Reflections on Ross Koppel’s lecture “Why is Health IT so hard to get right?”

His Subtitle: Examining the uneasy marriage of medication, computerisation and workflow.

Prof. Ross Koppel gave a lecture last night (17th April 2012) that I found interesting, entertaining and depressing all at the same time. It was great to hear one of the most eminent researchers in my field but where his experience gave reassurance and insight that this area is littered with problems to investigate it also led one to question the inertia of the area and if any great progress could be made. For years it seems people have been pointing out problems with healthcare IT but still the message is that we are failing patients, failing clinicians and failing to learn.

Here it seems appropriate to quote Liam Donaldson’s eloquent tweet: “To err is human, to cover-up is unforgiveable, to fail to learn is inexcusable.”

Basic Usability Issues
Ross gave lots of examples of health IT problems he had come across, some only frustrating users whilst others have led to patient harm and death. When he said that he could go on for months about these sorts of problems it didn’t seem like he was exaggerating. Some memorable examples he described included:

  • Separating data that needed to be interpreted together to be meaningful e.g. the two figures needed to analyse blood pressure were displayed in all sorts of ways, some easier to interpret than others, and the worst cases had pages of text and data between these two values that should be together. This creates extra work and cognitive effort for clinicians.
  • Someone guessing a patient’s weight to enter into a prescription as IT system required a weight, the weight did not effect the prescription and the clinician could not get the patient’s weight quickly. However, does another professional attending to this patient later in the process know this weight is only a guess when it is critical to their prescription?
  • Health IT in some areas now force nurses to scan the patient’s barcode on their wrist and then the medication’s barcode within about a minute of each other to check it’s right before treatment. However, there was one story of the medication’s barcode on a multipack in the fridge a long down the corridor. Here the nurse’s workaround is to print extra patient barcodes and stick them on her arm. To save time she has lots of different patient barcodes on her arm so the potential for confusion increases.
  • The patient is prescribed 1 x 20ml pill of something. The clinicians haven’t got 1 x 20ml so use 2 x 10ml instead but the computer that checks the barcodes won’t accept this as it’s not the same and the treatment cannot proceed. The clinician shouts at the machine but it maintains that it isn’t the same and won’t allow the treatment.
  • Prescriptions being grossly over- or under- estimated as grams and kilogram fields are confused with pounds and ounces.

The Problem of Predictability
From the examples above it is clear that there are problems here, so why do these problems exist? Some of it is basic usability and inexcusable i.e. the companies designing and selling this stuff should be doing a better job. However, deeper issues can occur because healthcare is a very complex workplace with lots of different groups working together under pressure to achieve good outcomes. For example, there are pharmacists, consultants, nurses, healthcare assistants and many other specialists trying to coordinate their expertise and actions. This complexity makes predicting how technology will be used within and between these different groups difficult. Even the physical location and other real world constraints impact the effectiveness of the IT system (as we saw in the fridge example above). Also, with so much technology it makes predicting how any one piece will interact with all the other pieces difficult. Ross told an example of patients’ scans disappearing between two systems. Both systems were working perfectly but the way one handled files conflicted with the other so they were lost in the gap between the two systems. This problem took weeks to uncover and fix at great expense. Ross told us that a typical design cycle takes over 18 months so your problem would unlikely be resolved until that time. Also, if you had a really good idea that added great benefit to the product then the vendor might see it as a great step forward to be saved for the next version so it’d sell better – in the meantime you’re left without the change you desire.

Barriers to Learning
So even if we are left in a position that problems are hard to predict then at least we would like a system that learned fast and learned hard once we find them, but this is not what we have. Actually we have barriers to sharing information about problems and we have larger forces at work that make change slow and expensive. The large vendors of health IT have written into their contracts that problems cannot be shared with others outside of the hospital! Ross told us of a case where someone reported a problem and the vendor said that they’d never heard this unusual thing before when in fact the reporter’s friend had told them the same thing a few weeks back – at best the vendor was mistaken and at worst they were lying. Ross also reminded us that the majority of issues are not reported, instead people put up with the systems, find workarounds and manage as best they can.

Some Recommendations
After explaining these issues there was a reminder that health IT had brought many benefits, and things were improving, but we’re not doing the best we can. Amongst other things that escape me now Ross recommended:

  • To get vendors to agree on ways to display and input data so it follows sensible practice.
  • More reporting of issues and an independent body to oversee data collection. It seems unacceptable that people who sell the products are also the ones that collect problems with them too. This inhibits learning and plays to a fundamental tension between portraying a positive image for selling purposes and being realistic about the problems that have been and will be encountered.
  • We need more people with a keen eye for observation as we know that errors, inefficiencies and usability issues are underreported so we need people to walk the wards and uncover these issues.

Conclusion
Health IT has brought benefits but we could be doing a better job. The good news is that this area and patient safety seems to be growing and attracting more attention, but it is a hard nut to crack and progress will be slow. There has been a report made to congress that recommends some of the things that Ross put forward in his lecture. However, there are some recommendations that are themselves hard to implement. For example, even though there is gross underreporting of issues the NPSA’s NRLS used to (and perhaps it still does) get thousands of reports a month – many of the reports do not contain a sufficient level of detail to fully understand the incident or subsequent learning and relatively few people are assigned to monitor this mass of data. How are we to properly manage this system when cuts are being made? Furthermore, how might we manage it when we encourage people to report even more issues? If we have more people to walk the ward with a trained eye to notice errors, inefficiencies and usability issues then where are these to come from? Perhaps if we had a team of people in every hospital then they could manage the error reporting process locally, which would take the pressure off a centralised system (I think there might already be ‘risk’ departments and processes within hospitals that do something similar already??). Furthermore, would hospitals benefit from a closer working relationship with universities to help uncover and investigate issues? We get the picture that there is almost too much data to manage and the healthcare sector is already stretched. I work in academia but many researchers are put off working with the NHS because of the laborious ethical approval processes that are enforced on people who want to do observational studies. People commonly share war stories of ethical approval taking between 6 months and 2 years to clear. This does not make for an easy life, particularly where project grants and PhDs have traditionally been for 3 years, so easier data is sought elsewhere in domains with less challenges. Even we if we create recommendations for improvement then how do we make software vendors more responsive to problems that are uncovered within and across different hospitals? Perhaps we need to encourage more openness and competition so those that have fewer problems and are more responsive are rewarded with more contracts. However, being too responsive might cause issue in itself if multiple updates are fired at systems on a regular basis. Then there is the further issue of clearing updates with regulatory bodies to make sure the updates are safe by having the appropriate tests and paperwork approved.

I think the concluding message has to be: We are making progress, there are problems, but we need to be doing a better job for the sake of patient safety. This will likely bring gains in efficiency too that should be welcomed when services are increasingly stretched.

Ross Koppel’s new book is out soon “First do less harm: Confronting the inconvenient problems of patient safety” – I haven’t read it but might order it after this thought provoking lecture.

This is not meant as an authoritative blog, and I have probably got some details wrong, it’s just my perspective in response to Ross’s lecture. Perhaps others can contribute comments to put right what I have got wrong or share their own perspective.

Finally, is there a more positive story we could share about the successes of Health IT, usability, error reporting and learning so we can look up to it and see the sort of thing we are aiming for?

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: