Back to the top of this file.   Go back to last location

I. Introduction

A. Why CPOE Now, After All These Years?

"What is your CPOE strategy?" Over several months of posing this question to hospital information technology executives, we have yet to run into one who did not have at least the beginnings of a plan. In recent meetings we have attended, the Healthcare Information and Management Systems Society summer meeting in Las Vegas in June 2002, heavily focused on computerized physician order entry (CPOE), and the Association of Medical Directors of Information Systems meeting in Ojai, Calif., in July 2002, focused almost entirely focused on CPOE.

Physicians are acutely aware of CPOE, and, as might be expected, are deeply divided in their views. While researching this study and writing the report, we’ve asked virtually every physician we’ve encountered in both our business and personal settings what they thought of CPOE. A clinical professor we visited at UCLA said she’d be thrilled to have a CPOE system, citing ease in entering complex orders and having them available and universally understood. On the other hand, a highly respected community-based specialist expressed the view that CPOE is a management tool to control physicians’ decisions (a topic discussed in this report). Also, as reported in Inside Healthcare Computing (IHC, 10/28/02), we ran into a physician who seems to dislike computers so much that he refused to accept one of the authors of this report as a patient because the only available formats for the relevant diagnostic images were digital–as a CD or as files on the computer hard drive of a diagnostic center 50+ miles away from his office. This M..D. demanded “the” film (of which there is one) or no visit.

Back to the top of this file.   Go back to last location

In October 2002, Sheldon I. Dorenfest & Associates, which surveys hospitals, said that in interviews of the first 441 interviewed for the 2002 survey, patient safety and CPOE initiatives are "motivating large growth in plans to purchase new clinical systems." Regarding CPOE systems, Dorenfest says these first surveys show:

Dorenfest says that purchases of certain other new clinical software systems are "skyrocketing," especially in PACS, surgery, radiology, emergency room, pharmacy, and document imaging. (Sheldon I. Dorenfest & Associates, Ltd., 515 N. State St., Suite 1801, Chicago, IL 60610, (312) 464-3002, fax (312) 464-3090.)

Why did this focus on--and possibly rush to implement -- computerized physician order entry (CPOE) come about? After all, physician order entry has been around since El Camino Hospital went live with the Technicon system in 1971 with physician order entry for pharmacy orders, and has been in wide use in major teaching hospitals for well over a decade.

An often-cited reason is that the heat has been on to reduce medical errors since November 1999, when the Institute of Medicine released what is widely perceived to have been a landmark report, To Err is Human. This report codified what has been increasingly obvious to perceptive patients for a decade or more -- that there are too many medical errors despite the dedication of care professionals. While the report calls for a “systems approach” to reducing medical errors, we believe its facts lend themselves equally in support of our view that the more fundamental issue is this: as the financing of health care has deteriorated since the late 1980s, so has the quality.

However, regardless of whether a systems approach, or reversing the slow fiscal starvation of caregivers, is the more fundamental response, the release of the report was extremely timely.

Soon after the IOM spoke up, an organization called the Leapfrog Group was funded by the Business Roundtable, a trade association of leading industrialists. Leapfrog leapt to, and has acted on conclusions including these--

1. medical costs can be significantly reduced by reducing medical errors; and
2. CPOE is a central, valid tool to reduce medical errors.

Leapfrog began economically pressuring health care providers in 2001 to implement CPOE. Its method was to seek to steer patients covered by its sponsor industrialists to hospitals with information systems which provide CPOE with clinical alerts to physicians.

We believe that there is ample reason to be skeptical that Leapfrog's core goal is to improve care -- if, for no other reason, than the fact of who is paying for it to speak up. The Business Roundtable, as far as we can tell, has never left the forefront of those entities which, in the political arena, wage constant war against corporate spending on health care benefits for employees. We believe that--

1. Leapfrog's focus on medical errors is, quite validly, a focus on reducing the high costs resulting from medical errors; but that--
2. A likely greater interest of Leapfrog’s sponsors in CPOE is as a management tool to clamp down on physicians’ consumption of hospital resources covered by health insurance; and that
3. There is good reason for provider organizations to embrace this motive – to make care more effective and also to head off its potential for abuse.

For more on this view, see the "Our Observations" section at the end of this report. However, regardless of its motivations, Leapfrog has emerged as a highly visible, and therefore important, source of pressure to implement CPOE.

However, many CIOs tell us they are implementing CPOE primarily because good patient care is their organizations’ mission, and they simply cannot tolerate the notion of inflicting avoidable harm. User reports gathered for this study strongly support the view that the reduction of medical errors and faster treatment turnaround (and thereby, less avoidable harm to patients) are both motivating factors, and indeed likely, if you implement CPOE well.

Some CIOs also see CPOE as a cost of staying in business. If you don't implement CPOE now or in the near future, you won't have to worry about finding funds for that new building, because you won't have any patients to put in it, said Ken Clarke, CIO at Central Maine Medical Center, which is in the selection process. "Who," he asks, "wants to go to a hospital that is unsafe?"

Another cited reason: in 5-10 years, hospitals will be bursting at the seams even more so than they are today, said Greg Walton, CIO for Carilion Health System. Hospitals that do well financially will make a priority out of shortening lengths of stay. A way to reduce lengths of stay is making treatment and information available to the patient as soon as possible. A well-functioning CPOE system helps meet that goal, he believes.

However, building the perfect system is, in the words of First Consulting Group's medical errors guru David Classen, M.D., "a very high-risk, difficult initiative." You need deeply committed top executives, and close synchronization between your top clinical leadership and your top executive leadership. You need to go to great pains to engineer your software for flexibility, so that if, during the course of implementation, you realize that the net effect of your efforts is to dump heaps of new work on someone, you can apply a just-in-time corrective measure, so you don't suddenly have a revolution on your hands. You need to decide up front how you will measure success, so that down the road when the going gets tough, as it surely will, you will have a way to convince your clinical staff to continue.

Moreover, Dr. Classen and others note, making the process of care faster and more "efficient," like a computer-managed industrial production line, can be a direct pathway to worse, not better, outcomes. See "Many Hospital Pharmacies Shut Off Patient Safety Software Alerts" below, and the "Our Observations" discussion at the end of this study, for additional views on this topic.

Back to the top of this file.   Go back to last location

B. Questions And Issues Addressed In This Report

This study has revealed to us that while institutions have succeeded at CPOE have taken varied paths to successful CPOE implementations, most have done a great many things similarly, and all have had to find ways to surmount the same hurdles.

A small portion of this report focuses on the "why" of CPOE. To a degree, your course of action will vary depending on whether or not you're planning to use CPOE to influence or control resource consumption decisions by physicians in situations in which such decisions are not, strictly speaking, questions of erroneous or correct medical practice, but of resource consumption. As we have said, we believe that addressing resource consumption is the hidden agenda behind some CPOE advocacy. It is also a significantly more ambitious undertaking than that of reducing errors. If you take it on, it will put CPOE implementation smack in the middle of major governance issues, and political issues both internal to your organization and national in scope.

However, in at least the early going, judging from what implementers have told us for this report, your course of action (that is, successfully getting CPOE working and stable without a physician or nursing rebellion) will be largely the same regardless of how far you intend to eventually push the potential of a CPOE system.

Most of the content of this report deals with this early-going "how-to," and the "how to avoid our mistakes" commentaries of organizations which needed to make mid-course corrections in their CPOE implementations. These are the issues and questions this study seeks to address:

• The real costs of CPOE above and beyond software, hardware, and implementation.
• How many order entry devices you will really need.
• Who should be on the CPOE physician advisory team.
• Challenges and hidden costs of CPOE in best-of-breed information systems
• Why CPOE may be easier to implement at a rural hospital.
• User self-assessments of what they did right
• What went wrong at some sites, emphasizing avoiding troubles they encountered.
• Implementers' views on why and how they succeeded, and benefits derived.
• How much physician order entry is done on wireless devices.
• Whether care providers are using departmental and individual order sets.
• How and how widely "live" on-line decision support is used.
• How many FTEs and which job skills are needed to implement a CPOE system.
• What early implementers would do differently if they were to do it again.
• The greatest benefits and payoffs of CPOE systems.
• How much training you should expect to do.
• Convincing physicians of benefits to them.
• Why CPOE screens will need to vary for some departments.
• Whether CPOE systems really pay for themselves.
• Clinical decision support: why careful management is needed.

Back to the top of this file.   Go back to last location

Closing Notes:

(1) This report steers clear of evaluating or specifically comparing vendors. Some of the comments herein may be useful to you in drawing conclusions about individual vendors and systems, but this is not intended to be a vendor-to-vendor comparison of capabilities. Most of those interviewed for this report are early adopters. Thus, they both had to discover and deal with unforeseen new-system problems, and had the benefit of extra vendor attention. Neither a glowing report nor report of trouble should be assumed to be a likely universal experience with that particular system.

(2) This report is written without footnotes. We are journalists, not academicians. We write in news style, in which most of the time, the attribution is contained within the paragraph in which a source is quoted. We believe that this makes for easier, less distracting reading.

II. Overview: Key Lessons Learned On Implementing CPOE

This overview quickly summarizes some of the broad lessons your colleagues have drawn from their implementations of computerized physician order entry:

Back to the top of this file.   Go back to last location

1. Be prepared to have to tailor CPOE screens and work flow to some departments' special needs or face system rejection in those departments.

A key crossroad: when you witness physician resistance, it may well be just resistance to change by physicians who are accustomed to doing things their own way. However, a highly likely alternative possibility is that the CPOE system, as designed or implemented, truly interferes with the ability of physicians to get their jobs done.

Asif Ahmad, CIO at Ohio State University Medical Center, winner of the 2001 Davies Award for outstanding computer-based patient record implementation, goes so far as to say that more frequently than not, when a CPOE implementation runs into trouble, "It is the CIO [who] is the problem." Most CIOs don't understand clinical process well enough to design a workable implementation. (For more information, see the Ohio State Case Study below.)

Back to the top of this file.  Back to last location

2. Take care in selecting a physician advisory committee:

1. Mr. Ahmad suggests including ancillary department physicians, not just surgeons and internists. You need physicians who think analytically and understand workflow, not necessarily MDs who are technically proficient with a computer.
2. Gary Strong of Fairview Health Services in Minneapolis, Minn., has recruited early users by specialty area, focusing those specialties which, by their nature, have the greatest potential for disastrous medical errors.
3. Holland, Mich., Community Hospital recruited physicians who are closely aligned with hospital management to test the system. They are likeliest to speak directly about what they think of the system, and to be forgiving if does not work perfectly right out of the box.
4. Meditech's Hoda Sayed-Friel recommends including a champion, a pessimist, and a devil's advocate, because that way, all physicians on your staff will feel represented. The IT department may find it easiest to communicate with physicians who are in roles of administrative leadership. They are likeliest to be willing to tell you if your system stinks, and you may find it easiest to tell them if their expectations are unrealistic.
5. Bill Van Dornik, director of information systems at Holland Community, used physician administrators to beta-test the first version of new QuadraMed CPOE software. They didn't like it, and they told him so, without walking away from the project.
6. However, if your committee is comprised only of physicians who are in positions of administrative leadership, you could be in trouble, because rank-and-file physicians do not necessarily respect administrative leaders. Similarly, you want to send your top clinicians on site visits, said Bart Lally, an MD with El Camino Hospital, Mountain View, Calif., which was shopping for a system to replace its 31-year-old TDS (now Eclipsys) installation as this section was being written. "They are your real users." And if they resist? Tell them that if they don't like the change, this is their chance to be part of the process.

Back to the top of this file.  Back to last location

3. Clinical alerts should not be too weak, but should not be too intrusive.

Plan on testing them with caregivers to achieve a balance.

Ohio State has implemented the Siemens Healthcare Dashboard, which fires alerts for abnormal results. The OSU implementation "triages" the alerts, so the MD can tell when an out-of-range result or other event is truly urgent.

(Also see "A Cautionary Tale: Pharmacy Alerts Shut Off.")

Back to the top of this file.  Back to last location

4. Honestly assess your organization's readiness for CPOE.

Here's a yardstick: look at your organization's track record for medical chart completion. Look at your rules, and look at what actually occurs. The greater the gulf between the two, the harder time you will have enforcing CPOE. (This comment came from an audience member at a seminar. We suspect that he was onto something, because he drew wide applause from a mostly physician audience.)

The group of physicians staffing the emergency room at Mount Carmel St. Ann's, Columbus, Ohio, picked up physician order entry on the A4 HealthMatics system easily, said Frank Orth, DO, group president. All 19 ED docs enter their own orders, along with 8-10 physician assistants. Training was fairly easy. Today, physician order entry is the single most popular application among St. Ann's ER physicians. One reason is that an interest in using cutting- edge technology to improve patient care is within the group's culture. But getting your average hospital's general medical staff on board with CPOE? He thinks that would be a much greater challenge. (See more details below in our Mount Carmel Case Study below.)

Back to the top of this file.  Back to last location

5. Rapid response to physicians' problems is a key to success.

A corollary is:

6. Build on what you already have--and especially what you understand well and can adjust in house.

If you have a long relationship with a vendor, consider sticking with it. Most implementers say that the success of a CPOE implementation is one-quarter the system you choose, and three-quarters the way you implement it. To the extent that you already know the vendor and the system, you may be way ahead of the game. See, for example, the Case Studies below of Lehigh Valley Hospital and Health Network, Allentown, Pa.., and St. Francis Care, Hartford, Conn.

Your vendors will not be able to make major system modifications as rapidly as your medical staff wants. Therefore, you will be much better off if your information technology department can make some changes in house and promise quick turnaround. For example, perhaps partly because of its extensive experience with IDX LastWord, Lehigh Valley Hospital and Health Network's IT department is able to respond rapidly and effectively to physician requests.

Back to the top of this file.  Back to last location

7. Integrated systems are probably your best bet.

Many organizations which have CPOE manually re-enter their orders into their pharmacy information systems, said William Bria, M.D., an Assistant Professor of Medicine at the University of Michigan School of Medicine in Ann Arbor, and a longtime leader in medical informatics. That is inefficient, less safe, and obviously best avoided.

David Classen. M.D., of First Consulting Group offers another reason to avoid non-integrated systems: they lack integrated decision support. Bruce Elkington, CIO at Overlake Hospital Medical Center, Bellevue, Wash., adds: maintaining interfaces is difficult and expensive, and many organizations will install Meditech CPOE for the simple reason that "it's installable."

8. Be sure that sure you have enough data entry devices.

Physicians are angered if they want to make rounds, and can't place orders because there's a line at the terminal. Remote access, allowing physicians to place orders from out of the ward and even from out of the hospital, is a big plus. However, there is no hard-and-fast figure on how many terminals make an ideal number; see the discussion of the numbers in case studies and in the survey response section.

Back to the top of this file.  Back to last location

9. Wireless data entry is widely used, but far from universal; PDAs are not a device of choice for wireless CPOE.

Most of the wireless devices used are PCs or laptops on carts which are moved through hallways. Physicians widely use personal digital assistants (PDAs), but the use of PDAs for CPOE is virtually nonexistent, probably because as devices, they're too simplistic for complex order sets. Microsoft’s Table PC software and the various hardware packages that go with it are getting some attention from a couple of CPOE vendors, however.

10. Get the strong commitment of executive and clinical leadership.

Be sure there is a single, shared vision for CPOE. We would add: vision is only 20-20 from the street. Share it with the troops–your clinical users, from physicians and nurses to departmental allied professionals--and listen to what they have to say about it.

Back to the top of this file.  Back to last location

11. There is more than one way to win physician support.

Some institutions, with the power do to so, don't win it--they force CPOE on physicians. See the section below, "Six Views on Winning Physician Acceptance of CPOE" below.

12. Return on investment: hard, replicable data is lacking; big savings are quite possible but present political and governance challenges.

Much of the discussion of benefits to date has to do with reduction of errors due to legibility of orders, error reduction due to decision-support in medication ordering, and improved time turnarounds to treatment. It is our view that these improvements, while clear-cut and useful, will eventually be seen as among the relatively trivial functions of clinical systems with CPOE.

CPOE does offer a clear and unambiguous pathway to significant hard-dollar savings. The chief benefit of CPOE may well not be in reducing errors and speeding turnaround of treatment, but as a management tool to standardize and limit resource consumption by physicians. However, a key question looms: to what degree does your organization's management have the clout with physicians to take this path? A second question: who will ultimately control the resource-consumption decision-tree--the caregiver organization or the payer?

More information: See the "Return on Investment" subsection immediately below. See also the benefits discussed by some early users in Case Studies and in the survey responses in Section IV, and especially the Vanderbilt responses at the end of that section. See the "Our Observations" section for our discussion of the political, governance, and resource-control issues.

Back to the top of this file.  Back to last location

Return On Investment From CPOE:
Anecdotal Information Is Real, But $$ Data Is Soft

Big Possible Savings:
Provider Management CAN Use CPOE To Change Physicians' Resource Use

Most of the information we have seen to date on cost savings directly attributable to CPOE is either soft on numbers or anecdotal. Some savings are based on a single-department system with more limited costs; some stated savings are for single applications which comprise only part of the overall CPOE or clinical system package. In many cases, improvements in turnaround time on care and reduction in errors can be documented, but no one has enumerated the savings.

