Evaluation of context-specific alerts for potassium-increasing drug-drug interactions: A pre-post study
Introduction
One key cause of preventable adverse drug events (ADEs) are drug-drug interactions (DDIs) [1]. Computerized physician order entry (CPOE) with built-in clinical decision support systems (CDSS) have the potential to prevent medication errors and consecutive ADEs at the very moment of prescribing [[2], [3], [4], [5], [6], [7]]. Yet, the evidence on the impact of CDSS on patient outcomes remains scarce [8,9]. It is well established that CDSS for DDI checking are often overly sensitive generating excessive alerts with low specificity leading to alert fatigue and high override rates, often exceeding 80% [[10], [11], [12], [13], [14], [15], [16]]. The main problems are the low specificity of the alerts and their perceived lack of clinical importance [4,12,16,17]. This makes it difficult for clinicians to distinguish between clinically significant and insignificant alerts leading to both types of alerts being overridden which compromises the primary objective of patient safety [4,13,18,19]. Integration of patient characteristics in the clinical decision support (CDS) logic was suggested to improve alert specificity [20,21]. Specifically, linkage and follow-up of laboratory values with the CDS rules was proposed [[22], [23], [24], [25], [26], [27]]. However, just displaying laboratory values in the alert did not significantly improve the alert adherence in high-risk patients [28].
We have encountered the same problem of low specificity and high override rates in our hospital [29,30]. Of all DDI alerts generated by the CDSS from the 1 st of January 2010 till the 30th of June 2011, 72.1% were alerts for the risk of hyperkalemia due to the interaction between potassium-sparing diuretics and potassium supplements, with an override rate of 85.7% [31]. Hyperkalemia is a serious and potentially life-threatening electrolyte disorder caused by an imbalance in potassium homeostasis and is associated with increased mortality and adverse cardiovascular effects such as cardiac arrhythmia and cardiac arrest [[32], [33], [34]]. Uijtendaal et al. found that DDI-induced hyperkalemia occurred in 10% of hospitalized patients who were prescribed at least one potassium-increasing drug [24].
Context-specific alerts for potassium-increasing DDIs with patient-specific risk assessments for hyperkalemia were developed as part of our CDSS. Alerts for low-risk patients were not shown to the physicians to improve the specificity, as suggested by Duke et al. [28]. The main objective of this study was to investigate whether these context-specific alerts reduced the alert burden and had a higher alert acceptance compared to alerts without context-specific rules. Because it is important to improve the efficiency of the DDI alerting system without compromising patient safety, the effect of the optimized CDSS on the patient outcome, occurrence of hyperkalemia, was also examined.
Section snippets
Design and setting
This pre-post study was conducted at the UZ Brussel, a 721-bed tertiary university hospital in Brussels, Belgium. Our hospital information system Primuz is a homegrown fully integrated system, meaning that all information from different sources (e.g. CPOE, CDSS, laboratory values, etc.) is stored in the same integrated database [26,35,36]. The intervention was the hospital wide implementation of a context-specific DDI alerting system, discussed in detail elsewhere [29]. In this study, the focus
Characteristics of study population
The patient characteristics in the pre- and post-intervention period are provided in Table 1. Significant differences were found for sex (P = 0.001), the frequency of diabetes (P = 0.038), the distribution of magnesium (P = 0.018) and eGFR values (P = 0.005), and for the administration of extra potassium supplements (P < 0.001), extra potassium-sparing diuretics (P = 0.011), NSAIDs (P < 0.001), and calcineurin inhibitors (P = 0.048). There was also a significant difference for the distribution
Discussion
The optimized CDSS which uses context factors for the individual risk assessment of hyperkalemia significantly reduced the alert burden without a significant difference in occurrence of hyperkalemia. The intervention converted 92.8% of the alerts – which would have been fixed level 1 alerts in the old CDSS – into level 3 alerts which are not shown to the physicians. This means a significant reduction of the alert burden, which was our primary purpose. Since most studies only measure alert
Conclusion
In conclusion, we succeeded to reduce the DDI alert burden of physicians without compromising patient safety by reducing the number of alerts shown to the physician by 92.8% without a significant difference in the occurrence of hyperkalemia. This study demonstrates the proposed strategy seems effective to get a grip on the delicate equilibrium between over- and under alerting many institutions struggle with. Further research into the development, optimization and evaluation of context-specific
Authors’ contibutions and acknowledgements
PC and AD participated in the design of the study. KM and KG were responsible for the data acquisition. KM and PC performed the data analysis and interpretation. The Interfaculty Center for Data processing and Statistics from the Vrije Universiteit Brussel was consulted for the data analysis. KM drafted the manuscript. All authors critically evaluated the manuscript and gave their final approval before submission. The study was carried out with the support of Wetenschappelijk Fonds Willy Gepts
Declaration of Competing Interest
The authors have no conflicts of interest regarding this study.
