Order Entry UX - First principles: Order + Action
Post five - in a series on Order Entry User Interfaces.
This post continues the previous post looking at order entry from First Principles.
Order turn knowledge into action. But there are two perspectives on how this happens
Perspective 1: the order does not participate in the process of knowledge or action.
Perspective 2: the order overlaps and participates in both knowledge and action.
Today we will look at the connection between Order + Action.
PART 1:
Orders + Actions
Separate or integrated steps?
The default for most electronic health records and most paper based systems, is to view an order only as an intermediate messenger between a clinicians knowledge into action. The previous post goes into detail regarding why the order participates in the process of Knowledge. The same arguments can apply between the connection of Orders + Action.
In a model where the Order is separate from Action - when it is only a messenger service - we have fragmented thinking and fragmented care.
A clinician orders a medication, the medication is dispensed. A clinician orders a lab test, the test is done. A clinician orders three imaging tests, 20 in-house lab tests, 5 send-off lab tests, and a consult. All 29 items are done individually.
The problem is that an order is never an action in isolation.
The result of an order, the action, is always connected to a purpose. The purpose is connection to a problem that is being solved. Did the medication work? What did the test show? Was the problem solved? Does something else need to be done? Was the result of the action verified? checked? reported on?
We can only understand if the order fulfilled its purpose, if we view the order as a step within an integrated workflow. Therefore the action needs to be tied closely to the initiating process of knowledge. This must happen in an integrated way.
However, both health care in general promotes a model that physically and mentally separates the processes of knowledge | order | action. This results in sloppy thinking and fragmented care. Unfortunately the electronic health record only makes this fragmentation worse.
Let us look at examples of fragmentation of knowledge | order | action vs integrated knowledge + order + action.
PART 2:
Examples: Order | Action, fragmented
eg 1. Those blood pressure pills.
The physician prescribes a patient new blood pressure pills for their hypertension. This is only a small part of the workflow. The important question is. Does the patient take them? What are the side effects? What is the effect on the patient’s blood pressure? Does the dose need to be increased? Does the dose need to be increased again? What is the target blood pressure for this patient? When will a second drug be added? Does this patient need their kidney function re-checked?
As you can see, a simple care intervention such as seeing that a patient’s blood pressure is high, and prescribing medication for it is only a small first step in the workflow that results. The traditional medical model would book this patient in for a follow up (typically at some arbitrary interval), and the physician would then think about some of the questions above.
There are two problems with this. (1) Why is the entire workflow after the medications ordered not standardized? Why are the follow up questions, dose titrations, labwork not part of the initial order? Why does the physician have to do all of this? Can’t the clinic nurse carry out a a multi-step workflow over the next two months with the patient by phone?
The reality is that most clinics don’t take a structured approach to the post ‘action’ step. Even though, that is where most of the work is required.
Takeaway: an initial ‘order’ is only a small first step part of a much more complex workflow. Ideally that workflow should be part of the original order. Because orders lack integration into workflows, the time to resolve the knowledge problem (the patient’s blood pressure is high) is slow. This is proven by the very slow time to get patient’s to appropriate blood pressure titration targets.
eg 2. Typical follow up of tests
Your physician orders 2 imaging tests, 20 in-house labs, 5 out-of-house labs, and a consult. How is this tracked? Do they have an explicit check-box for each of these items in their clinical note? Do they check off each item that returns and records the date it came back? Do they have an alert system if they have been waiting too long for an item? Do they have different expected times for different types of orders to come back depending if they are ‘regular’ vs ‘urgent’? Generally, the answer is no. How then does a clinic function with hundreds or thousands of patients each with dozens of tests? By luck.
The physician may book the patient to come back after an arbitrary time interval - ‘come back in three months’. Or after what they think may be the slowest test, ‘come back after the MRI’.
At that follow up encounter, often the physician will open your chart, look through to see what has come in since the last encounter, and quickly eyeball their last note to see approximate that what has returned is what they ordered. But this comparison is more an approximation and than exact match.
Ideally, the system would track every test that returns. Know when the complete ‘set of tests’ has been received. At this point the system would alert the clinician to review this information in bulk and assess if further tests should be ordered, or if the patient should be seen for follow up, or if a note can be written now to the referring physician. The system should also have alerts that trigger if specific tests have not been received within the expected time frame, therefore prompting a review of the chart prior to the entire set of tests being completed.
Takeaway: the ordering of tests needs to be closely tied to the original reason they were ordered. Otherwise they get lost, take too long to come back, or are never followed up on. Patient follow up should can also be triggered by complete return of the investigations.
eg 3. The Inbox - Paper
Your physician likely has (or had) a paper inbox. The hundreds (and thousands) of reports they receive each week are placed in a pile. The report typically contains the exact information for the test. The set of labs from that day. The imaging result from that day.
However, this system by default is a system that is fragmented. The test was ordered because of a specific problem, or to follow up of a past test. The imaging was ordered to compare it to the past.
Receiving the result in a way that is fragmented to the reason it was ordered, means it is hard to properly assess why it was done.
Your physician likely eyeballs the result. And if it looks ‘ok’ - meaning they see no obvious red flags - they initial the page as reviewed and move on. (I’ve previously written a two part post on the problems with reference ranges & abnormal flags Part 1, Part 2)
Often inpatient tests may come back weeks after a patient has been in hospital. The physician may not remember exactly who the patient was, or why something was ordered. Long as the order looks ‘ok’, they move on. The alternative - to pull the patient’s paper record from the hospital achieves. To flip through it to find why something was ordered. To find the previous results to compare it to. All this is too much work. Fortunately, not correlating a result to the original tests is ‘safe’ ‘statistically’ ‘enough’ times.
Takeaway: the result received must easily be connected to the reason for ordering
eg 4. The Inbox - Digital
Electronic Health Records unfortunately have not fixed the broken processes in the above examples.
Instead, the electronic health record has (once again) replicated the physical world through a tool called the ‘Inbox’ or ‘In Basket’. This is a place where clinicians receive ALL the results they ordered.
And like paper, each result is disconnected from its purpose. Now true, if you want you can open the result. Click through to open the patient chart. Click to find your note where you ordered something. Click to load past results. And then view the result in connection to past information. But that is a lot of work.
In fact. Even reviewing the inbox is a lot of work.
I have heard of physicians who admit to deleting their entire inbox without reading it. The volume that comes into it, and the difficult of reviewing it, is too great. (Yes, I realize this is illegal). As evidence that this isn’t a rare practice - Atul Gawande writes about this in his December 2018 article, “Why Doctors Hate Their Computers”.
Sadoughi told me of her own struggles—including a daily battle with her Epic “In Basket,” which had become, she said, clogged to the point of dysfunction. There are messages from patients, messages containing lab and radiology results, messages from colleagues, messages from administrators, automated messages about not responding to previous messages. “All the letters that come from the subspecialists, I can’t read ninety per cent of them. So I glance at the patient’s name, and, if it’s someone that I was worried about, I’ll read that,” she said. The rest she deletes, unread. “If it’s just a routine follow-up with an endocrinologist, I hope to God that if there was something going on that they needed my attention on, they would send me an e-mail.” In short, she hopes they’ll try to reach her at yet another in-box.
Atul Gawande, “Why Doctors Hate Their Computers” Dec 2018, The New Yorker
Takeaway: the result of an order, must seamlessly be connected to the workflow it is part of.
PART 3:
Examples: Order + Action, integrated
Fixing the disconnect between orders and actions is as much a change to the practice of medicine as it is having better tools to use.
A systems approach is required. One that is rigours and that views care as an integrated workflow with multiple steps, checks and balanced, and that helps guide patient’s safely in the right direction.
Let’s look at two examples where clinicians have integrated knowledge + order + action together.
eg 1. Hematology Clinic
I was speaking with a friend about the systems he used to have in place prior to the electronic health record. He is a hematologist.
