North York General Hospital (NYGH) is a 423-bed community teaching hospital affiliated with the University of Toronto, in Toronto, Ontario. Under its clinician and IT leadership, North York last year became the first “stage 6” community teaching hospital in Canada, according to the electronic health record (EHR) schematic developed by HIMSS Analytics, a division of the Chicago-based Health Information and Management Systems Society (one small rural hospital in Canada had previously achieved stage 6; and another Canadian hospital, an academic medical center, achieved stage 6 at the same time as NYGH).

North York is also the first Canadian hospital to implement computerized physician order entry (CPOE) that includes regularly updated evidence from the medical literature embedded into physician workflow. The clinicians at North York are using the evidence-based order set solution from the Los Angeles-based Zynx Health, embedded into their Cerner Millennium core EHR. As of late last autumn, medicine, surgery, and critical care, which represent the bulk of the hospital’s inpatient care volume, were live on the Zynx solution, with the remaining units in the hospital set to go live in the coming months.

Jeremy Theal, M.D., North York’s director of medical informatics, and effectively the hospital’s CMIO, spoke recently with HCI Editor-in-Chief Mark Hagland regarding his organization’s journey to date around evidence-based clinical decision support. Below are excerpts from that interview.

What was your organization’s preparation like for the rollout of the Zynx solution?

We didn’t use interim paper-based order sets with Zynx; we went directly into electronic. We had a core order set build team, comprised of four physicians and four pharmacists, as well as clinical informatics experts, which customized order sets to 80-percent completion, and we then had clinical subject matter experts review those order sets. Front-line clinicians from all scopes of practice were a part of the process the whole way through.

Jeremy Theal, M.D.

Were there any nurses on that team?

Not on the core team, but our clinical informatics analysts have past nursing backgrounds. And the first step with the prototype order sets was to have nursing, allied health, OT/PT [occupational therapy/physical therapy], radiology, etc., review the order sets. As a next step, we had our specialist physicians review the order sets. For example, for our pneumonia order sets, we had our respirologists review it, and so on.

What were the biggest challenges, if any, in preparing for the implementation of the new order sets?

For us, the biggest challenges were that we had a hospital whose culture, prior to this implementation, had not particularly been of the type where people were used to using standardized orders. We had a few standardized order sets, but for the most part, doctors wrote their own orders. So the standardization of care was a new thing for most people. We wanted to treat that situation carefully; we didn’t want people blaming the system. Because of all that, we were very careful to have the docs review the order sets, so that they could have direct input into how it was designed.

Were there any specific challenges for you personally in all this?

Our core vendor was Cerner; all the large core clinical vendors are based in the United States. There are some differences in terms of medical practice in Canada; so we had to create some custom workflows, as well as ask Cerner for Canadian-specific code changes. Those changes had nothing to do with the Zynx element; in fact, we found the Zynx content to be applicable to Canadian practice, with no significant modification required.

Can you mention one example of where differing practices in medicine have affected workflow, and thus, the EHR?

The way that our emergency department admissions are designed is that, when a patient comes into the emergency department, a new encounter is created for them, which is like an episode of care. Normally, in the States, all the care of that patient, even after being admitted as an inpatient, is part of that episode of care. But in many hospitals in Canada, the process is different due to government reporting requirements. In those hospitals, when the patient is admitted to hospital, the emergency department encounter is closed, and a new one is created. And that posed a problem in Cerner, because the orders would have been entered on the emergency encounter, and would not carry forward to the inpatient encounter. To solve this problem, we had to create special software patches in collaboration with Cerner, as well as to train our doctors on a new patient admission workflow.

Did the acceptance of Zynx and of evidence-based ordering come easily among physicians at your hospital?

Well, the reason we didn’t have any major problems was that we prepared very diligently for CPOE, emphasizing that this system was being created by clinicians, for clinicians, and we involved clinicians in every step of the design process. And we emphasized the potential of evidence-based order sets and CPOE for improving patient safety and saving lives; we shared with them articles from the literature that demonstrated those benefits. We made sure they knew that although their workflow would be slower at first, the result would be better patient outcomes, since evidence is built into the new workflow. And it’s hard to argue with that, since they all want the best for their patients.

So overall, then, the physicians have embraced it?

Yes, we have 100-percent physician adoption, and 94 percent of physician orders are directly entered by physicians into the system. The remaining 6 percent are verbal or are made by telephone; you can’t completely eliminate that, since doctors are not always near a computer or have their hands free to enter orders. And in terms of the medication-specific orders, physicians directly enter 86 percent of those into the system. That’s significantly above the Leapfrog standard, which calls for 75 percent of orders to be entered directly into the system by physicians.

Now that you’ve been doing this for over a year, do you have any thoughts on lessons learned?

Well, CPOE is sometimes treated as a bit of a check-box kind of process in some hospitals, particularly in the States, where you have meaningful use driving tight implementation deadlines. But the thing is, CPOE can be built well or built poorly, and the end result can be quite different in terms of its impact on the care of patients. You need to look at how your system is going to mesh with your current clinical workflows, because if you just perpetuate existing workflow problems, the new system will be blamed for pre-existing poor outcomes.

What’s more, if you don’t look carefully at the staffing, skills, and workflows necessary to mobilize the evidence contained in CPOE and order sets, you’ll miss key opportunities to improve patient care. For example, in terms of stroke, there’s clear evidence in the literature that a patient should undergo a swallowing assessment within the first 24 hours. The evidence and the corresponding order for swallowing assessment were in our new stroke order set for CPOE; however, we couldn’t meet the standard of assessment in less than 24 hours, because we have a limited number of specialists available in speech language pathology, particularly on weekends.

So we trained some front-line nurses to do a swallowing screening test, which is not as in-depth as a swallowing assessment, but it does help to determine which patients are most at risk. And now that we’ve trained a group of nurses who can perform that test, we are able to meet the evidence-based standard of assessment in less than 24 hours. The patients determined to be high-risk by the nurses are in turn referred for specialty assessment by speech language pathology.

And that’s an example of where if we had just plunked that evidence into the order set without examining the process, we would not have improved patient care, since there was an underlying gap that had to be addressed in the staffing, skills, and workflow. So unless you look at each order set and examine the processes involved, you’re just creating a system that doesn’t create the patient benefit that it’s supposed to.

So one of the key lessons learned here is to rethink processes, and to optimize them, and not simply automate past processes? 

Absolutely. So often, the temptation is simply to automate without thinking things through. And once you’ve implemented CPOE, it’s not an end-point, it’s part of a journey. You have to constantly measure what your system is doing well and what things it’s not doing well, so that you can make continuous improvements. We have a business information system that we’ve attached to our Cerner system, which generates regular metrics that help us to look at our patient outcomes and decide where we need to focus our improvement efforts next.

Do you have any explicit advice for CMIOs and CIOs?

Really, the last couple of points I’ve made are the important ones. First to implement CPOE not for the purposes of automation, but rather to build evidence into decision-making at the point of care and construct the associated hospital workflows, in order to ensure that the best care reaches the patient. Second, realize that once CPOE is turned on, it’s not the end of the journey; rather, it’s the beginning of a process to continually look at performance metrics and create clinical care process improvements that result in better patient outcomes.