Some savings and returns on investment from CPOE systems discussed in this report are:

1. Several survey respondents report payoffs and benefits in both the aggregated survey responses and in the Vanderbilt survey response in the Survey section below.
2. See the Mount Carmel St. Ann's case study for some details on savings. Officials there have estimated that direct savings will pay for their Emergency Department CPOE system in two years. The savings listed seem to be quite concrete and measurable. Note, however, that this is a single-department system.
3. Gary Strong, CIO of nine-hospital Fairview Health Services in Minneapolis, Minn., has said that Fairview expects a return of 27% per year on its $27 million investment in a clinical system with CPOE. (We did not obtain details for this report.)
4. A rare event to date–time saved in entering orders: at Lehigh Valley Hospital and Health Network, Allentown, Pa., IDX LastWord, seems to work well. The word in "hallway conversations," according to Allentown's Physician Liaison Don Levick, M.D., a pediatrician, is that proficient physicians can enter orders in LastWord in slightly less time than it takes to hand-write orders.
   
   This seems to be significantly better than what users are reporting from other CPOE systems, so it is a valid question to ask: Is this really likely to be replicated by other LastWord users? We will ask IDX for other sites that might be having a similar experience; we hope to provide an update in a future edition of this report and in an issue of Inside Healthcare Computing.
5. Ken Clarke, CIO at Central Maine Medical Center, predicts that his organization will see these three kinds of cost savings:
   • a reduction in adverse drug events;
   • a reduction in drug costs; and
   • a reduction in losses from payer audits.

Here is his logic:

• An adverse drug event costs $2,000-$4,700 (depending on which study you believe). Other studies suggest the number of adverse drug events an organization is likely have in the absence of CPOE. Five hundred per 25,000 is a pretty good number, he said. It is a reasonable assumption that CPOE will reduce adverse drug events.
• He notes that studies at Partners Healthcare have shown that physicians will choose the less costly alternative if presented with equally beneficial drugs. Central Maine plans to seize the opportunity to capture these savings by inviting physicians from Partners Healthcare to come to Maine and discuss how they approached medication substitution.
• You should boost insurance pay-up rates because the system will capture documentation on orders.

Mr. Clarke does not anticipate savings in FTEs, although he does expect to free up nurses to spend more time on patient care.

Back to the top of this file.  Back to last location

III. Case Studies

These 14 Case Studies of individual implementations are largely based on interviews which Inside Healthcare Computing Editor Suzanne Corrales conducted in 2002. For these Case Studies, all of those interviewed were some identical basic questions. However, a core purpose was to let interviewees discuss what they, not we, deemed significant. As a result, the Case Studies are wide-ranging, do not offer a common data set or answers to all the same questions, and follow no specific presentation template.

We sought to augment these Case Studies with a more formal survey asking exactly the same questions on 15 CPOE topics. The survey respondents were offered anonymity, and were allowed to respond in writing as well as in voice surveying, conducted by researcher Rachel Ross. The survey section appears below the 14 Case Studies.

Back to the top of this file.  Back to last location

The Case Studies are of these institutions:

1. Ohio State University Medical Center, Columbus, Ohio (Siemens Invision)
2. Fairview Health Services, Minneapolis Minn. (Eclipsys Sunrise)
3. Holland, Mich., Community Hospital (Per-Se Patient1)
4. Cottage Health System, Santa Barbara, Calif. (Eclipsys Sunrise at center of multiple vendors)
5. University of Texas M.D. Anderson Cancer Center, Houston, Tex. (IKnowMed CPOE with Per-Se Patient1 clinical information system)
6. Lehigh Valley Hospital and Health Network, Allentown, Pa. (IDX LastWord)
7. Great Plains Regional Medical Center, North Platte, Neb. (QuadraMed Affinity)
8. Multi-Site Survey by an IS executive (various systems)
9. Mount Carmel St. Ann's, Columbus, Ohio (A4 HealthMatics)
10. St. Francis Hospital Health Care System, Hartford, Conn. (IDX CareCast)
11. Montefiore Medical Center, Bronx. N.Y. (IDX LastWord)
12. Stanford, Calif., Hospitals and Clinics versus Lehigh Valley Hospital and Health Network’s experiences with CPOE screens in the same vendor's system--a comparison. (IDX LastWord)
13. Cambridge Memorial Hospital, Cambridge, Ontario, Canada (Meditech Provider Order Entry)
14. Overlake Hospital Medical Center, Bellevue, Wash. (Meditech Provider Order Entry)

Back to the top of this file.  Back to last location

Case Study 1: Implementation of Siemens CPOE Succeeded House-Wide...Except In Medical ICU, Where the Fixes Aren't Fully Satisfying Docs

Ohio State University Medical Center
1375 Perry Street
Columbus OH 43210
Phone: (614) 293-8000
Size: 923 total beds
CEO or Administrator: R. Reed Fraley
Chief MIS Executive: Asif Ahmad, CIO
Medical Director: Hagop Mekhjian, M.D.
CPOE System: Siemens Invision
Sources of information: Asif Ahmad, CIO

1. Tailor CPOE screens and work flow to each department's needs.

The #1 reason that CPOE projects fail has nothing to do with shortcomings in the technology or the medical staff. According to Asif Ahmad, CIO at Ohio State University Medical Center, winner of the 2001 Davies Award for outstanding computer-based patient record implementation, "It is the CIO that is the problem." Most CIOs, analysts and other IT employees do not understand clinical processes well enough to design a clinically workable implementation. When problems do arise, they chalk them up to physician resistance, rather than recognizing that they have a real issue on their hands.

Mr. Ahmad cites a problem in his own implementation: the CPOE needs of the medical ICU and the surgical ICU turned out to be so different that it was necessary to take the medical ICU down for six months and redesign the implementation (see below). If he hadn’t understood the problem, "I would have had a failure." To help understand, 40% of the people in Mr. Ahmad's 170-person IT shop have clinical backgrounds. Mr. Ahmad's background is in biomedical engineering. He also has an MBA.

2. Which physicians to involve in managing an implementation.

Physician support is a different story, he said. The physicians in the project don't need to be technical at all. They must be open-minded, analytical people, who can discuss workflow and care paths in a global way. "These are often some of the best physicians." Often hospitals box themselves into a corner with physician teams that are not sufficiently broadly based. "Too many surgeons and internists, not enough ancillaries," he said. OSU went live March, 2000 on a pilot in the transplant unit. It went live in all units in a "big bang" in May, 2000. The beauty of a so-called “big bang” is that it saves the cost of running double systems and the adoption rate is higher because, from a physician point of view, there is no switching between systems.

Implementation was preceded by two years of awareness training. Mr. Ahmad estimates that he personally participated at least monthly. Physician training was mandatory for the medical staff, and was a condition of medical center privileges.

Mr. Ahmad discounts the lore suggesting that CPOE implementation is relatively easy for academic medical centers because they have more power over their staffs than community hospitals. OSU's largest-volume service by DRG, women and infant care, is wholly dependent on community Ob/Gyns, he said.

3. Clinical alerts: should you? How and how many?

OSU's system is Siemens character-based Invision. As an older system, it is more stable than some newer-generation products. However, it is also based on a financial system, which makes it less flexible, he said. OSU has implemented the Siemens Healthcare Dashboard, which fires alerts for abnormal results. The OSU implementation "triages" the alerts, so a physician can tell when an out-of-range result or other event is truly urgent.

As the recipient of last year's award, Mr. Ahmad sat on this year's Davies award committee. His opinion: so far, there aren’t yet any new systems that fill the bill. "I don't see huge success stories out there," he said. "I see huge promise."

4. Medical ICU -- the exception to a successful house-wide implementation.

Buried in all of the well-publicized successes of the Ohio State University Hospital's implementation of Siemens Invision CPOE was one significant initial failure: the medical ICU.

The system went up house-wide in June 2000, following pilot testing on a transplant unit. It came down in the medical ICU one month later, after medical staff complained that it interfered with patient care so badly that it was creating a potentially dangerous situation.

This is the system that worked well enough to earn OSU a Nicholas E. Davies award, the most prestigious award in healthcare information technology. How could something work so well elsewhere throughout a 586-bed hospital, and fail so miserably in a 37-bed ICU?

Inadequate training was one problem, said Steve Hoffman, M.D., a pulmonologist who prefaces every conversation on this topic by noting that he does not consider himself computer-smart (but who was obviously up to his neck in this project). Insufficient hardware was another.

However, the biggest problem was one that many organizations could slam into: the workflow in University Hospital's medical ICU was fundamentally different from the workflow in most other inpatient areas. In most areas of the hospital, physicians evaluate one patient, then sit down and enter as many as 10-12 orders for that patient, then go on to the next patient. However, in University Hospital's MICU, care is far more random, and is more likely to be based on multiple, disjointed orders, rather than order sets given together.

Patient care was being delayed because it was taking too long to get orders into the system. Dr. Hoffman and others hoped that administration would relent, and decide that they did not have to use the system after all. Instead, they were given six months and considerable support from the information technology department to fix the problems.

Here are some steps that helped:

• They created order "pick lists." For example, a physician can choose "antibiotics with renal failure" and call up a list of appropriate orders.
• They created protocols. Rather than write a large series of orders, a physician now writes a single order to place a patient on a protocol–for example, a ventilator-weaning protocol.
• They created a direct route to patient orders of a particular type. Physicians can now call up a patient in the system, see all current ventilator orders, and modify them.

The re-implementation came off on schedule and with minimal difficulties, he said. Vastly more data entry hardware was crammed into the ICU, and 24 nurses were trained as "super-users," compared with the three who had been trained six months earlier.

However, even these fixes did not completely solve the problem of time-consuming order entry, he said. Residents and interns were still hand-writing orders, and then transcribing them into the system as of mid-2002. They still spent more time with the computer, and less with the patient.

Second problem: changing orders with transfers in or out of the ICU

A second significant problem was found: transfers in and out of the ICU were not handled well within the system. Treatments change when a patient transfers from one unit to another. That means that some nursing orders become inappropriate or irrelevant because the patient's physical surroundings have changed or because the medical interventions they support have been changed or stopped.

On a paper-based system, the orders are taken off the clipboard, and that's the end of that. However, in an electronic system, open order go into a queue and sit there until someone cancels them. The system does include commands for canceling orders upon transfer, but the sequence of steps was complex, and in the early days, interns and residents rarely got it right. As a result, nursing was beset with open orders that did not make sense and could not be carried out. That problem was still not entirely solved as of mid-2002, he said.

His attitude toward the system is best described as lukewarm. He said he does believe that CPOE can work in a medically complex environment, such as an ICU. There is even some evidence that it speeds up patient treatment in the ICU:

• Turnaround time on radiology orders appears to have been cut from about 7.5 hours to about 4.5 hours. (That figure is based on a comparison of the electronic system with a retrospective review of only about 20-25 paper charts on which the time or the order was noted, but even the small sample size may point to a valid conclusion.)
• Turnaround time on lab orders was reduced from a 31-minute wait from time of order to time of draw, to 24 minutes elapsed. That number is based on a comparison of 683 handwritten lab orders generated by the medical ICU after the system was taken down in the Medical ICU, and 1,100 electronic orders generated in the same time period by the Surgical ICU, which had kept it running.
• Turnaround time on medication orders was reduced from five hours to two hours, based on a comparison of 46 paper records from prior to the implementation and 70 after. (NOTE: this study was conducted on the transplant unit by the nurse who championed the project.)

These studies were not done prove the system's value, but to attempt to dispel provider concerns that it was slowing treatment down, he said. He also said he can tell empirically that turnaround times on X-ray and echocardiograms is vastly improved under the new system. However, "We have no idea why." As to improvements on laboratory and medication order turnaround times, he was guarded, saying only, "We clearly confirmed" that "things did not slow down."

"I think we can say for sure that the system did not add to the time," he concludes. If there was an improvement, it came "with a good deal of angst."

Back to the top of this file.  Back to last location

Case Study 2: CIO of 9-Hospital Chain Expects Alert Capability To Make CPOE Attractive To Community Physicians

Fairview Health Services
2450 Riverside Avenue
Minneapolis MN 55454
(612) 672-6300
Size: eight hospitals
CEO or Administrator: David Page
Chief MIS Executive: Gary Strong, CIO
CPOE System: Eclipsys Sunrise
Sources of information for this report:Gary Strong, CIO, (612) 627-6339

Gary Strong, CIO of nine-hospital Fairview Hospital & Healthcare Services in Minneapolis, Minn., is leery of turning on physician order entry for 1,700-bed Fairview-University Medical Center, because he believes that he knows what is coming: a thousand residents will want full use of every speck of functionality that Eclipsys' Sunrise Physician Order entry offers, and want it now. "These people are computer-literate," he said. Also, they are accustomed to the bugs, annoyances, and freezes of the Windows operating system. For them, "We have to be so ready."

At Fairview's community hospitals, he expects a different response. Community physicians tend to have a greater depth of medical knowledge than hospital interns, and they tend to be less tolerant of the headaches computer system dish out. They will have to expect to spend two minutes electronically entering orders that might take one minute to hand-write, Mr. Strong said. That presents a bigger challenge.

But their appetites are being whetted by an interesting source. Fairview also operates a large ambulatory care practice, which is rolling out Epic Systems' EMR. Physicians who have used Epic on the outpatient side are likely to be more receptive to Sunrise on the inpatient side because they are already familiar with the benefits of an order entry system.

Mr. Strong has given sales pitches for the system at numerous physician gatherings. He figures he's made at least 100 presentations and recruited a core group of 125 early adopters. He pitches CPOE on two grounds:

• that it will make patients safer; and
• that it will make care more efficient.

He has recruited early users by specialty area, focusing those specialties which, by their nature, have the greatest potential for disastrous medical errors.

The early adopters are "anxious to get on," he said. They seem to agree that the extra time they will spend on order entry will be compensated in the long run. "But we haven't hit the masses yet. That's where the walls will go up."

He mentions two factors that will help tear the walls down:

• First, and foremost, is the systems' ability to generate on-line alerts. For example, Fairview expects to set up its Eclipsys system to alert physicians if a piece of equipment a test order requires is out of order.
• Second, Fairview is rolling out first by giving physicians access to electronic results, then taking away paper results. Forcing physicians to look up results on line may help encourage them to enter orders electronically.

The estimated cost of the project is $27 million, including hardware, software, installation, and training across four organizations. The anticipated return on that investment is 27% per year. "We are talking about very direct savings," he said.

Back to the top of this file.  Back to last location

Case Study 3: A Pilot-Site Approach: IT Department Takes A Neutral Position , Lets Physicians Decide

Holland Community Hospital
602 Michigan Avenue
Holland, Mich 49423
(616) 392-5141
Size: 230 beds
CEO or Administrator: Judith Newham
Chief MIS Executive: Chuck Kohlruss, CIO
Chief of Staff: Roger Phillips, M.D.
CPOE System: Per-Se Patient1 Computer-based Patient Record System
Sources of information for this report: Bill Van Doornik, MIS Director, (616) 394-3449

If CPOE were generally faster than handwritten orders, implementing it at 230-bed Holland Community Hospital, Holland, Mich., might be a no-brainer. But MIS Director Bill Van Doornik has concluded that it will take physicians an extra 30 minutes per day to enter their orders electronically through the Per-Se physician order entry system--or, for that matter, any electronic order entry system. And, unlike teaching hospitals, Holland cannot compel physicians to do much of anything. "I don't have the tools to ram this down anybody's throat," Mr. Van Doornik said.

With one failed beta project under its belt, Holland's IT staff is avoiding coming out a "loser" if its medical staff ultimately rejects the system by taking a fairly neutral position on whether it should be adopted. IT will provide "the best system we can," he said. Whether or not it is rolled out will be entirely up to the discretion of executive management and the medical staff. "Hopefully, that is an empowering message," he said.

In a departure from the philosophy followed by many organizations, Holland deliberately recruited physicians who are closely aligned with hospital management to test the system. They are likeliest to speak directly about what they think of the system, and to be forgiving if does not work perfectly right out of the box, Mr. Van Doornik said.

Designing Screen Flows to Meet Physician Needs: Branching Structure for Order Sets

So far, the approach seems to be working. Holland Community beta-tested an earlier version of the Per-Se physician order entry system, found the screen flow cumbersome, and sent it back for a rewrite, he said. Holland is working with Per-Se on a branching structure for physician order sets, so that a basic order template will customizable for multiple orders, Mr. Van Doornik said. The project is still alive and Holland plans a second beta test.

Back to the top of this file.  Back to last location

Case Study 4: Best-of-Breed Approach: How Do You Integrate Clinical Decision Support?

