Medication-related Clinical Decision Support in Computerized Provider Order Entry Systems: A Review
Medication-related Clinical Decision Support in Computerized Provider Order Entry Systems: A Review Gilad J. Kuperman, MD, PhD, Anne Bobb, RPh , Thomas Payne, MD, FACP , Anthony J. Avery, MB, CHB, DM , Tejal K. Gandhi, MD, MPH , Gerard Burns, MD, MBA, FACS , David Classen, MD, MS, David W. Bates, MD, MSc J Am Med Inform Assoc. 2007; 14:29-40
This group gives a nice overview of where Clinical Decision Support (CDS) stands now when it combines with Computerized Provider Order Entry (CPOE). The authors go on to describe where the friction occurs between clinicians and the systems, and where research is needed to smooth, improve, and make safe the interaction.
Clinical Decision Support exists at two levels, basic and advanced. Basic CDS usual covers simple drug ordering and the primary drug-drug, drug-allergy, and duplicate drug order checking. Advanced decision support includes functions such as dose adjustment for age, weight and laboratory parameters, and total day or lifetime dose checking for drugs such as chemotherapy drugs. Advanced CDS may also include ‘corollary’ orders for certain drugs or procedures. Advance decision support may also include antibiotic choices for infections, but this borders on linked information.
The authors then explore the problems of CDS as it exists now. Excessive irrelevant alerting “is highly prevalent and a major disruptor of clinicians workflows.” (1) One study looked at “drug allergy alerts based on whether they were ‘definite’ or ‘possible’, [and] found that both groups showed decreasing compliance, from approximately 51% to 27% (definite alerts) and from 46% to 20% (possible alerts)”(2) over time.
Clinicians are trained to think in certain ways. They have their own knowledge and their own way of organizing knowledge. Designers of EHRs do not know how to fit alerts and warnings into the knowledge base and how best to present information from the computer.
One of the most frequent and least useful class of alerts is duplicate drug warnings. These can fire when inpatient drugs are reordered from outpatient drug lists, when a drug taper is ordered, when basal and coverage insulin is ordered, when analgesics are combined and when topical and oral meds are ordered. These sorts of alerts “interrupt clinicians’ workflow and cause frustration when the ‘duplicate’ was already considered. Excessive inappropriate alerting may lead to desensitization to all classes of alerts” (1)
The most useful part of the article is the ‘recommendations for future work”.
Provider organizations should decide how to present different sorts of alerts for highest effectiveness and monitor the affects of the alerts in their systems.
Application vendors must design alerts to be the least interrupting to clinician workflows and vendor tools should allow customization that endure through each new version of the product.
The knowledge base vendors should also allow customization of the database and allow retention across upgrades. Alerts should be “concise and actionable”. The alerts should be linked to the evidence.
National policy makers need to enable sharing of CDS. Now each provider organization has to produce its own, literally reinventing the wheel. Ownership needs to be worked out and a national databank needs to be developed such as by the NIH or the National Library of Medicine.
The authors’ most specific list of recommendations is for the research community, from programmers to behavioral scientists:
To what extent does alerting impact on clinician behavior and patient outcomes?
What is the optimal way to present alerts to prescribers?
How can clinicians’ sense of satisfaction with alerts and other kinds of decision support be increased, i.e. so clinicians find decision support useful and not annoying?
When does “alert fatigue” happen?
What is the best way for organizations to share alert knowledge?
How can commercial medication knowledge bases be edited to yield clinically valuable knowledge bases?
Where there are multiple presentation modes, which mode is most appropriate for any given alert?
Which member of the health care team -- for example, physician, nurse, pharmacist, other – is the best recipient of any kind of alert?
Should physicians and pharmacists see the same drug-related alerts? (1)
(1) Gilad J. Kuperman, MD, PhD, Anne Bobb, RPh , Thomas Payne, MD, FACP , Anthony J. Avery, MB, CHB, DM , Tejal K. Gandhi, MD, MPH , Gerard Burns, MD, MBA, FACS , David Classen, MD, MS, David W. Bates, MD, MSc . Medication-related Clinical Decision Support in Computerized Provider Order Entry Systems: A Review. J Am Med Inform Assoc. 2007; 14:29-40
(2) 11 Abookire SA, Teich JM, Sandige H, Paterno MD, Martin MT, Kuperman GJ, Bates DW. Improving allergy alerting in a computerized physician order entry system. Proc AMIA Symp. 2000; 2-6.
