Safety Still Compromised by Computer Weaknesses
Our 2005 field test results clearly show that, with few exceptions, little improvement has occurred during the past 6 years with pharmacy computer systems to safeguard against medication errors. Alarming findings from our 1999 field test demonstrated that pharmacy computer systems in this country were vastly unreliable when used to detect and correct prescription errors or pharmacy order entry errors (see the February 10, 1999 issue). And now, 6 years later, we must draw the same conclusion, especially since our experience during recent consults in healthcare organizations confirms the results of the 2005 field test.
In 2005, many respondents' computer systems performed poorly and were unable to detect potentially serious or fatal errors. All of the test elements were associated with actual errors or hazards reported to ISMP and published in this newsletter. In many areas, pharmacy computer systems in 2005 performed less reliably than in 1999 when tested with a new set of erroneous orders that corresponded to the 1999 test elements (see Table 1). Chillingly, only 4 of the 182 systems tested in 2005 were able to detect all the unsafe orders presented in the field test.
Although the 2005 field test was not a scientific study, it's frightening to think that many of our nation's pharmacy computer systems may still be incapable of detecting orders for medications that exceed a safe maximum dose. For example, substantially less than half of the systems tested detected a serious overdose of carbamazepine for a 4-year-old child or an adult. Only 1 in 4 systems detected a significant overdose of lomustine based on the patient's body surface area (BSA), or a potentially fatal overdose of methotrexate for a patient being treated for rheumatoid arthritis. Less than half of the systems were able to detect a potentially deadly order for conventional amphotericin B dosed according to its liposomal form. Also, fewer computer systems tested in 2005 offered dose alerts based on the patient's age or weight/BSA, when compared to the systems tested in 1999.
While most of the pharmacy computer systems tested in 1999 were able to intercept orders for drugs to which patients were allergic, less than half tested in 2005 alerted staff when an order for FLUZONE (influenza virus vaccine) was entered for a patient allergic to eggs. Additionally, there was little improvement between 1999 and 2005 with detection of a wrong route of administration. Only about a third of the systems flagged an order for intrathecal vincristine, which has resulted in fatalities. The results were similar to the poor performance in 1999 for detecting an order for an oral suspension by the IV route, an error that has also resulted in fatalities. One new area tested in 2005 also fared poorly. Just 1 in 5 computer systems were able to intercept the entry of a contraindicated drug based on the patient's diagnosis or condition (pregnancy).
Equally distressing, when the pharmacy systems detected unsafe test orders, an average of 9 in 10 systems allowed the users to override these serious warnings. In most cases, the warnings could be bypassed simply by pressing a function key. The chance of overlooking these important alerts is further heightened when considering that clinically insignificant alerts routinely appeared on the screens of 3 in 4 systems tested in 2005.
Some improvement was observed with detecting contraindicated drugs or doses based on lab results. Compared to 1999, today more pharmacy systems are directly interfaced with the laboratory system and can automatically alert staff according to current lab values (see Table 2). Despite this improvement, just 1 in 3 systems alerted when entering an order for NEUTRA-PHOS-K (potassium and phosphorous) for a patient with an elevated serum potassium, and less than half alerted when entering an order for metformin 500 mg BID for a patient with an elevated serum creatinine.
Respondents' current pharmacy systems performed better than in 1999 with detecting duplicate therapy, and three quarters of the systems tested this year were also able to detect a significant drug/drug interaction. However, just one quarter were able to detect a clinically significant drug/herbal interaction.
Eighty-eight percent of the computer systems tested in 1999 were able to produce computer-generated medication administration records (MARs), and in 2005, 81% said they were actually being used. There was only slight improvement in nurses' perceptions that they were safer than handwritten MARs (see Table 2 in the PDF version of the newsletter). Most systems were also able to provide reports of drug warning overrides and allowed staff to build alerts for serious error-prone situations. However, even today, only half of the systems tested allowed the use of tall man letters to differentiate look-alike drug names, and just 9% allowed users to change the font and color to highlight look-alike drug name pairs.
Failure to update technology may be one explanation for the general lack of improvement in pharmacy computer systems between 1999 and 2005. More than half (56%) of the participants in our field test were currently using a pharmacy system that was at least 5 years old, with no recent upgrades. In fact, 38% had been using the same pharmacy computer system for 8 years or more without upgrades. Newer pharmacy system technology (replacement or upgrades) might have yielded different results.
These findings suggest that perhaps hospitals should prioritize the upgrade and replacement of outdated pharmacy systems. Improved pharmacy technology could help to reduce the risk of serious patient harm from medication errors. Yet, complex self-programming and an unrealistic time commitment necessary to achieve optimal results may prohibit full use of the most advanced pharmacy technology. Thus, pharmacy system applications must be user friendly, alert only when clinically appropriate, allow for user-defined alerts, follow user workflow, ensure that upgrades and maximum capabilities can be achieved easily, and must not be cost prohibitive.
The drug information provider's software and untimely installation of updates may be another explanation for poor field test results. If the vendor deems certain types of alerts unimportant, they may be absent; or if the content is not current or adequate, important alerts won't appear. For example, the drug information vendor's software may not alert staff to unsafe single (one-time) doses. While the system may contain dose limits for drugs with a routine frequency, it may contain limited information for single doses. Consequently, an alert for a single dose that exceeds safe limits may not appear, and the user may not be informed that a dose check was not performed.
As many hospital pharmacy computer systems stand today, pharmacists cannot consistently rely on this technology to detect potentially harmful medication errors. While you may think that your computer system can detect unsafe orders, don't be lulled into a false sense of security.
We thank all who participated in the 2005 pharmacy computer field test. We plan to use the results to continue our efforts to promote improved pharmacy computer technology for more effective recognition of clinically significant subtherapeutic and excessive doses, allergic and cross-allergic reactions, drug class and ingredient duplication, drug interactions, contraindicated drugs or routes of administration, and drugs that need dose adjustments for specific diseases or other clinical situations.