By Dr Nahum Kovalski
The title of this post is most often attributed to Mark Twain. But apparently, this may be an error. If you view this link, you will see that Charles Dudley Warner was most likely the creator of this meme.
In any case, the message of this statement relates to the futility in trying to change certain things that are deemed beyond our purview. Most likely, the day will come when we can alter the weather, or at the very least, create very large enclosures that provide for a year-round 23°C, low humidity, minimal air pollution environment. At that point, if everyone is talking about the weather, then the person in charge [the Minister of Weather or the Minister of Localized Environmental Affairs, Actions, modifications and Travel and Culture and Tulips] should do something about it. And if not, that’s why we have elections.
At the time that this statement became popular, “weather” was likely considered to be beyond human manipulation, even amongst atheists. So if you didn’t like rain, you were far better off taking the train and then horse and buggy and moving to the sunnier edge of Spain. These days, many things that we treat as being nearly divine in origin and beyond human intervention, are in fact very amenable to change.
I follow a large number of websites that describe medical updates, and allow physicians and other healthcare workers to comment on medical technology. On one such site, the commentator starts off with the following statement “I just refuse to use the EHR in my institution”. He goes on to describe the failings in the specific system he uses and summarizes his comments with a call to his brethren to join together in the revolution and to “all refuse to use these” EHRs.
The commentator implies that returning to the traditional hand-written-on-paper-medical notes, would eliminate the fragmented clinical chart in its electronic version. The commentator states that “we no longer have a linear history of our patients” but rather lots of data points that do not coalesce to describe the patient in his or her entirety.
Firstly, this is a classic example of people having a warped sense of nostalgia when faced with change. I remember as a resident, dealing with a patient who had a kidney stone lodged at a very difficult to reach location. One of the older senior urologists stated that once upon a time, he would just have opened the patient’s belly, found the stone and literally cut the stone out. To listen to this older surgeon, it was easy to wonder why we ever bothered to stop doing things the old way. Was it just that progress was being imposed upon surgeons, and all doctors? Was someone in the hospital getting a kickback every time the hospital purchased one of the “minimally-invasive” equipment sets?
In this day and age, it doesn’t take long to look up papers and collect statistics. And it turns out, despite the talented hands of the surgeons of yesteryear, there was a very high failure rate when stones were removed in an open fashion (by making big incisions). In this particular case, given the position of the stone, the risk of serious kidney damage was ridiculously high, around 50%. Considering that the recovery time and infection rates were all improved with the new less-invasive technologies, suddenly, nostalgia was clearly old-school.
The same rules about nostalgia apply to paper charts. Paper works great until you can’t find a misplaced chart, or X-rays go missing or someone spills coffee.
I have written far too often about EHR’s and how they could be dramatically improved. The EHR that I wrote has been a success but unfortunately, it’s innovation has slowed dramatically since I left. The benefits of having a person with a combined technology and medical background, leading the team of software developers, was dramatic. There are not many people with my perspective.
I really do not like to self aggrandize. But sometimes, you have to recognize things for what they are. EHR’s from multibillion dollar companies continue to suffer from a fundamental lack of appreciation of physician needs. The majority of EHR’s are built on an old-school principle of monolithic mega-monstrosity systems that make it very difficult to update the software [lest a bug sneak in].
In the past, I have described multiple approaches to a fundamental redesign of EHR’s. There is no question that this is a moonshot scale program. But once you have the right people and software in place, along with whatever other tools you might need, systems would fall into place very quickly.
I will already admit that some non-digital things are almost by definition, impossible to replicate in their simplicity. If there are three patients in a clinic, and one of them has a kidney stone, it is very hard to create a system that is as simple or even simpler than calling out to the nurse and saying “nurse, please, give the patient with the kidney stone, 75 mg of Voltaren”.
There is a tremendous amount in this example which speaks to the unique power of our human brains to understand context. For example, let’s say there are two or more nurses working at the moment. Somewhere, somehow, at some time in the past, someone decided that there has to be a chain of command that is set into action when the doctor calls out the order.
Writing a command to a computer, such as “get_available_nurse” is by no means impossible (but is still harder than yelling “nurse”). From the computer’s side, this is very difficult to do. The computer would have to determine which nurses are in the clinic at the moment [and consider whether one or more of the nurses are five minutes from the end of their shift, so that they do not want to start a new procedure]. The computer would also need to know which nurse is certified in giving this pain medication. Not all nurses are certified in all procedures. And if it turns out that none of the nursing staff is capable of giving the injection, the computer system will have to decide how to deal with this situation [refer it back to the doctor, tell the doctor to wait until the next nurse comes on to shift within the next 15 minutes, call in a nurse from home or from a nearby department and so on].
