Tailoring of alerts substantially reduces the alert burden in computerized clinical decision support for drugs that should be avoided in patients with renal disease

Acute kidney injury associated with nephrotoxic drugs in critically ill patients: a multicenter cohort study using electronic health record data

Background

Nephrotoxic drugs frequently cause acute kidney injury (AKI) in adult intensive care unit (ICU) patients. However, there is a lack of large pharmaco-epidemiological studies investigating the associations between drugs and AKI. Importantly, AKI risk factors may also be indications or contraindications for drugs and thereby confound the associations. Here, we aimed to estimate the associations between commonly administered (potentially) nephrotoxic drug groups and AKI in adult ICU patients whilst adjusting for confounding.

Methods

In this multicenter retrospective observational study, we included adult ICU admissions to 13 Dutch ICUs. We measured exposure to 44 predefined (potentially) nephrotoxic drug groups. The outcome was AKI during ICU admission. The association between each drug group and AKI was estimated using etiological cause-specific Cox proportional hazard models and adjusted for confounding. To facilitate an (independent) informed assessment of residual confounding, we manually identified drug group-specific confounders using a large drug knowledge database and existing literature.

Results

We included 92 616 ICU admissions, of which 13 492 developed AKI (15%). We found 14 drug groups to be associated with a higher hazard of AKI after adjustment for confounding. These groups included established (e.g. aminoglycosides), less well established (e.g. opioids) and controversial (e.g. sympathomimetics with α- and β-effect) drugs.

Conclusions

The results confirm existing insights and provide new ones regarding drug associated AKI in adult ICU patients. These insights warrant caution and extra monitoring when prescribing nephrotoxic drugs in the ICU and indicate which drug groups require further investigation.

Modulators Influencing Medication Alert Acceptance: An Explorative Review

OBJECTIVES

Clinical decision support systems (CDSSs) use alerts to enhance medication safety and reduce medication error rates. A major challenge of medication alerts is their low acceptance rate, limiting their potential benefit. A structured overview about modulators influencing alert acceptance is lacking. Therefore, we aimed to review and compile qualitative and quantitative modulators of alert acceptance and organize them in a comprehensive model.

METHODS

In accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guideline, a literature search in PubMed was started in February 2018 and continued until October 2021. From all included articles, qualitative and quantitative parameters and their impact on alert acceptance were extracted. Related parameters were then grouped into factors, allocated to superordinate determinants, and subsequently further allocated into five categories that were already known to influence alert acceptance.

RESULTS

Out of 539 articles, 60 were included. A total of 391 single parameters were extracted (e.g., patients' comorbidity) and grouped into 75 factors (e.g., comorbidity), and 25 determinants (e.g., complexity) were consequently assigned to the predefined five categories, i.e., CDSS, care provider, patient, setting, and involved drug. More than half of all factors were qualitatively assessed (n = 21) or quantitatively inconclusive (n = 19). Furthermore, 33 quantitative factors clearly influenced alert acceptance (positive correlation: e.g., alert type, patients' comorbidity; negative correlation: e.g., number of alerts per care provider, moment of alert display in the workflow). Two factors (alert frequency, laboratory value) showed contradictory effects, meaning that acceptance was significantly influenced both positively and negatively by these factors, depending on the study. Interventional studies have been performed for only 12 factors while all other factors were evaluated descriptively.

CONCLUSION

This review compiles modulators of alert acceptance distinguished by being studied quantitatively or qualitatively and indicates their effect magnitude whenever possible. Additionally, it describes how further research should be designed to comprehensively quantify the effect of alert modulators.

An e-Delphi study to obtain expert consensus on the level of risk associated with preventable e-prescribing events

Aims

We aim to seek expert opinion and gain consensus on the risks associated with a range of prescribing scenarios, preventable using e‐prescribing systems, to inform the development of a simulation tool to evaluate the risk and safety of e‐prescribing systems (ePRaSE).

Methods

We conducted a two‐round e‐Delphi survey where expert participants were asked to score pre‐designed prescribing scenarios using a five‐point Likert scale to ascertain the likelihood of occurrence of the prescribing event, likelihood of occurrence of harm and the severity of the harm.

Results

Twenty‐four experts consented to participate with 15 pand 13 participants completing rounds 1 and 2, respectively. Experts agreed on the level of risk associated with 136 out of 178 clinical scenarios with 131 scenarios categorised as high or extreme risk.