Cottage Health System
Pueblo at Bath Street
Santa Barbara CA 93105
(805) 682-7111
Size: 3 hospitals. Flagship: Cottage Hospital, 445 beds
CEO or Administrator: Ronald C. Werft, CEO
Chief MIS Executive: Alberto Kywi, CIO
Medical Director: Robert A. Reid, M.D.
CPOE System: Mixed, with Eclipsys Sunrise Clinical Manager (SCM) as the core CPOE, interfacing with IDX radiology, McKesson pharmacy, Misys laboratory, and others.
Sources of information for this report: David Classen, M.D., Vice President, First Consulting Group; Alberto Kywi, CIO, Cottage Hospital.

When a best-of-breed shop implements computerized physician order entry with on-line decision support, how does it integrate clinical decision support?

For example, suppose that the physician enters an order, deals with a series of decision- support questions, and then passes the order to the pharmacist on line. Using an entirely different system, the pharmacist is confronted with a different series of decision-support questions, and presumably no history of how the physician acted on them. The pharmacist then releases the order to the nurse, who is confronted with yet a third set of decision-support questions.

This concern has been raised by, among others, David C. Classen, M.D., who discussed it in a presentation at the June, 2002 Healthcare Information and Management Systems Society (HIMSS) meeting in Las Vegas. Although orders can be integrated among disparate systems, Dr. Classen said he has not found any examples of integrated alerts, saying, "Technically, it is much more difficult."

Dr. Classen, a vice president with First Consulting Group, is a key architect of the CPOE testing protocol of the Leapfrog Group, and is an associate professor of medicine at the University of Utah, giving him both credentials and clout on CPOE. Dr. Classen implied in his HIMSS presentation, that vendors are putting significant pressure on First Consulting Group to remains silent on the issue of the difficulty of integrating clinical alerts in a best-of-breed environment. "Despite many pleas to the contrary," FCG is discussing it, he said.

We raised the question with Alberto Kywi, CIO for Santa Barbara Cottage Health System, which uses IDX radiology and McKesson pharmacy, Misys laboratory, and will launch a pilot for Eclipsys Sunrise CPOE in December. His answer: Eclipsys Sunrise will become Cottage Hospital's definitive decision support system.

Mr. Kywi said that pharmacy decision support is being built into Eclipsys Sunrise 5.3, and at this writing, it was due out at the end of 2002. He believes that eventually, the functionality in Sunrise will allow Cottage to use it in place of the McKesson pharmacy system.

Cottage will also implement the McKesson robotics system, which includes the Robot-Rx distribution system, the AcuDose-Rx cabinets, and the AcuScan-Rx Dose medication administration record. The plan is for orders to flow from Sunrise to pharmacy to the robot to AcuScan, which feeds back to Sunrise Clinical Manager.

Implementation to start in post-surgical unit

Santa Barbara Cottage Health System expects to begin piloting its Sunrise Clinical Manager with CPOE in a post-surgical unit by the end of this year.

Why post-surgery? CIO Kywi explains: SBCH surgeons like to try new things. Post-surgical patients usually don't go anywhere from the unit except home, so a patient's record will not be as fragmented between electronic and non-electronic systems as it might otherwise be. Finally, surgeons use a significant number of standing orders. From a physician’s point of view, the workload tends to be lighter if he/she can reuse orders.

For Cottage, CPOE represents the next logical step in a long-term information technology plan approved by the Board of Directors in 1996. The plan, which includes 16 initiatives, was developed by Mr. Kywi the year after he became CIO. He avoided using consultants in order to build credibility with management and the board. In earlier phases, Cottage upgraded its network, replaced an outdated McKesson Series patient financial management system with Eclipsys Financial Management (formerly SDK), and installed a successful picture archiving and communications system (PACS.)

The Cottage Board is driven by a need to both improve care and cut costs. (Because it's in a California earthquake zone, the health system is faced with seismic retrofitting requirements so large that it expects to entirely rebuild its flagship hospital, Santa Barbara Cottage). The board is buoyed by the success of its other IT projects, particular its DR Systems PACS. It has committed to fully funding the project, he said. Cottage is a teaching hospital, but most physicians are not employed by the health system.

Following the pilot, all three hospitals will go up on Sunrise results, nursing documentation, and CPOE for laboratory, medications, and radiology, he said.

Additional reasons for moving to CPOE:

• The HIPAA Data Security and Privacy rules. An order is more secure if it flows from the physician directly to the ancillary department than if it is handwritten and passes through other hands before reaching its destination.
• California law. A new law requires hospitals to adopt error reduction plans which make use of technology. CPOE is one component of the SBCH plan.
• Convenience. If the MD writes an order, he or she must turn it in to the nursing station and hope it gets entered. With Sunrise, physicians will be able to enter orders and monitor results from anywhere.
• Speed. When a handwritten order is turned in, there can be a significant delay before it is relayed to its destination. All physicians round at about the same time, and unit clerks are sometimes inundated. This can increase length of stay, and is a sore spot with physicians. If they can enter their own orders on their own timetables, the delay can be eliminated.
• Chart availability. With Sunrise, multiple users can look at the chart at the same time.

Mr. Kywi characterized Eclipsys Corp. as a firm that is "struggling" with issues of integration, of both its products and its business units. On the whole, they are succeeding, he said. "They are a very good company to work with," he said.

Back to the top of this file.  Back to last location

Case Study 5: IKnowMed Cancer Treatment Support CPOE System Being Piloted

University of Texas-M.D. Anderson Cancer Center
1515 Holcombe Boulevard
Houston TX 77030
(713) 792-2121
Size: 518 total beds
CEO or Administrator: John Mendelsohn, M.D
Chief MIS Executive: Richard Pollock, CIO
Medical Director: David Calendar, M.D.
CPOE System: IKnowMed and Per-Se
Sources of information for this report: Richard Pollock, CIO

M.D. Anderson Cancer Center, Houston, Tex., went live in the fall of 2002 on a pilot of IKnowMed EMR, a new product designed to support cancer treatment. A month into it, Anderson had a five-physician IKnowMed CPOE project at a breast clinic; nurses also use the product to document chemotherapy administration at an outpatient center.

"We're very pleased," said Richard Pollack, CIO. "It's been well-received." Anderson plans to roll IKnowMed out across its other outpatient clinics and to integrate it with its Per-Se Patient1 inpatient data repository. IKnowMed will be used for inpatient chemotherapy orders and for structured physician documentation; all other ordering will be through Per-Se, he said.

What's particularly interesting here:

In a world of tight budgets and cutthroat competition, Per-Se and IKnowMed are stepping up to the plate with a level of systems integration that many vendors cannot achieve even within their own organizations. Both vendors are rewriting code so their systems will parse each other. Instead of interfacing the EMR components on select data elements, Anderson expects to end up with a system that allows any data element to be pulled into either system, he said. The effort will be commercialized and available to other sites, he said.

M.D. Anderson was a very early Cerner Millennium site at a time when Cerner was having a hard time meeting its vision. That project failed three or four years ago (IHC,10/12/98; 5/15/00). The staff has bounced back and is "more than ready" to get back into the EMR implementation business, Mr. Pollack said. The experience defies popular wisdom that it takes many years for organizations to recover from a failed implementation. Mr. Pollack noted that the project predates his arrival, so possibly new blood at the top was part of the solution.

Working with Per-Se, Anderson has embarked on a project to unlock some 25 years of data in generations of legacy laboratory information system databases. The project involves exporting the original data in flat files, building screens in Per-Se to reflect the original presentation, and having Per-Se map the legacy repository. Anderson will eventually replace the legacy repository with Per-Se's Patient1, an M-based (MUMPS) clinical data repository, and its Decision1 Oracle-based research database. Anderson started work in the spring of 2002 and expects to finish by late 2002 or early 2003. The pace depends on the speed of the legacy repository and on Anderson's decision to clean data as it goes.

Why choose Per-Se? Mr. Pollack offered three reasons. First, it runs on MUMPS, allowing it to handle large volumes of data with an efficiency and reliability other database management systems can't match. Second, the Per-Se system is extremely flexible, which meant M.D. Anderson could replicate its legacy data presentations by creating hundreds, even thousands, of different screens in Per-Se. A less flexible system would have meant "tremendous compromises." Finally, Anderson happened to have several people who were extremely experienced in Per-Se systems on its staff.

Mr. Pollack: (713) 745-7960.

Back to the top of this file.  Back to last location

Case Study 6: Community Hospital Acclimates Physicians to CPOE With Gentle Persuasion, One-On-One Training, Customization of System Capabilities

Lehigh Valley Hospital and Health Network
c/o Lehigh Valley Hospital
Cedar Crest & I-78
Allentown PA 18102
(610) 402-8000
Size: 3 hospitals, 3 health centers; Lehigh Valley Hospital, 644 beds
CEO or Administrator: Elliott J. Sussman, M.D.
Chief MIS Executive: Harry Lukens
Medical Director: John Longenhagen, M.D.
CPOE System: IDX LastWord
Sources of information for this report: Don Levick, M.D., Pediatrician and physician liaison

Just how do you persuade community-based, attending physicians to enter their own orders?

Lehigh Valley Hospital and Health Network, Allentown, Pa., set an expectation that all physicians would be trained on the new system, with the understanding that no one would be forced to use it. Unit clerks continue to be available for physicians who are unwilling to enter their own orders. However, physicians are under constant, gentle pressure to get with the program. When they do, they can count on a lot of support.

Pediatrician Don Levick, MD, physician liaison to the Lehigh Valley CPOE project, explains how this works:

Suppose you have a physician who just completed training, and it's 7 p.m., and he's writing orders. A trainer approaches him and says, "Wouldn't you like to enter them electronically?" He says, "No. I want to go home." So the trainer says, "OK – we'll do it in the morning."

So far, it is working pretty well at this 644-bed community hospital. A trauma step-down unit went live on IDX LastWord in late 2001; three medical surgical units have gone live beginning in February, 2002. As of this report, about half the physicians are entering their own orders. About 60% are entered by community- based attending physicians; 40% are by residents.

Other factors that help:

• Lehigh Valley has been a LastWord site since 1993. Physicians are already accustomed to looking up results in the system. Labs and transcriptions are printed out and placed on the chart, but it is often easier to use the ubiquitous on-line record. Also, electronic results are available first. Medication administration records and vital signs are also available on line.
• Perhaps partly because of its extensive experience with LastWord, Lehigh Valley's IT department is able to respond rapidly and effectively to physician requests. For example: a physician wanted an order change for supplemental oxygen to include a titration order and the ability to choose a target level to wean the patient to. The department had the change ready within a few days. (Implementation took longer because the change had to be cleared with affected ancillary departments, and support staff and educators had to be notified, so they could communicate the change to the medical staff.)
• Medical staff leadership paved the way for the CPOE project with multiple meetings, presentations, and demonstrations. Physicians on the design team met twice per month, from 6:30-8 a.m. Coffee, muffins and bagels were always available. Free bagels and donuts were also available to nurses and physicians in the kitchen area of the unit as units went live. Project supporters received recognition at meetings.
• The project is well-staffed. Since the inception of the CPOE system in October, 2000, Dr. Levick has devoted about half of his time to it. A senior information systems manager spends most of his time on the project, as do four analysts and three educators. The team provided support 24 hours/7 days for several weeks after each unit went live, as well as ongoing support.
• All training is one-on-one. It takes about two hours.
• The system itself, IDX LastWord, seems to work well. The word in "hallway conversations" is that proficient physicians can enter orders in LastWord in slightly less time than it takes to hand-write orders.
• Data entry is by desktop computers, wireless laptops on carts, or wireless sub-notebook computers that are carried by some physicians. Lehigh has also made LastWord available on Citrix, meaning that it can be used with a thin-client terminal.

Back to the top of this file.  Back to last location

Case Study 7: Hospitals Lacking Competitors May Find Installing CPOE Easier

Great Plains Regional Medical Center
601 West Leota Street
North Platte NB 69101
(308) 534-9310
Size: 116 beds
CEO or Administrator: Lucinda Bradley
Chief MIS Executive: Kevin Olson
Medical Director: Janet Bernard, M.D.
CPOE System: QuadraMed Affinity (CPOE development site)
Sources of information for this report: Robert Buckland, M.D., (308) 534-5370; Kim Dyer, COO

There's "no question" that Great Plains Regional Medical Center will get good cooperation from physicians on its computerized physician order entry project, said Robert Buckland, MD. How can the physician chair of the information system committee for a 116-bed hospital in the middle of nowhere make an assertion like that?

"Middle of nowhere" may be precisely the point. Big-city physicians may feel no ties to the hospitals they practice at, and be less willing to extend themselves for the good of the organization. But the 60 or so physicians who practice in the North Platte area understand that their fates are wrapped up with the future of Great Plains Regional.

Some 20 physicians turned out for a meeting at which QuadraMed Corp., the vendor, showed a prototype of the system. Half a dozen have agreed to essentially donate the time it will take to get the prototype ready for a beta test; 12-15 are likely to participate in the six-month test. "They're excited," he said.

Geography is facilitating the project in another way as well. With such a small medical community, the overwhelming majority of physicians' offices are a stone's throw from the hospital, and are already tied to it by fiber optic cables. Adding order entry and results retrieval from offices should be relatively straightforward. Viewing is through QuadraMed's browser-based results viewer, Clinician Access.

But the main reason he expects the project to succeed is his belief that QuadraMed is poised to do things right functionally. "If you've got a good system, getting cooperation is easy," he said.

A longtime Affinity client, Great Plains is the development site for QuadraMed Affinity physician order entry. Applications are being built on the Affinity Cache database; a change to Oracle at a later date is likely, he said.

Great Plains is installing medication order entry, laboratory, and radiology, and it won't roll out until the entire system is built, said Kim Dyer, VP, operations. The reason: part of the problem with offering computers to physicians is that they are often also compelled to consult paper medical records as well. "Until we have it all electronic, they are not going to be all that interested in using it."

Great Plains is also installing Antibiotic Assistant, a decision support system from Theradoc, Salt Lake City, Utah. The system recommends appropriate antibiotics on the basis of lab values and gives underlying literature to support its recommendation. QuadraMed could be set up to provide the same decision support, but the purchase saves Great Plains the work of setting it up. Ms. Dyer said she was "leery" of licensing software "telling physicians how to do clinical practice," but found that they see it as "a great tool."

Back to the top of this file.  Back to last location

Case Study 8: Early CPOE Sites Don't Necessarily Have Electronic Pharmacy, Medication Systems In Place

Organization: several, based on literature review and survey by the head of a HIMSS committee.
CPOE System: various.
Sources of information for this report: Joe Miller, chairman of a HIMSS Special Interest Group for the Computer-Based Patient Record and manager of IS for Christiana Care Health System.

Few organizations had pharmacy systems working when they started their computerized physician order entry projects. Also, the most common decision support rules among early CPOE sites had more to do with efficiency and data compliance than they do with prevention of adverse drug events. For example, alerts on incomplete orders, or incomplete data are fairly common.

This information comes from an early 2002 review of 10 CPOE projects that have passed the pilot stage. The results are surprising because reduction in errors, particularly medication errors, is one of the most frequently cited reason for implementing physician order entry. However, most decision support to reduce medication errors can work only if the organization also has the patient's medication available electronically.

Organizations thinking they need a full electronic medical record in place prior to implementing CPOE "may be setting the bar to high," said Joe Miller, chairman of the Healthcare Management and Information Systems Society's special Interest Group for the Computer-Based Patient Record and manager of information services for Christiana Care Health System.

The organizations he studied were: Sarasota Memorial Hospital, Sarasota Fla.; Montefiore Medical Center, Bronx, N.Y.; Trinity Health, Farmington Hills, Mich.; Maimonides Medical Center, Brooklyn, N.Y.; Ohio State University Medical Center, Columbus, Ohio; Boston Medical Center, Boston, Mass.; Abington Memorial Hospital, Abington, Pa.; Rush Presbyterian-St. Luke's Medical Center, Chicago, Ill.; University of Illinois at Chicago Medical Center; and Alfred I. Dupont Hospital, Wilmington, Del.

Four are academic medical centers, five are teaching hospitals, and one is a community hospital. In a world in which many CPOE projects never make it past the pilot stage, this list includes some roaring success stories. All are past the pilot stage and more than half claim 100% CPOE, he said. Sarasota, the community hospital, has about 25% physician order entry. He based his conclusions on literature about the projects, and on his own discussions and surveys. Not all of his questions were answered by all organizations, he said.

Organizations implemented CPOE for a variety of reasons. Some were worried about medical errors. Others were responding to competitive pressures. Still others saw is as part of a broader re-invention effort that accompanied a merger. All their projects pre-dated Leapfrog.

Successful organizations varied in their basic approach to CPOE. Some scrapped existing systems entirely in favor of implementing totally new, integrated system. Others upgraded what they had. Boston Medical Center went straight from 100% paper-based orders to CPOE without ever having electronic orders entered by nursing or unit clerks. Abington Memorial achieved 40% CPOE, waited a while, then made a strong push for 100% compliance.

