teaser
John E Murphy
PharmD
Professor
Department Head/Associate Dean
University of Arizona College of Pharmacy
Tucson, Arizona
USA
E:[email protected]
Prevention of medication misadventures caused by drug–drug interactions (DDIs) is a common role of pharmacists in both hospital and community settings. Unfortunately, the computerised alerting systems used in many pharmacies and in computerised prescriber order entry (CPOE) systems are plagued by excessive alerting of unimportant or irrelevant interactions. At the same time, research has shown that these same systems occasionally miss important DDIs. Computer alerting systems and more traditional reference texts have also been shown to be somewhat inconsistent in evaluating the potential severity of DDIs, and little information is given on the incidence and prevalence of clinically significant interactions that lead to morbidity and mortality. To avoid alert fatigue responses (the quick overriding of interactions messages) and to maximise the catching and preventing of serious DDIs, new research and refinements of the alerting systems will be required.
Discussion
There is potential for a DDI to occur any time a patient consumes at least two medications. In many developed countries, it is not uncommon for a patient to see several specialists and be prescribed multiple medications in both hospital and ambulatory care settings, leading to an even greater possibility of a DDI. In addition, the worldwide pharmaceutical industry continues to create many new products each year, and the mean age of the population in many countries is increasing, both of which can lead to the use of more medications per individual.(1) It is fortunate that the number of clinically important interactions actually received by patients is small.(2,3) For those receiving a drug pair that leads to morbidity and subsequent hospitalisation (4,5) or mortality, however, the small numbers are of little relevance.
Identification of clinically important drug–drug interactions by pharmacists and physicians is difficult, due to the very large number of drug products, limited evidence of DDI severity and incidence, differences in criteria used to grade interactions and subjectivity in assignment of ratings. This is particularly true when clinicians are asked to identify interactions of moderate to major importance by memory.(6) A number of determinants have been identified that increase the chance of patients receiving a drug pair listed as interacting.(7) These include: more primary care physicians and dispensing pharmacies; too many and too few signals; use of technicians for initial screening and thus knowledge, instructions and supervision of these individuals; and workload. Understanding these determinants can help in designing systems to overcome misadventures caused by DDIs.
The pharmacy profession has long used technology to improve drug safety, including screening for potential DDIs. In-hospital software or personal digital assistants (PDAs) may be used to provide messages regarding DDIs and assist pharmacists in identifying potential drug-related problems in hospital settings. However, research on the alerting systems has shown that they are not functioning as well as they might. For example, studies indicate that screening programs and references do not always keep up with the evolving knowledge of DDIs and that important interactions can be missed.(8,9) Simple alerting systems may not provide adequate information for clinicians trying to make appropriate decisions on treatment options, while sophisticated alert systems still often lack important patient data and the clinical reasoning that is part of the prescribing process. Variability in assessment of clinical significance due to conflicting criteria in the literature, the lack of relevance of many alerts that lead to a signal-to-noise ratio that diminishes the effectiveness of important alerts, and lack of validation for screening criteria for included alerts all contribute to the problem.
Studies have shown that pharmacists and physicians routinely override the majority of DDI and other alerts presented to them and that most alerts are either insignificant or not relevant to the patient.(10–14) Other studies have shown that potentially serious interactions may also be ignored.(15) However, many pharmacists and physicians believe DDI alerts are very important, and some believe that it should be harder to override potentially serious DDI alerts.(10,14)
To improve the detection and prevention of serious DDIs, which includes reducing the number of irrelevant or insignificant alerts, there should be a better assessment of risk. For example, if a DDI causes only minor side-effects, even in large numbers of patients, or the interaction is well known and easily managed, an alert should perhaps not be generated. On the other hand, if significant morbidity or mortality occurs with a DDI, even if it occurs in a small number of patients with specific risk factors, this must generate an alert. This alert will preferably provide information on the risk factors and subsets of patients so that clinicians can make appropriate decisions on management.
Bergk et al developed an algorithm for risk determination using four decision layers: severity, manageability, risk/benefit assessment and patient-related risk factors.(16) Using this approach, they found that, of 881 interacting drug pairs, 15.0% were of major severity, but 76.5% of these were manageable. Only 23.5% of the major interactions (3.5% of the total 881) offered no management options and, therefore, were considered to be DDIs that should be avoided. Malone and colleagues took a different approach and developed a list of 25 well-established,well-documented DDIs in order to study how frequently important interactions occur.(17) Similar approaches could be taken by software developers to highlight DDIs that should be difficult to override for prescribers, pharmacists and pharmacy technicians.
Conclusion
There is much to improve in the screening of DDIs. Program writers must do a better job of determining potential significance (incidence, severity, mechanisms) and relevance (eg, whether the patient is still taking medications, or has never received combination in the past), so fatigue does not occur from too many alerts. Evidence-based medicine approaches to grading reports of interactions should be used to help rate the understanding of outcomes and incidence. For example, controlled trials should have greater weight than case reports.
Screening programs must account for situations when the patient is no longer receiving the drug, the patient has tolerated the drug combination in the past or the course of therapy will not lead to a problem (irrelevant alerts). The programs must do a better job of presenting the truly important interactions and making it harder to override them, as well as providing decision support information and links to more information that will help pharmacists and physicians to determine appropriately the relevance of an interaction for a specific patient. Users of alerting systems must clearly understand that there are many factors that can affect the relevance of an interaction in an individual, and when technicians are used in the initial screening of alerts they must clearly understand policies associated with notification of the pharmacist and must be adequately supervised. The same would hold true for ward clerks and nurses who might be charged with entering prescribers’ orders in a CPOE system.
There are many partners in the DDI alerting and screening process, including prescribers, pharmacists, pharmacy technicians, nurses, other healthcare providers, patients, and developers/marketers of DDI screening software. All must work together to continue the evolution of alerting systems and prevent important DDIs from reaching the patient.
References
- Rosholm JU, et al. Dan Med Bull 1998;45:210-3.
- Guedon-Moreau L, et al. Eur J Clin Pharmacol 2003;59:689-95.
- Malone DC, et al. Am J Health-Syst Pharm 2005;62:1983-91.
- Juurlink DN, et al. JAMA 2003;289:1652-8.
- Hamilton RA, et al. Pharmacotherapy 1998;18:1112-20.
- Weideman RA, et al. Am J Health-Syst Pharm 1999;56:1524-9.
- Becker ML, et al. Drug Safety 2005;28:371-8.
- Abarca J, et al. J Am Pharm Assoc 2004;44:136-41.
- Hazlet TK, et al. J Am Pharm Assoc 2001;41:200-4.
- Murphy JM, et al. Am J Health-Syst Pharm 2004;61:1484-7.
- Armstrong EP, Denemark CR. J Am Pharm Assoc 1998;38:149-54.
- Abarca J, et al. Community pharmacists’ perception of computerized drug-drug interaction alerts. Accepted J Am Pharm Assoc 2005.
- Armstrong EP, Markson TJ. J Am Pharm Assoc 1997;NS37:315-20.
- Magnus D, et al. J Clin Pharm Ther 2002;27:377-82.
- Cavuto NJ, et al. JAMA 1996;275:1086-7.
- Bergk V, et al. Clin Pharmacol Ther 2004;76:85-96.
- Malone DC, et al. J Am Pharm Assoc 2004;44:142-51.
Resources
University of Washington Drug Interaction Database W:depts.washington.edu/ventures/UW_Technology/Express_Licenses/DIDB.php
Medscape Multi-Drug Interaction Checker
Check for drug-drug interactions in a regimen of two or more drugs.
W:www.medscape.com/druginfo/druginterchecker?src=google”
