Please provide a literature review of studies assessing medication warning optimization in the electronic health record in general patients as well as pediatric and oncology populations.

Comment by InpharmD Researcher

Studies of medication warning optimization in the electronic health record generally report reduced alert burden and improved clinician responses after removal of low-value warnings, adjustment of severity and triggering criteria, incorporation of patient-specific information, and selective use of noninterruptive decision support. These efforts commonly involve multidisciplinary review, clinician feedback, and ongoing monitoring. Pediatric evidence emphasizes customized dosing rules that account for weight, age, clinical context, and rounding, while oncology evidence emphasizes regimen-specific dosing safeguards, cumulative-dose monitoring, and treatment-plan compatibility. Across populations, the literature supports continued refinement of warnings to improve clinical relevance, although evidence is stronger for alert and prescribing-process outcomes than for reductions in patient harm. Oncology safety findings frequently reflect comprehensive prescribing systems rather than warning changes alone. Please see the summaries and tables below for a more detailed overview.

Background

General population:

Several reviews have evaluated strategies to optimize medication-related electronic health record (EHR) alerts. A 2023 scoping review of 16 post-implementation medication-alert optimization studies identified six primary interventions: alert inactivation, severity reclassification, addition of contextual information, adjustment of triggering thresholds, provision of additional information within alerts, and encounter-specific suppression. Alerts were commonly selected for optimization based on firing frequency, clinician or pharmacist feedback, or review of clinical significance, generally through multidisciplinary processes involving pharmacists and physicians. Alert inactivation reduced alert quantity by 1% to 25.3%, while severity reclassification reduced alert quantity by 1% to 16.5% in 6 of 7 studies and increased alert acceptance by 4.2% to 50.2%; high-severity alerts reached 100% acceptance in studies reporting this outcome. Overall, alert quantity or frequency decreased in 10 of 15 studies reporting these outcomes, and 6 of 7 studies evaluating acceptance or override rates reported improvement. Clinical error outcomes were reported in only 4 studies and either decreased or showed no significant change where sufficient comparative data were available. [1]

Furthermore, a 2021 systematic review of hospital CDS alert governance similarly found that all 8 included studies used multidisciplinary committees to guide alert selection or optimization, with 7 studies specifically addressing medication-related alerts. Optimization processes additionally incorporated clinician feedback, alert firing and override data, literature and drug references, and, in some institutions, visual dashboards. Specific modifications included deactivating or downgrading low-value DDI alerts, tailoring alerts to clinician type, suppressing alerts in selected clinical contexts, modifying trigger criteria, and redesigning alert presentation. Six studies evaluated modifications after implementation, with 5 reporting reduced alert rates. Individual interventions included removal of 802 of 875 moderate DDI alerts at one health system, contributing to a 27% reduction in weekly medication alerts; severity-based DDI recategorization that reduced alerts by 55% and pharmacist DDI-review time by 45%; and deactivation of 63 DDI alerts at a pediatric hospital, reducing pharmacist alerts from 58.74 to 25.11 per 100 orders and provider alerts from 19.73 to 15.11 per 100 orders. [2]

Complementing these findings, a 2022 review of interruptive alert stewardship described several approaches for optimizing alert effectiveness, including identifying high-volume alerts and alerts disproportionately affecting specific patients or clinicians, incorporating end-user feedback and anomaly detection to identify malfunctioning or poorly designed alerts, and applying human-factors principles to alert design. Alert redesign strategies included clearly communicating why an alert was displayed, incorporating relevant clinical information, providing directions for the expected user action, keeping alert text concise, and clarifying alert acknowledgment options. Alert performance could be evaluated using multiple measures, including firing frequency, acceptance or override rates, appropriateness of responses, intended clinical actions, and downstream clinical outcomes rather than relying on a single measure. The review also emphasized multidisciplinary governance, representative end-user involvement, establishing criteria for when an interruptive alert is warranted, standardized alert design, and preimplementation validation or background testing, with continued monitoring of alert performance and unintended consequences after implementation. Collectively, these reports support an ongoing, multidisciplinary approach to medication-alert optimization that prioritizes clinically relevant alerts, reduces low-value alerting, and monitors for unintended consequences following modification. [3]

Pediatric:

The 2024 American Academy of Pediatrics (AAP) clinical report recommends pediatric-specific EHR medication decision support that incorporates pharmacologic databases with biometrically appropriate and technically deliverable dosing. Medication clinical decision support (CDS) should clearly display recommended dosing parameters, rounding factors, calculation steps, relevant biometric data (eg, weight, body surface area, gestational or chronologic age), and special dosing considerations; verify that biometric data are sufficiently current; and account for age- and gestational age–dependent dosing recommendations. The AAP’s 2008 technical report, which was retired in June 2025, provides additional detail, recommending pediatric computerized provider order entry (CPOE) with weight/body surface area–based dosing; pediatric-specific single-dose, daily-dose, and cumulative-dose ranges and alerts, including lifetime cumulative doses for chemotherapy; contextual dosing decision support for neonatal, renal, oncology, and other clinical states; and automated calculations and dose limits for larger patients. The retired report also emphasizes that CDS rules should prioritize high-impact alerts to avoid alert fatigue and that pediatric drug-dose ranges and alerts should be informed by pediatric clinical expertise. [4], [5]

Oncology:

According to the American Society of Health-System Pharmacists (ASHP) guidelines on preventing medication errors with chemotherapy and biotherapy, CPOE should incorporate clinical decision support with appropriate alerts for allergies, drug interactions, duplicate therapies, and maximum doses, consistent with the institution’s overall alerting strategy. Chemotherapy-specific decision support may also incorporate patient parameters such as laboratory values and comorbidities, while dose and administration limits should be established and active alerts used when defined chemotherapy limits are exceeded. The guidelines recommend continued review and modification of CPOE after implementation using provider feedback, safety and near-miss data, with changes considered through a multidisciplinary process involving clinicians and information systems staff. [6]

