As a medical resident at the Hospital of the University of Pennsylvania (UPenn), Dr. Scott Halpern spent hours at his hospital\u2019s computer terminals searching for the right tests and medications for his patients. But Halpern would often become so frustrated with the system\u2014which was slow and required specific language for each request\u2014that he would give up and stop using the system when he could find another way to take care of the patient. Halpern is much happier with his hospital\u2019s newer, more user-friendly order-entry system. But he still sees problems. One of the biggest, he says, is the annoying alerts that constantly pop up onscreen as he orders a patient\u2019s dosage. "I honestly haven\u2019t paid attention to a pop-up alert in years," says Halpern, who like many doctors believes that alerts should be limited only to those that might help avoid a serious medical error. "I just click right through them as quickly as possible and I think most doctors do the same thing," he says. Those pesky alerts were designed to prevent medication errors, but because they pop up so often, many are ignored. Halpern\u2019s frustration with such a poorly designed feature reflects an ongoing struggle with computerized drug-order systems at hospitals across the country. Computerized physician order-entry, commonly known as CPOE, holds great promise to improve patient safety as it radically changes the way that doctors, nurses and hospital employees do their jobs. CPOE is still in its early phases\u2014only 4 percent of U.S. hospitals are using the systems according to consultancy Klas Enterprises\u2014but research shows it can improve patient safety. Studies at Brigham and Women\u2019s Hospital, where informatics leaders developed their own system in the 1990s, revealed that CPOE cut medication errors by 80 percent. And nationwide adoption of CPOE could save $44 billion a year in reduced costs from radiology, laboratory and medication errors, according to a study by the Center for Information Technology Leadership. Doctors such as Halpern agree the new systems are superior to the pen-and-paper method of prescribing medication, which can lead to misunderstandings and transcription errors. But early experiences with CPOE show that success involves much more than plugging in the software. Those at the forefront of CPOE adoption agree that systems are expensive and difficult to implement in hospital environments. And a recent study performed at UPenn raised alarm by claiming that CPOE can actually increase the potential for medical errors. Indeed, experts agree that CPOE can introduce new risks if not designed and implemented correctly, or does not fit in smoothly with a hospital\u2019s particular "work flow." CIOs ready to invest millions of dollars in CPOE need to make sure that IT staff work closely with their medical counterparts to design the systems and provide extensive training for those who will use it. They should also partner with their vendor to customize the system for their own specific needs. And CIOs need to monitor the CPOE closely to make sure that glitches are fixed before they can cause unexpected medication errors. CPOE projects are like many tricky enterprisewide implementations, and CIOs from health care and other industries can learn important lessons about change management from hospitals that have gotten CPOE right, including Brigham and Women\u2019s, Duke University Medical Center and Health System, Intermountain Health Care in Utah, and St. Joseph Health System in Orange County, Calif. "The simple truth is that CPOE is not a turnkey solution," says Brian Strom, chair of the department of biostatistics and epidemiology at the UPenn medical school. "Getting it right takes a tremendous amount of monitoring and tailoring. No one expected Word 1.0 to be perfect, so it\u2019s not surprising that CPOE 1.0 isn\u2019t perfect either."To Err Doesn\u2019t Have to Be Human Until the turn of the 21st century, only a few major hospitals had attempted to design and implement CPOE systems. But then in November 1999, the Institute of Medicine published a report that galvanized the nation\u2019s medical community. The report, "To Err Is Human," put forth some disturbing figures. Avoidable medical mistakes kill anywhere from 44,000 to 98,000 people a year\u2014more than breast cancer, highway accidents or AIDS. The report also said that more than 7,000 deaths are caused by medication errors. CPOE quickly rose to the forefront of health IT systems as a promising means of preventing medical errors. Large vendors such as Cerner and McKesson developed and updated CPOE systems, and hospitals large and small began to adopt the new technology. The road has not always been smooth for those adopting the systems, however. In January 2003, in a high-profile case that spooked the medical world, doctors at Cedars-Sinai Medical Center in Los Angeles, unhappy with the extra time it took them to enter orders on the computer, staged a rebellion and forced the hospital to shelve the CPOE portion of a $32 million implementation project after three months of use. Then, in March of this year, researchers at UPenn published a study in the Journal of the American Medical Association (JAMA) documenting that their CPOE system, an early model from Eclipsys, could cause 22 types of medication error risks. The researchers, led by Ross Koppel, a sociology professor at the UPenn School of Medicine, surveyed staff using the CPOE and shadowed doctors, nurses and IT staff to see how they used the system. One risk discussed in the study was the difficulty staff had in discerning which patients the doctor was ordering for, because the CPOE display was fragmented and required switching between multiple screens. Such clunky features increased the risk for faulty orders on a regular basis, the study found. The