The long and winding road to computerized physician order entry: massive, 8-year project to bring CPOE to Thomas Jefferson University Hospital is nearly complete.

Founded in 1825, Thomas Jefferson University Hospital is a 932-bed multi-facility hospital located in downtown Philadelphia that admits more than 40,000 patients a year and cares for over 80,000 patients annually in its emergency department.

Throughout the 1970s and '80s Jefferson developed--largely in-house--a variety of information systems that remained in the background from the clinician's perspective. Paper charts, written orders, printed laboratory reports and consultations remained virtually unchanged over the preceding decades.

By 1996, Thomas Jefferson University (TJU) and Thomas Jefferson University Hospital (TJUH) reached a watershed in their approach to information systems. A steering committee with broad representation across the hospital and university recommended scrapping the home-grown custom designed systems that supported admission, discharge and transfer (ADT), laboratory results reporting and other functions, and replacing it with an off-the-shelf commercial system.

Because the hospital had never installed a computer order system of any kind, the committee also recommended taking the bold leap of bypassing a generic order system and implementing a computerized physician order entry system (CPOE).

Furthermore, they believed strongly from the outset that in order to reap the benefits of this approach, CPOE would have to be mandatory for every patient and every physician. Eight years later, Jefferson is in the final months of realizing this goal.

Why CPOE?

In the early 1990s, the great information technology juggernaut was in full swing. Daily newspapers published charts showing the geometric rise of Web sites, which in turn boasted of their volume of "hits." Microsoft launched an aggressive and comprehensive strategy that encompassed operating systems, business applications, Internet browsers and more. E-mail was moving from an office-based to a global entity with the advent of AOL and the iconic "you've got mail" reaching deep into popular culture.

Physicians were, if not all early adopters, at least conversant with information technology. On one level, physician innovators had helped develop computed tomography, computer-assisted interpretation of electrocardiograms and pulmonary function tests and other applications. In general, physicians had completely abandoned manual searching through the index medicus in favor of "MedLine," "Grateful Med" and other online programs. They demonstrated that if information technology were convenient and helpful, they would use it

At Jefferson University Hospital, the aging mainframe systems that had so well supported the business functions of the hospital for so long were at the end of their useful life expectancy, and the institution was at a crossroads with respect to information technology.

One path led to the replacement of current functionality, perhaps with client-server architecture to supplant older technology. A second approach involved changing the model of hospital information systems at Jefferson from a primarily financial, administrative and clerical one to one that put clinical information and processes first.

Perhaps the weight that tipped the scales was the admonition from the information technology consultant (engaged by administrative leadership to guide the institution through the process): that physicians, through the order sheet, controlled up to 70 percent of the cost of medical care.

Information systems provided, for the first time in history, a potential to intervene, standardize, direct and constrain the physician order on an institutional scale. If a clinical information system could also offer the physician a perceived benefit in patient care and practice, then there would be a double benefit: relationships with the community physicians could be strengthened without incurring the cost of purchasing practices, and unnecessarily costly ordering behaviors could be curtailed.

These were the two driving forces that led the institution toward a potentially risky undertaking, one that had a tarnished and inconsistent record of success: the complete transition from paper to electronic ordering.

There was a third consideration: Jefferson's success could provide a competitive advantage.

Laying the groundwork

Clearly, strong leadership for the project from the CEO of the hospital and the dean of the medical college was an absolute prerequisite for success. They recognized that this was a huge project that required reassignment of major responsibilities, one that would not succeed with the usual project plan and assignment to an implementation team. Existing staff needed to devote the majority of their effort over several years to the project, and new employees with novel skill sets would need to be recruited.

From the outset, the ground rules were clear. Far beyond simple automation of existing procedures, the mandate was to assess the current state of clinical processes (registration, results reporting, patient education, etc.) and then utilize information technology to support their redesign.

[ILLUSTRATION OMITTED]

The CEO and the dean signaled the importance of the undertaking and their commitment to its success by initiating several visible and decisive actions.

1. A large information technology consulting firm was engaged to guide the entire selection process, demonstrating an institutional commitment of resources.

2. Several half-day and all-day "retreats"--attended by physicians, administrators, nurses and other clinicians, information systems (IS) personnel and department representatives--sent a clear signal that this was not "just another project."

3. The appointment of a steering "triumvirate" including the university vice president for human resources (representing business and administrative functions), the hospital vice president for planning (representing clinical administration) and the associate dean for academic affairs (representing the physician faculty) established accountability and responsibility at a senior level.

4. Finally, capital and operating placeholders in the budget were communicated early in the project.

Over a six-month period, the process included:

* Creation of a vision statement

* Redefinition of existing hospital processes (such as scheduling) to accommodate the vision

* Definition of the functional requirements by the working group

* Request for proposals and evaluation by the working group including financial evaluation of vendors

* Invitation to selected vendors to present to the working group

* Prioritization and subsequent site visits by members of the working group

* Final prioritization and contract negotiations

One of the results that emerged from the multidisciplinary working groups was an explicit articulation of the functional requirements at a high level. Seven years later we have achieved some of these goals, while others still elude us.

Other strategic principles included:

1. Buy rather than build

2. Stay on the leading but not "bleeding" edge of technology

3. Partner with a vendor that had demonstrated success in a complex academic medical environment

Five systems emerged as potential candidates and PHAMIS Lastword (later acquired by IDX) was chosen as the vendor. Site visits at Lastword user sites allowed a multidisciplinary Jefferson team to gain a clearer picture of real-world limitations and advantages of the vendor's product.

In the midst of this process, two momentous local events occurred: TJUH separated from the University and joined the Jefferson Health System, and TJUH merged with Methodist Hospital. Both required reevaluation of decisions because other members of the merged entities had relationships with other information system vendors. Part of the process included providing justification for the decision to select Phamis.

Following contract signing in 1996, the implementation plan became a reality. The first phase of the project was to increase clinician interaction with the system by providing widespread access to results reporting for the physicians, and employing the system as the "front end" for patient registration and ADT functions.

A large and complex implementation team encompassing six projects was created to oversee this phase (Figure 1) including 34 people working as project leaders, systems analysts, technical analysts and client leaders. In addition, physician and other advisory workgroups regularly reviewed the progress and guided the output of the six projects.

On one night in January 1999, the major reporting and business functions of the main frame system were retired and TJUH went live with Lastword on approximately 1,000 new devices throughout the entire hospital for results reporting. This implementation required network, server and other planning by information systems and training of the entire staff of TJUH using Lastword for admission, transfer and discharge, registration, medical records, results and pharmacy. Nearly every physician, nurse and other clinician was affected.

Although physician training initially was conceived as a structured supervised session, it soon became clear that physicians wanted a brief orientation, a reference, and access to immediate problem resolution.

Because one of the deciding factors in vendor selection was a transparent user interface, we adopted this "minimalist" approach. Laminated help cards describing task-oriented instructions were attached to each workstation, a Web site was created for more detailed instruction and optional hands-on classes were held in the early mornings and late afternoons adjacent to the dining facility.

For the vast majority of physicians, this approach to training was a success. For the rare physician who desired even more specific training, the IS department scheduled one-on-one sessions.

Recognizing that the move to physician order entry would require an additional year of planning, early implementation of this "results" phase allowed the users to become familiar with navigating the interface while the basic physician and other clinical workflow remained unchanged.

The final push to CPOE

From the outset it was clear that CPOE would challenge almost all of our everyday routines in patient care. Even a superficial attempt to document existing workflow revealed the imprudence of automating existing processes.

For example, nearly every order under the existing paper system flowed through the nurse or unit clerk, who would then send a paper copy to an appropriate department, schedule a test using the telephone or enter a nursing care order into the "Cardex."

CPOE permitted the instantaneous transmission of orders to the receiving department, bypassing the unit clerk and potentially speeding delivery of care. To accommodate the formidable task of redesigning care to achieve efficiency and safety, we created a workflow group that met biweekly for most of a year; its membership, included a social worker, physician, nurse, pharmacist, hospital administrator, business manager and others.

Current departmental workflow was documented and posted on a Web site, and then the anticipated new workflow with CPOE was presented to the group. After each department went through this process, the group looked at a series of scenarios that described a hypothetical patient moving through a hospital stay that crossed departments and settings.

[FIGURE 1 OMITTED]

From the beginning, we were careful not to "oversell" CPOE. Nothing is faster or more convenient than writing an order on a piece of paper, handing it to a clerk and saying, "Do this," However, in addition to the well-recognized safety issues associated with paper orders, there are many tangible benefits to the physician as well.

For example, orders can now be entered from an office, an operating room or any nursing station without having to spend time searching for a patient chart. Hospital-approved order sets provide comprehensive, effective access to high quality care consistent with national guidelines and local standards.

Finally, the automatic generation of rounding lists, access to a longitudinal record of results and automatic allergy and drug interaction checking all improve the physician's efficiency and effectiveness.

We decided that a pilot must precede more general implementation, so we faced a critical question: should we implement on a nursing unit basis or a department basis?

The advantage of a unit basis included the nursing staff living only in a CPOE world, without the need to switch between computer and paper chart. The disadvantage was that departments such as pharmacy, dietary, radiology and others would need to live in two worlds, as they would continue to receive paper orders from existing units and computer orders from CPOE units.

We decided that, primarily for patient safety reasons, we would implement on a unit basis. Similarly, we decided to make CPOE mandatory on a given unit, so that the nursing staff would only need to utilize computer order. To accomplish this, we removed all paper order sheets from the CPOE unit.

Another working group provided critical input to the CPOE process: the clinical design team, composed of attending physicians and house staff from a variety of specialties, nursing, pharmacy and other clinicians. This group vetted, edited and approved screen designs as information systems constructed them, effectively anticipating physician resistance to inelegant or objectionable designs.

Unlike the approach to education for conversion to Lastword results, the seriousness of potential adverse consequences dictated a more rigorous educational program for CPOE. Physicians were required to attend a formal 90-minute training session and to demonstrate competence using a computer-aided instruction tool before receiving password authorization to use CPOE.

Recognizing that the inevitable bugs and slips would be a source of difficulty for our physicians, we planned to implement nursing functions associated with CPOE (medication administration record, verbal orders, flowsheet charting and others) three weeks before CPOE to allow nurses to become accustomed to the system; we hoped that the majority of frustration would be limited to the physician staff.

Every nurse received up to eight hours of targeted instruction, generating the expense associated with more than 8,000 nursing hours of effort. House staff were required to attend training sessions, and new house staff were trained as part of their initial orientation.

A general medical unit was the first pilot unit, which began nursing flowsheet documentation and CPOE in September 2001. House staff from medicine and family medicine admitted patients with a wide variety of medical conditions to the unit. We reasoned that this would be a worst case scenario with which to test the robustness of our planning.

In light of the additional stress physicians faced in learning a new way to care for patients, we wanted to avoid frustration associated with waiting for devices. To address this, we provided several "excess" wireless portable systems in additional to the fixed devices that could later be redeployed to other areas with minimal cost.

After three months of fine-tuning on the pilot unit, two other units quickly followed. Eight months later in a single conversion that encompassed an additional 227 beds across all clinical departments, every non-monitored bed was converted to CPOE by November 2002.

Currently, CPOE is the only option for order writing on 16 nursing units encompassing 350 beds, and in two post-anesthesia care units. The first critical care unit went live in January 2004, with the remainder of the hospital's 757 beds scheduled for June of this year.

Anticipated and unanticipated benefits were realized, including the elimination of handwriting as a source of potential error and computerized checking of drug allergies, drug interactions and appropriate dosing ranges.

The creation of order sets permitted not only efficient ordering of complex sequences of tests, but also facilitated best practice using evidence based medicine to select the most effective therapies.

Time delays between ordering and other processes (such as scheduling or performing tests) were minimized through the elimination of handoffs. For example, the order for a chest X-ray now is received in the radiology department instantaneously. Copying the order sheet, carrying it to the radiology department and manually entering it into the radiology information system have been eliminated.

In addition, the system requires the ordering physician to include specific information, such as the ICD-9 code associated with the indication for ordering, and the creatinine value for radiologic tests with intravenous contrast. Both of these features decrease the need for calls back to the ordering physician, and improve efficiency as well as safety.

An internal study found near elimination of certain error-correcting pharmacist intervention (for example, wrong doses of medications), but an increase in interventions needed to address duplicate orders. It is anticipated that this problem will be minimized when residents become more familiar with the CPOE system.

The near future holds the promise of additional functionality, such as outpatient problem lists, automated discharge summaries and electronic prescription writing. By this June, we anticipate that every order written by physicians at Jefferson will be computerized.

As systems become increasingly easy to use and powerful, there will undoubtedly occur a revolution in medical information systems as CPOE is widely adopted. The inherent complexities of medicine, the high costs of and scant resources for implementing CPOE, the lack of standardization of CPOE systems and the "tradition" of selecting and educating physicians delayed CPOE's arrival.

IN THIS ARTICLE ...

Examine the detailed planning and training involved with making the switch to CPOE, and see how one major hospital successfully made the conversion.

Additional Resources

Bates DW, Gawande AA. "Patient Safety: Improving Safety with Information Technology." N Engl J Med 2003, 348:2526-2534

Dexter PR, Perkins S, Overhage JM, Maharry K, Kohler RB, McDonald CJ. "A Computerized Reminder System to Increase the Use of Preventive Care for Hospitalized Patients." N Engl J Med 2001, 345:965-970

Senholzi C and Gottlieb J. "Pharmacist Interventions after Implementation of Computerized Prescriber Order Entry." American Journal of Health-System Pharmacy. 2003,60:1880-2.

RELATED ARTICLE: High-level functional requirements articulated in 1996

Complete patient care information whenever/wherever patients receive care

[ILLUSTRATION OMITTED]

Cost and clinical information to be coordinated to optimize care

Electronic databases to identify unintended variation, desirable outcomes and best practice

Ability to link providers, settings, and patients

Data structured to be patient-centered across settings, rather than encounter-based

Patients having common registration across settings

Universal access to data for monitoring quality and resource use

RELATED ARTICLE: Critical decisions that contributed to success

Mandatory CPOE for all patients on a CPOE unit

[ILLUSTRATION OMITTED]

Redesign of all workflows associated with physician orders

Creation of a team for screen building with significant physician involvement

Implementation of nursing CPOE functions preceded physician functions

Task-oriented mandatory physician training

24 by 7 support for 3 weeks following go-live

Provision of adequate devices to prevent frustration from access issues

Forthright promises of benefits and disclosure of disadvantages

Jonathan E. Gottlieb, MD is senior vice-president for clinical affairs and chief medical officer at Thomas Jefferson University Hospitals, Inc. in Philadelphia, Pa. He can be reached by phone at 215-955-8433 and by e-mail at [e-mail]

[ILLUSTRATION OMITTED]

By Jonathan Gottlieb, MD