A 2015 literature review examined the implementation, safety outcomes, and challenges of electronic prescribing for intravenous and oral chemotherapy. Implementation approaches included modifying existing CPOE systems, acquiring chemotherapy-specific prescribing software, and developing institutional prescribing modules. Common safeguards included standardized regimen templates, automatic dose calculation and adjustment, relevant patient and laboratory data, pharmacist validation, and drug-drug interaction, duplicate-therapy, and dose-limit alerts. Across intravenous chemotherapy studies, electronic prescribing generally reduced prescribing errors and improved documentation. One sequential comparison reported error rates of 4.2% with handwritten orders, 1.5% with preprinted orders, and 0.1% with CPOE; another found lower error rates with computerized prescriptions than spreadsheet prescriptions, 11.8% versus 20.4%, with a relative risk of 0.57 (95% confidence interval, 0.47-0.72). An interrupted time series study found a relative risk of 0.06 (95% confidence interval, 0.03-0.10) for dose-calculation errors after electronic prescribing. However, some studies identified new errors, including treatment-plan and dose mismatches caused by interface problems, failure to carry dose reductions into later cycles, and administration changes resulting from incomplete documentation. [7]

Oral chemotherapy interventions incorporated drug-specific maximum doses, frequency restrictions, conservative defaults, daily or weekly cumulative-dose alerts, monitoring orders, and treatment-guideline links. One before-and-after study found approximately 69% lower odds of prescribing errors, with an odds ratio of 0.31 (95% confidence interval, 0.11-0.89), and elimination of observed dose-limit and methotrexate-associated errors. In another study of 6,673 oral chemotherapy prescriptions, dose-limit warnings occurred in 6%, but only 4% of prescriptions generating warnings were aborted. The review identified cost, workflow changes, training, interoperability, and unsuitable alerts as implementation challenges. The authors recommended multidisciplinary process assessment, failure modes and effects analysis, repeated order-set testing, phased implementation, and ongoing review to remove low-value alerts and address newly introduced errors. They concluded that electronic prescribing generally reduced errors, particularly calculation errors, but methodological limitations and short follow-up restricted conclusions about its overall effect on cancer-care safety and quality. [7]

A 2011 implementation report details Memorial Sloan-Kettering Cancer Center’s multiyear transition from paper chemotherapy orders to customized computerized provider order entry (CPOE) across inpatient, outpatient, and regional sites. A multidisciplinary team of oncologists, nurses, pharmacists, operational managers, and systems specialists collaborated with the software vendor to reproduce established ordering and verification processes while adding electronic safeguards. Regimen-specific forms grouped linked medication orders and incorporated automatic calculations using body surface area, weight, flat dosing, or area under the curve; dose caps and rounding; percentage-based dose modifications with documented reasons; relevant laboratory results; and restrictions on administration routes and units. Orders entered by fellows or nurse practitioners required attending approval, followed by nursing and pharmacy verification, and modified orders restarted the verification sequence. Safety features included allergy, drug-drug interaction, duplicate-order, weight-change, and previously discontinued drug warnings, with maximum-dose checking and cumulative anthracycline alerts available in the pharmacy system. Order sets underwent detailed multidisciplinary review, scenario-based testing, and two clinical pilots, with feedback incorporated before sequential rollout by disease service. Prescribers received individual training before access was granted, and electronic ordering became mandatory after rollout. [8]

According to the report, the institution had created more than 1,250 adult and 466 pediatric order sets; in 2010, 73,225 order sets were entered by more than 416 practitioners across eight sites, supporting more than 225,000 chemotherapy doses. Reported incomplete orders decreased from 197 in 2008 to 77 in 2009, and the authors reported elimination of problems caused by illegible or lost orders, although those problems had not been formally tracked. Ongoing review of actual and near-miss events prompted additional warnings for weight changes exceeding 10% and unmet pretreatment requirements, such as pregnancy testing. Order-set use was reviewed every six months, and obsolete sets were inactivated. Limitations included the absence of formal comparative safety and alert data, cumulative-dose information residing only in the pharmacy system, and continuing challenges with carboplatin dosing as renal function changed. The authors concluded that successful implementation required close vendor collaboration, standardized workflows, clinician champions, training, and continuous monitoring and refinement. [8]

Finally, a 2025 case study describes the identification and redesign of a malfunctioning clinical decision support (CDS) alert intended to ensure that children enrolled in therapeutic cancer trials received the corresponding research chemotherapy treatment plan. Implemented in Epic in 2018, the alert compared treatment-plan tags with disallowed tags in the patient’s research study record. However, “SickKids” had been automatically entered as a disallowed tag in all research records while also being assigned to every institutional chemotherapy plan. Consequently, warnings appeared even when the correct research plan was selected or when patients were enrolled only in ancillary studies, such as biobanking. The malfunction persisted for nearly six years before a physician reported it. [9]

Applying alert logic to treatment-plan application reports, the investigators estimated that warnings occurred with 93.4% of treatment-plan applications, corresponding to 1,039 alerts in 2023, or approximately 17.3 alerts per provider annually among 60 providers. The alert was absent from standard vendor alert-monitoring reports, limiting its detection and accurate quantification. The team removed the conflicting tag from existing research records and disabled its automatic addition, stopping the warning. Rather than rebuilding an interruptive alert, they manually linked compatible chemotherapy plans to each research study and configured a default “research” tab to preferentially display those plans during ordering. Research plans were also hidden from searchable options for patients not enrolled in the corresponding study. The authors emphasized clinician engagement, user acceptance testing, planned performance reviews, and passive CDS approaches that guide selection within the workflow. Clinical harm was not established, and the report did not quantify postimplementation changes in incorrect treatment-plan selection, workflow, or patient outcomes.[9]

Literature Review

A search of the published medical literature revealed 11 studies investigating the researchable question:

Please provide a literature review of studies assessing medication warning optimization in the electronic health record in general patients as well as pediatric and oncology populations.

Level of evidence

C - Multiple studies with limitations or conflicting results Read more→

Please see Tables 1-11 for your response.