Imagine the same system dealing with 30 patients on a ward, three regular staff physicians, 2 to 3 physicians rotating through the department, 4 nurses, 2 nursing students and the background noise that comes from all of the equipment spread out across the 15 rooms in the department. Now imagine the exact same scenario of a doctor sticking his head out of a room and saying “can you please give a Voltaren shot to this kidney stone patient here in the room I am in right now”. Is this easier than going to a computer station, finding the patient, typing the order and so on? Of course, it’s easier. But the potential for errors is staggering.
As I noted in my last blog post, hundreds of thousands of people die every year in the US from medical errors. There is absolutely no question that a good proportion of those errors could be eliminated by taking the time to enter information into an EHR and then having the EHR doing all the necessary background checks to make sure that there is no contraindication to the medication and there is no conflict between medications and that the patient didn’t already receive the medication and that the patient isn’t undergoing a procedure the next day which renders the medication high risk and so on.
Why are computers still so hard to design for? Shouldn’t you just be able to give the computer a list of information you need, and it should figure out what to do with it? Yes – this would be nice. But the design of a computer system is actually a very subtle thing. You have to consider what your specific users will like and how they most often interact with the system.
Computers still mostly follow preprogrammed paths of logic. If you forgot to consider one possible path, the software can crash. When designing new systems, you have to spend A LOT of time considering what the computer will do/suggest when its internal logic hits an “I don’t know what to do” road bump. Machine learning technology, which should be incorporated as quickly as possible into every EHR, can simulate human thinking and figure out a solution even though there is no specific code explaining what to do for the case.
Let’s say that our Urology computer utility gives instructions on how to deal with stones of varying size and location. But it turns out that this particular kidney stone is of a size that the programmers never considered. Badly written code could just crash. Better code could crash out BUT say that it crashed for the specific case of a stone of size X in location Y. Even better, would be a system that asks a special module what to do for such a patient given the stone’s parameters. A machine learning system could look up papers in the medical literature and cases from patient files from the hospital, and LEARN how to handle such a case. The computer could then make a solid suggestion to the doctor rather than just crash out or say it does not know.
The answer to all of this is not to stop using EHR’s. The answer to all of this is not to go back to paper and pen. The answer to all of this is to say that the billions of dollars that have been invested so far in the development of various versions of EHR software have not produced the necessary end product. In a free market, what should happen is that a vendor should step up with a product that is clearly superior to what is presently on the market, and basically force everyone else out.
But that’s not happening because CFOs think in terms of year-to-year budgets and have great difficulty scrapping projects that have already been paid for, at least partially. It is very difficult for CFOs to come and say that, after having spent five hundred million dollars on an installation of a computer system, that the whole thing is going to be scrapped and replaced with a totally different five hundred million dollar system. Such a thing can be done, and occasionally it is done. But not often. And not many CFOs survive long in such an environment.
Building a fundamentally new EHR is a huge scale project, that I would love to have seen picked up by Google. I’ve mentioned in the past that redesigning EHR’s such that they are effectively collections of apps, just like your phone is, would be the most effective solution. If you have a particular app on your phone or hospital system, then the data for that app will be saved in a preset location. If you don’t have the app, it won’t be.
When it comes to security and privacy, whenever a user is registered for activity on a particular ward or in a hospital, this could be set by a central manager and immediately, all of the appropriate hospital specific apps would be downloaded to the online environment and the phone of the doctor.
If a phone gets lost, you could even go to any other phone, type in your name, your code and additional identifying information which would immediately freeze your mobile account. You would then go to tech support and they would transfer your remote access to another phone.
When you think of an EHR as a collection of these tiny units of functionality, you get the same amazing device that we all carry in our pockets – the device that helps us manage our lives and is tremendously flexible. But when you start an EHR project with the plan to treat every possible disease ever encountered and which will ever be encountered, you very quickly get to a ridiculously complex system.
Someone needs to build this kind of lego-like system. Would other types of systems work? Sure. But what I am describing is straightforward, incredibly flexible, amenable to varying sizes of teams, easily available, easily expandable, and a lot more positive things. And there is no reason why such a system should cost hundreds of millions of dollars.
Thanks for listening.
Originally published in: