In the original format of an Admission History and Physical, the “History of Present Illness” spoke to the single problem for which the patient needed to be hospitalized. The “Past Medical History” described problems the patient might have had in the past, but by definition, were behind them. In the days when patients were admitted for a single acute illness, the distinction was clean.
The Edwin Smith Papyrus, an ancient Egyptian text on the surgical management of trauma.
As medicine became progressively more successful turning once-fatal conditions in to manageable chronic diseases, the distinction between “Past Medical History” and the patient’s current problems became fuzzy. When a patient with type 2 diabetes mellitus is admitted to the hospital with urosepsis (excuse me, sepsis of urinary origin), is the diabetes part of the “history of present illness” or the “past medical history”? The patient isn’t being admitted because of their diabetes, and yet, it’s wrong to say it’s part of their past medical history because its an ongoing, usually lifelong problem requiring active management. In practice chronic medical illnesses started to show up in both places – the HPI would begin “Ms X is a 75 year-old female with type 2 diabetes admitted for urosepsis”, but the diabetes would appear again in the PMHx. For patients with multiple chronic medical illnesses, the HPI one-liner got packed with a comma-separated list of ongoing conditions, which then line up duplicatively in the PMHx, there joined by medical problems truly in the past.
In an electronic health record this fuzziness becomes obvious because a problem-based EHR (Epic for example) forces the physician to put their nickel down. Is the Past Medical History really in the past? The actively managed Problem List is integrated in to documentation, order writing, signout, and many other clinical functions. The diabetes belongs on the “Problem List”, not the Past Medical History, because it is an ongoing, not past, condition.
I just finished discussing this at UCLA as the Medicine housestaff are building up their inpatients’ problem lists for the first time. At UCSF we’ve tried to move the culture towards using the Problem List for all the patient’s active medical issues, including chronic issues, and reserving the Past Medical History for items actually in past. I remember pitching this for the first time to our Medicine residency program leadership, who looked at me like I had two heads. (in the most friendly and collegial way) Today’s patient is likely to have more present problems than just one. It’s a culture change, but it’s more accurate, more concise, and consistent with the original distinction between the HPI problem(s) in the present and past medical problems in the actual past.
Every EHR go-live marks their support staff in some colorful way so they’re easily identified. At UCSF we had our support team in green polo shirts. At UCLA we’re in spiff red fleece vests. As a Cal graduate, wearing Stanford colors on a UC campus will always disorient me, but it works.
At midnight tonight UCLA Medical Center will go live with CareConnect, their name for their implementation of the Epic electronic health record. With this, four of the five University of California Medical Centers (Davis, San Diego, San Francisco, and Los Angeles) will have the same EHR infrastructure.
A handful of us from San Francisco, including my colleagues Carolyn Jasik, Ellen Weber, David Robinowitz, and I, are spending four days here helping support their clinicians and their go-live team. The two institutions have a lot in common, and several physicians from UCLA visited UCSF during our go-live to provide support and meet with our leadership team. It will be fascinating to see the similarities and differences in our respective builds and in how the clinicians work through their first days with the system.
On September 1st UCSF Medical Center turned on electronic health information exchange with our Epic electronic health record. It’s an important step forward and one of the features of electronic health records I’m most enthusiastic about. It has the power to improve health not just at our own institution but wherever our patients go. We had hoped to enable this as part of our June 2nd inpatient big-bang go-live, but decided to wait 90 days to make sure some final details were fully hammered out.
Epic has two levels of electronic health information exchange, “CareEverywhere” for exchanging information between Epic customers, and “CareElsewhere” for more limited exchange with a non-Epic EHR. For now we’ve turned on CareEverywhere, connecting us with participating northern California providers like UC Davis, Stanford, and Palo Alto Medical Foundation, although there’s no geographic limit within the United States for where records can be shared. (With their particular sense of humor, Epic presented this week at their annual meeting on the future “Intergalactic” sharing of health records, emphasizing the point with photos of the Curiosity Rover and The Netherlands) Our first exchange was at 10am on the 1st when we electronically received records for an ill youth hospitalized at UCSF who had previously received care at Stanford, and we had a dozen exchanges in the first 7 days.