Conclusion

We identified 131 extreme or high‐risk prescribing scenarios that may be prevented using e‐prescribing clinical decision support. The prescribing scenarios represent a variety of categories, with drug–disease contraindications being the most frequent, representing 37 (27%) scenarios, and antimicrobial agents being the most common drug class, representing 28 (21%) of the scenarios. Our e‐Delphi study has achieved expert consensus on the risk associated with a range of clinical scenarios with most of the scenarios categorised as extreme or high risk. These prescribing scenarios represent the breadth of preventable prescribing error categories involving both basic and advanced clinical decision support. We will use the findings of this study to inform the development of the e‐prescribing risk and safety evaluation tool.

Categorization of Medication Safety Errors in Ambulatory Electronic Health Records

Preventable medication errors continue to affect the quality and consistency in the delivery of care. While numerous studies on medication safety have been performed in the inpatient setting, a review of ambulatory patient safety by the American Medical Association found that medication safety errors were the most frequent safety problems in the outpatient arena. The leading cause of ambulatory safety problems, adverse drug events (ADEs), are common, with estimates of more than 2 million ADEs each year in the ambulatory Medicare population alone, and these events are frequently preventable. We conducted an environmental scan that allowed us to create our own categorization schema of medication safety errors in electronic healthcare records (EHRs) found in the outpatient setting and observed which of these were additionally supported in the literature. This study combines data from the California Hospital Patient Safety Organization (CHPSO), with several key articles in the area of medication errors in the EHR era. Method: To best utilize the various EHR ambulatory medication events submitted into CHPSO’s database, we chose to create a framework to bucket the near misses or adverse events (AEs) submitted to the database. This newly created categorization scheme was based on our own drafted categorization labels of events, after a high-level review, and from two leading articles on physician order entry. Additionally, we conducted a literature review of computerized provider order entry (CPOE) medication errors in the ambulatory setting. Within the newly created categorization scheme, we organized the articles based on issues addressed so we could see areas that were supported by the literature and what still needed to be researched. Results: We initially screened the CHPSO database for ambulatory safety events and found 25,417 events. Based on those events, an initial review was completed, and 19,242 events were found in the “Medication or Other Substance” and “Other” categories, in which the EHR appeared to have been a potential contributing factor. This review identified a subset of 2,236 events that were then reviewed. One hundred events were randomly selected for further review to identify common categories. The most common categories in which errors occurred were orders in order sets and plans (n=12) and orders crossing or not crossing encounters (n=12), incorrect order placed on correct patient (n=10), orders missing (n=8), standing orders (n=8), manual data entry errors (n=6), and future orders (n=6). Conclusion: There were several common themes seen in this analysis of ambulatory medication safety errors related to the EHR. Common among them were incorrect orders consisting of examples such as dose errors or ordering the wrong medication. The manual data entry errors consisted of height or weight being entered incorrectly or entering the wrong diagnostic codes. Lastly, different sources of medication safety information demonstrate a diversity of errors in ambulatory medication safety. This confirms the importance of considering more than one source when attempting to comprehensively describe ambulatory medication safety errors.

The use of a clinical decision support tool to assess the risk of QT drug-drug interactions in community pharmacies

Introduction:

The handling of drug–drug interactions regarding QTc-prolongation (QT-DDIs) is not well defined. A clinical decision support (CDS) tool will support risk management of QT-DDIs. Therefore, we studied the effect of a CDS tool on the proportion of QT-DDIs for which an intervention was considered by pharmacists.

Methods:

An intervention study was performed using a pre- and post-design in 20 community pharmacies in The Netherlands. All QT-DDIs that occurred during a before- and after-period of three months were included. The impact of the use of a CDS tool to support the handling of QT-DDIs was studied. For each QT-DDI, handling of the QT-DDI and patient characteristics were extracted from the pharmacy information system. Primary outcome was the proportion of QT-DDIs with an intervention. Secondary outcomes were the type of interventions and the time associated with handling QT-DDIs. Logistic regression analysis was used to analyse the primary outcome.

Results:

Two hundred and forty-four QT-DDIs pre-CDS tool and 157 QT-DDIs post-CDS tool were included. Pharmacists intervened in 43.0% and 35.7% of the QT-DDIs pre- and post-CDS tool respectively (odds ratio 0.74; 95% confidence interval 0.49–1.11). Substitution of interacting agents was the most frequent intervention. Pharmacists spent 20.8 ± 3.5 min (mean ± SD) on handling QT-DDIs pre-CDS tool, which was reduced to 14.9 ± 2.4 min (mean ± SD) post-CDS tool. Of these, 4.5 ± 0.7 min (mean ± SD) were spent on the CDS tool.

Conclusion:

The CDS tool might be a first step to developing a tool to manage QT-DDIs via a structured approach. Improvement of the tool is needed in order to increase its diagnostic value and reduce redundant QT-DDI alerts.

Plain Language Summary

The use of a tool to support the handling of QTc-prolonging drug interactions in community pharmacies

Introduction: Several drugs have the ability to cause heart rhythm disturbances as a rare side effect. This rhythm disturbance is called QTc-interval prolongation. It may result in cardiac arrest. For health care professionals, such as physicians and pharmacists, it is difficult to decide whether or not it is safe to proceed treating a patient with combinations of two or more of these QT-prolonging drugs. Recently, a tool was developed that supports the risk management of these QT drug–drug interactions (QT-DDIs).

Methods: In this study, we studied the effect of this tool on the proportion of QT-DDIs for which an intervention was considered by pharmacists. An intervention study was performed using a pre- and post-design in 20 community pharmacies in The Netherlands. All QT-DDIs that occurred during a before- and after-period of 3 months were included.

Results: Two hundred and forty-four QT-DDIs pre-implementation of the tool and 157 QT-DDIs post-implementation of the tool were included. Pharmacists intervened in 43.0% of the QT-DDIs before the tool was implemented and in 35.7% after implementation of the tool. Substitution of one of the interacting agents was the most frequent intervention. Pharmacists spent less time on handling QT-DDIs when the tool was used.

Conclusion: The clinical decision support tool might be a first step to developing a tool to manage QT-DDIs via a structured approach.

Computerized provider order entry-related medication errors among hospitalized patients: An integrative review

The Institute of Medicine estimates that 7,000 lives are lost yearly as a result of medication errors. Computerized physician and/or provider order entry was one of the proposed solutions to overcome this tragic issue. Despite some promising data about its effectiveness, it has been found that computerized provider order entry may facilitate medication errors.The purpose of this review is to summarize current evidence of computerized provider order entry -related medication errors and address the sociotechnical factors impacting the safe use of computerized provider order entry. By using PubMed and Google Scholar databases, a systematic search was conducted for articles published in English between 2007 and 2019 regarding the unintended consequences of computerized provider order entry and its related medication errors. A total of 288 articles were screened and categorized based on their use within the review. One hundred six articles met our pre-defined inclusion criteria and were read in full, in addition to another 27 articles obtained from references. All included articles were classified into the following categories: rates and statistics on computerized provider order entry -related medication errors, types of computerized provider order entry -related unintended consequences, factors contributing to computerized provider order entry failure, and recommendations based on addressing sociotechnical factors. Identifying major types of computerized provider order entry -related unintended consequences and addressing their causes can help in developing appropriate strategies for safe and effective computerized provider order entry. The interplay between social and technical factors can largely affect its safe implementation and use. This review discusses several factors associated with the unintended consequences of this technology in healthcare settings and presents recommendations for enhancing its effectiveness and safety within the context of sociotechnical factors.

Development, Implementation, and Evaluation of a Personalized Machine Learning Algorithm for Clinical Decision Support: Case Study With Shingles Vaccination

BACKGROUND

Although clinical decision support (CDS) alerts are effective reminders of best practices, their effectiveness is blunted by clinicians who fail to respond to an overabundance of inappropriate alerts. An electronic health record (EHR)-integrated machine learning (ML) algorithm is a potentially powerful tool to increase the signal-to-noise ratio of CDS alerts and positively impact the clinician's interaction with these alerts in general.

OBJECTIVE

This study aimed to describe the development and implementation of an ML-based signal-to-noise optimization system (SmartCDS) to increase the signal of alerts by decreasing the volume of low-value herpes zoster (shingles) vaccination alerts.

METHODS

We built and deployed SmartCDS, which builds personalized user activity profiles to suppress shingles vaccination alerts unlikely to yield a clinician's interaction. We extracted all records of shingles alerts from January 2017 to March 2019 from our EHR system, including 327,737 encounters, 780 providers, and 144,438 patients.

RESULTS

During the 6 weeks of pilot deployment, the SmartCDS system suppressed an average of 43.67% (15,425/35,315) potential shingles alerts (appointments) and maintained stable counts of weekly shingles vaccination orders (326.3 with system active vs 331.3 in the control group; P=.38) and weekly user-alert interactions (1118.3 with system active vs 1166.3 in the control group; P=.20).

CONCLUSIONS

All key statistics remained stable while the system was turned on. Although the results are promising, the characteristics of the system can be subject to future data shifts, which require automated logging and monitoring. We demonstrated that an automated, ML-based method and data architecture to suppress alerts are feasible without detriment to overall order rates. This work is the first alert suppression ML-based model deployed in practice and serves as foundational work in encounter-level customization of alert display to maximize effectiveness.

Reducing Interruptive Alert Burden Using Quality Improvement Methodology

BACKGROUND

Increased adoption of electronic health records (EHR) with integrated clinical decision support (CDS) systems has reduced some sources of error but has led to unintended consequences including alert fatigue. The "pop-up" or interruptive alert is often employed as it requires providers to acknowledge receipt of an alert by taking an action despite the potential negative effects of workflow interruption. We noted a persistent upward trend of interruptive alerts at our institution and increasing requests for new interruptive alerts.

OBJECTIVES

Using Institute for Healthcare Improvement (IHI) quality improvement (QI) methodology, the primary objective was to reduce the total volume of interruptive alerts received by providers.

METHODS

We created an interactive dashboard for baseline alert data and to monitor frequency and outcomes of alerts as well as to prioritize interventions. A key driver diagram was developed with a specific aim to decrease the number of interruptive alerts from a baseline of 7,250 to 4,700 per week (35%) over 6 months. Interventions focused on the following key drivers: appropriate alert display within workflow, clear alert content, alert governance and standardization, user feedback regarding overrides, and respect for user knowledge.

RESULTS

A total of 25 unique alerts accounted for 90% of the total interruptive alert volume. By focusing on these 25 alerts, we reduced interruptive alerts from 7,250 to 4,400 per week.

CONCLUSION

Systematic and structured improvements to interruptive alerts can lead to overall reduced interruptive alert burden. Using QI methods to prioritize our interventions allowed us to maximize our impact. Further evaluation should be done on the effects of reduced interruptive alerts on patient care outcomes, usability heuristics on cognitive burden, and direct feedback mechanisms on alert utility.

Impact of a clinical decision support system for drug dosage in patients with renal failure

Background A clinical decision support system (CDSS) linked to the computerized physician order entry may help improve prescription appropriateness in inpatients with renal insufficiency. Objective To evaluate the impact on prescription appropriateness of a CDSS prescriber alert for 85 drugs in renal failure patients. Setting Before-after study in a 975-bed academic hospital. Method Prescriptions of patients with renal failure were reviewed during two comparable periods of 6 days each, before and after the implementation of the CDSS (September 2009 and 2010). Main outcome measure The proportion of inappropriate dosages of 85 drugs included in the CDSS was compared in the pre- and post-implementation group. Results Six hundred and fifteen patients were included in the study (301 in pre- and 314 in post-implementation periods). In the pre- and post-implementation period, respectively 2882 and 3485 prescriptions were evaluated, of which 14.9 and 16.6% triggered an alert. Among these, the dosage was inappropriate in respectively 25.4 and 24.6% of prescriptions in the pre- and post-implementation periods (OR 0.97; 95% CI 0.72-1.29). The most frequently involved drugs were paracetamol, perindopril, tramadol and allopurinol. Conclusion The implementation of a CDSS did not significantly reduce the proportion of inappropriate drug dosages in patients with renal failure. Further research is required to investigate the reasons why prescribers override alerts. Collaboration with clinical pharmacists might improve compliance with the CDSS recommendations.

Determining Inappropriate Medication Alerts from "Inaccurate Warning" Overrides in the Intensive Care Unit

OBJECTIVE

This article aims to understand provider behavior around the use of the override reason "Inaccurate warning," specifically whether it is an effective way of identifying unhelpful medication alerts.

MATERIALS AND METHODS

We analyzed alert overrides that occurred in the intensive care units (ICUs) of a major academic medical center between June and November 2016, focused on the following high-significance alert types: dose, drug-allergy alerts, and drug-drug interactions (DDI). Override appropriateness was analyzed by two independent reviewers using predetermined criteria.

RESULTS

A total of 268 of 26,501 ICU overrides (1.0%) used the reason "Inaccurate warning," with 93 of these overrides associated with our included alert types. Sixty-one of these overrides (66%) were identified to be appropriate. Twenty-one of 30 (70%) dose alert overrides were appropriate. Forty of 48 drug-allergy alert overrides (83%) were appropriate, for reasons ranging from prior tolerance (n = 30) to inaccurate ingredient matches (n = 5). None of the 15 DDI overrides were appropriate.