All longitudinal patients had a results flowsheet. The sheet had at its top
- the patient’s name and age
- their diagnosis (eg. Thrombocytopenia, thrombocytosis, auto-immune hemolytic anemia)
The rows consisted of
- specific lab tests being tracked related to their condition
- medications relevant to their condition
- other relevant data connected to their condition (eg imaging, life events, surgeries, etc)
The columns consisted of
- dates when new labs would be written in
- dates when medication changes or significant events would be entered
The workflow was straightforward. The clinic nurse would receive every new test result on a patient. They would then record in a new column on the patient’s flow sheet the relevant new information.
The competed patient’s flowsheets were handed to the physician. A paper-clipped held together the flowsheet, the original new results, and a sticky note.
In one glance, the physician could see, the patient’s name, remind themselves about the condition the patient has that is being managed, see their most recent lab results, and how they trend to the past labs, medications, surgeries, and events.
It is very easy now to make a complete and informed decision using the result of their last order. They could write on the sticky note - something such as how to decrease the medication, increase the medication, add a medication, or continue the the existing plan. The clinic nurse would then follow up on this and liaison with the patient.
This system is safe, fast, logical, and just makes sense. All of healthcare should be like this. However, this is truly the exception to the norm.
Unfortunately, when the clinic migrated to an electronic health record the above process was broken. Results now are dumped into a digital inbox. To understand why they are done, one has to open the patient’s digital file, scroll back to find your last note, read the last note, open other windows to try and compare the lab results, open other windows to find (if you can) trends of their past medications and surgeries, and then open other windows to place an order. It is crazy that this is the norm. This is now an example order oder disconnected from action. A return to fragmented care
Takeaway: the action (eg. results) need to be closely connected to the order, and the past knowledge
eg 2. The dermatology follow up
Another example of an integrated system (yet again, on paper).
A dermatologist I knew would record in a book every single biopsy he took. Write the patient’s name. Site. and then in a column what he thought the pathologist report would say (eg. Basal Cell. Melanoma).
The dermatologist would then record every pathology result that came back in this book. He had two symbols. One symbol was placed beside the original row if he got the diagnosis correct. The second symbol indicated the result came back and his prediction was wrong.
He took pride in being almost always correct in his ability to anticipate what the pathologist would say. In addition, “When I see my predications are becoming incorrect, I will know it is time to retire”.
This system is safe: he always knew exactly what results are outstanding. Because he wrote down what he was anticipating the biopsy would show - he knew which outstanding results to follow up on more urgently.
It also allows one to match their prediction to reality. I have not yet across another physician who so closely tracks his predictions vs outcomes. This is unfortunate, because without doing so we cannot get better.
Takeaway: connecting action + orders + knowledge is safer for patients, and allows clinicians to improve their skill and see their weak areas
Conclusion
We have looked at several examples of how typical healthcare operates in a fragmented environment. Where test knowledge + orders + action are disconnected from each other. The order is viewed only as a messenger between to separate domains.
The previous post looked at the advantages of integrating ordering into the process of step of clinical knowledge. How the order can help guide the clinician in writing the order, and verifying that it is correct.
This post has shows several exemplary examples of how orders and action can be linked more closely. Connecting order and action emphasis how an order is only a step within a large workflow and process. The the action that results from the order is closely tied to the reason for the order itself.
Although this post focused a lot on ordering of tests, almost every type of order - a medication, an intravenous infusion of fluid in the hospital, a nutritional order, is part of a workflow. It originated because of an issue / problem, and requires a follow up and assessment at a next step.
This linkage doesn’t have to be complicated. As we have seen, thoughtful physicians have come up with paper based ways to do this. Efficient and effective care systems are actually pretty easy to design. But it takes a mindset that embraces a level of operational efficiency that mirrors that of a high performing organization. (Not the cottage industry of health care today).
As electronic health records expand their scope, let us not build order entry user interfaces that promote and encourage the fragmentation of care. Instead let us design integrated systems from the start. This will create order entry user experiences that are faster, more efficient, more reliable. Better for both patients and physicians.