Optimizing drug-dose alerts using commercial software throughout an integrated health care system
Design	Case Report
Objective	To develop and evaluate generalizable strategies to optimize drug-dose alerting
Methods	All drug-dose alerts were activated silently in the electronic health record (EHR) and evaluated over 3 months using default EHR settings and vendor-provided drug-dose values. Alerts were categorized by alert type, care setting, and pediatric-only versus nonpediatric-only departments. A 2-step optimization approach consisted of system-level changes followed by drug-specific changes. System-level strategies included turning off entire alert categories or incrementally increasing alert-firing thresholds; final thresholds were selected by 3 physicians and a clinical pharmacist to balance reductions in presumed false-positive alerts with potential patient risk. Drugs accounting for >1% of all alerts were subsequently identified for drug-specific modification of alert-firing thresholds. Self-reported drug-dose errors from the health system’s patient safety reporting system were also identified during the study period.
Duration	Over 3 months in 2013
Outcome Measures	Reduction in drug-dose alerts; maintenance of clinically significant alerts
Results	Drug-dosing alert category	Baseline drug-dose alerts, % (n)	Optimization of drug-dose alerts, % (n)	Decrease in drug-dose alerting, %
	Minimum drug-dose daily dose alerts (removed)	24 (24,508)	0 (0)	100
	Minimum drug-dose frequency alerts (removed)	10 (10,330)	0 (0)	100
	Maximum drug-dose duration alerts (removed)	5 (4,972)	0 (0)	100
	Maximum drug-dose single-dose alerts (increased to 125% of threshold)	23 (23,539)	42 (19,503)	17
	Maximum drug-dose daily dose alerts (increased to 125% of threshold)	23 (24,183)	45 (21,052)	13
	Maximum drug-dose dose frequency alerts (increased to more than 2 doses/day of threshold)	16 (16,566)	14 (6,433)	61
	Among 834,911 medication orders, 104,098 silent drug-dose alerts were generated at baseline (12% of orders). Removing minimum daily-dose and minimum-frequency alerts eliminated 24% and 10% of alerts, respectively; increasing maximum single- and daily-dose thresholds to 125% reduced these alerts by 13% and 17%, respectively, and changing frequency alerts to >2 doses/day above the recommended frequency reduced these alerts by 61%. Overall, the 6 system-level interventions reduced the alert rate from 12% to 5% (46,988 alerts), and subsequent drug-specific modifications to the 22 highest-alerting drugs further reduced the rate to 3% (25,455 alerts). Drug-dose alerts were most prevalent in inpatient and pediatric-only settings. Of 32 reported drug-dosing errors, 3 involved medication ordering and would have triggered an alert both before and after optimization; the remaining 29 were administration-related and would not have been affected by drug-dose ordering alerts.
Study Author Conclusions	Computerized provider order entry (CPOE) drug-dose alerts using commercial EHRs and drug information can be optimized using system-level and drug-specific approaches, significantly decreasing alerts while maintaining alerts for potentially clinically significant overdoses, thus improving patient safety.
Critique	The study demonstrated substantial reductions in alert burden across multiple care settings; however, alerts were evaluated silently, clinician acceptance was not measured, and the potential for increased false-negative alerts or patient harm after optimization was not objectively assessed.

Table 1 References:
[10] Saiyed SM, Greco PJ, Fernandes G, Kaelber DC. Optimizing drug-dose alerts using commercial software throughout an integrated health care system. J Am Med Inform Assoc. 2017;24(6):1149-1154. doi:10.1093/jamia/ocx031

A systematic approach to optimize electronic health record medication alerts in a health system
Design	Quasi-experimental study
Objective	To evaluate the effectiveness of a systematic, streamlined approach to optimize drug–drug interaction alerts in an electronic health record for a health system
Methods	A multidisciplinary medication-related clinical decision support committee reviewed and modified alerts using three primary strategies: modification of existing knowledge-database alerts, suppression of drug–drug interaction and duplicate-therapy alerts within order sets, and drug–drug interaction tiering and reclassification. Changes were reviewed and approved through a systemwide committee process. Alert frequency, severity, and responses were compared across 29 weeks preintervention and 52 weeks postintervention, with modification/acceptance and acknowledgment/override rates and alerts per 100 medication orders evaluated.
Duration	June 2016 to January 2018
Outcome Measures	Primary: Weekly overall, modification, and acknowledgement rates of medication alerts after drug–drug interaction reclassification Secondary: Sub analysis of types of medication alerts (drug–drug interaction and duplicate therapy alerts) and alert use by providers (pharmacist and prescribers)
Results	Following the intervention, weekly inpatient medication alerts decreased from 68,900 to 50,300 (p< 0.001). Drug–drug interaction alerts decreased significantly, whereas duplicate-therapy alerts increased (p< 0.02 for both). Weekly alert acknowledgment increased from 11.8% to 13.7% (p< 0.001), and alerts resulting in medication-order modification increased from 5.0% to 7.3% per 100 medication orders (p< 0.001), primarily due to increased modification following drug–drug interaction alerts. Pharmacist alert acknowledgment or modification increased after the intervention (p< 0.001), whereas prescriber response remained unchanged (p> 0.4) despite an increase in total medication alerts presented to prescribers.
Study Author Conclusions	A committee-led systematic approach to optimizing drug–drug interactions facilitated a significant decrease in the overall number of alerts and an increase in both medication alert acknowledgement and modification rates.
Critique	The study demonstrated a significant reduction in the number of alerts and an increase in alert acknowledgment and modification rates, highlighting the effectiveness of a systematic approach. However, the study was conducted at a single large academic center, which may limit the generalizability of the findings to other settings. The reliance on a multidisciplinary committee with dedicated IT support may not be feasible in all health systems. Additionally, the long-term sustainability of the committee's impact and the clinical significance of suppressing certain alerts require further investigation.

Table 2 References:
[11] Bhakta SB, Colavecchia AC, Haines L, Varkey D, Garey KW. A systematic approach to optimize electronic health record medication alerts in a health system. Am J Health Syst Pharm. 2019;76(8):530-536. doi:10.1093/ajhp/zxz012

Differences, Opportunities, and Strategies in Drug Alert Optimization—Experiences of Two Different Integrated Health Care Systems
Design	Retrospective study across two health care systems
Objective	To report on comprehensive drug alerting rates and develop strategies across two different health care systems to reduce the number of drug alerts
Inclusion Criteria	All alerts across all care settings
Exclusion Criteria	None
Methods	Standardized reports compared drug alert rates between two systems. Each system used an integrated team consisting of the Chief Medical Informatics Officer, pharmacy informatics specialists, and physician informatics specialists to review alert frequency and clinical utility, with final implementation decisions made by the pharmacy and therapeutics committee. Optimization included suppression of entire alert categories, modification of system-level settings, and drug-specific changes. At MetroHealth System (MHS), interventions included disabling minimum-dose and incomplete-information dose alerts, increasing maximum single- and daily-dose thresholds to 125%, increasing the dose-frequency threshold to >2 doses/day above the initial threshold, and modifying alerts for the 22 highest-alerting medications. At CaroMont Health (CH), interventions included suppressing undetermined-severity and moderate-interaction alerts; daily-dose-exceeded, daily-PRN-exceeded, and frequency-exceeded alerts; duplicate-therapy and duplicate-medication alerts; drug–drug interaction alerts within order sets; and alerts across transitions of care. Alerts for calcium gluconate and iron sucrose injection were also disabled because institutional dosing differed from the drug database.
Duration	December 2016 to February 2017
Outcome Measures	Drug alerting rates; identification of principles for drug alert optimization
Baseline Characteristics		Drug alerts	Drug alerts (%)	Alert/100 orders
	Duplicate therapy	63,305	35	21
	Drug–drug	53,650	30	17
	Drug allergy	20,979	12	7
	Duplicate medication order	20,119	11	7
	Dose	14,199	8	5
	Lactation	5,258	3	2
	Pregnancy	1,603	1	1
	TPN	134	0	0
	Abbreviations: TPN, total parenteral nutrition. Baseline alert data were available across all alert categories for CH but not for MHS, limiting direct comparisons between the two systems.
Results		CH alerts	CH alerts/100 orders	MHS alerts	MHS alerts/100 orders
	Duplicate therapy	55,685	19	22,392	2
	Drug–drug	45,336	15	207,998	17
	Drug allergy	20,646	7	72,986	6
	Duplicate medication order	18,887	6	99,791	8
	Drug–disease	N/A	N/A	17,492	2
	Dose	7,768	3	41,004	3
	Lactation	109	0.1	1	0
	Pregnancy	199	0.1	1,747	0
	TPN	131	0.1	1,975	0
	At MHS, 465,286 alerts fired across 1,189,445 drug orders, corresponding to an alert rate of approximately 39%. System-level and drug-specific optimization reduced dose alerts from 12% to approximately 3% of medication alerts. At CH, 179,247 alerts fired during the preoptimization period (58 alerts/100 orders); optimization reduced duplicate-therapy alerts from 21 to 19, dose alerts from 7 to 5, and drug–drug interaction alerts from 17 to 15 alerts/100 orders. Overall, CH alerts decreased from 179,247 to 148,763 (58 to 51 alerts/100 orders), removing 30,484 (9%) unfiltered alerts. Following optimization, drug–drug interaction, drug-allergy, and duplicate-medication-order alerts remained among the highest alert categories across both systems.
Study Author Conclusions	A team-based, systematic approach to optimizing drug-alerting strategies can reduce the number of drug alerts, but alert rates still remain high. Additional tactical guidelines and recommendations need to be developed to enhance out-of-the-box clinical decision support for drug alerts.
Critique	Based on lessons learned across the two health systems, the authors recommend a multidisciplinary approach using a trusted drug-information source, standardized reporting, silent alerting before implementation, system-level followed by targeted drug-specific modifications, and ongoing review with end-user feedback. Limitations included evaluation of a single EHR vendor, lack of baseline MHS data across all alert categories, and no assessment of the impact of reduced alert burden on override rates.

Table 3 References:
[12] Saiyed SM, Davis KR, Kaelber DC. Differences, Opportunities, and Strategies in Drug Alert Optimization-Experiences of Two Different Integrated Health Care Systems. Appl Clin Inform. 2019;10(5):777-782. doi:10.1055/s-0039-1697596

Creating Meaningful Alerts and Reducing Alert Fatigue: Strategies Implemented by Informatics Pharmacists to Optimize Dose Range Checking Alerts in a Multihospital Health System
Design	Single-center, retrospective quality-improvement/implementation study
Objective	To highlight the impact that informatics pharmacists can make by implementing different strategies to decrease nuisance alerts and create clinically meaningful dose range checking (DRC) alerts that guide clinicians in their practice
Methods	Following activation of DRC alerts for 3,623 medication groupers, 623 of the highest-alerting groupers were selected for clinical review. Twenty-one pharmacists independently reviewed DRC content and established minimum and maximum single- and daily-dose limits according to age and renal function, using internal policies/guidelines, prescribing information, drug databases, and peer-reviewed literature. Completed DRC tables were incorporated into Cerner Millennium, with a 10% variance applied to weight-based dose limits. Weekly audits identified frequently firing and overridden alerts, and pharmacists could report clinically inappropriate alerts. Optimization focused on three interventions: modifying order sentences for high-alerting medications (including dose, route, and age/weight filters); modifying rule logic to suppress alerts for discontinued orders; and suppressing five DRC error messages related to multiple groupers, missing/invalid order dose, missing/invalid creatinine clearance, missing/invalid patient weight, and inability to convert units. The changes applied to both inpatient and outpatient dose checking. Pasted markdown
Duration	July to September 2018
Outcome Measures	Reduction in the number of overridden DRC alerts; improvement in the quality of DRC alerts
Results	Month-year	Number of DRC alerts	Number of orders	Ratio (%)
	January-2019	14,731	224,721	6.6
	February-2019	11,768	190,354	6.2
	March-2019	11,190	197,422	5.7
	April-2019	12,028	242,282	5.0
	May-2019	8,560	192,296	4.5
	June-2019	8,638	194,362	4.4
	July-2019	9,997	239,576	4.2
	August-2019	7,408	198,916	3.7
	September-2019	9,212	252,826	3.6
	October-2019	7,569	211,590	3.6
	November-2019	7,463	207,644	3.6
	December-2019	7,892	211,480	3.7
	January-2020	10,500	270,980	3.9
	February-2020	8,312	218,136	3.8
	March-2020	5,893	183,754	3.2
	At baseline in June 2018, an average of 11,452 DRC alerts were overridden per week, with a peak of 13,183. Following implementation beginning in July, overridden alerts decreased to 4,796 by the first week of September, representing a 63% decrease over 3 months. Oxycodone-acetaminophen generated the most overridden alerts during one baseline week (n= 600), with approximately 88% attributed to missing/invalid weight and multiple-grouper alerts. Overridden DRC alerts subsequently decreased from 14,731 (6.6% of 224,721 orders) in January 2019 to 5,893 (3.2% of 183,754 orders) in March 2020, a 60% decrease in absolute number.
Study Author Conclusions	By reducing the number of DRC nuisance alerts and improving the clinical content of DRC alerts, informatics pharmacists can contribute to lowering alert fatigue and improving providers’ trust in clinical decision support (CDS) alerts.
Critique	The study effectively demonstrates the role of informatics pharmacists in reducing alert fatigue through systematic strategies. However, it lacks a direct correlation between specific interventions and their impact on the rate of overridden DRC alerts. Further analysis is needed to assess the impact of alert suppressions on adverse events and patient outcomes. Additionally, the study does not include data on the number of order sentences updated or alerts triggered on discontinued orders.