CONCLUSION

The "Inaccurate warning" reason was selectively used by a small proportion of providers and overrides using this reason identified important opportunities to reduce excess alerts. Potential opportunities include improved evaluation of dosing mechanisms based on patient characteristics, inclusion of institutional dosing protocols to alert logic, and evaluation of a patient's prior tolerance to a medication that they have a documented allergy for. This resource is not yet routinely used for alert tailoring at our institution but may prove to be a valuable resource to evaluate available alerts.

Better specification of triggers to reduce the number of drug interaction alerts in primary care

OBJECTIVE

Drug interaction alerts (drug-drug and drug-disease interaction alerts) for chronic medications substantially contribute to alert fatigue in primary care. The aim of this study was to determine which events require (re)assessment of a drug interaction and whether using these events as triggers in clinical decision support systems (CDSSs) would affect the alert rate.

METHODS

Two random 5% data samples from the CDSSs of 123 community pharmacies were used: dataset 1 and 2. The top 10 of most frequent drug interaction alerts not involving laboratory values were selected. To reach consensus on events that should trigger alerts (e.g. first time dispensing, dose modification) for these drug interactions, a two-step consensus process was used. An expert panel of community pharmacists participated in an online survey and a subsequent consensus meeting. A CDSS with alerts based on the consensus was simulated in both datasets.

RESULTS

Dataset 1 and 2 together contained 1,672,169 prescriptions which led to 591,073 alerts. Consensus on events requiring alerts was reached for the ten selected drug interactions. The simulation showed a reduction of the alert rate of 93.0% for the ten selected drug interactions (comparable for dataset 1 and 2), corresponding with a 28.3% decrease of the overall drug interaction alert rate.

CONCLUSION

By consensus-based better specification of the events that trigger drug interaction alerts in primary care, the alert rate for these drug interactions was reduced by over 90%. This promising approach deserves further investigation to assess its consequences and applicability in daily practice.

Technologic Distractions (Part 1): Summary of Approaches to Manage Alert Quantity With Intent to Reduce Alert Fatigue and Suggestions for Alert Fatigue Metrics

OBJECTIVE

To provide ICU clinicians with evidence-based guidance on tested interventions that reduce or prevent alert fatigue within clinical decision support systems.

DESIGN

Systematic review of PubMed, Embase, SCOPUS, and CINAHL for relevant literature from 1966 to February 2017.

PATIENTS

Focus on critically ill patients and included evaluations in other patient care settings, as well.

INTERVENTIONS

Identified interventions designed to reduce or prevent alert fatigue within clinical decision support systems.

MEASUREMENTS AND MAIN RESULTS

Study selection was based on one primary key question to identify effective interventions that attempted to reduce alert fatigue and three secondary key questions that covered the negative effects of alert fatigue, potential unintended consequences of efforts to reduce alert fatigue, and ideal alert quantity. Data were abstracted by two reviewers independently using a standardized abstraction tool. Surveys, meeting abstracts, "gray" literature, studies not available in English, and studies with non-original data were excluded. For the primary key question, articles were excluded if they did not provide a comparator as key question 1 was designed as a problem, intervention, comparison, and outcome question. We anticipated that reduction in alert fatigue, including the concept of desensitization may not be directly measured and thus considered interventions that reduced alert quantity as a surrogate marker for alert fatigue. Twenty-six articles met the inclusion criteria.

CONCLUSION

Approaches for managing alert fatigue in the ICU are provided as a result of reviewing tested interventions that reduced alert quantity with the anticipated effect of reducing fatigue. Suggested alert management strategies include prioritizing alerts, developing sophisticated alerts, customizing commercially available alerts, and including end user opinion in alert selection. Alert fatigue itself is studied less frequently, as an outcome, and there is a need for more precise evaluation. Standardized metrics for alert fatigue is needed to advance the field. Suggestions for standardized metrics are provided in this document.

Electronic Alerts for Acute Kidney Injury

BACKGROUND

Acute kidney injury (AKI) often takes a complicated course if diagnosed late and undertreated. Electronic alerts that provide an early warning of AKI are intended to support treating physicians in making the diagnosis of AKI and treating it appropriately. The available evidence on the effects of such alert systems is inconsistent.

