So people who come after us may only see things only in those terms. It is hard to give practical examples on the spot.

TA
As we were discussing yesterday Root-cause-Analysis requires data that pushes you in certain directions.

EK
The simplest example what is codified is important and defines our understanding for instance looking at accounting systems only through the profit lens.

NB
I think that has perhaps happened with the P systems that Tony talked about yesterday but the construction of risk in the early days was drug interactions. This was not a big a big agenda and not a clear good-bad thing. Now because we are able to research risk in different ways & we know now that misprescribing is one of the biggest sources of errors so we will prioritise a system that stops you giving silly doses and we work back from those needs. Drug interactions are less and less important because drug companies use their DIVINE process screen out drugs to see if they interfere with major enzyme systems. So this type of error is being designed out by the drug producers. You will continue to get interactions because drugs have opposite effects but if you have a case of a difficult diabetic hypertensive then you are using drugs that interfere with one another but still drug interactions dominate the discourse about errors.

BF
When US hospitals select a new form of automation or a computer system they develop a design specification with turnaround times and response times etc. I wonder if the same thing is done here in the UK? Can we devise a screening tool for recommendations about safe computer systems that are not out there yet?

TC
The screen shots that you were showing yesterday tells us that there is a lot of rather elementary stuff to be done it is not long but short specifications that are needed.

TA
That is one of the things that the national I.T. programme is coming up with - pretty detailed specifications for a safe system down to screen design and alert messages with a big red cross in the middle of the screen that can only be overridden by doing X. Yesterday we were talking about the amount of consultations being done by locums who are going from system to system. If we could have some generic way of training people so that you do not have to spend four years working with a system before you realise that the most serious alert displayed is 'Hazard!!'

TC
It is interesting just to go back and conjecture why was the software created that way with that modest hazard alert.

TA
I guess that it was a pretty basic system that came before 'Windows'.

TC
Who was doing it and what were they interested in? Was it a programmer who felt professionally unqualified and had been told to be quiet because they were just a programmer and so felt that a modest hazard was as far as they could trespass on your territory?

TA
The hazard alert is very interesting because research has showed that a yellow warning meaning a risk of death gets the same arousal as a red warning that means a slight danger.

have made a mistake. Sometimes people violate particular rules and protocols and do so for the best of intentions. Of course there are exceptions and the Shipman case has caused a widespread blame to be attached to all doctors' mistakes. There is this widespread assumption that if the mistake has a really bad outcome then with enough control of the situation this could have been avoided. I hope that by the end of this talk I will have persuaded you that this is not the case. With the wisdom of 20-20 hindsight you can look back and realise what could have been done differently but at the time you do not have all the information and know all the options that were later apparent.

I want to put error in context. One famous psychologist said that all psychology is social psychology because peoples' behaviour always is set in a social context. The individual rather than the organisational environment is the focus when an error occurs. I shall talk about modern theories, about the high performing individual and then to widen the picture using Reason's organisational accident model. We start with three types of errors - those associated with skills, rules and knowledge. We use skills automatically and we do not have to think about them because they require very little conscious effort e.g. experienced drivers. Rule-based activities requires some learning and therefore you still have to think a little bit about what you are doing but knowledge-based activities require the most effort because you have to make judgements based on no prior knowledge. Nothing can be done automatically and there are no rules that give the correct results and so much mental effort is necessary. 

What Reason did was to take these error categories and divided them into two further categories. The first are errors of execution, sometimes memory lapses but also due to interruptions. For example a doctor might write the wrong drug on the prescription form when interrupted. Another example is prescribing a specific dose for a drug when you are not so familiar with the dosages prescribed and so make a knowledge-derived mistake. This is because you don't understand the relationship between dose and effect for this particular drug. A familiar situation in I.V. therapy is the lack of standardised choice of pumps. You find that one pump does not work so you get another from the cupboard and make a mistake in administration because you do not know how to use the equipment. Now I want to talk about the individual and how we perform and the factors that influence this. We need a certain level of arousal or stress to perform but if you are feeling tired or are not very awake you are not going to perform too well. When the stress is too much then performance also falls. Doctors on call may not be fully awake by the time the have got to the wards but if we have had breakfast and a shower and travelled to work we are fully awake. The doctor may be hungry or perhaps not feeling 100%. Emotional factors that are preoccupying attention can also affect performance and other aspects are confidence and motivation. Doctors are expected to act confidently in front of patients that they have never met before; the culture of medical education to throw people in at the deep end. The cognitive factor - the thought processes that inform decision-making for instance about the patient diagnosis and the related choice of drug - is also important. Novel tasks may be particularly affected by the environment in which doctors, nurses and pharmacists work. Finally, there is stress, which affects all these aspects.

There is quite a lot to Reason's Organisational Accident Model and I will start in the middle, which talks about tasks and errors, this is where the individual is working. Someone makes an error, which may be a slip, violation or breaking rules (which some people may do routinely). We have already heard about the routine omissions of patients' weights but there are contributing conditions - more about these later. Right at the far end we have organisational and corporate culture, which contribute latent conditions and whilst far away from the sharp end do affect decisions about staff numbers and types, training and equipment. Latent conditions are in turn constrained by economic and political factors that also play an indirect role in accidents or incidents. The 'Swiss Cheese' model illustrates the opportunities for error. There are layers of defensive barriers such as alarm bells or warning signs. These barriers are not necessarily electronic i.e. nurses checking drugs. These defences open up from time to time and when are all lined up in the model there is a serious accident and lining up symbolises the simultaneous breaching of all defences. I want to emphasise the difference between active and latent failures. The former are failures of people working at the sharp end and quite often these have immediate consequences and in contrast latent failures result from decisions taken at high level about staffing, for example. These can be quite dormant for a while. What can we do with all of this? Charles Vincent (UCL.) came up with this framework for analysing adverse events and critical incidents using Reason's model. The individual task and contributing factors were looked at to produce this list.

First are patient factors because patients react differently from one another and to individual medical conditions. There may be language problems and so the patient may not tell you that he/she has an allergy. Some patients also want to be involved in decisions made about their care. Informed consent is now a CNST. requirement yet quite often patients just want to be looked after by someone they can trust.

Now onto task factors such as protocols and decision-making aids that is where electronic prescribing can help. Let us look at protocols and their ready availability for checking. I have heard of instances where two people were required to check and to administer a controlled drug, one says 'I will give it' say because of time pressures, the other then says 'I want to check it because that is in the protocol' and the first person then replies 'don't you trust me?' In aviation, pilots before takeoff will call the checks and cross-checks and they do not see that as double-checking. Doctors assume that the pharmacist will check the prescription without seeing that as personal distrust. Supposing there was a paediatric nurse who does not feel comfortable with drug calculations, assuming that her colleague can calculate correctly when having to check the dose - again a huge potential for error and these examples link into communication in teams and leadership (such as failure by consultants to mention details of the dose to junior colleagues).

The environment - the workload - I am not sure how many of you are aware of the reduction in doctors' hours over the past few years. There is the old school of thought that because doctors are absent for long periods of time they are less aware of patient care. It is the balance that is necessary between not overworking doctors yet providing consistent care. Some of the earlier comments about communication in this meeting are particularly important. Workload is affected by staff turnover and sickness throwing more burdens on remaining staff. My earlier example of lack of equipment standardisation is another burden because progress in technology can then bring unnecessary complexity. The NPSA is looking into both standardising pumps and staff training in their use.

Everybody has mentioned organisation, management and politics, money and resources. There is also the safety culture at the organisational and unit level. So if you hear of a person on another ward who was disciplined for a mistake this may then affect your behaviour and what you report if an incident occurs. Finally there are quite a lot of regulatory bodies that inspect every few years - CNST. and CHI.

Now I want to consider errors - are they person or system generated? A person who makes a mistake is deemed careless or reckless so the person-centred approach is to change behaviour, provide training or remove the individual and replace with another and people who are insufficiently trained are going to make mistakes. In contrast the systems view looks at all the factors I have talked about and is based on the idea that we are fallible and to reduce errors the system of work must be changed My view is that it is not one or the other it is everything; all the factors and their interactions must be examined, so if one factor in the system is changed it is going to have a knock-on effect. Thus the removal of one type of error by introducing new systems will introduce the potential for other types. We need to look at the risk assessment to find out what kinds of errors exist and then put in barriers to prevent them from occurring.

In summary, I hope I have put the nature of error in context and considered the interactions and relationships between the factors causing errors. I want to introduce one extra point also made by Linda. Observation is required to find out what is happening using techniques like task analysis of what people do. However remember that observations are snapshots at a particular time and set of circumstances.