Here are other details he learned:

• Several organizations warned that the physicians who are most attracted to a CPOE project sometimes lack clinical influence. It is crucial to involve physicians who are respected clinical leaders among their peers, he said. Leaders of successful projects "selected very carefully who was involved," he said.
• Virtually all of these organizations offer physicians order sets, which are groups of orders that can reduce the amount of time it takes the physician to enter his or her own orders. Academic institutions tend to do departmental order sets, he said. Teaching hospitals and community hospitals often allow personal order sets as well.
• Classroom training was common among early adopters. (This flies in the face of advice that says "physicians won't show up for classroom instruction.") Only one institution used one-on-one training, and one used computer-based training. Most offered around-the- clock support for two to four weeks following go-live.
• Several organizations reported that additional training sessions were necessary weeks or months after go-live. The reason: nurses were back to entering orders because physicians were demanding it.
• User access to input devices was a big concern. Many started out with one PC for every three to four beds on medical surgical units, and ended up with two beds per PC.
• Few had wireless systems when they started their projects, but most have wireless today. Wireless is a great convenience in CPOE because it allows an organization to easily expand the number of data entry units in the early stage, when users tend to be slow with their orders, he said. Later, as people speed up, the devices can be transferred to the next new unit.
• Many did studies highlighting "early wins" to their organizations. Several focused on reductions in time elapsed between the time the order is written and the time the order is filled.
• Virtually all said that top-level commitment and strong clinical leadership were essential to project success. They "did not locate the authority for the project in information technology."
• Many found it useful to create a simple, direct vision statement of 10-20 words which summarized the goals of their electronic medical records projects. Vision motivates people to behave in ways that are not necessarily in their own interest, and it helps coordinate the actions of hundreds of individuals who are involved in a project, he said. Mr. Miller, (302) 324-3523.

Back to the top of this file.  Back to last location

Case Study 9: Mount Carmel St. Ann's Didn't Buy Its Emergency Department CPOE System To Save Money--But Is Saving A Bundle

Emergency Physician Acceptance Is High;
However, Some Design Aspects Annoy Other Hospital Physicians
Vendor is addressing that problem

Mount Carmel East, West Health & St. Ann's
739 West State Street
Columbus OH 43222
(614) 234-5000
Size: 989 total beds, including:
St. Ann's Hospital of Columbus
500 South Cleveland Avenue
Westerville OH 43081
(614) 898-4000
Size: 175 total beds
CEO or Administrator: Joseph Calvaruso, Mount Carmel CEO; Kirk Hummer, St. Ann's Administrator
Chief MIS Executive: Cindy Sheets
Chief of Staff: Gerald Girardi, M.D.
CPOE System: A4 HealthMatics Emergency Department System
Sources of information for this report: Frank Orth, D.O., president of the physician group which staffs the emergency room.

Mount Carmel St. Ann's of Columbus, Ohio, anticipates that it will take less than two years to pay back the cost of its A4 HealthMatics emergency department system, said Frank Orth, D.O., president of the physician group which staffs the emergency room, in an interview with the Inside Healthcare Computing staff for this report. Here are the returns on the investment:

• Reduced transcription costs. It formerly cost $450,000/year to transcribe reports that originated in emergency. Now, 95% of that cost is saved because physicians document on line using Fujitsu hand-held notebooks.
• Reduced insurance denials. Mount Carmel set its system up to force physicians to make coding choices that carry Medicare reimbursement. Test denials, which previously cost tens of thousands of dollars per month, have dropped to zero, he said. Dr. Orth said he guesses that the savings on denied tests will at least equal the savings in transcription costs.
• Improved charge capture. As with reduced insurance denials, dropped charges for nursing supplies have become less common.

Here's what's so interesting: St. Ann's implemented an emergency room system as a way to gather data to support performance improvement, not to save money, said Dr. Orth, who is also physician facilitator for performance improvement for the emergency department.

A4's HealthMatics ED went live at St. Ann's in mid-2001. St. Ann's chose the system first for its reporting capabilities and second for its flexibility. "We can customize it on site without spending a lot of money," he said.

St. Ann's deployment includes all major modules: nurse triage, nursing documentation, patient tracking, physician documentation, order entry, and results reporting.

Screen design/flow issues did cause some pain

Installation was "painful" because of unpredictable issues with interface design and workflow. The only real remedy would have been a full-blown test environment, with the ability to mock up all interfacing ahead of time. That would have been prohibitively expensive, he said.

The system will benefit by some tweaking, but all modules are in production and working well, and the vendor has been remarkably good at taking suggestions from the medical staff and incorporating them into upgrades, Dr. Orth said. Allergy checking, order screening for completeness and logic, and certain changes to the way the chart is presented will be available in the future. Overall, support has been "very good," he said.

Physician acceptance within St. Ann's busy (40 beds, 65,000 annual visits) emergency department has not been a huge obstacle, principally because of the group's culture. All 19 ED physicians enter their own orders, along with 8-10 physician assistants. Physicians groups of 3-4 were trained on order entry. Getting ER physicians to attend the sessions was not terribly difficult, he said, but "I'd hate to try" to persuade a general medical staff, he said.

Physicians place lab, x-ray and pharmacy orders electronically. A year after installation, physician order entry is the most popular component of the system with emergency physicians, nurses and secretaries. The reason: orders are no longer lost or delayed before being entered in the system, and secretaries no longer track down physicians with questions about handwriting ambiguities or disallowed orders.

When a physician places a lab order, it appears immediately on the ED tracking board, where it notifies nursing that a blood draw is needed, and a label prints out of the printer. Sometimes the nurse is there with the needle before the physician has left the patient's bedside, Dr. Orth said.

It is almost impossible for structured data templates to capture the nuances of every possible patient scenario. To overcome this limitation, St. Ann's allows dictation of parts of the history and physician if the case just won't lend itself to the system. Physicians use this feature in about five percent of cases. Over-use is discouraged by the six-hour or so delay before reports are transcribed and appear back in the system.

St. Ann's also uses HealthMatics to force selection of appropriate dosages. Amounts are set by the medical staff.

One argument which helps ease acceptance of order entry: when reluctant physicians complain about becoming "clerks," Dr. Orth reminds them that they are judged by the quality of the data going in to the system, and asks whether they wouldn't prefer to be in control of it themselves.

Also, St. Ann's was able to redesign order entry screens for the 12 most common diagnoses to "be at least as fast as dictation." The system administrator, a nurse who was trained by HealthMatics but has no IT background, did the work. St. Ann's emergency department uses departmental templates, but physicians are free to develop their own.

Problem: rest of the medical staff dislikes the long reports

HealthMatics is available system-wide through Mount Carmel's browser-based front end for dictated reports and certain other patient data.

If the system is so terrific and so cost-effective, then why isn't every hospital in the country rolling it out? One factor is the resistance of other hospital physicians. They complain that ED records, which formerly comprised a page and a half of dictation, now run eight pages or more. Their resistance has been "a little messy," Dr. Orth acknowledged. A4 has helped address the problem by re-organizing the data. A true fix, putting a summary report at the top of the chart, was to become available within a couple of months after the interview for this report.

Another hurdle for non-emergency physicians is that if an emergency patient is to be admitted, and an attending physician wants to order a medication or test for that patient, the order must be entered by an ED physician or other trained system user.

Back to the top of this file.  Back to last location

Case Study 10: One of The First CPOE Hospitals Is Retiring Old TDS (Eclipsys) System For IDX CareCast And Is Developing Clinical Pathways

St. Francis Hospital Health Care System
114 Woodland Street
Hartford CT 06105
(860) 714-4000
Size: 617 beds
CEO or Administrator: David D'Eramo, Ph.D.
Chief MIS Executive: Catherine Szenczy
Chief of Staff: Fred M. Gabrielle, Jr., M.D.
CPOE System: IDX CareCast
Sources of information for this report: Catherine Szenczy, senior vice president and CIO

An organization which is moving to its second generation of CPOE has chosen the relatively new IDX CareCast system to replace its old Eclipsys (formerly TDS) 7000. St. Francis Care, Hartford, Conn., talked to "a lot of vendors" and "a lot of clients" before settling on IDX.

One reason for the decision was that CareCast offered a platform and an architecture which the hospital concluded would provide fast response times and high reliability, said Catherine Szenczy, senior vice president and CIO. CareCast, introduced in February, 2002, runs on Tandem Nonstop Himalaya, just as its predecessor, IDX LastWord, does. The old Eclipsys 7000 runs on an IBM mainframe, which is also fast and highly reliable, she said. "Several-second response times" were simply out of the question, she said. Cerner Corp. was the other finalist.

Another reason for choosing CareCast: integration was crucial, primarily because of the importance of integrating clinical decision support with order entry.

St. Francis is serving as the development site for Care Paths, a clinical pathways system, in CareCast. The system St. Francis envisions will allow physicians to change the patient's clinical pathway and adjust the underlying order without having to leave the care path to go back into order entry, she said.

St. Francis will use Cap Gemini Ernst & Young's Rapid Design Methodology to develop the Care Paths, she said. The development plan is for the entire 150-member end-user design team to meet in three 3-day sessions to plan the entire system. At each session, they're being asked to make a new set of decisions, she said. The system will be prototyped based on these suggestions to be sure it is really what the users had in mind, she said.

The meetings, held by IDX, are spaced about six weeks apart. Cap Gemini's approach was first used by St. Francis two years ago with great success to plan its e-health strategy, she said. While getting people to a meeting for three consecutive days requires enormous planning, it is more effective and less difficult than coping with the constant interruption of a long series of meetings, she said. IDX and CGEY were to attend all sessions and provide a transcriptionist to document all decisions.

Roll-out is to three facilities: St. Francis Hospital Healthcare System, St. Francis Hospital, and the Medical Center Mount Sinai Campus.

Bristol Hospital, a 108-bed affiliate in nearby Bristol, Conn., with no history of CPOE, will also participate in the project, replacing its McKesson Series 2000 with IDX Carecast. Bristol will go up on order communications for nursing and unit clerks with electronic resulting. The idea, Ms. Szenczy said, is to ease physicians into CPOE through the availability of an electronic medical record. Bristol will participate side-by-side with St. Francis in the nine days of planning meetings, but there is no firm timetable for implementing CPOE at Bristol.

St. Francis is a teaching hospital, staffed by community physicians and 100 employed physicians, who train some 200 residents. It installed the original predecessor to the EC 7000 about 30 years ago. What made for a successful CPOE program 30 years ago:

• Extremely strong top-level support. The CEO and senior management made it clear that they were fully committed to physician order entry, and that if physicians intended to admit patients to St. Francis, they would have to go along with the program.
• The first thing senior administrative support did was make sure that senior medical management wholeheartedly endorsed the project. Whether they talked to administrative management, or medical management, physicians heard exactly the same message.
• System selection involved a very large group of physician who participated in the advisory committee, the selection process, and site visits. They willingly gave time to the project because they understood that if they did not participate, they might be stuck with a system they did not like.
• The caliber of St. Francis physicians made a big difference. They are highly motivated to partake in cutting-edge technology. "That is very relevant to our success."

Back to the top of this file.  Back to last location

Case Study 11: Step-By-Step 3.5-Year Rollout, 100% CPOE at Montefiore ...Clinical Systems Director Offers Many How-To Tips

Montefiore Medical Center
111 East 210th Street
Bronx, NY 10467
(718) 920-4321
Size: 1,060 total beds
CEO or Administrator: Spencer Foreman, M.D.
Chief MIS Executive: Jack Wolf
Medical Director: Brian Currie, M.D.
CPOE System: IDX LastWord
Sources of information for this report: Dorrie Napoleone, Director of Clinical Systems, (718) 405-4092. (This report is based on both an in-depth interview and Ms. Napoleone's responses to our survey; most other user reports in this study are based on one or the other.)

Teaching hospital Montefiore Medical Center just finished rolling out the IDX LastWord system with CPOE to the last of its nearly 1,100 beds in the fall of 2002.

The implementation, which began in March, 1999, has cost $25 million, including hardware, software, and out-of-pocket vendor implementation fees. That amount doesn't include the substantial amount of in-house effort.

The first unit was a general medical unit, chosen because there was, and still is, good physician support on that unit. It "went up and stayed up," said Dorrie Napoleone, Director of Clinical Systems.

We asked what she would have done differently. Her answer: "It took three years to roll out the system to serve 1,100 beds. I would have liked to have hired more staff so that we could have done it in two years. [But] [o]ur [core] strategy we would repeat again: unit by unit--being successful as you go."

System selection approach

Montefiore's management assembled a committee of 25 people to select a system. About half were physicians. This committee also included nursing, ancillary department staff members, and administrators. The physicians were from a diverse variety of services--surgery, pediatrics, internal medicine--who were well-respected by colleagues, and even a new graduate from a residency program who had just joined the staff. Management wanted the decision to carry weight after it was made.

The entire group went on "road trips" to learn about systems. They had marching orders to look for a system that would provide high quality, reduced cost, and high physician utilization of the system.

They saw five vendor systems in action: IDX LastWord at Michael Reese, TDS at New York University, and Cerner, HBOC (now McKesson), and the SMS (now Siemens) system at a time when these last three really have much.

The eventual choice was LastWord. This may or may not have been the entire group's basis for making a selection, but Ms. Napoleone says, "For hospitals approaching this project, I recommend that you purchase and install a system that has 100% up time. This is critical when you're trying to go electronic. You can’t afford down time at all."

(One of the benefits which IDX touts about LastWord and its successor, CareCast, is that they run on high-reliability Tandem mainframe hardware.)

After selection: talk it up to build support

After the selection of LastWord, two well-respected physicians were chosen to be champions to speak up for the system. They spoke at as many existing Montefiore staff and employee forums as possible--perhaps--25 times. They went on grand rounds and attended departmental meetings, giving updates.

Going live, getting physicians to do CPOE: remote access a plus

Montefiore forces physicians and nurses onto the system. "We take away the paper, so there's no choice for the physicians but to use and master the system. This rule went into effect: a physician could not use the system if he/she did not sign on for training, and couldn’t place an order without training. A benefit was provided at the same time: the physician received remote access and the ability to place orders remotely -- a big plus for the physician. "They really like remote access." They now give very few verbal orders.

Another important success factor, Ms. Napoleone said, is this: start at a location where you will have the greatest chance of success. Once you get 1-2 units under your belt, it is hard for a nay-sayer to claim that the system "never would succeed."

Also, it's very important to pay attention to the details as you go live. For example, the first nursing unit had a lot of dialysis patients. The implementers had to be closely engaged to review orders. "People don’t like to tell you what they have been doing wrong."

Building and implementing the system--more how-to details:

The in-house "build" and implementation team is about 13 FTEs. "We employ a couple of physicians for about 20% of their time..."Because they're still practicing medicine, they have more clout within the physician community. We don't want them to become the "IS doctors" and lose their standing in the practice of medicine. We supplement their salaries for the time they put in with IT, but they can continue to practice. We have five nurses, one pharmacist, and the remainder of staff go in the programmer/analyst bucket...We found that it is easier to teach clinical people about technology than the other way around."

The group included 1-2 people from nursing whose principal tasks were to do workflow analysis for nursing units. They would identify the uniqueness of the department's or unit's needs, and send the screens back to a couple of analysts to do screen tailoring.

Then, at go-live, they would take away paper from nurses. They gave support 24 hours a day, 7 days, for at least two weeks after each go-live. In the ICU, typically that support was for 3 weeks; it varies by number of beds and census.

The clinical CPOE system was a bigger change for nursing than for physicians because of all the products that were brought up for nurses at once. Each nurse had a four-hour class; physicians had 1.5 hours.

After going live, you are still implementing

Montefiore goes back 3-4 months after each unit is live and asks what users need and want. "You need to solicit user feedback." LastWord has a powerful screen builder tool, so they can make 95% of needed modifications without going back to the vendor.

Integration with pharmacy, nursing helps build support

Another success factor: integration. Montefiore brought up the LastWord pharmacy system before implementing CPOE.

Your best chance for success, Ms. Napoleone said, is to have order entry completely integrated with pharmacy and with medication administration so that the physician, the pharmacist, and the nurse are always on the same page when looking at a medication order. "I feel very strongly" that integration there is a key. Montefiore currently has an older Community Health Computing (CHC) lab system, which it's replacing with a Triple-G system. The radiology system is IDXRad.

In failed implementations she has studied, they didn't provide the clinical data feedback to physicians. They'd have to place orders and drop to the paper world for results.

Ratio of data entry devices to patient beds

Montefiore has about one device for every 2.5 beds in all units except the ICU, where it's one-to-one. "It's adequate, but I would love to have more devices," she said. "Part of the issue is what else you use the devices for. If it's only for CPOE, then the ratio is OK. But if the staff is doing e-mail and research on these same devices, then the ratio needs to be higher."

Decision support: another big benefit

"We use a combination of out-of-the-box rules checking for allergies, drug interactions, and duplicate checking," Ms. Napoleone said. "We write some rules beyond these just mentioned using the rules engine that comes with the IDX product."

She said that Montefiore has found "tremendous benefits" in reducing prescribing errors through the system's allergy alerts, drug interactions, warnings on incorrect routes, and dosing alerts--along with the fact that orders were perfectly legible.

"With each unit we brought up, we've seen more than a 50% reduction in medication errors. That's immediate payoff. We've also seen considerable reduction in turnaround time (approx. two hours less) from order placement to the patient receiving the medication or procedure. This time savings translates into short LOS because we're treating patients faster, and they're getting better quicker."