METHODS

We employed the PRISMA recommendations for systematic literature reviews to identify relevant articles in the PubMed, Scopus, and Web of Science databases. All of the studies that were retrieved were independently assessed by two of the authors with respect to the methods of computer-assisted electronic alert systems and their effects on process indicators and clinical endpoints.

RESULTS

16 studies with a total of 32 842 patients were identified. 8.5% of admitted patients had community-acquired or hospital-acquired AKI, with an in-hospital mortality of 22.8%. Fifteen electronic alert systems were in use throughout the participating hospitals. In 13 of 15 studies, alarm activation was accompanied by concrete treatment recommendations. A randomized controlled trial in which no such recommendations were given did not reveal any benefit of the alert system for the patients. In controlled but non-randomized trials, however, the provision of concrete treatment recommendations when the alert was activated led to more frequent implementation of diagnostic or therapeutic measures, less loss of renal function, lower in-hospital mortality, and lower mortality after discharge compared to control groups without an electronic alert for AKI.

CONCLUSION

Non-randomized controlled trials of electronic alerts for AKI that were coupled with treatment recommendations have yielded evidence of improved care processes and treatment outcomes for patients with AKI. This review is limited by the low number of randomized trials and the wide variety of endpoints used in the studies that were evaluated.

Computerized Clinical Decision Support: Contributions from 2015

OBJECTIVE

To summarize recent research and select the best papers published in 2015 in the field of computerized clinical decision support for the Decision Support section of the IMIA yearbook.

METHOD

A literature review was performed by searching two bibliographic databases for papers related to clinical decision support systems (CDSSs) and computerized provider order entry (CPOE) systems. The aim was to identify a list of candidate best papers from the retrieved papers that were then peer-reviewed by external reviewers. A consensus meeting between the two section editors and the IMIA editorial team was finally conducted to conclude in the best paper selection.

RESULTS

Among the 974 retrieved papers, the entire review process resulted in the selection of four best papers. One paper reports on a CDSS routinely applied in pediatrics for more than 10 years, relying on adaptations of the Arden Syntax. Another paper assessed the acceptability and feasibility of an important CPOE evaluation tool in hospitals outside the US where it was developed. The third paper is a systematic, qualitative review, concerning usability flaws of medication-related alerting functions, providing an important evidence-based, methodological contribution in the domain of CDSS design and development in general. Lastly, the fourth paper describes a study quantifying the effect of a complex, continuous-care, guideline-based CDSS on the correctness and completeness of clinicians' decisions.

CONCLUSIONS

While there are notable examples of routinely used decision support systems, this 2015 review on CDSSs and CPOE systems still shows that, despite methodological contributions, theoretical frameworks, and prototype developments, these technologies are not yet widely spread (at least with their full functionalities) in routine clinical practice. Further research, testing, evaluation, and training are still needed for these tools to be adopted in clinical practice and, ultimately, illustrate the benefits that they promise.

Lower alert rates by clustering of related drug interaction alerts

OBJECTIVE

We aimed to investigate to what extent clustering of related drug interaction alerts (drug-drug and drug-disease interaction alerts) would decrease the alert rate in clinical decision support systems (CDSSs).

METHODS

We conducted a retrospective analysis of drug interaction alerts generated by CDSSs in community pharmacies. Frequently generated combinations of alerts were analyzed for associations in a 5% random data sample (dataset 1). Alert combinations with similar management recommendations were defined as clusters. The alert rate was assessed by simulating a CDSS generating 1 alert per cluster per patient instead of separate alerts. The simulation was performed in dataset 1 and replicated in another 5% data sample (dataset 2).

RESULTS

Data were extracted from the CDSSs of 123 community pharmacies. Dataset 1 consisted of 841 572 dispensed prescriptions and 298 261 drug interaction alerts. Dataset 2 was comparable. Twenty-two frequently occurring alert combinations were identified. Analysis of these associated alert combinations for similar management recommendations resulted in 3 clusters (related to renal function, electrolytes, diabetes, and cardiovascular diseases). Using the clusters in alert generation reduced the alert rate within these clusters by 53-70%. The overall number of drug interaction alerts was reduced by 11% in dataset 1 and by 12% in dataset 2. This corresponds to a decrease of 21 alerts per pharmacy per day.

DISCUSSION AND CONCLUSION

Using clusters of drug interaction alerts with similar management recommendations in CDSSs can substantially decrease the overall alert rate. Further research is needed to establish the applicability of this concept in daily practice.