Table 4 References:
[13] Choukroun JF, Lee K, Rey A. Creating Meaningful Alerts and Reducing Alert Fatigue: Strategies Implemented by Informatics Pharmacists to Optimize Dose Range Checking Alerts in a Multihospital Health System. J Pharm Technol. 2022;38(6):319-325. doi:10.1177/87551225221117152

Transitioning Ineffective Medications on Hold Alert from Interruptive to Noninterruptive Alert to Decrease Alert Burden
Design	Pre–post intervention cohort study
Objective	To evaluate if transitioning a high-firing medication on hold alert from interruptive to noninterruptive would change provider practices
Methods	The medication-on-hold alert, which triggered when at least 2 medications had been held for >48 hours, was transitioned from an interruptive alert to a noninterruptive display in the patient storyboard. Provider practices were evaluated during the 6 months before and 6 months after the transition, including median time to medication resumption, number of patients whose medications were not resumed before discharge, and the most common medications not resumed. Electronic health record (EHR) data were used to assess alert activity, medications placed on hold, and hold duration. Medication-related safety reports were also searched using predefined medication-hold terms, with identified reports manually reviewed.
Duration	6 months before and after transitioning the alert (April 2022–September 2022 and October 2022–April 2023)
Outcome Measures	Number of actions taken by clicking on the alert; median hold duration of medications; frequency of medications never resumed; reporting of medication-on-hold safety events
Results		Interruptive alert	Non-interruptive Alert	p-Value
	Any action taken	33,623	305	NA
	Median hold time (all held meds)	21.5 hours	21.1 hours	0.83
	Median hold time (>48-hour meds)	81.5 hours	85.6 hours	0.22
	Unique patients with meds never unheld	43% (4,697 patients)	43% (4,910 patients)	NA
	During the 6-month interruptive-alert period, the alert fired 33,632 times across 11,164 hospital admissions lasting >48 hours (3.0 firings/encounter). After transition to a noninterruptive alert, alert clicks decreased from 33,632 to 305 (0.02 clicks/encounter), while 315,032 passive firings were recorded.
Adverse Events	There was no increase in reported medication-on-hold safety events, with 13 verified medication-hold errors before versus 14 after the transition; none were attributed to medications being held for >48 hours.
Study Author Conclusions	The initial interruptive medication on hold alert was not effective and contributed to a high volume of alerts in our institution. Transitioning the medications on hold alert from an interruptive to a noninterruptive alert reduced potential alert fatigue without significantly impacting clinical outcomes. These findings highlight the need for careful evaluation of clinical decision support (CDS) alerts to balance clinical utility and provider alert burden. Alerts that do not affect the desired clinical outcome should be redesigned or retired.
Critique	The study effectively demonstrated a reduction in alert fatigue without adverse effects on clinical outcomes. However, it was limited by the inability to track clinician interactions with noninterruptive alerts and the reliance on safety event reporting, which may underestimate associated safety events. Future studies should include mechanisms to track clinician responses to noninterruptive alerts to better understand their impact on clinical decision-making and patient care.

Table 5 References:
[14] Knake LA, Kettelkamp JM, Bronson A, Meyer N, Hacker K, Blum JM. Transitioning Ineffective Medications on Hold Alert from Interruptive to Noninterruptive Alert to Decrease Alert Burden. Appl Clin Inform. 2025;16(4):848-854. doi:10.1055/a-2632-0605