"Nobody wants to go back to paper," Ms. Napoleone said.

Back to the top of this file.  Back to last location

Case Study 12: Stanford Missteps And Screen Design Issues Led To Early Dissatisfaction; Bumpy CPOE Start; Lehigh Valley Faced Some Of Same Problems Differently

Stanford Hospitals and Clinics
Stanford CA 943045300 Pasteur Drive
(650) 723-4000
Size: 663 total beds
CEO or Administrator: Mike Petersen
Vice President of Information Systems: Russ Peckenpaugh
Medical Director: Michael Smith, M.D.
CPOE System: IDX LastWord Version 4
Sources of information for this report: Russ Peckenpaugh, Stanford's VP of information systems; also, comparative results of a LastWord CPOE implementation from Don Levick, M.D., and physician liaison at Lehigh Valley Hospital and Health Network. (For more on Lehigh Valley, see Case Study 6 above).

Two things went right for Stanford in the implementation of computerized physician order entry in its IDX LastWord system, according to a 2002 conference presentation by Russ Peckenpaugh, Stanford's VP of information systems:

1. LastWord has an integrated pharmacy information systems which he said was "absolutely key" to the project.
2. Data entry was through a wireless network. Physicians' use of low-end laptops on carts during rounds has worked well, he said.

The list of things Stanford would do differently was longer.

The good news: a year after the go-live, most physicians have accepted its benefits and are entering their own orders. However, the beginning was bumpy. In a commendable act of bravery, aimed at helping other organizations avoid repeating Stanford's mistakes, Mr. Peckenpaugh has gone public with the "do this differently" list:

• Stanford did not mandate training. The result was "residents on units who did not have a clue" what to do with the system, which was not intuitive.
• Stanford didn’t deal well with nursing expectations. Nursing believed that nursing entry of verbal orders would decline once physician order entry was live. In fact, physicians continued giving nurses verbal orders to enter at the usual rate. The problem was particularly acute in surgery, where physicians would announce their orders, then rush off to meet a surgery schedule.
• Stanford did not go live on nursing documentation prior to physician order entry. Without nursing documentation active, it is difficult to manage the nursing work list, and it limits the amount of is feedback physicians receive about the orders they have placed.
• Stanford rolled out the entire organization in approximately six months, at times bringing one unit per week live. "We could have easily spent twice that long" building order sets and stabilizing the units, he said.
• Stanford had difficulty balancing the tension between ancillary departments, which wanted as much order detail as possible and physicians who wanted to keep screens bare-bones. "We ended up with too much information on the screens and too many detail screens required."

There were also some IT-related bumps. For example, if an order is not added to the patient profile in the Pyxis drug dispensing software by the time the nurse needs to administer the medication, the nurse will perform what is called a manual override. The nurse then needs to go into Pyxis or IDX (the systems are interfaced) and manually assign the order to the physician. If the nurse chooses the wrong physician, the physician will rightfully refuse to sign the order. The order must then be researched and resolved to get it signed by the right physician. "This required a well thought-out plan," which was not in place initially.

LastWord screen flow was a big challenge for Stanford physicians. For example, Stanford's system did not include a screen a physician could consult to see all of the orders he had placed on a patient before finalizing them, he said. The problem has been addressed in CareCast, IDX's next-generation clinical system, he said, but Stanford runs on Version 4 of LastWord. "We are working with IDX to bring the functionality" to V4, he said.

Lehigh Valley Hospital and Health Network: Better Experience With CPOE on LastWord

Don Levick, M.D., physician liaison for the CPOE project at LastWord shop Lehigh Valley, had quite a different experience. The majority of Lehigh physicians find LastWord screen flow "intuitive and acceptable," he said. He described Lehigh's order entry process as follows:

First the physician chooses orders from multiple lists. The lists depend on how a system is configured: Lehigh's are divided into medication, laboratory, imaging, and nursing. Once orders are selected, they appear in an "unprocessed orders list." The physician then goes back in and fills in order detail. For example, he or she would give the reason for an order for a radiology study. Then the physician enters the order. At that point in the process, the system issues any decision-support flags.

"There is no final window" for reviewing all orders, but if the physician wants to see what he or she just did, there is another route for it. All active orders on a patient appear in an "order profile screen," which, when chronologically sorted, gives a view of all ordering activity. Lehigh Valley Hospital and Health Network physicians find that the system gives them "adequate control" over order entry flow and display, he said.

Back to the top of this file.  Back to last location

Case Study 13: Canadian Hospital Takes A Careful, Consensus Approach; Sees Reduction in Medical Errors As Principal Benefit of CPOE

Cambridge Memorial Hospital
700 Coronation Blvd.
Cambridge, Ontario, Canada
Phone: 519-621-2333
Size: 227 beds
CEO or Administrator: Helen A.I. Wright
Chief MIS Executive: Rena Burkholder, Chief Information Officer
Chief of Staff: Jim Snider, M.D.
CPOE System: Meditech Magic Patient Order Management
Sources of information for this report: Rena Burkholder, Chief Information Officer, (519) 621-2333, ext. 1395; Cheryl MacInnes, MLT, ART, MBA, Program Manager, Diagnostics and Therapeutics

CIO Rena Burkholder of Cambridge Memorial Hospital figures that she recently witnessed the end of an era: the last physician to refuse to use a computer to look up patient medical information has finally retired, five years after Cambridge began its electronic medical record project.

What worked best to encourage physicians to do computer lookups? "...bring them on when they are comfortable, and not fight," she said.

For Ms. Burkholder, the event held an important message that is relevant to Cambridge's CPOE project: left to their own devices, physicians will get there eventually. "What works best is bringing them on as they are comfortable, and not fighting them."

Cambridge plans to involve seven hospitalists in a two-year pilot project on Meditech Magic Patient Order Management (POM). The project is supported by an $89,000 CDN grant from the Ontario Hospital Association's Change Foundation. As in the United States, CPOE is widely seen in Canada as a way to reduce medication errors. Medication errors are believed to kill 500-700 Canadians annually. Placing decision support at the physician's fingertips is the principal allure of the project, Ms. Burkholder said.

Capital costs are $91,600, including $89,000 for software. The two-year operating expenses are $152,400, including the cost of evaluating the project. The system goes live in November. Initial rollout is pharmacy order entry only, with lab and radiology to follow.

Meditech Patient Order Management (POM) allows physicians to create their own custom order groups with no support from the information technology staff. Meditech has also added fuzzy logic to help speed order entry. Physicians have only to input a few letters of a patient's name, before the system will start guessing at whose record the MD wants to see. "I have been thrilled," she said. Her comments seem to challenge the view that Meditech has a very limited physician order entry capability.

Roll-out will be by speciality. Each speciality will be evaluated prior to rollout to address gaps between what physicians require and what the system can do. If a speciality is not satisfied, the project will not proceed to another specialty area until it is -- “period” -- she said.

Do lab and radiology order entry fit into the plans? Yes. "[w]e hope to move to lab and radiology once medication ordering is complete, said Ms. Burkholder.

These steps should also help ease physician acceptance:

• Cambridge will pay hospitalists at their regular rates of pay for the extra time they devote to developing the system.
• In support of the project, the pharmacy department has agreed to manually document the amount of time it spends on the telephone with physicians clarifying orders. Ms. Burkholder believes that these logs will help prove to physicians the utility of CPOE.
• Cambridge pulled together some 50 relevant articles on the benefits of CPOE and presented its medical staff with a summary and the original research so that they could see for themselves the risks and benefits. It was very well received.

Here are the areas in which the challenges lie:

• Working out a process for getting patient allergies entered into Meditech by someone who understand the difference between a sensitivity and a true allergy. Ideally, allergies will be noted by a nurse and affirmed if necessary by a physician or a pharmacist.
• Making sure that staff support is in place and that the system is not missing some critical component before it is rolled out to each specialty area. A researcher will conduct focus group sessions with physicians to ensure that they are satisfied with the results of development efforts.

Back to the top of this file.  Back to last location

Case Study 14: Overlake Starts Pilot Testing; Blending CPOE Orders With Paper Charts Is A Key Transition Workflow Issue

Overlake Hospital Medical Center
1035 116th Avenue NE
Bellevue WA 98004
Size: 337 beds
Phone: (425) 688-5000
CEO or Administrator: Ken Graham, President and CEO
Chief MIS Executive: Bruce Elkington, Chief Information Officer
Chief of Medical Staff: Tom Castle, M.D.
CPOE System: Meditech Magic Patient Order Management
Sources of information for this report: Bruce Elkington, CIO, (425) 688-5000

Overlake Hospital Medical Center launched a pilot test of CPOE using Meditech Magic CPOE in October, 2002. The original July starting date was delayed while Meditech did some upgrades recommended by clients who were CPOE early adopters, said Bruce Elkington, CIO. The module will “snap on” over Overlake’s existing pharmacy and medication administration record systems.

Overlake is pilot-testing the system with hospitalists, who write a significant portion of Overlake's medication orders. The implementation staff will include a pharmacist, a nurse, a half-time IS person, and a physician, who will devote eight to ten hours per week to the project.

The nurse, who is still being recruited, will help with order set development and workflow issues, and participate in training, he said. One key workflow issue: blending CPOE with a paper chart. One knotty question raised by the dual system is this: if some physicians place electronic orders and others hand-write orders, how will the nurse, the respiratory therapist, and others on the care team know whether an order has been placed on a patient?

Training will be by the pharmacist and the nurse, who will provide some instruction and will also follow physicians around until they get the hang of the system.

Data entry will be by wireless device; Overlake has considered Pocket PCs, Microsoft Tablet devices, PCs on carts, and wall-mounted PCs he said. The hospital has no solid answer, and may end up using more than one type of device.

Overlake expects to implement medication decision support, including dose checking by patient size, drug interactions, therapeutic checking, allergy alerts, and flags for high-cost drugs.

Mr. Elkington firmly believes he is at a natural advantage in his CPOE project because Overlake years ago chose Meditech Magic, a closely integrated system. Other organizations are spending "many times" what Overlake is because they face the problem of integrating the underlying systems, he said. He believes that CPOE is more common among Meditech sites than it is among hospitals that are building on competing legacy systems “because (Meditech is) installable."

Overlake's main motive is improvement of patient care, he said. Overlake is situated in Bellevue, Wash., where major employers are driving participation in Leapfrog Group reporting. “Leapfrog is there,” he said, but we’d be doing this anyway...My family comes here too.”

Back to the top of this file.  Back to last location

IV. Additional User Reports: Survey Results

As noted above: most of those interviewed by Editor Suzanne Corrales for the Case Studies above were asked some identical basic questions. However, a core purpose was to let interviewees discuss what they, not we, deemed significant. As a result, the Case Studies are wide-ranging, do not offer a common data set or answers to all the same questions, and follow no specific presentation template.

However, for the section below, we asked system implementers to respond to one standard set of questions in e-mail and voice interviews (with some, but little, opportunity to move off-topic).

Researcher Rachel Ross obtained survey results from clinical implementers or information technology executives at 10 hospitals. The majority asked not to be identified; therefore, in this section, none is identified, with these exceptions:

1. Information from her interview with Montefiore Medical Center is combined with the information obtained by Editor Suzanne Corrales in the Montefiore Case Study above and is not included in the aggregated responses below.
2. Vanderbilt University Medical Center spent years developing its own unique and, judging from what we have heard about it, highly capable system, and then licensed the system to vendor McKesson for an ongoing commercial adaptation. Because no other user information is available yet about the system, we concluded that the Vanderbilt survey response may be of great interest to potential buyers of a CPOE system. Therefore, the survey response from Karen Hughart, Vanderbilt's director, systems support services, is not included in the aggregated responses, but is presented separately below them.

As some vendors have few CPOE clients, it was necessary to avoid associating the exact size and CPOE vendor with a particular institution’s response in order to better preserve their anonymity. This does, unfortunately, make some responses a bit less useful than they might be (for example, in making judgments about the size of an implementation team). However, given the limitations of promised anonymity, this is the best we could do for this edition.

Survey Questions And Aggregated Responses:

Questions asked on the survey are in boldface. Responses are in regular type.

Back to the top of this file.  Back to last location

1. Types of organizations from which we received survey information:

One respondent is a community hospital with fewer than 300 beds, with no residency program. One is a community hospital with fewer than 200 beds, with a residency program. One is a community hospital with more than 400 beds, with a residency program. One is both a community hospital and an academic medical center with more than 500 beds. One is a military hospital with more than 400 beds, with a residency program. Three identify themselves as academic medical centers; of these, the smallest has more than 500 beds, and the other two have more than 800 beds.

2. Status of implementation:

The responding community hospital without a residency program is just planning its pilot test. One of the community hospitals with a residency program was just going live on its pilot test at the time of the survey response. All of the others are live in all or most departments.

3. Which systems these respondents are using:

Two community hospitals are using the Meditech Provider Order Management System; two are using Per-Se Patient1. The military hospital has the CliniComp system, CIS. The academic medical centers are using Eclipsys Sunrise Clinical Manager, Per-Se Patient1, and two use Siemens Invision. As noted, Vanderbilt, for which responses are broken out separately, uses Whiz, an in-house-developed system first rolled out in 1994.

Back to the top of this file.  Back to last location

4. What department did you use for pilot testing? Was it the right choice? Why or why not? How did you chose it? How long did your pilot test last?

5. When did you go live on your first non-pilot unit?

Community hospital with no residency program, with fewer than 300 beds, planning pilot: "Proposing the Hospitalist Group - (provide care to patients who have no family physician)." Planned go-live of pilot is April, 2003.

Community hospital with residency program, fewer than 200 beds, currently live on its pilot program: "We selected one patient floor and one physician. The physician was selected due to the variety of patients, the close proximity of patients, and most importantly his willingness to participate in the pilot. Selecting one physician permitted us to focus in one area and correct problems as they occurred in a central area. We feel this worked well as the pilot uncovered some nursing procedure changes. These were easier to manage on one floor than if we had attempted numerous nursing areas. Our pilot is ongoing for the last week." Go-live on other units was two weeks away at the time of this survey.

Community hospital with residency program, over 400 beds: "We didn't have a pilot unit. We've had CPOE since 1996, and it was always "live." We started with resident order entry in 1996, and the resident population (medical and surgical) now average 95% direct order entry.

"For the attending population, we concentrated on physician direct order entry in September, 2000 for a specific unit in our Inpatient Behavioral Health Division...They were a smaller, more controlled group of physicians (mainly psychiatrists) whose order sets were standardized according to their specifications, and the medication order sets were not as complex as the medical or surgical attendings.

"...We (IS Department) provided a minimum of 3 weeks of 24x7 support for both the physicians and the nurses. We had an issues list, and we worked with those groups to correct their issues for at least the next 6 months."

Community hospital with residency program which is also an academic medical center, 500+ beds: "Pilot test with Lab system about 10 yrs. ago. Lab dept. was chosen because their contract with CHC (Community Health Computing, a lab system vendor at the time) was ending and it was a logical jump at the time. Test lasted for 6 months."

Go-live beyond the pilot: "Six months after lab dept. went live with CPOE, we went live with the order entry piece and then the pharmacy piece."

Military hospital with residency program (respondent had worked on a CPOE implementation at a prior military hospital): "At the previous facility, I used L&D (labor and delivery) & NICU (neonatal intensive care) for 6-weeks. However, the hospital stopped deploying CPOE after I left. Nurses are still, for the most part, transcribing from paper orders. The NICU & L&D are still the only units using CPOE. These units were chosen because the population is somewhat contained, and business processes don't filter to the other units with transfers, etc. Reasons are patient care continuity and SAFETY!

"...[M]y current employer (also a military hospital) initiated CIS with the providers entering orders from the very beginning. When CIS was implemented, so was CPOE."

Academic medical center with 500+ beds: "Piloted on a general medicine unit to ensure the largest breadth of goods/services ordered (particularly medication orders). It was the right choice. Pilot lasted about 3 weeks." This hospital went live on first non-pilot unit in February, 2001.

One of the academic medical centers with 800+ beds: "We did our pilot test in the transplant unit because it represents a good mix of all kinds of patients. It's like a mini-hospital that enabled us to address all ancillary areas, and iron out related medical issues. It was the right choice because of the variety of patients. Pilot lasted a good three months before rolling it out to rest of our departments...We went live in all units in May 2002.

"After our success with the pilot in the transplant department, we didn't want to go unit by unit in a piecemeal way. We went fully live in all units, all ancillary areas the same day. Sort of a big bang approach! Part of the reason people fail at implementation is because they do it unit by unit and the system has less adoption. If you go unit by unit, doctors have to order meds in two different kinds of systems. The frustration in the healthcare environment is stressful enough, but if you put two different processes in front of a physician, forget it, they won't use it. Once burned, twice shy, so to speak. They won't use it. We have to keep it simple."

The other academic medical center with 800+ beds: "We piloted using the inpatient General Medicine Unit. This unit represented a wide cross-section of ordering patterns that allowed us to identify and remediate the majority of issues quickly. Pilot testing lasted for approximately one month, followed by a phased implementation in other inpatient care areas.