References (43)
Patient safety and computerized medication ordering at Brigham and Women’s Hospital
Comm. J. Qual. Improv.
(2001)- et al.
Preventable medication errors: identifying and eliminating serious drug interactions
J. Am. Pharm. Assoc. Wash. (Wash)
(2001) Overriding of drug safety alerts in computerized physician order entry
J. Am. Med. Inform. Assoc.
(2006)Medication-related clinical decision support in computerized provider order entry systems: a review
J. Am. Med. Inform. Assoc.
(2007)Improving acceptance of computerized prescribing alerts in ambulatory care
J. Am. Med. Inform. Assoc.
(2006)Turning off frequently overridden drug alerts: limited opportunities for doing it safely
J. Am. Med. Inform. Assoc.
(2008)What, if all alerts were specific - estimating the potential impact on drug interaction alert burden
Int. J. Med. Inform.
(2014)Pilot evaluation of an optimized context-specific drug–drug interaction alerting system: a controlled pre-post study
Int. J. Med. Inform.
(2015)High-priority and low-priority drug–drug interactions in different international electronic health record systems: a comparative study
Int. J. Med. Inform.
(2018)Framework for a clinical information system
Int. J. Med. Inform.
(2000)
Critical drug-drug interactions for use in electronic health records systems with computerized physician order entry: review of leading approaches
J. Patient Saf.
Effect of computerized physician order entry and a team intervention on prevention of serious medication errors
Jama
Improving recognition of drug interactions: benefits and barriers to using automated drug alerts
Med. Care
The epidemiology of prescribing errors: the potential impact of computerized prescriber order entry
Arch. Intern. Med.
Outpatient prescribing errors and the impact of computerized prescribing
J. Gen. Intern. Med.
Effectiveness of computerized decision support systems linked to electronic health records: a systematic review and meta-analysis
Am. J. Public Health
Effects of computerized decision support system implementations on patient outcomes in inpatient care: a systematic review
J. Am. Med. Inform. Assoc.
A systematic review of trials evaluating success factors of interventions with computerised clinical decision support
Implement. Sci.
Characteristics and override rates of order checks in a practitioner order entry system
Proc. AMIA Symp.
Physicians’ decisions to override computerized drug alerts in primary care
Arch. Intern. Med.
Drug safety alert generation and overriding in a large Dutch university medical centre
Pharmacoepidemiol. Drug Saf.
Cited by (8)
Evaluation of an optimized context-aware clinical decision support system for drug-drug interaction screening
2021, International Journal of Medical InformaticsCitation Excerpt :This pre-post study was conducted in a 721-bed university hospital in Brussels, Belgium. The hospital information system Primuz is a homegrown system with integrated CPOE and CDSS, discussed in previous papers [19,20]. The intervention was the hospital wide implementation of an optimized context-aware CDSS for DDI screening in hospitalized patients.
Added value of patient- and drug-related factors to stratify drug-drug interaction alerts for risk of QT prolongation: Development and validation of a risk prediction model
2023, British Journal of Clinical PharmacologyPrescribers’ reported acceptance and use of drug-drug interaction alerts: An Australian survey
2022, Health Informatics JournalModulators Influencing Medication Alert Acceptance: An Explorative Review
2022, Applied Clinical InformaticsPractice-enhancing publications about the medication-use process in 2020
2022, American Journal of Health-System PharmacyThe clinical decision support system AMPEL for laboratory diagnostics: Implementation and technical evaluation
2021, JMIR Medical Informatics