Like always, we only share health information between institutions after getting written permission from our patient. The large majority of patients want their health information shared electronically with other physicians and hospitals when we need it to provide safe and appropriate care, as long as we are sharing securely. In my experience, my patients are surprised to learn even major hospitals have largely remained isolated islands of information. When I collect permission from a patient to obtain their records from a hospital across town, patients are usually surprised and discomforted to learn I didn’t have access to it already. Health care is far behind other industries in this kind of information integration, and fixing this in a hurry is a centerpiece of the federal government’s standards for health IT implementation.
For the last few decades health records have been shared primarily by telephone and fax. We call the primary physician’s office (for example) and if we actually reach the physician immediately, we usually get their best recollection of the patient off the top of their head, followed by more complete information by fax hours or days later if at all. If the patient was recently hospitalized, getting that hospital record requires work from that hospital’s medical records department, seldom a 24/7 operation, and it arrives as a thick, grainy, often disordered, fax-of-a-copy-of-a-scan of the original record. This helps, but the information has to be manually transcribed in to our own record, which is only as accurate and complete as any 10-fingered process.
With electronic health information exchange, sharing patient records is more secure and more accurate. The electronic point-to-point connection between institutions is encrypted and the identity of the patient is confirmed electronically between the EHRs. The patient’s health information arrives immediately in our EHR instead of on the tray of a fax machine some unknown number of hours later. A physician on our receiving end reads and validates the electronically exchanged information before incorporating it in to our own record. The exchange is at the level of data instead of pieces of paper, and so discrete information like medication lists, drug allergies, problem lists, and other pieces of history can be synchronized between the institutions.
Unfortunately we’re still not able to exchange information with San Francisco General Hospital or Kaiser Northern California, two providers with whom we share many patients. Kaiser Northern California has been on Epic for years, but does not to participate in electronic health information exchange. San Francisco General is moving fast on implementing its own electronic health record, and we look forward to connecting with them when the capability on both sides is ready.
We’ve been live on Epic now for a little over 1 month. Our newly-minted interns started work on June 21st and a flock of new upper-year resident physicians and fellows began July 1st. One of the most enjoyable aspects of practice at UCSF is the phenomenal quality of students and trainees we attract, and as hoped, our new trainees have taken to the EHR and computerized provider order entry especially smoothly. In some respects they are more comfortable with the workflows than the senior trainees who directly supervise them because the new arrivals have no prior expectations from how things at UCSF used to work. We also have the advantage that something like one-third of our new trainees come from a hospital system that was itself an Epic customer.
Total call volume to our “provider” (physician, NP, and PA) and general help desks, starting with the go-live date.
The interaction between Epic CPOE and our lab and radiology systems continue to have some challenging wrinkles. Epic offers an integrated laboratory system called “Beaker” and a radiology system called “Radiant”. (Epic likes to give cute names to its software components) For reasons of project scope among others, we chose to stay with Sunquest and IDX/Rad for lab and radiology (respectively) for the time being. Although these are each leading systems and widely used elsewhere, the workflow integration between Epic and these ‘outside’ systems remains a work in progress in edge-case scenarios.
The only workflow to date we’ve backed out of is using Epic to satisfy the CMS requirement for an attending physician to document their face-to-face evaluation of an inpatient to qualify them for home care. We built this in Epic as an ‘order’ with all the required elements, and the Case Manager could tee this up (‘pend’ it in Epic jargon) for the attending’s review and signature. For reasons of workflow and the competition for attention, we’ve backed off on having this be electronic for now and reverted to the paper form.