UPenn study generated heated discussion and dozens of editorials in medical journals. Critics castigated the study for failing to compare data from the pen-and-paper era and for focusing on what many consider an out-of-date system that is not representative of systems widely in use today. But CPOE experts say that, nonetheless, the study highlighted some important truths about the difficulty of implementing CPOE. And the UPenn researchers say that more recent studies on a newer system show that while the doctors and nurses prefer it to the former system, the new system has not solved all the original problems that could lead to medication errors. Most notably, the structure and format of the screens can still cause medical staff to prescribe drugs for the wrong patient, the researchers say. All of this debate comes as hospitals are facing increasing pressure from the government and patient safety groups to install systems such as CPOE and electronic medical records (EMRs), which replace paper files by placing medical and patient records on a computerized system. The U.S. Medicare system recently announced plans to offer lower-priced medical records software to doctors. (See "Feds Sweeten the Pot" on Page 80.) So far, slow adoption of CPOE and EMRs has been attributed to lack of funding and physician resistance. The difficult experience at Cedars-Sinai highlights the scope and amount of preparation needed to implement large-scale clinical systems. A move to CPOE can have a serious impact on the way that doctors and nurses do their work, and those hospitals that have been successful with CPOE have worked to avoid the kinds of disruptions that have made clinicians shy away from using the systems. "CPOE will be the biggest clinical change initiative most hospitals have ever undertaken," says Dr. David Classen, who helped design the CPOE system at Intermountain Health Care in Salt Lake City and is now a consultant on CPOE to The Leapfrog Group, a Washington, D.C.-based health-care consortium.How to Get It Right the First Time Hospitals going through this massive change need to first focus on how doctors do their work to avoid inconveniencing them with additional tasks, says Asif Ahmad, CIO of Duke University Medical Center and Health System. Ahmad, who implemented CPOE at Ohio State University before coming to Duke, says the key to success lies in working closely with the doctors and nurses to help design the system, even if using vendor-built software. At Duke, Ahmad cochaired the CPOE task force with the hospital\u2019s chief medical officer, and within his IT department, there are six physicians as well as 30 nurses and four pharmacists reporting to him. "It\u2019s imperative that CIOs in hospitals build technical and clinical teams," he says. "People responsible for building or overseeing CPOE systems need to understand the clinical system." Although Duke bought its CPOE system from McKesson, doctors helped customize it to fit in with their way of doing things. When a doctor or nurse clicks a button to admit a cardiac patient, for example, the system shows them tasks and treatments such as administering aspirin. "If physicians help design CPOE, and it\u2019s done the right way, it\u2019s not going to cause errors," Ahmad says. Duke\u2019s CPOE system, which went fully live on all 590 patient beds as of June 20, includes 259 "order sets," specific instructions for standard medical procedures that were jointly developed by physicians. "The idea was to get all the cardiologists in one room and tell them they can\u2019t leave until we agree on a set of directions on how to manage patients," he says. "Sometimes there are personal preferences but we have to agree to make the system work." During the first nine months of the process, he held regular meetings with doctors and nurses to map workflow for each speciality and unit of the hospital. As a result of these meetings and close consultation with medical staff, the team of doctors and IT staff were able to avoid design problems they viewed as potentially harmful to patient safety. For example, the design team made sure that information on a single patient remains on the same page or screen and staff cannot swing to another patient\u2019s page until the orders have been completed for the initial patient. "If you display more than one patient on the same screen, you have a greater potential for error," Ahmad says. Ahmad also says doctor input pushed the IT team to create a "single sign-on," which allows physicians to look at test results on one system and then move into the CPOE system without having to log out and log in again. And in order to combat "alert fatigue," Ahmad says Duke is working on building a corollary system that will send most alerts outside of CPOE. Then, doctors will be able to do their work without the nuisance of constant pop-ups telling them about interactions that are not dangerous to the patient. A select group of pharmacists, doctors and nurses monitoring this corollary system will be able to respond to serious alerts.Big Bang Versus Gradual Approach Benjamin R. Williams, CIO and senior VP of strategic innovation at St. Joseph Health System, an hour south from L.A., agrees that a close relationship with the medical staff is key to CPOE success. Williams watched closely as doctors rejected the system at nearby Cedars-Sinai and has carefully structured the CPOE initiative at his own network of community hospitals to involve doctors from the start. He designated champions of the system among the medical staff to convince more reluctant doctors to use the system. He also decided to phase the CPOE rollout over a period of several years and encourage doctors to use it, rather than force them to do so all at once, which is what Cedars-Sinai did. Phasing it in has also allowed his team to make fixes and respond to doctors\u2019 