Improving Allergy Documentation and Clinical Decision Support in the Electronic Health Record (EHR)
Design	Multi-step study with quantitative and qualitative analyses, involving real-time EHR clinical decision support tool implementation
Objective	To improve allergy documentation by developing a comprehensive value set, implementing an innovative allergy reconciliation module within the electronic health record (EHR), redesigning drug allergy alerting mechanisms, and distributing outputs to healthcare institutions and the research community
Methods	Historical drug-allergy alert data from Mass General Brigham (MGB) and UCHealth were analyzed to develop a tiered alerting approach for penicillin, cephalosporin, and opioid alerts. Alert override rates and reaction severity were evaluated to identify which alerts should be interruptive, with opioid alerts for non–immune-mediated reactions considered for silent or noninterruptive presentation. Penicillin and cephalosporin alert logic incorporated existing clinical pathways and historical alert data. New alert rules incorporated allergy group, cross-sensitivity group, reaction severity, and alert history, and alerts involving multiple reactions were redesigned to display a single alert for the most severe reaction. The tiered alerting mechanism was implemented in a simulated web application, with alert layout, wording, and timing iteratively reviewed and revised by an interdisciplinary team.
Duration	May 7, 2018 - April 30, 2022
Outcome Measures	Allergy documentation quality, use of coded versus free-text reaction documentation, allergy-reconciliation algorithm performance and use, drug-allergy alert override patterns, and estimated reduction in alerts with the proposed tiered alerting approach
Results	Across MGB and UCHealth, 54.5% of penicillin and 55.1% of cephalosporin alerts were overridden, with override rates of 89.8%–93.9% for penicillin alerts triggered by cross-sensitive class matches. Approximately two-thirds of opioid alerts were overridden, including >700,000 interruptive alerts for 50,527 patients involving morphine-analogue opioids; codeine and morphine alerts were most frequently overridden. Most reactions triggering opioid alerts were non–immune-mediated, while approximately one-fifth were immune-mediated. At MGB, override rates for low-severity allergy entries were 36.7% for penicillins and 39.1% for cephalosporins, compared with 6.0% and 6.8%, respectively, for higher-severity reactions. Informational and interruptive alert prototypes incorporating the proposed tiering approach were developed and tested through clinical scenarios in a web-based simulation environment, with scenarios reviewed by a multidisciplinary team.
Study Author Conclusions	The study concluded that enhancing allergy documentation and clinical decision support in EHRs is feasible and effective. The implementation of comprehensive picklists, reconciliation modules, and tiered alerting mechanisms can improve documentation accuracy, reduce alert fatigue, and enhance patient safety.
Critique	The study's strengths include its comprehensive approach to improving allergy documentation and alerting in EHRs. However, limitations include the lack of real-time testing for the alert tiering mechanism and the challenges in fully integrating the tools within existing EHR systems due to vendor restrictions. The study also faced potential confounding variables due to the COVID-19 pandemic during the implementation phase.

Table 6 References:
[15] Zhou L. Improving allergy documentation and clinical decision support in the electronic health record: final report. Agency for Healthcare Research and Quality; 2022. Accessed September 30, 2026. https://digital.ahrq.gov/ahrq-funded-projects/improving-allergy-documentation-and-clinical-decision-support-electronic-health/final-report

Optimization of drug–drug interaction alert rules in a pediatric hospital’s electronic health record system using a visual analytics dashboard
Design	Cross-sectional study N= 2,391,880 medication alerts evaluated
Objective	To develop and evaluate an electronic dashboard of hospital-wide electronic health record medication alerts for an alert fatigue reduction quality improvement project
Methods	A visual analytics dashboard was developed to monitor medication-alert volumes and override rates and was used to identify frequently triggered drug-drug interaction (DDI) alerts for review. The Clinical Decision Support Committee prioritized high-frequency DDI alerts, and 10 pediatric clinical pharmacists independently reviewed selected alerts in pairs using Medi-Span, with comparison to Lexicomp and Micromedex and primary literature when needed. DDI alerts considered clinically insignificant were recommended for deactivation; disagreements were resolved by pharmacy-group consensus, and controversial alerts were reviewed with the physician groups most likely to encounter them. Recommendations underwent Clinical Decision Support Committee review and Therapeutic Standards Committee approval. Three successive interventions deactivated clinically irrelevant DDI alert rules; no other alert types were systematically modified. Alert and clinician-response rates before and after the interventions were evaluated for all medication alerts and for DDI alerts that remained active and unchanged throughout the study, with medication safety events monitored as a balancing measure.
Duration	January 31, 2011 to January 26, 2014
Outcome Measures	Reduction in alert and override rates; decrease in medication serious safety event rate
Baseline Characteristics		Pharmacists	Providers
	Baseline median alert rate	58.74 alerts/100 orders	19.73 alerts/100 orders
	Baseline median override rate	95.14 overrides/100 alerts	84.22 overrides/100 alerts
Results		Pharmacists	Providers	p-value
	Alert rate after interventions	25.11 alerts/100 orders	15.11 alerts/100 orders	<0.001
	Override rate after interventions	84.38 overrides/100 alerts	84.91 overrides/100 alerts	0.16
	A total of 63 DDI alert rules were deactivated across the three interventions (2, 42, and 19, respectively), primarily affecting pharmacists. For DDI alerts that remained unchanged, pharmacist override rates decreased from 93.56 to 85.71 overrides/100 alerts (difference, 7.85%; 95% confidence interval [CI] 6.63%–9.07%; p< 0.001).
Adverse Events	The medication serious safety event rate decreased from 0.18 to 0.08 events/10,000 adjusted patient-days, with no serious safety events reported in association with the deactivated DDI alert rules.
Study Author Conclusions	An alert dashboard facilitated safe rapid-cycle reductions in alert burden that were temporally associated with lower pharmacist override rates in a subgroup of DDIs not directly affected by the interventions; meanwhile, the pharmacists’ frequency of selecting the "cancel" option increased. We hypothesize that reducing the alert burden enabled pharmacists to devote more attention to clinically relevant alerts.
Critique	The study effectively demonstrated the use of a visual analytics dashboard to reduce alert fatigue and improve medication safety. However, the study was limited by its setting in a single hospital and the lack of generalizability to other EHR systems. Additionally, the study did not address the potential for missing relevant alerts after deactivation of certain rules.