We’ve been live on Epic in inpatient now for two weeks and we’re cautiously very happy with our results. Direct entry of orders by physicians is stable at just over 90%, with the balance being orders written on paper in settings we planned for (chemotherapy and pediatric TPN) and verbal or telephone orders. We’re digging in to our live data on verbal and telephone orders to see how they cluster and how we can continue to reduce them. Revenues remain within the margin of variation, and our near-term clinical metrics (door-to-floor, average length of stay, etc) are unchanged to slightly improved. The very interesting clinical outcomes, like rates of medication error or risk-adjusted mortality, await more data.
Our knottiest workflows in the system are what Epic calls (a little strangely) “Hospital Outpatient Departments”. These are facilities that serve both inpatients and outpatients, like interventional radiology and the endoscopy suite, and so require a mix of inpatient and outpatient workflow and software. If the patient needs full anesthesia for the procedure, as is often the case with child patients, you have a triple-intersection of complexity. In our paper-based prior existence this was all smoothed over by smart people with lots of institutional knowledge and the right relationships. They knew how to get anything done. In preparing for an integrated EHR we put a great deal of effort in to analyzing this work prospectively, but the magnitude of change brought by automation has had unintended effects. No surprise, and we’re crunching through all those processes again with the benefit of experiencing the system in real life, and the problems continue to look solvable.
Meanwhile our clinicians continue to ask deeper and more interesting questions about the system, moving from “How do I get my job done?”, to “How do I manage this complex discharge?”, and on to “Our Division wants to start publishing custom packages of SmartLinks and we want to send people to Wisconsin. Whom do we talk to?”
Still in its early stages at UCSF, one week old as of today, is inpatient documentation centered heavily around the Problem List. This has been used by some providers in the ambulatory setting over the last year, but is now being used nearly universally on hospitalized patients. Rather than writing an assessment and plan as free text at the bottom of an admission, progress, or consult note, each relevant item on the hospital Problem List will now have a unique assessment and plan note written specifically about that item.
This has several potential beneficial effects:
1) Sharpening a physician’s diagnostic thinking.
2) Improved continuity-of-care and ability to review a patient’s chart. Currently, if a patient named Bob has COPD and you want to find out what has been done over the last ten years for his COPD, you have to sift back through every old note in the chart, looking for the ones where COPD was discussed. Very time-consuming. With the Problem List approach, Epic allows you to “View All Notes” for any given item on the Problem List. This means that with a single click, you can conjure up every assessment and plan note written specifically about Bob’s COPD, whether from the ED, ambulatory clinic, or inpatient setting, going back through time.
We think this is a very powerful tool and are excited to see it begin to improve care.
Its our 7th day live on Epic in our inpatient areas, and at 1522 we lost commercial power. The rumor so far is that PG&E has had a substation go down and power is out for several blocks at least. Our clinical areas are equipped with multiple layers of emergency power and are fine, as is our phone bank for supporting clinicians, but here at the Command Center in UC Hall we’re in the dark for now.
Update – the word on Twitter is that power is out all the way to 19th Avenue (we are at 3rd Avenue) with 11,000 customers blacked out.
We’re five days in to our go-live and continuing to maintain all our expected workflows on the system. Our CPOE rate hit 90% today, the balance being orders written on paper for expected reasons (Pediatric TPN and Chemotherapy) and telephone or verbal orders. Our ED door-to-floor time is about expected, neither better (yet) nor measurably worse. Revenues are within the margin of variation.
Below are the trouble tickets called in to our service desk broken out by category. As anticipated the large majority of issues are users whose security is not set up perfectly for their role. Security settings in Epic are very customizable and, it has to be said, enormously complicated. When I took the Epic security class in Madison it reminded me of the nephron.
Next in the stack is “optimization” which for now doesn’t count as trouble, and third on the list is printing. With the incredibly powerful things we can do with technology, like order and document extracorporal membrane oxygenation on a premature infant clinging to life, I find it remarkable that printers everywhere continue to be such a pain. No amount of rigor and testing (and we did a lot of testing) here or elsewhere can seem to entirely decomplexify them.