Most commercial pharmacy order verification systems either provide or allow for the integration of clinical decision support (CDS), with drug interaction and allergy alert screening being among the most commonly used. However, it has been well established that if the burden of these safety alerts is too high, the ensuing alert fatigue will cause clinicians to bypass both important and unimportant alerts.
Based on staff feedback that our clinical decision support system was generating an unacceptable amount of alerts that were deemed unimportant, we decided to closely review how the system was performing and set our objectives to reduce the frequency of non-value added clinical alerts in our pharmacy order verification system, develop a process to review the clinical significance of alerts provided by our content vendor, and assess the impact of suppressing select alerts moving forward.
In order to calculate the impact of this project, baseline alert frequency statistics were collected prior to any alert suppression. After the month-long baseline assessment period, pharmacists were given instruction on how to mark alerts for review within the order verification system. After a sixty-day period, all clinical alerts that were marked for suppression were accounted for and a 30-member panel of clinical pharmacy specialists then reviewed them to determine whether they qualified as non-value added. If deemed clinically insignificant by the panel, the alert was designated for suppression in the pharmacy order verification system. It also was determined that most alerts would be suppressed at the generic drug identification level within the order verification system, however some alerts may be suppressed at the NDC level in order to provide more specificity to the filter. One such example was a decision to only suppress heparin flush NDCs for an anticoagulant-anticoagulant class interaction. This prevented the pharmacist from seeing an alert each time a heparin flush was ordered for a patient that may also have had an active anticoagulation order.
Results of Suppression
The initial review of the baseline data revealed a significant number of alerts that were triggered by missing or invalid NDC numbers. Those alerts were deemed to be of no clinical value and were systematically suppressed. This resulted in a decrease in NDC error alerts of 39%. The first round of clinical alerts reviewed by our panel of clinical specialists included 57 different drug interaction or allergy alert pairs. Of those reviewed, 55 pairs (eg, beta-blockers and oral hypoglycemic agents) were approved to be suppressed. After review of all approved alert pairs the total number of individual warning alerts (eg, metoprolol and glyburide) suppressed was 1016. This decreased the number of overall alerts from 150,902 (an 85% alert rate) in July 2010 to 107,770 (a 52% alert rate) in December 2010. This decrease represented a net reduction in total number of clinical alerts received by pharmacists of 33%.
Ramifications of Alert Suppression
Pharmacists reviewing orders using a computerized order verification system in an acute care setting receive a significant number of clinical alerts. Many of these alerts may be generated due to system level mismatch between the hospital catalog NDC number files and CDS vendor NDC number files. These types of alerts are generally deemed non-value added by clinical staff and can therefore be minimized with ongoing surveillance by information systems staff. However, even with removal or correction of these types of systems-level avoidable alerts, there are still a significant number of potentially non-value added clinical alerts supplied by commercial decision support vendors.
The process for reviewing the clinical significance of such alerts and subsequently suppressing those deemed unnecessary throughout the verification system required more internal resources than expected. This was largely due to the process our information systems staff members were required to use in order to ensure that all medications within a given alert class were adequately captured. Additionally, it was discovered that if vendor-supplied warning identification numbers were changed in the future, those alerts previously suppressed might reappear. This is a limitation any institution considering suppression of clinical alerts in information systems should consider. Nonetheless, we determined that reducing the number and frequency of non-value added clinical alerts improves the likelihood that pharmacists will be better equipped to review and respond to those important alerts that may require further patient care intervention.
Brian Pinto, PharmD, MBA, is the assistant director of medication use policy and clinical decision support at The Johns Hopkins Hospital in Baltimore.
Pauline Newman, PharmD, is an information systems pharmacist at The Johns Hopkins Hospital.
Enter our Sweepstakes now for your chance to win the following prizes:
Just answer the following quick question for your chance to win:
Entries are limited to one entry per person in any active sweepstakes.
Thank you for your entry.