To enhance safety during the drug ordering process, Vinay Vaidya, M.D., CMIO, Kelly Basfield, director of clinical applications, and the clinical applications team at Phoenix Children's Hospital undertook a detailed and targeted project to build a robust pediatric dose range checking solution within their electronic health record (EHR). Because of the strategy, goals, breadth, and impact of the “Strive for Zero Prescription Errors” initiative, it has been recognized as the second-place winner of the IT Innovation Advocate Award, jointly sponsored by Healthcare Informatics and the Association of Medical Directors of Information Systems (AMDIS) and awarded at the HCI Executive Summit in May.
Because there wasn’t a plug and play solution that could be easily incorporated into the Phoenix Children's EHR (their vendor is the Chicago-based Allscripts), the clinical applications team built its own solution. In pediatrics, patients can weigh as little 10 ounces or as much as 200 pounds for adolescents, so because of this, dosing can also run the gamut. “Safety was the overriding factor for us, and because of the variability of children’s weights,” says Basfield, “that increases the likelihood that mistakes can be made, and the smaller the weight you don’t have much room for error.”
Phoenix Children's went live on its EHR in 2002, and gradually rolled out computerized physician order entry (CPOE) hospital-wide in 2009. With about 1,100 drugs in its formulary, Vaidya says that less than 10 percent of the drugs had dose range checking built into them. For the project’s phase one of implementation that began in October 2010, the team analyzed the frequency of drugs ordered since 2002, which yielded close to 750,000 orders. The next step was targeting the 100 most ordered drugs, as well as the most high-risk drugs. Basfield notes that this pre-analysis of ordering habits, rather than taking a different formulary approach (i.e. alphabetically), really enhanced the breadth of coverage in phase one.
Creating No-‘Nuisance’ Drug Alerts
To create drug alerts within the CPOE system, Vaidya separated the configuration from the programming logic by creating a separate reference table. “That was the critical defining moment, we completely took [out] drug by drug configuration, and for our 1,100 drugs there is one single custom code that looks up all the drugs,” says Vaidya. Vaidya says his team identified a custom code that would identify the drug ordered; evaluate the order to see if it fell within the dosing ranges in the reference table. If the order did not fit within the ranges, the system would fire a soft stop alert; or if the dose was so high, it would initiate a hard stop.
Vaidya also realized that the alerting configuration had to be sensitive and strike the right balance so that alerts wouldn’t be seen as nuisances. “When you’re about to alert a physician in real-time, you need to be very careful about at what level do you alert them,” Basfield says. “Part of what Dr. Vaidya came up with was a alarmingly high dose strategy that allowed us to figure out at what point do we want to alert a physician, which is very different than alerts at the normal dosing range.” Vaidya notes that since his team had access to dosing patterns, it could set appropriate limits to correspond with these patterns.
“You often hear in discussions and literature that hard stops are not good; find a workaround,” says Vaidya. “Death and taxes are the only two hard stops in CPOE. We challenged that notion. We feel that if misused hard stops can bring system down, but if used judiciously with multidisciplinary buy-in, with data to support, they can be extremely powerful.”
The clinical team devised a two-tiered system of alerts to target the physician at the point of order entry. If the physician orders a drug that exceeds a soft stop limit, a hard stop alert is provided with instructions on how to resolve the problem, along with normal dose ranges for reference. In more than 50 percent of the cases at Phoenix Children’s, alerts have resulted in the physician either cancelling or correcting the order. If the physician wants to proceed, a reason must be provided. That reason is then sent to the pharmacy system through the interface so that the pharmacist doing the clinical verification can understand the physician’s intent.
Dose Range Checking Go-Live
Within four months, 604 drugs—80 percent of ordered medications—were mapped in phase one and the new system went live in February 2011. Vaidya remembers the excitement in the command center watching the queries his team pre-set to analyze how frequently alerts were firing and what effect they were having. During hour three, he says, the team saw a glitch in the programming before users even reported it and made the appropriate changes, and since then hasn’t really modified the system much. In the first month, Phoenix Children’s had a 3 percent alerting rate for all orders, of which 2 percent were soft stops, which proved to be good balance for providers, Vaidya says. The initiative has also resulted in an 80 percent reduction in follow-up phone calls from the pharmacists back to physicians, which at five minutes per call can add up.
Since go-live, there have been no prescription overdose errors; and at the six-month mark, Phoenix Children’s will do a formal review and analysis of the initiative. The next phase of the initiative will expand to include 100 percent of all medication orders and will continue modifying the programming logic. The next targets will be alerts for highly ineffective dosing, as well as renal, gestational, and eventually chemotherapy dosing.