When doctors take patients off of a prescription medicine, it is often for a good reason. But pharmacists don't always get the memo. A new study finds that more than 1 in 100 discontinued prescriptions were filled by the pharmacy anyway, putting some patients at serious risk.

In the U.S., pharmacists filled more than 3.7 billion prescriptions in 2011. With so many prescriptions and refills—and our still largely human- and paper-based prescribing system—there are bound to be mistakes. Pharmacists may overlook drug interactions, dispense inappropriate medications, or commit other little-understood errors.

One such underappreciated problem area is the process of taking patients off medication. While errors in initial prescribing have drawn much attention, potential for error when doctors order a prescription to stop also looms large. And electronic health records, which have helped to minimize medical errors in other areas, might be partly to blame.

These electronic communiqués might be giving some doctors—and patients—a false sense of efficacy. Doctors might assume that when they make a note on a patient's electronic health record to stop a prescription the pharmacy will automatically get the message as it does when they first prescribed that medication. This, however, is not always the case, wrote Adrienne Allen, of the North Shore Physician Group, and Thomas Sequist, of Brigham and Women's Hospital in the new paper, published online November 19 in Annals of Internal Medicine.

To find out how often the pharmacies continue to dispense meds the doctor no longer ordered, Allen and Sequist analyzed electronic health records of 30,406 adults in a Massachusetts health system whose doctor had discontinued a drug to treat a high-risk condition such as high cholesterol, hypertension, diabetes, blood coagulation or platelet aggregation. Some 83,900 medications were discontinued during the course of a year. Nevertheless, pharmacists still dispensed 1,218 of these prescriptions after they were discontinued. The most common drug that pharmacists dispensed after a doctor canceled the prescription was metoprolol (Lopressor or Toprol), which is often prescribed to treat high blood pressure after a heart attack and which can have harmful drug interactions with other commonly prescribed drugs.

In a subset of medical records, a computer analysis flagged more than a third (34 percent) of the improper dispensations as creating a "high risk of potential adverse events" such as a harmful reaction, potential drug interaction or suspect lab test result, the researchers noted. And manual assessment verified that potential harm actually occurred in at least 12 percent of cases.

Patients receiving these drugs were more likely to be taking more medications, older, enrolled in Medicare and black. Additional medications make it more likely that a patient will suffer an adverse drug interaction if they take an unintended prescription (especially if a doctor has subsequently prescribed a similar drug to take the discontinued drug's place). And older adults might be less likely to notice a mistake.

One limitation of the study is that the researchers could only study the 52 percent of discontinued prescriptions that were filled at participating health care system pharmacies; unaffiliated pharmacies might have even higher inappropriate dispensation rates. Additionally, the researchers only studied a limited number of drugs. Adding other drugs to the analysis would likely increase the number of discontinued prescriptions dispensed, even if the risk of side effects might be lower.

They researchers see promise for filling this communication gap in the future. Electronic health records offer an opportunity to track these missteps, and adding more direct communication with pharmacies about prescription discontinuation should help avoid these errors.

For now, however, the new technology is often not as powerful as many doctors think it is. So some of the responsibility will continue to lie with the patient. Officials would be wise to help "increase patient awareness of their medication list," the researchers concluded. That is, until the computers can just do it for us.