"The project was initiated in Spring 1993 and went live on our first non-pilot unit in Summer 1996."

Back to the top of this file.  Back to last location

6. When you went live, did you have --

• A--a pharmacy system that was integrated with your physician order entry system?
• B--a medication administration record that was integrated with your CPOE system?

Community hospital with no residency program with fewer than 300 beds, planning pilot: does have integrated pharmacy, does not have integrated meds administration. it's "planned for future but not feasible now since we need hand-held devices for nursing."

Community/military hospitals with residency programs:

• Three did have integrated pharmacy, one did not.
• Three did have integrated meds administration, one did not (a slightly different subset than the three above).

Academic medical centers:

• Two did have integrated pharmacy, one did not.
• One did have integrated meds administration, one did not.

(Others didn't respond.)

Back to the top of this file.  Back to last location

7. Of the total number of physicians who practice at your hospital, what percentage currently enter their own orders through your CPOE system?

Community hospital with no residency program with fewer than 300 beds, planning pilot: "too early to tell."

Community hospitals with residency programs: the one which has fewer than 200 beds is "still piloting with one physician," and at the 400-bed hospital which has been live since 1996, "90-95% of docs do order entry online."

Community hospital with residency program/academic medical center, over 500 beds: "Residents are at 95% and the attendings are at 40%."

Military hospital with 400+ beds: "100% for inpatient. There are areas which have such complex order sets [that] they remain on paper (ex: Chemo, Dialysis). We are trying to upgrade to accomplish this also."

Academic medical centers:

• "99%"
• "100% mandatory. Everybody's doing it."
• "approximately 99% of residents."

Back to the top of this file.  Back to last location

8. What is your ratio of data entry devices to patient beds? Is it adequate?

Community hospital with no residency program, fewer than 300 beds, planning pilot: "We have almost one device for every patient bed and anticipate an increased demand once we move beyond the pilot group, which is a very small group of physicians."

Community hospital with a residency program, fewer than 200 beds, currently in pilot: this hospital is using the pilot project to test the appropriate number.

Another community hospital with a residency program: "1 PC for every 2 beds at our 600-bed hospital. But that is just in the hospital. There are more PCs in the clinics."

Community hospital with residency program, over 400 beds: "We have bedside, point of care devices in every patient room in addition to having several devices at every nursing station and substation. We can use more at the nursing stations and areas where the MDs congregate."

The military hospital with 400+ beds: "1 to 1 on all inpatient units, plus some! Providers want more. Not all physicians have access to CIS on their office PCs and/or in the outpatient clinic areas. However, they DO desire this. We are working on it."

Academic medical center with 500+ beds: its ratio is 1 to 2.5 beds. Its response did not comment on adequacy.

One academic medical center with 800+ beds: "Hard to say because our stationary PCs are used for so many different computerization systems. But at every nursing station there are a least a minimum of two PCs, that are used for CPOE, among other things."

The other academic medical center with 800+ beds: "The ratio of data entry devices to patient beds is approximately 1:4 on the inpatient units. However, a positive aspect of CPOE is off-unit access to orders which allows physicians to manage patient care from their office computers, or any network-connected computer."

Back to the top of this file.  Back to last location

9. What percentage of clinician order entry is on wireless devices?

The answers were all over the map:

One community hospital with a residency program: "100%--although they could use desktop if desired."

Another community hospital with a residency program: "Only 10%. Wireless devices used mostly in surgery dept."

Community hospital with residency program, over 400 beds: "We do not use PDAs for order entry, but we do have our home health RNs set up to do documentation and order entry on laptops that use CDPD wireless technology."

The military hospital with 400+ beds: currently, zero.

Among the academic medical centers:

• The 500+-bed medical center said zero.
• One of the two very large medical centers with 800+ beds said, "128 wireless laptops that roam around the hospital for physician order entry. We have 128 dedicated wireless devices to 650 staff beds."
• The other: "On the inpatient units, approximately 40% of the devices are wireless- enabled laptops affixed to rolling carts. In our outpatient setting, only a small percentage of devices in use are wireless. At this time, we do not support wireless hand-held devices such as PDAs or tablet computers."

Back to the top of this file.  Back to last location

10. What kind of training did you use for physicians--Computer based? Classroom? One-on-one? Combination? And what worked best?

Community hospital with no residency program, with fewer than 300 beds, planning pilot: "A combination of classroom, one-to-one and printed manual is planned."

A community hospital with a residency program: "We did one-on-one training. This provides a better training method and works well trying to accommodate physician schedules."

Community hospital with residency program, over 400 beds: "A combination of classroom and one-on-one. Training is a hurdle. Docs seem to be embarrassed to be seen by peers as "not knowing" something in a classroom situation. They don't want to be seen as being behind in any area. It doesn't matter what schedule of classes we offered--morning, noon, night, the docs stopped coming after a while.

"We found we got the best results from private sessions with docs. Also, we make rounds once a day to try to catch docs actually doing their order entry at the computer. Then we approach them, make a friendly connection through conversation, and try to create a "teachable moment" when they are on the job, so to speak.

"We also provide a separate phone line, just for docs, to get help for non-urgent questions. They can leave voice mail if they are just wondering how something works. We e-mail them back (documentation of response) and/or call. Docs seem to like the conversation piece. For urgent needs, we have a 24-hour beeper help line. If they are stuck on the system and need to prescribe meds right away, they beep us and we help them on the spot."

The military hospital with 400+ beds: "A combination of classroom and a lot of e-mail infomercials--physicians are extremely reluctant to come to a classroom. At the last hospital, I had super-users (nurses), who understood the system due to the transcription of paper, teach the physicians for 6-8 weeks prior to going live."

The community hospital which identifies itself as also an academic medical center, 500+ beds: "We used a combination of all three formats. Sometimes we have no other choice due to the doctor's schedule. We usually need to do one-on-one with ER doctors, for example, because they're working around the clock. We also provided Internet self-paced learning. But the classroom format proved to be the most popular, and worked best because it offered more hands-on training."

The academic medical center with 500+ beds: "a combination of classroom and 1:1."

One of the academic medical centers with 800+ beds: "There were mandatory computer classes for doctors to retain privileges at the hospital. We staggered the training throughout the implementation period. We used a combination of mandatory computer classes and we also trained "super users" from each department, and several were trained to roam the halls and find out if a physician was having trouble, and provide one-on-one help to physician. So we always had one super user per floor but within a month, the demand plummeted because the system is so easy to use, people got used to it right away. A lot of hospitals that provide only classroom training make a mistake. It's the real-life scenarios that require the most help, and that is the best place to apply the learning emphasis."

The other academic medical center with 800+ beds: "A combination of classroom and one-on-one. Medical students are taught CPOE in classroom sessions as part of their introduction to UTMB. We also arrange for one-on-one training at the convenience of the medical staff, either as full training sessions or "refresher" courses."

Back to the top of this file.  Back to last location

11. Did you go live with departmental order sets? Departmental and personal? Has this changed since you went live? Does it depend on the department? Which uses which?

Community hospital with no residency program, with fewer than 300 beds, planning pilot: "Both-- we will go live with personal (physician created) order sets, and some order sets that have been predetermined."

The community hospital with a residency program which was in its pilot test at the time of the August, 2002 survey: "We are using both departmental and physician order sets. It's to early in our process to tell if changes will be necessary."

Community hospital with residency program, over 400 beds: "We use a combination of departmental and personal order sets. Any order sets are kinetic and we make adjustments to personal order sets when requests come through. Revisions are very common requests from the physicians. As they use the system, they are finding what works easily for them versus what is too cumbersome. Overall, many of the changes made for departments are because of regulatory requirements, especially in the pharmacy area."

Community hospital with residency program which is also an academic medical center, 500+ beds: "Combination. It really does depend on the department. Changes are bases on the coding books. If they change or add new codes, we've got to update the order sets. If you're a doctor that's a specialist, you have your own personal order sets designed. We accommodate them based on policy and procedure. It's really very user friendly, that is the key. Otherwise doctors won't use it at all."

Military hospital with 400+ beds and one academic medical center: both replied that they did go live with departmental order sets, but didn't elaborate.

The academic medical center with 500+ beds: “yes” to departmental order sets.

One of the two very large academic medical centers with 800+ beds: "Order sets are organized to be service-specific and then they're ailment-specific. So there's specific order sets for cardiology, and another one for NICU. We have about 400+ different order sets for various ailments and conditions. We encourage practicing evidence-based medicine, best-practices medicine to contain costs and ensure the patient gets the best possible care. Physicians do have ability to modify, but that doesn't alter it permanently. We have accountability to our patients to ensure they get the best possible care so when a physician wants to change something, we go back to evidence-based medicine. If they really believe that their way of practicing medicine is better, they present it to the ongoing order set evaluation committee. So, if it makes sense, we change it for the baseline."

The other academic medical center with 800+ beds: "Yes, we utilize departmental order sets extensively. We do not support personal order sets. The use of departmental order sets and protocols was a key factor in acceptance and subsequent success of CPOE."

Back to the top of this file.  Back to last location

12. What use are you making of on-line decision support?

• A--most rules are for efficiency and data compliance;
• B--out of the box medication checking rules (vendor wrote the rules);
• C--Adverse drug event checking (Organization had to write the rules);
• D--Other?

Community hospital with no residency program with fewer than 300 beds, planning pilot: "Our organization will do some tweaking of the rules" (presumably a standard set of vendor-supplied decision-support rules).

The community hospital with a residency program which is still in its pilot: "We are using the vendor-supplied interactions and adverse drug checking."

The community hospital/academic medical center, 500+ beds: "Most rules are for efficiency and data compliance. (e.g., Is the order complete? Logical? For example, system might flag pregnancy test on a man because it is illogical.) "We use Medi-Span, which is integrated with Patient1 pharmacy module. It has, for example, IV interaction alerts. Also, if a patient is lactose-intolerant, it will signal if a drug has a lactose derivative. But you have to set the different settings. If you want it at the most sensitive, then it will bleep at anything. But if you just want the most critical, then it will only bleep at the most critical reaction."

Community hospital with residency program, over 400 beds: it is using out-of-the-box medication checking rules provided by a vendor, its own adverse drug-event checking rules, and also something called "criteria engineers," which the respondent describes as "a kind of program logic that we build into the system as part of our HIS system."

The military hospital, 400+ beds: "First Data Bank (now Medi-Span) drug-to-drug, drug-to-allergen, duplications."

The academic medical center with 500+ beds: "Both vendor-supplied rules and custom-built."

One of the academic medical centers with 800+ beds: "We jointly developed rules with Micromedex. We were one of the development sites. Most of our rules were developed by us. Siemens didn't provide any rules in its software. We developed all of our checking rules internally, as a partner with Micromedex."

The other very large academic medical center: [the medical center] uses a combination of vendor and locally-written rules for medication interaction/allergy checking; ADE checking; medication dosage, timing and scales per unit validation."

Back to the top of this file.  Back to last location

13. How many FTEs were assigned to your project and for how long:

• Physician?
• Nurse?
• Pharmacist?
• Programmer/analyst?

Community hospital with no residency program, with fewer than 300 beds, planning pilot: it is planning on these figures:

• Physician: .48 FTE total, varies over 20 weeks;
• Nurse: "to be determined once rollout begins, initial impact will be very minimal;"
• Pharmacists: .53 FTE total, varies over 20 weeks
• Programmer/analysts: .53 FTE total, varies over 20 weeks.

The community hospital with residency program which is in its pilot program: "We had a team for the implementation but primarily our clinical analysts worked on the project. To estimate, I would say 1.3 FTE. The project length is hard to estimate since we had to wait mid project for our vendor to make some necessary changes. Total would be about 8 months."

The community hospital/academic medical center with 500+ beds: "On a go-live basis, 4 to 5 of each of the above. We did what we call "train the trainer." We train key personnel intensively so that when we're not there, they can troubleshoot and support the others in their department using it. We're a team, in a sense, but we don't have to be there because these users are the first line of defense. They can usually handle any IT or help a doctor when he gets stumped, but if they can't figure it out, then they're usually the ones to call us. It works well with these super-users because without them, the others would be calling us all the time and we wouldn't get anything done."

Community hospital with residency program, over 400 beds:

• -Physicians: none;
• -Nurses: 4 IS RNs;
• -Pharmacist: one part time in IS
• -Programmer/analyst: 1 lab analyst.

"There was no real time line - in 1996, when the residents began order entry, that project lasted about 9 months."

The military hospital with 400+ beds: "Initially, there were 5 assigned to deploy CIS, then reduced to 4. Now, we are hiring the database administrator (a 1st) to review order entry processes (overrides, etc.). We handle all our own training because there are so many ongoing changes to CIS, which have an impact on everyone."

One of the academic medical centers with 800+ beds: "During implementation, we had about 20 FTEs, ongoing super user groups, and about 10 consulting physicians who met with us about once every two weeks. Now that we're fully live, our support personnel is much less."

The other academic medical center with 800+ beds:

• Physicians "participated as members of Steering Committee."
• Nurses "participated as members of Steering Committee, varied as new areas were brought up on POE."
• Pharmacists "participated as members of Steering Committee."
• Programmer/analysts: "initially, a team of 12 programmer/analysts developed, implemented, and supported the system. Currently, a team of six provides enhancement and support."

Back to the top of this file.  Back to last location

14. If your project has been live long enough for you to see payoffs (benefits), what were the first you saw? How much time went by before you saw them?

Community hospital/academic medical center: "The payoffs are immediate in dollars saved. That shows within 6 months to a year. Everything is documented, [and so the] onus is on everyone who has access to the system. It documents and tracks what they do. The doctor can't say that the nurse didn't understand his writing. Even if the doctor can't spell, he selects from clear options, so there's no spelling errors. There's not a lot of room for major mistakes, so if one is made, it's clear who's responsible. The individual accountability is right there; each person in each department is held accountable. Some people are afraid of that, but it's good for the hospital, and especially for the patients."

Community hospital with residency program, over 400 beds: "The RNs like the fact that they do not have to deal with the bad handwriting or calling for clarification. Pharmacy likewise, but they are also more concerned with seeing fewer medication errors. They also like the fact that the docs see the drug interaction alerts and the duplicate check messages when they direct-order."

Military hospital with 400+ beds with residency program: "There is a more integrated operation with nursing, physicians, pharmacy, and even medical records and JAG--much more of a team effort in patient care and better understanding of how we all impact each other's processes."

The academic medical center with 500+ beds: "Immediately: legible orders, fewer phone calls to clarify order intent or order status, more complete orders, reduction in ADEs (adverse drug events). Later: increase in IV to PO (intravenous to patient oral) conversions."

[Editor's note: while this respondent didn't give an indication of the cost significance of conversion of intravenous medication to a patient taking a medication orally, we did find this piece of anecdotal information at the University of Kentucky Hospital's web site: "The conversion of each patient receiving IV famotidine to the oral form just one day earlier is projected to save the hospital approximately $8,000." -- http://www.mc.uky.edu/pharmacy/dic/interchange/famotidineIC.htm.]

One of the academic medical centers with 800+ beds: "We saw payoffs in the pilot phase. They were better access to care and length-of-stay reduction (LOS). At 6 months post-live, we saw more benefits:

• Turnaround time from prescribing a medication to getting it into the patient improved by 66%, reducing it from about 7 hours to about 3 hours.
• Radiology exam turnaround improved by 45%. Finishing the tests, getting the results into the system improved 25%.
• In crucial areas, like cardiology, LOS reductions were favorably measured. There was a 25% reduction in medical errors administration.
• Transcription errors went to ZERO."

The other academic medical center with 800+ beds: "Initial payoffs seen almost immediately were:

• use of printed requisitions to ancillaries resulted in less confusion and time spent transporting requests;
• standardized data collection resulted in fewer call-backs to the floor for additional information;
• clinical practice guideline enforcement;
• standardized processes for medication dose, dose timing, and dose scales per unit;
• standardized prescription writing and prescribing (units written out, metric);
• rounds reporting, interdisciplinary rounds report, checkout listing for cross coverage;
• off-unit access to orders;
• legibility of physician orders; and
• increased compliance with required fields in requisitions and orders."

Back to the top of this file.  Back to last location

15. All things being equal, knowing what you know now, what would you have done differently?

Community hospital with residency program, over 400 beds: "We should have had a better rollout plan. The hospital administration mandated to the "captive" MDs that they would need to do direct order entry hospital-wide beginning in January, 2002, but we weren't prepared with all the doctors' order sets or education. Doctors resisted--there is a large cultural change that must be taken into consideration--and you need enough resources to be a presence on the units to support the docs."

Community hospital/academic medical center: "The initial idea was to copy and scan all charts on line, but because we're in NY, there are strong unions, and this would have eliminated lots of medical clerks' jobs. There was a big fight about that, so we decided to do it slowly. But eventually, it will be paperless--that's our goal. You won't need the paper chart. It will all be electronic. We have a back-up generator, we've got the necessary safety mechanisms in place."

