The 21st century Doctor- A reimagining of hospital care

It was Tuesday morning when I met Fiona (not her real name), a bright 30 year old lady who had come into hospital over the weekend with abdominal pain. She had been diagnosed with Crohn’s Disease six years earlier and so was on our ward for a suspected flare up. Crohn’s Disease is an autoimmune disease of the gastrointestinal (GI) tract which causes widespread inflammation to the tissues as a result of an overactive immune system. It’s a subtype of a family of diseases known as inflammatory bowel disease and often has episodes of acute inflammation, which vary from mild symptoms managed at home to life threatening needing admission and sometimes surgery.?

Fiona was not new to this, having lived with Crohn’s Disease for six years; she had been back and forth from hospital because her condition was difficult to control. We had already started on biologic therapy- targeted molecules that reduce inflammation- six months ago to try and control her condition. The most important thing for us to figure out that morning was whether this was working.?

I reviewed her history. “How are you doing?” I began. An iPad was by her side and she had an overnight bag of essentials on the floor by her bed. “Good”, I thought, “she’s prepared to be in hospital for a few days at least.” She felt reasonably well, her symptoms had started towards the end of last week with bad abdominal pain. She initially tried to manage it at home, but the symptoms became severe enough that she decided to come into hospital. She hadn’t opened her bowels in a couple of days; a potentially warning sign. Next, I reviewed blood tests and vitals signs. Her inflammatory markers were not particularly elevated and her vitals were within normal ranges. Finally imaging, she’d undergone a CT scan on admission which had shown a thickened terminal ileum, which was possibly causing a blockage in the gut, but had been seen by the surgeons who had advised for a watch and wait approach for the time being.?

Blood tests aren’t the best for looking at inflammation in the gut, so Fiona needed more detailed imaging to see what was going on. We had a couple of options. The first is with a colonoscopy, a procedure where we pass a special camera up the back passage into both the large bowel and the last part of the small bowel. We often take samples and send them off to the lab to look at the tissue under a microscope and assess the levels of inflammation.? Results however take 5-7 days to come back. The second-? much preferred test by patients- is a specialised MRI scan which looks particularly at the small bowel. We chose the MRI scan. We knew from previous scans that she had Crohn’s disease in her small bowel and her history and CT scan made us concerned that she may have a stricture; a narrowing in the bowe, causing blockage and eventually dysfunction.

?

I was hopeful about our plan for Fiona. We’d get the MRI scan which would point us in the right direction and then we could start putting together a plan. However, when my junior called me later that morning, she informed me the scan had been rejected.?

“Reason?” I asked,?

“Inappropriate request” she responded.?

“No matter,” I thought, “I’d just go down and discuss with the radiologist myself”.?

Rejected scan requests are par for the course in medicine. Typically you’re? required to add some information to justify the need for a scan you’ve requested.? It’s not unusual for that information to be incomplete or unclear, in particular because they’re requested by junior members of the team who may not have a clear understanding of why the scan is needed. One of the important roles of a radiologists is to prevent inappropriate scans from happening. All the more important given we blast people with radiation whenever we want to look at their insides. In Fiona's case however, an MRI scan was indicated. I had? guidelines to back me up and we’d also performed scans for numerous patients in the same condition as Fiona. It would be a short discussion.?

The radiologist on duty sat in her office, bespectacled, facing the computer. We knew each other professionally and being of similar grade, we had a good working relationship. It was obvious she was already busy. I was the third person in the office waiting for her time. Approving scans sometimes require superhuman time management skills, particularly when working in a busy hospital,? the phone seems to ring nonstop and it can often feel like every patient in the hospital needs some type of scan.

“Yeah I wasn’t sure about this one,” she began, when it came to my turn, “the MRI right?” I was impressed. She’d know what I’d come for.

“Yeah” I responded.

“Typically we need it to work out whether it’s inflammation or stricture. That determines whether we go for another biologic or we do something clever endoscopically or even surgically.”?

“Hmm, and you don’t get all that information already from a CT scan?”?

“No, not really.”

“It’s probably best if we speak to one of the GI radiologists just to double check”

”No problem, are they around?”

She promptly picked up the phone dialling the number for the office downstairs. I knew the consultant radiologist. He’d sat with us in many of our specialist GI meetings, and was also my personal go-to-guy when I needed some advice on imaging. No answer. After some digging around in schedules, she told me where he was. It turned out he wasn’t going to be in until later that afternoon. He was reporting in a cancer meeting which ran most of the morning. I’d have to come back later and rediscuss.?

Delays are somewhat inevitable when you’re working with multiple teams. Fiona had seen four sets of doctors by the time she was under our care. Each team of doctors with their own clinical opinions and plans. The nature of our system meant that she couldn’t just walk in and see a specialist. This is part feature, and part side effect of our healthcare. There aren’t enough specialists to see every patient from the moment they arrive at hospital, and so what happens is a compromise of sorts. Patients are first seen by ED (emergency department), before being referred to the relevant specialty if the decision is for admission. The medical team on duty will then refer her to the relevant specialty who reviews the patient and advises on a management plan. It functions as a system but can often leave a patient frustrated by the overlaps and delays in care.?

Latency is a problem well known to hospitals (I define latency as the time difference between when a task is requested and when it happens. For example, the time difference between when a request for an MRI scan is put on the system and when the patient has the scan).? So well known,? in fact, that we’ve started to see a program of “red days” and “green days” implemented on wards. A “red day” is a day where nothing clinically meaningful happens to the patient. The plan for the patient remains the same as it did the day before. There is no progression. A “green day”, on the other hand, is a day where there is clinical progression of the patient’s care. Something meaningful has happened that makes the patient a step closer to going home.?