Questions

LF We talk about knowing where errors occur, their severities and the proportions of inpatients and outpatients and whether they affect transcription or administration. However do we know have an analysis in terms of such root-cause categories as knowledge, communication or systems?

DB It varies a lot depending on the setting. In our inpatient study there were many knowledge-based errors as well as many ????? but in the outpatient study we had a hard time getting enough information to decide what was really happening.

BF We are just finishing an outpatient study and the root-causes depended on the point of entry process, 95% of errors had to do with order entry and this reinforces the need for better decision-support systems. We had process errors like the wrong drug or wrong strength we also had had problems to do with the technology because the system overrides did not provide enough information about what was going wrong. Maybe more information and education is needed in the system to deal with situations where for example a member of staff says 'You gave me the wrong drug' which turned out to be a cream instead of an ointment but the system didn't know this.

LF That was a knowledge issue?

BF Yes a combination of people failures and the barcode checking system that did not assess the package size or the form of the drug.

LF Are there any quick and simple ways to find out in a particular setting if the errors are knowledge-based or have another root-cause?

MW I think this is what the NPSA is trying to do - a campaign to investigate every kind of error not only the error but also the contributing factors such as equipment failures or low staff levels are reported back to the Agency. The outcomes for the patient are also included. The most serious incidents or those with great potential for harm will be investigated. However only when you have a great deal of data can you get information about what is happening and where it is happening. Staff changes - six-monthly (e.g. SHO's) and annually are new opportunities for errors.

TA (asking BF) Were the underling causes of order-entry errors knowledge-based or rule violations?

BF Most were related to incorrect entry of the instructions, particularly the incorrect code for the drug (these have to be remembered because there are not drop-down menus in the pharmacy software). I would say that these are the most frequent errors.

SG I used to work in transport before moving into healthcare. One of the big issues associated with road traffic safety was the concept known as risk migration - the safer you made the system the more individual reacted against that to make it unsafe again e.g. drivers drive faster when using their seat belt or motorcyclist corner faster when wearing a helmet. Does this happen in healthcare?

MW I don't know of any evidence but I can see why people would behave in the same way if they believe that prescribing errors would be picked up electronically.

NB It is called risk homeostasis in some quarters, is still contested in healthcare but is still worth hanging onto.

AJ One of the things that Linda said about root-cause analysis interested me because we are struggling with what the Trust is doing and with the NPSA stuff. Because there is a feeling that the person reporting the error should do the Root-Cause-Analysis (R.C.A.). What we know is that people report errors for many different reasons and quite often because they are on campaigns. So if you are really fed up because you cannot get an IV line then their rate of reporting and consequent delays in administering IV therapy is what gets reported. We should be clever enough to know that the rate may not have changed but what has is the interest in reporting and is a real issue. We are all encouraged that the reporter should do the R.C.A. The real problem is training people and what their view of the organisation is and how they understand it. So the value of the reporter doing the analysis is fraught with disaster. The problem with healthcare is the cultural and organisational factors such as the different professional approaches to the definition of error. What a nurse thinks is a serious error is completely different to the opinion of a pharmacist or a doctor. This particularly applies to the symmetry between cause and outcome. I have had to fight some real battles because a pharmacist made a simple dispensing slip and in my view most slips cause no adverse outcomes. So if that pharmacist does not have a track record of poor dispensing, they will suffer no consequences from me as their manager as if there had been no outcome at all (unless they do not take it seriously). I think there is a real problem with R.C.A. and who does it and therefore the value of it. One final point; if my memory serves me correctly we have some dispensing-error-work that shows that the busier you are the more likely you are to make an error is completely confounded by work that revealed when you are very busy you concentrate properly but when you are hanging around with not much to do and thinking about dinner tonight then the errors are higher. Bryony, have I got that right?

BD This was only a qualitative study that showed that dispensing errors happened during quiet times when people were thinking about the weekend.

NB Two related things when we looked at the causes of prescribing error. Wasn't knowledge one of the factors?

BD We interviewed doctors who had made prescribing errors and looked at the starting point. There were mostly slips or mistakes not many lapses, hardly any violations and very few purely knowledge-based mistakes. There were rule-based mistakes (knowing the rule but forgetting it or not knowing what the right rule was). Some of these were quite difficult to classify as either knowledge or rule based. So there were mainly slips or rule-based mistakes.

NB The follow-up point is that some people are questioning the usefulness of R.C.A. in the sense that we can deconstruct an event into steps to which we can apply consequences. However if we think of an event as the product of random sequencing of steps then the notion that there is one action that will solve the problem is suspect.

MW Looking for a cause can be tantamount to looking for someone to blame. There is an industrial technique that requires all steps, actions and thoughts are mapped and then examined to predict what can go wrong and what corrective actions can be taken. Whilst this may be possible in industry, healthcare has so many emergencies that there is not enough time to apply this technique.

LF There is a difference between retrospective analysis to establish cause and prospective analysis to predict effect. Has there been any work in the UK to use the risk reduction techniques from industry in healthcare that you have talked about?

MW We are at present looking at risk reduction techniques in the A & E department.

RD Two comments: I have always favoured the notion of ROUTE cause analysis I agree with your point that there has to be a single cause it is more constructive to understand the process by which this came about and the choices that were considered or not considered. The other point that you raised, Ann, is who does this because one of the things that has struck me over the past 30 years is how very ill-prepared are healthcare workers for working in organisations. This is because the dominant mode is the autonomous self-sufficient professional so pharmacists nurses and Uncle Tom Cobbley model themselves on this. So no one gives any thought to the obvious point that we are all working in an organisation so we need to think in terms of systems rather than individuals.

MW Attempts to provide education for groups of doctors and nurses have always failed because of the attitude of the doctors.

RD I don't think this is about interprofessional education but it is about how we teach the various groups about their work settings. For instance the medical sociology courses for undergraduates teaches them nothing about the sociology of organisations. Leadership and teamwork is sometimes provided but are not relevant to what it takes to work in a complex modern organisation This is how we need to think about and understand the issue. No major corporation would induct its graduates in that way.

IW Can I ask some questions? Have you seen changes in the attitudes in the health professionals that you have encountered in your research?

MW I think it is gradual but the problem is that when I go to conferences to give talks I am talking to the converted about the causes and practical actions to combat errors. However I don't meet the vast majority who hold different views but things are changing. A colleague of mine attended the NPSA conference of last week who said there are still the same issues of getting more people to report and receiving feedback about the reports. The word is spreading; different people are attending conferences while the original group is perhaps becoming bored. There are changes. How do we investigate these things - we have the information so what changes can we make and how can we evaluate the results?

IW Can the Metanetwork facilitate these changes?

MW I think so, there is a website.

DB A couple of quick points. I think that systems' effects and things that are at the blunt tend to get underrated because, for instance, it is hard to relate a specific effect to poor staffing. With respect to R.C.A.'s we never have the reporter doing the R.C.A. We have a group including someone who has some systems expertise. Lastly, we have been doing a study with the leading sleep expert we took interns and let some of then sleep and some others had them do the usual things. While we were doing this there were national requirement for working hours issued. Many hospitals including John Hopkins were in gross violation of the regulations ands were shut down. We were reporting about the impact of not being able sleep and surgeons were the most resistant because they believed that sleep deprivation does not affect performance. The empirical evidence from medicine is that you might be better off getting people to work longer hours because there are lower adverse events with fewer handovers and there is only one contradictory study. This is a very interesting area.

AJ The work we have been doing at our Trust shows that the major issue is communication at handover from one ward or shift to another and on-call. This applies to all professions and if we go into the European Time Directive we can expect more incidents because there are more shifts and handovers with very poor communication. So sleep is less of an issue than handover

MW I want to contradict that because there have been some studies in A & E departments that show that the triage determines the care of patients and affects the thought processes of doctors about what diagnoses to rule out. The care chosen goes down one track.

TC Handovers can be linked to the Swiss Cheese model and maybe handovers can be another barrier.

NB Thank you today we have seen some of the reality, the problems of the present and the future and just beginning to move into theoretical constructs which we will continue with tomorrow. The issue of culture in organisations is illustrated when I used to run the pharmacy at the national Heart Hospital - an 80-bedded tertiary referral unit - mainly populated by academic medics. Getting change was easy if you quoted papers and used a rational argument so I used the same approach at a large D.G.H. and I took the same changes that I had made (having copied the changes that Anne had made at St Mary's) to the Head of the Coronary Care Unit and he said 'what does old so-and-so do at Barts he has a sensible head on him? Does he do this? It was a completely different culture about perceived authority - an academic or a good chap but secondly it was clear that no one thought medicines were a dangerous area.