Military hospital with 400+ beds, with residency program: "I would have, if empowered to do so, insisted that the first hospital I worked at with CIS implement the system the way [the second one] did, and direct the physicians to put the orders in."

One of the academic medical centers with 800+ beds: "We used a mainframe-based, non-Windows system, which took almost 3 years of planning because we wanted to go live in all departments at once. Today's systems are equipment are much more flexible. Adoption would have been easier. Selling the idea and regular use to doctors with today's technology, which is web-based, robust, user-friendly, would have been easier. Vendors have matured a little. They used to build clinical systems on business platforms but now they're building them from scratch and designing them clearly for clinical settings. That's an improvement."

The other academic medical center with 800+ beds: "We would probably do very little differently. I would recommend compressing the implementation schedule as much as possible, as it was difficult for medical staff during transition between "POE Units" and "non-POE Units."

Back to the top of this file.  Back to last location

CPOE Survey Response: Vanderbilt University Medical Center

Vanderbilt University Medical Center
1660 21st Avenue South
Nashville TN 37232
Phone: (615) 322-5000
Size: 631 total beds, 609 acute
CEO or Administrator: Mark Penkhus
Chief MIS Executive: William Stead, M.D.
Chief of Staff: C. Wright Pinson, M.D
Medical Director: John Sergent, M.D.
CPOE System: In-house-built Whiz System
Sources of information for this report: Karen Hughart, Director, Systems Support Services, 615-936-6779

1. Type of institution:

Academic medical center.

2. Status of CPOE implementation:

Live in all departments for several years. (Began project in 1994.)

3. What kind of system are you using?

"An in-house system called Whiz, which the medical center sold to McKesson. The medical center will be installing the McKesson Horizon version of the system within the year."

4. What department did you use for pilot testing? How did you choose it? Was it the right choice? Why or why not? How long did your pilot test last?

"Our pilot units for Whiz Order were the Cardiac ICU and Stepdown units. We had strong physician leadership, a homogenous group of patients, and strong nursing leadership on the units. The patients on the two units transferred between each other so we didn't have such a big problem with patients going from implemented to non-implemented units by starting there.

"We also got very early experience with both ICU and lower acuity patients. In hindsight, this was a good choice, because we had an excellent chance for success. We had assurances that the Nursing and Physician leadership would not fold at the first signs of adversity. They made a commitment to work with us till we got it right. We implemented the next unit (Adult Myelosuppression unit) approximately 2 months after the go-live date in Cardiology."

5. When did you go live on your first non-pilot unit?

March, 1994.

6. When you went live, did you have --

A--pharmacy system integrated with your CPOE system?

Yes.

B--medication administration record (MAR) integrated with your CPOE system?

"No. We had an MAR computer generated from the Pharmacy Computer system that was manually completed. (We still do.)"

7. Of the total number of physicians who practice at your hospital, what percentage (estimate) currently enter their own orders through your CPOE system?

"There are approximately 600 attending physicians and an additional 300 residents and fellows. Approximately 68-70% of all orders are entered by physicians. Because we are a teaching hospital, the overwhelming majority of all orders entered by physicians are entered by house staff.

"The 30% of orders entered by others are entered by nurse practitioners and staff nurses (as verbal, telephone, protocol, or nursing orders not requiring MD countersignature; by pharmacists, respiratory therapists, dieticians, and other professional disciplines who are permitted by policy to enter verbal, phone, and protocol order for their own disciplines; and by clerical personnel who may enter written orders."

8. What is your ratio of data entry devices to patient beds? Is it adequate?

"We have 1-2 patients per computer in our ICUs, approximately 3 patients per computer in Stepdown areas, and 4-5 patients per computer in general care areas. There are also computers in ancillary departments, the cafeteria, etc. We do use some wireless laptops on rolling carts to enter orders during rounds in a few units.

"Especially on medicine units, which tend to have physicians present more hours of the day, we do have some wait times for computers during busy times of the day. We would put more devices in these areas if we had sufficient space. Except for these few areas, the numbers are generally adequate.

9. What percentage of clinician order entry is on wireless devices?

"We do use some wireless laptops on rolling carts to enter orders during rounds in a few units."

10. What kind of training did you use for physicians? What worked best?

"A combination. Except during initial rollout, we've trained all house staff in classroom training sessions. They give us one day during house staff and fellow orientation and we train 50-150 people each day once per year. We like to train groups of attending physicians, too, but don't have much luck with this. Most of them get trained in one-on-one or very small groups of 2-3 physicians at a time."

11. Did you go live with departmental order sets? A combination of departmental and personal order sets?

"Yes, we started from the beginning with 'best practice order sets' that an entire group was encouraged to use. We also had 'Private Whiz,' which allowed each user to develop up to ten

personal order sets for use when a 'best practice' order set was not available or to enter orders for patients not yet registered that could be 'played back' once the patient was registered."

Has this changed since you went live?

"We are still doing the same."

Does it depend on the department?

"Some departments are more proactive in maintenance of best practice order sets than others, so in cases where the order sets are out of date, they are not used as often."

12. What use are you making of on-line decision support?

"We have access to commercial (AHFI and Moseby) Drug References and hospital-developed monographs from within the order entry system. We have dose range checks for certain classes of drugs for adults and most all pediatric medications. We also have more sophisticated drug dosing decision support for titratatable drips, antibiotics, and several other types of drugs used by children that consider weight, age, and creatinine clearance (if lab data is available). There are lots of other drug-specific decision support tools incorporated into Whiz to address specific issues (e.g., an empiric antibiotic advisor that shows cost adjusted rankings for therapeutically equivalent therapies, interventions to advise users away from drugs that are non-formulary, etc.)

"Our radiologists have worked out several decision support tools to help users order the best test to diagnose a problem to help steer users away from ordering both CAT scans and MRIs when only one or the other is warranted. We have decision support to help users order restraints in a compliant fashion, easily initiate appropriate Latex precautions orders, implement fall prevention orders appropriately, target skin prevention and skin care orders based on skin risk assessment, specify reasons for blood transfusion, and special preparation of blood bank products, etc. "There's a LOT of decision support built into Whiz."

13. How many full-time employees were assigned to your project? For how long?

A--Physician: "For most of the rollout, the physician-programmer who created Whiz was on the project full-time. For the past two years when we have not been actively implementing new users, we have had several part-time MDs involved in the project, but none focus on POE full-time."

B--Nurse: "There have been 5-6 nurses focusing on Whiz as their sole or primary project since the beginning. This group is responsible for implementation planning, training, implementation support, and on-going support for all Whiz users. They also develop order sets and do some simple systems maintenance."

C--Pharmacist: "Two pharmacist spend very significant percentages of their time helping maximize pharmacy impact from Whiz. A third, a chemotherapy pharmacist, maintains all chemotherapy order sets."

D--Programmer/analyst: "The numbers have varied, and there has been regular involvement from biomedical informatics students throughout the project, but there are an average of 2 programmers and 1 analyst that work full-time on POE. Part-time help that might add up to another FTE develops modules and does other smaller pieces of work."

14. If your project has been live long enough, what were the first "payoffs" (benefits) you saw, and how much time passed before you saw them?

"We realized some benefits almost immediately. We are able to deliver first doses of medications 17% faster for ICU and 38% faster for general care patients if orders are entered directly by physicians. Orders are legible and we've been able to alter cumbersome work processes by taking advantage of the computer's ability to produce an accurate "snapshot" of active orders on demand. Nurses don't have to manually update Kardexes. We can review current orders when patients change service or level of care and make changes as needed instead of starting from scratch. Physicians, nurses, and other members of the team are all working from a common set of orders rather than from individual perceptions of what's active and what's not.

"Other benefits came along somewhat later after we had a solid user base. We generally think it's best to get people using the system, then give them data, and then work on interventions rather than start with new users with interventions to change practice dramatically from day #1.

"We have used data from Whiz to identify areas with over- or inappropriate utilization of services. We've then been able to target interventions to address these situations without adversely impacting quality of care. For example, our clinical experts have looked at patterns for ordering lab tests, portable chest x-rays, and other heavily-used services, and made changes in the system to curtail this. You can no longer order open-ended orders for these targeted services, but must instead review orders every 24 hours in most cases. This targeted way of cutting cost has been done with no identifiable impact on the quality of care, but has saved hundreds of thousands of dollars. We've been able to provide decision support to address quality and cost issues for a variety of services (some mentioned previously)."

15. All things being equal, knowing what you know now, what would you have done differently?

"We would have been more realistic about ongoing support needs. It takes a lot of time, expertise, and effort to support clinicians and to maintain up-to-date clinical systems. The more users and areas implemented, the heavier the ongoing support burden.

"We underestimated how different pediatrics was, and spent over 3 years trying to develop the medication ordering infrastructure to support their needs.

"We've been slow to respond to the increasing blur between inpatient and outpatient status and to develop the application to support both. Now that we are working with the ED, the Clinical Research Center, and other patient care areas that treat large numbers of outpatients, we are finally making system changes we've needed earlier to better support appropriate ordering for outpatients that are in inpatient beds for observation."

V. Additional Topics:

1. Six Approaches To Winning Physician Acceptance of CPOE
2. Hand-Helds: Many Physicians Do Like Them--But Only for Limited Purposes
3. A Cautionary Tale: Pharmacy Alerts Shut Off

Six Approaches To Winning Physician Acceptance of CPOE:

Five of the six organizations discussed below use persuasions based largely on helping physicians to find and appreciate the benefits of CPOE. In the sixth--a prominent hospital loaded with "physicians to the stars," one of whom made the cover of Time Magazine--the approach is, surprisingly, "You WILL use it." These hospitals are:

• Holland Community Hospital, Holland, Mich.
• Fairview Health System, Minneapolis, Minn.
• Overlake Hospital, Bellevue, Wash.
• Great Plains Regional Medical Center, North Platte, Neb.
• Cambridge Memorial Hospital, 277 beds, Cambridge, Ontario
• Cedars-Sinai Medical Center, Los Angeles, Calif.

Back to the top of this file.  Back to last location

Holland Community Hospital; Fairview Health System; Overlake Hospital

In most situations, entering most orders into a computer will take a physician extra time, which physicians will be loath to give up. So, if winning physician acceptance of computerized physician order entry (CPOE) is tough in a teaching hospital, where many physicians are employees, what should a community hospital do, when most physicians aren't employees?

Bill Van Doornik, MIS director at 213-bed Holland, Mich., Community Hospital, plans to install the Per-Se Patient1 CPOE system. He agrees with estimates that CPOE will take physicians an additional 30 minutes per day. Gary Strong, CIO at nine-hospital Fairview Health System, Minneapolis, Minn., guesses that it will add one minute of physician time to an order that would normally take two minutes to write. He's installing Eclipsys Sunrise.

The solution, they and other project leaders agreed, is to make the technology so compelling, the order sets so easy to set up, and the decision support so useful, that physicians begin to demand it. "You have to make the benefits visible," said Bruce Elkington, CIO, Overlake Hospital, Bellevue, Wash., who is installing Meditech Magic.

In some cases, circumstances will help. Overlake has a staff of hospital-based intensivists and hospitalists who are responsible for writing a significant portion of Overlake's orders. For some reason, it also has a disproportionate number of technically-oriented physicians on its OB/Gyn staff. It will beta-test physician pharmacy order entry in the ICU, and probably roll out first to Ob/Gyn, Mr. Elkington said.

Fairview Health System operates a large ambulatory care practice, which is rolling out Epic Systems EMR. Physicians who are already familiar with the Epic system come to the community hospital more willing and eager to computerize than they otherwise might be, because they are familiar with the benefits, Mr. Strong said. But he is also out there doing old-fashioned sales pitches. He estimates that he has participated in at least 100 meetings with physicians and physician groups about CPOE.

Mr. Elkington figures that he has a natural advantage because Overlake chose Meditech Magic, a closely integrated and relatively inexpensive platform, years ago. Other organizations are spending "many times" what Overlake is because they face the problem of integrating the underlying systems, he said. If you look at the First Consulting Group Leapfrog report, you notice that a lot of Meditech sites have CPOE, he said. "That's because it's installable."

Back to the top of this file.  Back to last location

Great Plains Regional Medical Center, North Platte, Neb.

Great Plains Regional Medical Center, 116 beds, North Platte, Neb., has a physician project champion who just happens to be married to the woman who was, until recently, the Great Plains CIO, Bonnie Buckland.

Not surprisingly, Robert Buckland, MD, sees plenty of advantages for physicians in CPOE. One obvious one: they'll be spared the annoyance of puzzling over each other's handwriting. As a surgeon, Dr. Buckland often walks into a patient room cold and picks up the threads of what another physician has started. Being able to consult a legible, well-organized medical record and tell at a glance what is going on will be a big improvement, he said.

Not being interrupted during the day for order clarifications, and not having to repeat orders or prolong patient stays because someone didn't understand what the MD meant in the first place are other major benefits for physicians, he said.

Back to the top of this file.  Back to last location

Cambridge Memorial Hospital, Cambridge, Ontario

277-bed Cambridge Memorial will try to document that benefit for its pharmacy and medical staffs. Cambridge pharmacists will manually log the amount of time they spend on the phone with physicians, both before and after installation of Meditech Magic CPOE, said Rena Burkholder, CIO.

Cheryl MacInnis, Cambridge director of clinical pharmacy, also conducted an extensive literature search on the benefits of CPOE. Ms. MacInnis summarized some 50 CPOE research articles and presented them, along with the original articles, to the medical staff. It was well-received, she said.

Cambridge will also try to boost physician acceptance by paying physicians at their regular rates of pay, for time they spend on the hospital's Meditech Magic CPOE project. Cambridge received an $89,000 grant from the Ontario Hospital Association's Change Foundation. The technical quality of underlying systems is the one critical success factor that is heavily within the control of the IT department, so project leaders are focusing on it. For example, Holland Community beta tested an earlier version of the Per-Se order entry system and found the screen flow cumbersome, Mr. Van Doornik said. Per-Se has rewritten the module using Java, and Holland plans a second beta test. "We are building them the very best system we can."

Integration of the underlying system is also key. Dr. Buckland, who is installing QuadraMed Affinity, and Mr. Van Doornik said having a unified, underlying system architecture is a crucial first step. "Without that, you just couldn't do this," Mr. Van Doornik said.

Mr. Van Doornik, (616) 394-3449; Mr. Elkington, (425) 688-5000; Mr. Strong, (612) 627-6339; Ms. Burkholder (519) 621-2333, ext. 1395; Dr. Buckland, (308) 534-5370.

Back to the top of this file.  Back to last location

Cedars Sinai Policy Suspends Physicians Who Don't Train For CPOE

In what seems to us to be a bold move, Cedars Sinai Medical Center is planning to suspend physicians if they don't finish either a 2-hour classroom training course or a self-paced computer assisted learning module on the hospital's new computerized physician order entry (CPOE) system before their units go live.

Physicians do receive a benefit--CME credits--for the training, which is set into stone as a requirement in Cedars Sinai's medical staff bylaws.

The strategy seems to be working. Virtually all physicians had completed the course by the time a pilot test went live in the fall of 2002. Of two who had not, one was retiring and the other had moved away, said Michael Shabot, M.D., medical director of enterprise information systems.

The new system, Patient Care Expert (PCX), is a home-grown patient management, patient accounting, and clinician order entry system. Physician use of the system is mandatory. Paper orders will not be accepted once the system goes live. Physicians who have not taken the training will not be able to give telephone orders, because their names will not be in the system. Cedars officials said that medical staff cooperation on the training requirement is actually fairly high because the system was built largely to their specifications.

Back to the top of this file.  Back to last location

Hand-Helds: Many Physicians Do Like Them, But Only for Limited Purposes
Most Are Using Devices To Consult EPocrates

Just because many members of your medical staff carry Palm Pilots doesn’t meant that they are necessarily ready to accept hand-held applications or would vote for hand-helds as the device of choice for CPOE. Those are our conclusion after listening to a presentation by Neil Smithline, M.D., and conversing with him briefly.

Dr. Smithline consults on hand-held deployments. He says the overwhelming majority of physicians who use hand-held computers use the simplest of all applications, EPocrates, the Physician Desk Reference, with automated drug interaction checking and regular on-line updates via the Internet. EPocrates is simple, convenient, and free to physicians. Some 170,000 physicians, or one out of five in the U.S., use it. Patient Keeper, another vendor, offers a $35 application that allows physicians to manually enter and store patient demographics, histories and physicals, lab data, and other clinical data on a Palm. Patient Keeper claims 45,000 users of that application.

Now take all other major vendors of wireless applications: Allscripts, MDeverywhere, e-Physician, and OnCallData--applications which do more and cost more. They may have 10,000-15,000 users combined, Dr. Smithline said.

His bottom line: physician adoption of hand-helds has been slow because physicians are not yet convinced that the benefits they offer are worth the money and the trouble. The more features hand-helds have, the more complex they are, and with complexity come costs and technical challenges.