The value of a healthcare system can be measured across three domains. The delivery of high quality care, in a timely manner, at a price that’s cheap. High quality healthcare is axiomatic (have you ever heard someone advocating for “low quality healthcare”?), but needs definition, a consensus that we all agree on. Timely because you don’t want to wait an eternity to access this “high quality care” and cheap because you don’t want to have to “donate” an organ every time you see a doctor. These three factors have long been seen as a trilemma. In other words, you can only optimise for two at a time.

Any technology that improves healthcare exponentially on one of these axes is an innovation which demands consideration to be implemented in a healthcare system. Deliver on two and it’s almost a certainty. Three is almost unheard of.?

We have a systemic bias towards looking for improvements in “healthcare quality”. Better outcomes for patients in terms of mortality (life or death) and morbidity (health or disease) so we often forget about the other two axes. But healthcare can be made better without a necessary change in patient outcomes. Just make it faster and cheaper?

In defence of latency?

There are sometimes very good reasons for latency. As a system that deals with lives and health; breaks that are put on the system can serve to reduce errors. Errors that can often be costly. The radiographer has to call and check that a person isn’t infectious before accepting them into the CT room. Failure to do so can result in the CT room having to be shut down for the rest of the day whilst it’s deep-cleaned, or worse- contagious bugs passing from person to person. With backups, we may go slower. But we ensure we get it right the first time. However, it can reach a point where the latency becomes maladaptive. The bureaucracy that attaches itself to every decision made in hospital can actually have the perverse effect of reducing decision making. Doctors now spend more time navigating the idiosyncrasies of their hospital system than making clinical decisions. An incredible waste of resources.

We thought that digitisation would be the solution; a panacea which reduces bureaucracy and increases healthcare quality.? But where it has improved care in certain areas,? it's become hopelessly complicated in others. As a junior, I still remember when prescribing medications was done on paper charts. One of my tasks as a weekend SHO, represcribe all the medications for a poor patient on the medical ward because the medication chart has run (charts only had slots for nurses to document that they’d administered medications for 14 days, if the patient was in any longer a new chart had to be written). There’s a special place in hell for doctors who leave this for the weekend SHO to do. E-prescribing software has largely eliminated the need for this. Two thumbs up!

On the other hand, much of the hospital software that we use today are behemoth 90s era software built as pre-internet closed systems which lack the flexibility of much of the newer and more integrated software. The result? I now open up a total of seven different systems to get a full clinical picture of a patient in front of me.?

More digitisation isn’t necessarily the answer, a patient already generates more data in one hospital visit than any one doctor can reasonably hold in their head. What we need is not so much data presentation , as data comprehension. Technology that helps us make decisions based on the data in front of us.?

Fiona eventually did get her scan, but it took until the afternoon for it to get approved and another 2 days before it happened. Meanwhile, we were doing our best to treat what felt like symptoms. My daily ward round started to feel a little monotonous, I didn’t have any update for her and felt like a robot repeating myself every morning. I could see the frustration in her eyes at the delay. We’d encountered a roadblock.

The 21st century Hospital- reimagined?

I want you to think again about my ward round with Fiona. This time, however, I'm wearing a small device with an installed large language model trained to provide medical care. The device scans the previous clinic letters and generates a summary of Fiona’s care to date. I can see at a glance the number of interactions Fiona has. It then records my interaction with the patient and generates a summary of the ward round. Next it develops a management plan for the patient based on my suspicion of an ulcerative colitis flare. Third-and this is where it gets good- it executes all of the action plans that are purely digital. An MRI request for the patient gets put in first time with the appropriate information.

The MRI request is sent across to another AI client that acts to vet the scan. My radiologist friend has access and can see which scans are being vetted and also where in clinical guidelines, the indication for the scan is outlined. They still have veto power of course, these AI clients are strictly parametrised so that they only approve the most straightforward of scans. They decide to veto and I receive a discrete notification telling me that my scan hasn’t been approved along with the indication. I may still have to find the radiologist to discuss, but that process has occurred much quicker than otherwise.?

Once vetted, the scan request goes to a third AI client which is an automated scheduler. The client checks everything. Presence or absence of O2, barrier isolation and patient mobility. This is already recorded on hospital patient systems and so would be trivial to access.? Everything else in the physical world happens as normal. Fiona still has to wait to go down for her MRI. If we decide to change her biologic, she will still need to be physically given the treatment and should she need surgery, she’d still need to be operated on by a surgeon. An AI can’t do any of that.

But what it does do is abstract a massive amount of bureaucracy without compromising safety. Importantly it should be flexible, so that we can let it take over as much or as little of the workflow that we want. Top down systems imposed with central planning and not responsive to change work less well than bottom up systems that allow for evolution over time. We’re already seeing AI clients starting to support in important areas,, my radiology reports now come back to me with a note below, “please note the radiologist has used artificial intelligence to generate some of this report”. At present, this may be no more than autocompletion of common phrases radiologists use in reports, but it’s rapidly scaling to be much more.??

We decided to change Fiona’s biologic after the MRI scan. It suggested ongoing inflammation in her small bowel which could respond to medical treatment. I entered her room a couple of days after starting treatment.?

“How are you getting on?” I asked,?

“Great,” she responded, and then added “Desperate to go home.”

“That’s fine, I think you’re ready to go home today actually. We’ll just get your paperwork ready and the medication you need and that be all. We’ll arrange some more imaging and an outpatient follow up for you to see how you’re getting on”

“When will that be?”?

“Well” I began, “How long’s a piece of string.”