concerns about the system\u2019s design. Dr. Thomas Hughes, an ob-gyn specialist at St. Jude Hospital, a part of the St. Joseph Health System, has served as a kind of liaison between doctors and IT staff. At one point, Hughes brought together a team of 41 doctors to help design the system, working with vendor Meditech. At St. Jude, 40 percent of admissions are now done using the system that includes CPOE, and about 60 doctors out of 280 are using the system regularly. A larger number use the system to check lab results, and Hughes hopes that he and his team will gradually be able to convince the rest to use computerized drug-ordering. Hughes said that by bringing in doctors who showed interest in using the system first, the hospital has been able to get started in a gradual and more positive way. Then, as the first enthusiastic doctors get used to the system, they will, he hopes, be able to convince more reticent ones to join them. While some hospitals, including Duke, have been successful with quick, "big bang" types of rollouts, many CPOE veterans say that approach is generally more risky than a gradual implementation. While installing the system one hospital unit at a time can create confusion\u2014especially if doctors and patients are moving between floors and wards that use CPOE and those that don\u2019t\u2014most people who have experienced a CPOE implementation advise hospitals to take it slowly. "The gradual approach can be painful," says John Glaser, CIO at Partners HealthCare System, which includes Brigham and Women\u2019s in Boston. "But given the risks, I\u2019d rather give myself the opportunity to pause if there are problems." To encourage doctors to use the system and to avoid inadvertent medication errors, Hughes suggests holding regular meetings and listening closely to doctor complaints. For example, the Meditech system in use at St. Jude\u2019s has nine levels of warnings on drug interaction. If all nine are activated on the system, everything a doctor orders will lead to a series of time-consuming pop-ups that are likely to be ignored. After consulting doctors, the hospital decided to activate only the two highest levels of warnings so that those using the system will read the pop-ups and possibly avoid making serious errors. Doctors also convinced the IT team and Meditech to change the system to remove a function that forced them to "deselect" items and tests they did not want to perform. With this modification, the doctors are now required to select only what they want to order, saving considerable time. Looking back at Cedars-Sinai\u2019s experience, Dr. M. Michael Shabot, Cedars-Sinai\u2019s medical director of enterprise information services, says that, at the time, there was little awareness of the difficulties involved in a large CPOE implementation. "Given the money and time that is spent on this system, it has to be meaningful to the physician," says Shabot. "The reason for it has to be very clear."A Technology in Its Infancy When the UPenn researchers published their CPOE study in JAMA, Williams was in the midst of his complicated CPOE project at St. Jude\u2019s. Williams acknowledges that after reading the JAMA study, he recognized some possible shortcomings in his own CPOE system. As a result, he recently requested the Meditech system provide a link between procedures and medication orders so that if the procedure (such as open-heart surgery) is canceled, the correlating medication order would automatically be canceled as well. In addition, his team is working closely with Meditech to improve integration of the pharmacy and physician order management to make sure that the orders are going through in real-time, without delay. In a written response to the JAMA article, circulated among his staff, Williams wrote: "The safety and effectiveness of CPOE arises from the understanding that it is primarily a process change, a major transformation in the procedures surrounding the delivery of patient care." Even those who question the study\u2019s methodology agree that it highlighted the stress that a CPOE rollout can put on a hospital and the need to constantly monitor and fix the systems. After implementing a homegrown system at Brigham and Women\u2019s, Dr. David Bates, chief of the division of general medicine, wrote in the Journal of Biomedical Informatics, "We routinely tracked errors and problems that were created, and made thousands of changes to the original application." If he had one thing to do over, Bates says he would have devoted even more resources to this area. "It is just impossible to get it all right at the outset, because the processes involved are so complex," he adds. And some say it is harder to make all of the necessary changes if you\u2019re installing a system designed by a vendor. Bates says, and other practitioners agree, that just as with any new technology, CPOE must be aggressively monitored and tinkered with if it is going to succeed. The Leapfrog Group\u2019s Classen is leading a project to create what he calls a "CPOE flight simulator," which hospitals and outpatient clinics will be able to use to test their CPOE and decision support systems to gauge whether they "prevent or cause harm" among patients. "What\u2019s clear is that not all medication errors are created equal," Classen says. "We need to find the ones that cause harm or death." CPOE experts hope such new tools will be able to test the effectiveness of the drug-ordering systems and help hospitals make them as foolproof as possible. "Saying that \u2019CPOE causes medication errors\u2019 is like saying that \u2019cars cause accidents,\u2019" Bates writes. "Of course CPOE can cause medication errors. Clearly we should strive to make it work better. Accident rates were high with cars early after their introduction, but few today want to go back to the horse and buggy."