Prescribing medications for patients in a neonatal intensive care unit is tricky business, but a new database-driven, Web-based physician order entry (POE) system at Brigham and Women’s Hospital in Boston is helping doctors avoid dosing mistakes when caring for their newborn patients.

The $1.5 million POE, based on InterSystems Corporation’s Caché post-relational database, automatically calculates prescription dosages based on each baby’s weight for infants as small as 2.5 pounds.

Steve Flammini, CTO of hospital corporate parent Partners HealthCare, believes the Brigham and Women’s POE is the first in the nation to automate complex dosing schemes in a neonatal intensive care unit (NICU), and a step forward in patient care.

“There was a lot of complexity of dosing in the NICU setting,” Flammini explained. “There was quite a window of vulnerability for making errors. At the end of the day, that’s what we develop these physician-based systems for—error reduction and quality of care.”

While Partners HealthCare is still evaluating the effectiveness, and cost, of the new POE, Flammini expects a similar ROI from earlier POE versions. A previous client/server network-based POE resulted in a 17 percent reduction in adverse drug reactions throughout the hospital, he explained, and savings of up to $10 million a year. Since both POE systems were developed internally at Partners, both the network and Web-based versions of the application are based on Caché.

The new benefits
The Web-based interface has several advantages over older client/server POE models. The system can be run on a variety of computers and operating systems—basically anything with a browser.

“A client/server requires a specific operating system with specific components,” explained Flammini. “The support burden and the whole general software maintenance scheme are superior under the Web model.”

The Web-based POE could eventually be used on handheld devices—a technology quickly catching on within the medical community. It also allows its designers to easily link to medical reference materials on the Web, Flammini noted.

Under manual POE approaches, there could be a considerable time lag between the time a doctor handwrote a prescription or test request and when a pharmacist or nurse was finally able to interpret it, due to handwriting deciphering.

“If a doctor wanted you to get some tests or some medical treatment or therapy, he scribbled something on this piece of paper or checked some boxes,” said Paul Grabscheid, InterSystems’ VP of strategic planning. “Some time later, perhaps, it was entered into a computer, and perhaps, some time later, it was reviewed by another human being, who might then have a question.

“Then they try to page the doctor, who might not be in the building. It creates a real problem to get this order changed, if it ought to be changed,” Grabscheid added.

The new POE system was rolled out in early October and Partners HealthCare is already planning to expand the technology to other departments. Flammini said the POE, which is being used by about 200 nurses and 20 doctors in the NICU, is among the hospital’s most popular technical advances with the medical staff.

“The NICU was important for us architecturally because it was our first foray into the Web environment specifically for ordering,” Flammini said. “The actual physician ordering, which has to be extremely reliable, high performance, and high availability, was a more challenging app to bring to the Web. Now that we’ve succeeded with that, we’re ready to move our entire order entry base over to Web technology.”

The implementation process
Though it’s Web-based, setting up the system was no small exercise. A team of eight in-house IT developers, clinical analysts, and NICU doctors worked on the project for more than a year. Much of that time was spent working on the requirements for an automated system, Flammini said.

“NICU was largely unautomated prior to this, and it took a lot of analytical work to gather the requirements for such an advanced system,” he said. “Once the requirements were finally nailed down, the development went pretty smoothly, and it gives some credence to the adage that the requirements and specs really are the hard part.”

The hospital chose InterSystems’ Caché database because it works better for an instant feedback system, Flammini said. “For ultra-high-performance, high-availability transaction processing, which is what these clinical systems are, we use Caché,” he said. “It’s very good at super high throughput, high performance, high availability—better than the relational databases.”

He explained that  Caché works better than the two relational database products, Oracle and SQL Server, that the hospital uses for other purposes. Caché is superior for high-availability, high-scale, transaction-intensive clinical systems. The other relation technologies are used for data warehousing applications, but not patient care apps.

Three programmers developed the complex model for NICU-based dosing and a set of tools to construct templates for ordering. Much of the rest of the programming time was spent on navigation and moving the POE from a tightly coupled client/server system to the Web-based system, with the emphasis on separating presentation from services.

“The programmers spent a lot time rewriting a lot of the core application infrastructure in terms of loosely coupled services,” Flammini said. “Developing order-entry systems isn’t new to us, but developing it for the Web was a fundamental architectural shift.”

The user experience
While the technology has a bit of a learning curve, Dr. Eric C. Eichenwald, a neonatologist at Brigham and Women’s Hospital, said the convenience is worth it.

“I think it is great although it does take some getting used to,” he said. “It is very feature-loaded, so it can be difficult to find what you want. The best parts are that the nutritional management is built in—so it saves a lot of time in calculations that used to be done by hand with our old handwritten-based system.”

The hospital provided a couple of computer-based training and interactive group training sessions for the doctors and nurses using the new POE, Flammini said, although doctors generally aren’t big fans of the group setting.

“The bottom line is that physicians tend not to seek very much training, until it’s time to use the system,” he explained. “Then, they tend to want to figure it out themselves, or from their peers, and lastly, from IS support folks who are available to train them. They’re quite busy, so this is understandable. However, we find that nurses generally want to be well-trained before using the system.”

The downside of the new system, Eichenwald said, is that it takes more time up front, with doctors having to sit down at a computer instead of quickly handwriting a note. “We are hoping that this will get better as we get used to the system,” he added.

InterSystems’ Grabscheid said the project meets several client needs.

“They [the medical organization] recognized they needed to do two things: One, they needed to build a system that was quite smart at the beginning and got smarter month by month in terms of feedback on these orders, and the other is it had to respond really quickly,” he said. Though it’s not a “learning system,” per se, it’s a system with clinical knowledge and rules bases embedded in it. The embedded clinical knowledge represents medical practice policy developed by medical leadership at the institution.

“If they were going to get the physician to sit in front of the computer and enter it himself or herself, which the physician didn’t want to do, and then have to wait what seemed like a long time to get a response, physicians just weren’t going to do it.”