Some physicians do use hand-helds for CPOE in a few hospitals. However, given Dr. Smithline's observations, and the relative difficulty of performing complex decision-tree tasks on a small device, we believe that at present, synchronized wireless devices of notebook size or smaller are more likely to end up being popular as data retrieval, limited prescription-writing, and transcription devices for physicians on the move than as full-blown order entry units. (There are, of course, exceptions, as noted below.)

In case you're considering providing wireless devices for your physicians, Dr. Smithline offers three basic considerations when looking at hand-held applications:

1. Operating system: The leaders are Palm and Windows CE. At present, Palm is by far the most popular with physicians.
2. Wireless vs. cradle synchronization: at the time of his presentation, Allscripts, MDeverywhere, and OnCallData offered wireless synchronization; Allscripts, MDeverywhere, e-Physician, MedAptus, and I-Scribe offered cradle synchronization. The advantages of wireless are that it is real-time, and that it reduces the load on devices with limited data storage. Wireless costs are coming down and response times are good, but security is often only so-so, he said.
3. Integration: if the hand-held is not integrated with the system that stores patient demographic data, it will not be used, he said.

Here's his rundown on the major hand-held applications, what they do, and what their limitations may be:

Prescribing: major vendors of hand-held prescribing applications include Allscripts, Touch Works, OnCallData, and e-Physician. While having physicians enter their own orders directly into a hand-held can theoretically help reduce medical errors, the system is only as good as the data that has been entered into it. Hand-held applications can, for example, provide drug allergy checking, but only if the patient's allergies have been entered into the system in the first place. Similarly, drug interaction checking works only insofar as patient medications are in the system.

Transcription: major vendors of hand-held transcription applications include Allscripts, Touch Works, MDeverywhere, and MedAptus. Early dictation systems were clumsy because the hardware did not support four-button functionality (fast-forward, rewind, etc.); more recently vendors have built this capabilities into the software.

Charge capture: major vendors of hand-held charge capture applications as of June, 2002, included Allscripts Touch Works, e-Physician, MDeverywhere, and MedAptus. Charges are less likely to be lost with such a system. Also, to the extent that the system builds in insurance company rules, charges are less likely to be rejected by the insurance company.

Six to one return on investment claimed

Studies publicized by MDeverywhere on its own system claim returns on investment of 6 to 1 and sometimes higher, but they have involved academic medical centers and may not hold true for small practices, Dr. Smithline said. His reasoning: while salaried physicians can afford to be casual about billing, physicians whose livelihoods depend on accurate billing are probably already doing a better job, and therefore may have less room for improvement from solely using a more convenient charge-capture device.

He predicts that hand-held applications will evolve from today's stand-alone systems to a more integrated approach. The evolution is already occurring. Examples include:

Patient Keeper has developed a common platform that includes modules for charge capture, lab data review, and problem lists, and it supports legacy applications. Cerner Corp. recently announced plans to build on the Patient Keeper platform.

Epic: arguably the leader in electronic medical record systems, Epic is developing a hand-held interface to its EMR. It will use either cradle or wireless synchronization. Charge capture is the first functionality it will offer.

Dr. Smithline, (415) 331-7407.

Back to the top of this file.  Back to last location

Microsoft Believes It Sees A Future For Tablet Devices

We do note, however, that just as we completed the first edition of this study, Microsoft held a press conference in which it announced the Tablet PC, a Microsoft name for its Windows XP Tablet PC Edition software running on a variety of third-party wireless devices. Two hospital software vendors, Allscripts and Eclipsys, announced physician order entry applications on this operating system and hardware. Here’s more information, from a Microsoft announcement (see http://www.microsoft.com/presspass/press/2002/nov02/11-07Tablet2002LaunchPR.asp):

"Tablet PCs are available in two primary designs: One features an attached keyboard and can be configured in the traditional laptop "clamshell" mode, and the other uses a variety of detachable keyboard designs in a so-called "slate" form factor. All Tablet PCs are designed to be a user's primary business PC, recognizing input from a keyboard, mouse or pen. Tablet PCs are powered by chips optimized for low power consumption and longer life from Intel Corp., Transmeta Corp. and Via Technologies Inc. Tablet PCs are available at retailers throughout the United States. Final retail prices are being announced by each manufacturer."

Whether the Tablet PC or something like it becomes a new CPOE device of choice, or turns into the latest clinical computing device to find its use as a paperweight like the AT&T PNut or a coat-rack like the IBM Clinical Workstation, only time and use will tell.

Back to the top of this file.  Back to last location

A Cautionary Tale: Many Hospital Pharmacies Shut Off Patient Safety Software Alerts Why? The Software Interferes With Getting The Job Done

In a test which may offer an implied warning for organizations implementing any clinical decision-support system, it was found that most pharmacists had turned off alerting systems and safety checks in their software because it interfered with getting their work done.

Perhaps it is arguable whether the response of a pharmacist to annoyances in his/her computer system necessarily imply that a physician, or a nurse, would behave similarly. However, it is not unknown for physicians to loudly object--and even to sabotage -- information systems which they perceive as a serious interference with their ability to effectively treat patients. (One of the co-authors of this report wrote about a case in which military physicians began unplugging terminals and reporting them down when superiors tried to force a faulty order entry system on them in the early 1990s.)

Here is what the test found:

Only four of 307 pharmacists found that their software issued alerts on all 10 dangerous errors run through the software as medication orders in the test.

"It was a real eye-opener," said David Classen, M.D., the First Consulting Group medical errors expert who shared the results at a recent meeting of the Association of Medical Directors of Information Systems. Dr. Classen is also helping to shape the Leapfrog decision support system test, which is expected to be available early next spring.

In addition, Dr. Classen said that Leapfrog Group evaluations of decision-support capabilities so far had turned up these observations:

• Most medication errors that harm patients are caused by errors in IV solutions, rather than errors in pill distribution. Therefore, it is possible that bar-code-driven systems at the bedside will reduce errors without actually reducing adverse drug events.
• It is possible for automation to increase medication errors. Dr. Classen said a new pharmacy robot at LDS Hospital somehow misfired, and 273 patients received the wrong medication.
• A study under way at Stanford and several other California hospitals is helping define traits of what he called "high reliability organizations" (basically, safe hospitals).

The most commonly observed characteristic of less safe hospitals: a relentless pressure to produce. "Nothing had a greater impact," he said.

Back to the top of this file.  Back to last location

VI. Our Observations

Summary: Is Your Principal Goal To Reduce Errors Or Costs And Resource Use?

CPOE can be principally a tool to reduce errors. Or, it can be principally a tool to reduce physicians' consumption of resources. Or, it can be a bit of both. However, control over physician use of resources can also be a political battleground -- an "HMO through the physician entrance"-- to the hospital. The authors argue below that the principal reason that the large industrial entities are expressing interest in CPOE via the Leapfrog Group is not to improve care, but to seek to standardize physician actions for the purpose of reducing resource use and costs. This is not necessarily an undesirable purpose. However, ground on which you, as a caregiver implementing software which these businesses will seek to appropriate as a tool of their goals, should tread with eyes wide open.

Discussion: CPOE Is The Common Tool Of Two Groups With Divergent Agendas

Two divergent core reasons exist for implementing CPOE. You -- that is, the medical and bean-counting managements of your care-providing entity--must decide which is/are yours.

As stated in our introduction above, the Institute of Medicine report, "To Err is Human," and the Leapfrog Group's push for CPOE are widely cited as mutually consistent, dovetailing pressures to implement CPOE. It’s our view that this is seriously faulty logic. Let us explain.

First, IOM and "To Err Is Human"...

The stated mission of IOM is to “advance and disseminate scientific knowledge to improve human health.” That is, to use science to improve medical care. "To Err Is Human" seeks to do so by reducing medical errors.

However, error reduction is only a partial strategy if the core purpose is to improve care. Vastly more care fails the sick and maimed because it is not given than fails because it is given improperly. Most of us understand intuitively that this is so. In fact, the IOM now seems to be attempting to re-steer the debate away from CPOE as a panacea in a new publication, “The Future of the Public's Health in the 21st Century,” announced Nov. 11, 2002. “The report calls for the federal government to lead a national effort to examine options and develop plans for making comprehensive and affordable health insurance available to everyone,” according to a National Academy of Sciences announcement about it.

It also intuitively seems likely that more adverse treatment outcomes arise in ways unlikely to be prevented by CPOE with decision support than occur in ways which are headed off by CPOE. For example, a patient the authors know had a tiny tumor which went undiscovered, apparently because the reading physician was just too busy to notice or to explore an anomalous spot on an MRI. In that same patient, different physicians failed to spot a cancer metastasis–in part because diagnostic activity for cancer is fragmented, and no one compared a tiny anomalous glow-spot on a PET scan image with the position of a lymph node on a CAT scan. These are two sets of real-life examples with which the authors of this report are closely familiar, and which would not be headed off by a CPOE system.

In another example, the final writing of this report was interrupted in October, 2002–first for a week by an injury to one of the authors, and then for another week by a series of treatment bungles. These adverse outcomes (“mistakes,” to the mere layman) do not appear anywhere in the medical records of the treating institution from whence they arose, but their repair cost more than another full week for a series of outpatient visits at a second institution (where, obviously, a medical record was created, but with no cause-and-effect link to the errors which made them necessary).

Thus, if there is one core activity needed to reduce medical errors, it is that of judiciously spending more money on health care on an ongoing basis. To reduce adverse outcomes, that money would likely best be spent on improved access to care, better diagnostic equipment, more intensive training of residents, more peer review and internal quality-assurance watchdogs, and above all, more time for clinicians to do their jobs well without being rushed. CPOE, while certainly useful, logically falls quite low on the list as a strategy if the objective is to improve care.

The fact is that CPOE is higher in priority as a tactic to “reduce medical errors” (which is just a subset of the larger strategy of improving care) because is offers that old (and as often discredited as not) computer argument: it shows the potential to be the most cost-effective, rather than the most actually effective, tactic for reducing adverse outcomes.

Second, Leapfrog...

We doubt that this goal, or course of action–broadly improving care--is within the marching-orders of the Leapfrog Group. The sponsors of Leapfrog are industrial organizations which who have a lengthy, unbroken, and, while quite legitimate, sometimes far too effective track record of seeking to reduce their health insurance costs. (Granted, to the extent that they are economically motivated, they are also motivated by a desire to reduce lost employee work days arising from medical errors. However, as in the cited situation involving one of the authors of this report–fixing medical errors is also usually an additional insurance-paid expense, and therefore comes directly out of the employer’s pocket in the end, where “lost productivity” due to absenteeism is one of those theoretical numbers pulled out of the air for its own political reasons.)

Put bluntly, the larger motive of the industrialists behind Leapfrog is to reduce consumption of health care resources, not increase their use, which would be needed to turn around the decade-plus-long decline into "sicker and quicker" discharges of hospital patients.

We do not mean to suggest that there is anything inherently unethical about these cost-cutting motives. Rather, we think that as one implements CPOE, it is extremely important to understand them for what they are because caregiver organizations should use CPOE to adopt some resource-saving strategies.

However, CPOE, in the hands of an overly money-focused management, can take treatment decisions too far down that path. If it is fair to characterize the worst bottom-feeders among HMOs (and other managed-care entities) as those which cut expenditures and maximize net income as the principal goal, and which elaborate medical justifications for reducing resource use as the path thereto, then a hospital management which over-uses CPOE to reduce physician consumption of resources can be characterized as having sneaked its own HMO in through the physician’s entrance.

In other words, a clinical system with CPOE can either focus solely on improving care, or it can focus principally on limiting resources. Or, third, it can provide half a loaf to each camp by catching the mundane medical errors and speeding turnaround, but at the same time strait-jacketing some physician diagnostic and treatment decisions with built-in decision trees. A clinical system with CPOE cannot take you all the way down both paths--improving care and limiting resource consumption--at once.

Back to the top of this file.  Back to last location

Whose hand will be on the steering wheel? Yours? Washington? Leapfrog? The HMO?

If a clinical system with CPOE and decision support can be used to limit resources, a second question is: whose hand is on the steering wheel?

We think the sponsors of Leapfrog view CPOE as a version of industrial Enterprise Resource Planning (ERP) software. Reducing errors does reduce costs--but reducing consumption of resources in the course of production, in a standardized manner, on a day-in, day-out basis--that is, requiring a hospital staff to behave like the production line of an industrial organization -- has the larger cost-reduction potential.

CIO Magazine describes ERP as --

"a sort of battering ram for improving the way your company takes a customer order and processes it into an invoice and revenue--otherwise known as the order fulfillment process. [ERP] takes a customer order and provides a software road map for automating the different steps along the path to fulfilling it.... With luck, the order process moves like a bolt of lightning through the organization, and customers get their orders faster and with fewer errors than before." (http://www.cio.com/research/erp/edit/erpbasics.html)

Note the terms at the core of the ERP process: "order," "customer," and “battering ram.”

In the case of a care provider, the identity of the "customer" is, obviously, a source of debate and confusion. The physician may be the "customer" in that it is the physician who decides what to order, places the order, incurs the cost, and demands fulfillment. However, it is the need (or perceived need) of the patient which is being fulfilled (or not fulfilled). As such, the patient is, in the most real sense, the "customer." However, a synonym for “customer” is the term, “paying customer.” Therefore, a third and even a fourth real "customer" are in the equation–the payer (an insurance company, Medicare, Medicaid. CHAMPUS, an HMO, etc.) and the industrial corporation which pays for all or part of its employees' health insurance. They are real "customers" of health care to the extent that it is from their pockets from which the principal payment arrives. This last group of "customers" is Leapfrog's constituency.

In non-HMO (and pre-CPOE) medical practice, more of the power of “customer” decisions lies with the patient and physician. The “paying customer” has a small say through pre-authorization and post-decision utilization review in the background, which modify the physician's ordering practices. In managed-care situations, the HMO is vastly more aggressive as the paying "paying customer," and influences utilization with the now largely-discredited practice of waving dollars in the face of the physician or his/her management for choosing to use fewer resources.

CPOE, when coupled with decision support, offers the opportunity for exactly the kind of programmed, cost-driven channeling of decisions that the sponsors of Leapfrog have always had as their principal health care agenda.

This use of software to guide, pressure, or force physicians into resource-conscious decision paths is, as stated, already happening in CPOE systems. It is prominent in Vanderbilt University Medical Center's Whiz system. Other academic medical centers and community hospitals with residency programs have also cited resource consumption decisions being among the core functions of their CPOE systems. Two of the respondent medical centers in the survey section of this report allude to greater control over physician-ordered resource consumption, one citing "clinical practice guideline enforcement" as a benefit, the other referring to an "increase in IV to PO (intravenous to patient oral medication) conversions," which can save significant amounts of money and reduce the risks of IV-caused infections.

In addition, we reported last year on presentations by John Halamka, M.D., CIO of CareGroup Healthcare System, a large Boston, Mass.-area physician group practice, who said that a software system which physicians use in a manner similar to a CPOE (and which could likely be integrated with CPOE) saves CareGroup $10 million a year by heading off referrals to physicians outside the group practice.

That system, which CareGroup calls Service Tracking and Referral System (STARS), is a CareGroup add-on to the New England Health EDI Network (NEHEN), a network which allows its members to send HIPAA-compliant electronic eligibility verifications, claim status inquiries, referrals, authorizations, and specialty referrals. The way it works is that all 535 of CareGroup's primary care physicians use NEHEN to obtain payer approvals to refer patients. If the caregiver seeks a payer approval to refer a patient to a specialist who is not a CareGroup provider, a STARS screen pops up, warning the primary-care physician that this out-of-network referral must be justified, that it will be reviewed, and that it may be delayed. The system invites the physician to refer to a CareGroup specialist. STARS issues a similar challenge if the primary care physician requests too many specialist referrals.

A referral, while not strictly speaking a medical "order," is, in effect, another kind of physician order. Thus STARS is a "decision support" tool, interjected into the process of arranging for further clinical care for purposes including maximizing revenues.

It is our view that as CPOE systems evolve, the error-reduction functions (misunderstood handwriting, drug interactions, wrong prescriptions, etc.) and the time-saving functions (faster movement of prescriptions, tests, and faster resulting treatment) will be seen as lower-level functions of CPOE. They'll always be quite important--but they will be the small stuff. The emerging principal function of CPOE systems will be as a software system to get more bang for the health care buck by heading off uses of expensive medical resources which health care management deems to be "unnecessary" or "cost-inefficient" or "wasteful," or whatever term management applies to its decision-making.

And as that happens, the core questions are: “Where do you draw the line?” and “Just who are you–clinical management the financial management, or an outside financial or regulatory influence?” That is, a CPOE system with decision support can readily be devised as a way to manage (and justify) a lower-quality, denial-of-expensive-procedures, faster-discharge version of hospital care. A clinical system with CPOE can be a system which not only turns the hospital into an industrial production line, but builds a bullet-proof record to medically justify it.

This emerging capability of clinical information systems with CPO to make resource-use decisions will require decisions by the implementing caregiver's money and medical managers:

What do you want your clinical information system to do?

Who is the "customer" of CPOE -- the patient, the physician, or the payer, or (we hope) some careful balance of all three?

Back to the top of this file.  Back to last location