Drug–drug interactions that should be non-interruptive in order to reduce alert fatigue in electronic health records
This is a review of Phansalkar, S., Van der Sijs, H. , Tucker, A. D., Amrita, D. A., Bell, D. S., Teich, J. M., Middleton, B., Bates, D.W. (2013) article Drug–drug interactions that should be non-interruptive in order to reduce alert fatigue in electronic health records 
Medication-related clinical decision support (CDS) allows healthcare providers to become aware of potential adverse drug-drug interactions before any medications are used. The intention of the system is to alert drug prescribers of potential adverse drug-drug interactions (DDI) but alerts are often ignored and overridden with high rates of up to 96% because system users believe the alerts lack content specificity.
The purpose of the research study detailed in the article was to determine whether or not it was possible to reduce the number of DDI alerts by classifying alerts as either critical, and thus necessarily interruptive, or non-critical, and thus unnecessarily interruptive. If the occurrence of interruptive alerts could be reduced then it was hoped that patient safety could be increased by decreasing the tendency for the alerts to be ignored by system users. Rather than give an alert for all drug-drug interactions, the researchers proposed that the system only alert the user for the highest priority drug-drug interactions. In a different study done by Khan BA et al. in Wishard memorial hospital, it was found that adding human intelligence to the CDSS alerts would improve the effectiveness of those alerts. Please check this link:Clinical decision support system and incidence of delirium in cognitively impaired older adults transferred to intensive care Khan BA et al. found out that interruptive alerts were given to ICU physicians when they added order in GOPHER for inappropriate anticholinergic drugs, order for urinary catheter, or order for physical restraint. They opined that added human intelligence to these alerts would have made a difference by reducing cognitive impairment in older adults in ICU. Please check this article: 
In order to conduct their analysis the researchers obtained logs from an Electronic Health Records (EHR) system in use at an academic medical center. A list of DDI alerts were then shown to a panel of 11 experts and each expert was asked to determine whether or not certain types of DDI alerts could safely be made non interruptive.
A list of the top 50 most frequently occurring DDI pairs were analyzed in detail by the panel and it was determined that 16 DDI pairs should remain interruptive but that 33 could safely be changed to non-interruptive alerts.
In an attempt to provide patient safety, electronic health record (EHR) systems contain clinical decision support sub component systems that alert users to potentially negative drug-drug interactions. Unfortunately, the system users find that they are alerted too frequently for DDI that are not severe enough to warrant attention but because the alerts interrupt the user workflow and require input in order to circumvent, the alerts cause what is known as alert fatigue. Because there are so many interruptions, the alerts cease to signal something that must be attended to but rather are seen as something that should be ignored. This defeats the purpose of the alerts, which is to interrupt a workflow that might potentially cause an adverse effect.Carspecken et al . conducted a study that observed a patient in the PICU who experienced complications as a result of an extended series of non–evidence-based alerts in the electronic health record.
In conclusion, the theory behind having a system that alerts users before harm is done is noble but when users are given alerts that are too vague or are not truly serious in nature they become more of a nuisance than a useful tool. When users begin to see the alerts as nothing more than an inconvenient interruption to their workflow they begin to find ways to circumvent the system. Only when interruptive alerts are truly meaningful are they in turn useful. When interruptive alerts are truly meaningful then system users will give them the attention they deserve. In a study done by Khan BA in Wishard memorial hospital, Interruptive alerts were given to the physicians when they type in the CPOE advising use of physical restraints,use of anticholinergics or a urinary catheter. The details are mentioned in this article:Clinical decision support system and incidence of delirium in cognitively impaired older adults transferred to intensive care
Lastly, the researchers had the right idea in that a system should only alert when necessary but their methodology could have been better. Their panel of experts only comprised 11 people and no mention of a physician being on the panel was made. Rather than only having pharmacists and pharmacologists on the panel, it would have been nice to have a more diversified panel including the users of the systems as well. In a different study by Khan BA et al., the results showed that diverting the alerts to the ICU nurses and other staff rather than the ICU physicians would have had more impact and the CDSS could have proven effective in cognitively impaired patients and reduced delirium. 
- Phansalkar, S., Van der Sijs, H. , Tucker, A. D., Amrita, D. A., Bell, D. S., Teich, J. M., Middleton, B., Bates, D.W. Drug–drug interactions that should be non-interruptive in order to reduce alert fatigue in electronic health records. Journal of the American Medical Informatics Association: JAMIA, 20(3), 489-493. doi:10.1136/amiajnl-2012-001089. Retrieved from http://jamia.oxfordjournals.org/content/20/3/489