Table 7 References:
[16] Simpao AF, Ahumada LM, Desai BR, et al. Optimization of drug-drug interaction alert rules in a pediatric hospital's electronic health record system using a visual analytics dashboard. J Am Med Inform Assoc. 2015;22(2):361-369. doi:10.1136/amiajnl-2013-002538

Appropriateness of commercially available and partially customized medication dosing alerts among pediatric patients
Design	Retrospective analysis N= 573
Objective	To evaluate dosing alert appropriateness, categorize orders with alerts, and compare the appropriateness of alerts due to customized and non-customized dose ranges at a pediatric hospital
Study Groups	All patients (N= 573)
Inclusion Criteria	Inpatient pediatric medication orders
Exclusion Criteria	Outpatient prescriptions, patients ≥18 years of age, and research protocols
Methods	The electronic health record (EHR) used vendor-provided pediatric dose ranges for >2,000 medications, with >600 dose ranges customized by the institution through modification or addition of parameters such as age or weight range, minimum and maximum dose, and route; customized ranges took precedence over vendor ranges, without changes to the dosing-alert rule logic. Each alerted order was reviewed using patient-specific clinical information, Lexi-Comp pediatric dosing recommendations, and institution-specific recommendations and classified as an appropriate or inappropriate alert based on dose appropriateness and availability of dosing recommendations. Alerted orders were also evaluated according to whether the alert resulted from a customized or noncustomized dose range and were compared across age and weight categories.
Duration	July 2011
Outcome Measures	Primary: Appropriateness of dosing alerts Secondary: Categorization of medication orders with alerts, comparison of customized vs. non-customized dose ranges
Baseline Characteristics		All patients (N= 573)
	Median age, years	6.1 (0-17.9)
	Female	289
Results		No. overall orders (% alerted orders)
	Justifiable dose, inappropriate alert	1,122 (58.0%)
	Correct dose, inappropriate alert	545 (28.2%)
	Incorrect dose, appropriate alert	155 (8.0%)
	No recommendations, appropriate alert	113 (5.8%)
	Among 47,181 inpatient pediatric medication orders, 1,935 (4.1%) generated 3,774 dosing alerts, and all alerted orders were overridden by the prescriber. Of alerted orders, 86.2% had inappropriate alerts (58.0% for justifiable doses and 28.2% for correct doses), while 13.8% had appropriate alerts (8.0% for incorrect doses and 5.8% with no dosing recommendations available). Appropriate alerts occurred more frequently with customized versus noncustomized dose ranges (19.7% vs 12.8%; p= 0.002); customized ranges also generated higher proportions of alerts for incorrect doses (11.2% vs 7.4%; p= 0.027) and medications without dosing recommendations (8.5% vs 5.4%; p= 0.034). Customization provided 33 additional appropriate alerts for incorrect doses and 25 additional appropriate alerts for orders without dosing recommendations that would not have occurred with noncustomized ranges.
Study Author Conclusions	The vast majority of dosing alerts were presented to practitioners inappropriately, potentially contributing to alert fatigue. Appropriate alerts occurred more often when alerts were due to customized ranges. Advances in dosing alerts should aim to provide accurate and clinically relevant alerts that minimize excessive inappropriate alerting. Medications requiring dosing adjustments based on clinical parameters must be taken into account when designing and evaluating dosing alerts.
Critique	The study highlights significant issues with dosing alert appropriateness, particularly the high rate of inappropriate alerts leading to potential alert fatigue. The retrospective design and single-center setting may limit generalizability. The study effectively identifies areas for improvement in CDS systems, but further research is needed to evaluate the impact of these alerts on clinical outcomes.

Table 8 References:
[17] Stultz JS, Nahata MC. Appropriateness of commercially available and partially customized medication dosing alerts among pediatric patients. J Am Med Inform Assoc. 2014;21(e1):e35-e42. doi:10.1136/amiajnl-2013-001725

Computerized Dose Range Checking Using Hard and Soft Stop Alerts Reduces Prescribing Errors in a Pediatric Intensive Care Unit
Design	Retrospective study
Objective	To evaluate prescription error rates in the pediatric intensive care unit and the pediatric cardiovascular intensive care unit using an enhanced dose range checking system with soft and hard alerts
Study Groups	N/A
Inclusion Criteria	All patients who received a medication ordered via CPOE in the PICU and pediatric cardiovascular ICU during the designated study periods
Exclusion Criteria	N/A
Methods	Retrospective before-and-after study evaluated a customized dose range checking system in pediatric intensive care and cardiovascular intensive care units. The intervention expanded medication coverage from 71 to 600 medications, selecting frequently prescribed drugs through review of five years of ordering data and identifying high-risk drugs through pharmacy review. Dose thresholds were established using published literature, historical prescribing practices, and physician-pharmacist consensus, incorporating weight-based limits and maximum adult doses. Warnings were tiered into soft and hard stops. Soft stops displayed usual dosing ranges and required acknowledgment with a documented justification to proceed; hard stops required discussion between an attending physician and pharmacist, who could enter the order with justification. Both alert types generated pharmacist email notifications, and submission of a hard-alert order generated a pharmacy consult. To reduce unnecessary warnings, the system suppressed repeat soft alerts when the same provider reordered the same dose for the same patient within 72 hours and incorporated medication-specific rounding buffers, with smaller buffers for higher-risk drugs. Two-month periods before and after implementation were compared for alert frequency and whether prescribers reduced, cancelled, or submitted orders unchanged.
Duration	Pre-implementation: July to August 2010 Post-implementation: July to August 2011
Outcome Measures	Prescription error rates before and after implementation, provider response to soft and hard alerts
Results		Before Implementation, %	After Implementation, %
	Single-dose alerts	0.3% (30/8601)	3.4% (419/12,482)
	Single-dose soft alerts	100% (30/30)	68% (283/419)
	Single-dose hard alerts	0% (0/30)	32% (136/419)
	Soft alerts - Dose reduced	10% (3/30)	25% (70/283)
	Soft alerts - Order cancelled	0% (0/30)	9.5% (27/283)
	Soft alerts - Submitted unchanged	90% (27/30)	66% (186/283)
	Hard alerts - Dose reduced	N/A	29% (39/136)
	Hard alerts - Order cancelled	N/A	64% (87/136)
	Hard alerts - Dose submitted unchanged with verbal pharmacist communication	N/A	7.4% (10/136)
Study Author Conclusions	The enhanced DRC system with hard alerts and pediatric clinical pharmacy input led to a change in prescriber behavior, preventing medication errors before reaching patients.
Critique	The study effectively demonstrated a reduction in prescription errors with the enhanced DRC system. However, the retrospective design and short duration may limit the generalizability of the findings. The study groups were not directly comparable due to differences in the number of medications monitored and the introduction of hard alerts in the enhanced system.

