Douglas Johnston, M.D., a cardiovascular surgeon, and director of the length-of-stay and throughput initiatives taking place at the Cardiovascular Institute, a division of the Cleveland Clinic Health System, Cleveland Oh., is a practicing surgeon who is very supportive of the medical documentation improvement efforts taking place at Cleveland Clinic. He has been working closely with William Morris, M.D., vice chairman of clinical systems, and a practicing hospitalist in the organization, and with Susan Belley, manager of coding and documentation improvement, on a project to improve the quality of physician documentation among doctors practicing in the Cardiovascular Institute there. A description of their work can be found in the May cover story of Healthcare Informatics.
Johnston spoke recently with HCI Editor-in-Chief Mark Hagland regarding his perspectives on the challenges of making physician documentation work for everyone. Below are excerpts from that interview.
How do you see the tension between free-text and structured documentation?
The challenge is finding ways, I think to think outside the traditional paradigm; and there’s a cultural barrier in that we are all trained as physicians and other providers to believe that documentation is a matter of putting certain verbiage down on paper, whether physical or virtual, and that that is an endpoint for the encounter, and that that is a narrative. It’s not a concept-oriented thinking, it’s a narrative-oriented thinking.
And actually, that’s not the way we think about patients. We end up synthesizing what’s really a very fluid process, regarding patient problems; and I think we already lose something right off. There’s a certain element of information that works best in free-text: Mrs. Jones is a very elderly lady, more frail than appears, etc. And, this [optimizing physician documentation] is more difficult because of that.
Douglas Johnston, M.D.
On the other hand, everything we’re measured on is in terms of diagnoses; and if we can’t find an accurate way to produce an accurate list of diagnoses, we’re thrown into a harsh light. And come to think of it, most of our encounters with patients are longitudinal.
So we’re thinking, what did I think was wrong with the patient before? What’s changed? Etc. And we don’t currently have the tools we need; right now, we produce a note. And I think that’s actually detrimental to the thinking process. What we really want is a heads-up, fighter-pilot view. And we’re certainly not there yet. But the first step of the way is to ask, what’s essential that we need to know? Figure out what’s wrong and what the essential interventions are. So if we think of documentation within that paradigm, it will work better.
So one should think about problem lists first?
Yes, that’s what we’ve set up here. We think it makes most sense in terms of serving patients. But we certainly have to meet external requirements. So phase one is providing the metrics and tools to support problem-oriented documentation—help facilitate problem lists and still facilitate a note for billing and regulatory purposes. But that note can be structured around the problems first—make the notes in order of APSO—assessment, plan, subjective, objective, instead of subjective, objective, assessment, plan (SOAP) [the traditional sequencing of note-making among physicians]. SOAP is about constructing a note around the way you interact with the patient; I call it the Florence Nightingale approach. Meet with the patient, get their story, construct a narrative and plan. But in reality, if we know the patient is a diabetic, we don’t necessarily start at zero. What’s important is that you can create a note that will immediately help your next colleague down the line.
The other thing we did was to structure it so that the problem list pulls automatically into the note. So that all the thinking work in terms of taking care of the patient is done within the problem list. And we’ve structured the process of rounds around the problem list. So when the residents and mid-level providers—NPs and PAs—do rounds, they open the notes, and they first get the problem lists.
Today, we need to order an echo—so the work structure is all around running through the problems. And as they wheel the rolling computer in there, they say, Mrs. Jones, this is what we’re going to do with you today. So the problem list also becomes the plan of care, and everything becomes transparent in terms of what we’re doing.
What would your advice be for CIOs and CMIOs, as they move forward on innovating in physician documentation?
Well, I think it would be a mistake to think about the problem list as just something you have to do just to meet meaningful use. We have the highest acuity of any hospital in the United States. We believe it’s both a clinical imperative to do things this way, and the only way to meet the increasing scrutiny from a billing and regulatory standpoint. So if we’re going to take care of these very ill patients well, we have to rigorously take care of them, and rigorously document how we take care of them. And the only way to do it well is to do it discretely and transparently.
So we took this on not as a way to meet meaningful use, but as a way to make the record friendlier to providers and patients. So we took the world’s largest heart center live with problem-centered charting within a week, after two months of training, and we went from 20-percent adoption to 96-percent adoption. And this is something that providers embrace as a clinically useful tool, once they get over their initial hesitance. And even our biggest naysayers have become our biggest supporters. Our chief of the ICU says it shaves an hour off his documentation time. So in all those ways it makes sense.
But the key for CIOs and CMIOs is understanding that this is a clinically useful tool, and provides a process of care change that improves care processes, and allows us to monitor in a much more granular and precise or accurate way, what’s going on. So in real time now, we can say who’s in the hospital and who has heart failure. Or who has a combination of heart failure and renal failure and is on a particular drug. Otherwise, it’s buried in the free-text of notes. It also allows us to identify patients who may be inappropriately identified as having had an adverse event; for instance, which ones have pressure ulcers?
In many cases, the patient may have an issue, but might have come into the hospital with it. Or it might be a minor problem but mis-coded. And so we get more consistency of documentation, and we can identify inconsistencies in real time. I mean, for example, we might appear to have had a bump in acuity in October in the past, but who remembers October? So we’re just beginning to the tap the data we can pull out of this effort, in real time.
One example: I’m in charge of the length of stay and throughput efforts for our heart center. And one challenge has always been dealing with this back-end administrative data. For example, if we knew someone was headed for a relatively long length of stay, we could plan for a number of different interventions: we could get case management and physical therapy involved earlier; could potentially pre-schedule them for a rehab or skilled nursing bed; or make certain kinds of arrangements for home care. But the key is knowing what’s going on beforehand. Ideally, we’d like to have an automatic set of triggers.
So we’ve created a predictive model that takes the problem list together with other data available from the EMR; and it’s not as good a tool as if you were able to ask a patient 100 questions, but it’s very good as an automated predictive model. We can pull back a predictor of greater than expected length of stay, longer length of stay, potential complication or need to go to a skilled nursing facility, by putting data in. So it solves one of the age-old problems, I have a huge list of patients to interact with today, and who do I focus on.
It’s essentially a risk stratification tool?
That’s right. And our intent is to start looking at a whole bunch of different types of data eventually, using this tool. There are all sorts of tools out there using nursing documentation, or admitting data plus sematocrit, etc., but incorporating physician documentation into this creates a much more powerful prediction model.
Do you think these kinds of tools will add significantly to the effectiveness of healthcare?
We hope so. I think we really have the chance in the next five to 10 years to make documentation and other information technology tools a mentor to clinicians rather than a hurdle. We need to make this the fighter-pilot version and not the Microsoft Excel spreadsheet, which it is now.
And the physicians are embracing these tools?
Remarkably, the physicians are embracing them very well; and the usability level is still low, but the adoption is still great; which tells me that the idea of doing problem-centered charting appeals to physicians, and works well. So the fact that we’re making progress with this tool set is a hopeful sign.