Table 9 References:
[18] Balasuriya L, Vyles D, Bakerman P, et al. Computerized Dose Range Checking Using Hard and Soft Stop Alerts Reduces Prescribing Errors in a Pediatric Intensive Care Unit. J Patient Saf. 2017;13(3):144-148. doi:10.1097/PTS.0000000000000132

Optimization of drug–drug interaction alert rules in a pediatric hospital’s electronic health record system using a visual analytics dashboard
Design	Cross-sectional study N= 2,391,880 medication alerts evaluated
Objective	To develop and evaluate an electronic dashboard of hospital-wide electronic health record medication alerts for an alert fatigue reduction quality improvement project
Study Groups	N/A
Inclusion Criteria	Medication alerts presented to physicians and pharmacists between January 31, 2011 and January 26, 2014
Exclusion Criteria	N/A
Methods	This quality improvement study used an interactive dashboard to identify and remove clinically irrelevant drug-drug interaction warnings at the Children’s Hospital of Philadelphia. The dashboard extracted electronic health record (EHR) data with a 24-hour delay and allowed users to examine alert frequency, override rates, triggering medications, patient locations, and provider types. The team prioritized frequently triggered interactions for review by 10 pediatric clinical pharmacists working in pairs. Each pharmacist independently evaluated the vendor’s interaction monograph, compared its severity and management recommendations with other drug references, and reviewed primary literature when needed. Disagreements were resolved through pharmacy consensus and consultation with affected clinical specialties. Override rates were not used to determine removal, because frequent overrides could reflect alert fatigue. Recommendations underwent clinical decision support and Therapeutic Standards committee approval. Three successive interventions deactivated 2, 42, and 19 interaction rules. Evaluation used interrupted time series analysis of alert rates and clinician responses, including responses to alerts that remained unchanged, with medication safety events monitored as a balancing measure.
Duration	January 31, 2011 to January 26, 2014
Outcome Measures	Primary: Reduction in alert and override rates Secondary: Decrease in medication serious safety event rate
Baseline Characteristics		Pharmacists	Providers
	Baseline median alert rate	58.74 alerts/100 orders (IQR 54.98–60.48)	19.73 alerts/100 orders (IQR 18.66–20.24)
Results		Pharmacists	Providers	p-value
	Alert rate after interventions	25.11 alerts/100 orders (IQR 23.45–26.57)	15.11 alerts/100 orders (IQR 14.44–15.49)	<0.001
	Override rate for untouched DDIs	85.71 overrides/100 alerts	Not specified	<0.001
Adverse Events	No serious safety events reported in association with the deactivated alert rules.
Study Author Conclusions	An alert dashboard facilitated safe rapid-cycle reductions in alert burden that were temporally associated with lower pharmacist override rates in a subgroup of DDIs not directly affected by the interventions. Reducing the alert burden enabled pharmacists to devote more attention to clinically relevant alerts.
Critique	The study effectively demonstrated the use of a visual analytics dashboard to reduce alert fatigue and improve medication safety. However, the study's generalizability may be limited due to its specific setting and EHR system. The reliance on reported patient events as a balancing measure may not capture all medication-related safety events. Additionally, the study did not include a control group for comparison.

Table 10 References:
[19] Simpao AF, Ahumada LM, Desai BR, et al. Optimization of drug-drug interaction alert rules in a pediatric hospital's electronic health record system using a visual analytics dashboard. J Am Med Inform Assoc. 2015;22(2):361-369. doi:10.1136/amiajnl-2013-002538

Feedback at the Point of Care to Decrease Medication Alert Rates in an Electronic Health Record
Design	Prospective, interventional study
Objective	To evaluate the feasibility of involving pediatric emergency department providers in identifying additional medication alerts that were potentially incorrect or intrusive and to assess the impact of modifications on alert salience
Inclusion Criteria	All users able to see medication alerts in the EHR at Cincinnati Children's Hospital Medical Center
Methods	A prospective, 34-month intervention at Cincinnati Children’s Hospital Medical Center used clinician feedback to identify inaccurate or bothersome medication warnings, focusing on pediatric emergency department dosing alerts. An “Alert Feedback” option was added to the existing override-reason dropdown, allowing users to flag the exact alert encounter for review and agree to follow-up contact for additional clinical context. Users received initial education, periodic reminders, and updates about resulting changes. A multidisciplinary group of physicians, informaticists, and pharmacists reviewed submissions weekly, considered user feedback and associated ordering data, and reached consensus on modifications, with final pharmacist validation. Users received individual feedback about the decision, and approved dosing-rule changes applied hospital-wide, so proposed changes had to be appropriate across clinical settings. Examples included adjusting ipratropium and cetirizine dose limits to accommodate clinically acceptable dosing. Evaluation assessed participation, submission types, and alert frequency and salience, defined as the proportion prompting corrective action, using 12 months of data before and after modification when available. The process modified 75 dosing alerts, with complete comparison data available for 47.
Duration	March 6, 2014, to December 31, 2016
Outcome Measures	Primary: Feasibility and adoption of alert feedback implementation Secondary: Alert rate and salience of alerts changed as a result of the process
Results		PED Users	All Departments
	Number of feedback responses	49	245
	Average reviews per month	4	12
	Dosing related alerts modified	75	Not specified
Study Author Conclusions	User-centered, context-specific alert feedback can be used for selecting suboptimal, interruptive medication alerts. This approach should be used in parallel with a data-driven one to capture alerts that are not the most common but pose a significant risk or strain on users.
Critique	The study provides a novel approach to addressing alert fatigue by involving end-users in the feedback process, which is a strength. However, the study's reliance on voluntary feedback may introduce bias, as more proactive users are likely to participate. Additionally, the study did not specify the number of participants, which limits the ability to generalize the findings. The complexity of collecting historical EHR data on alert rates also poses a limitation.

Table 11 References:
[20] van Camp PJ, Kirkendall ES, Hagedorn PA, et al. Feedback at the Point of Care to Decrease Medication Alert Rates in an Electronic Health Record. Pediatr Emerg Care. 2020;36(7):e417-e422. doi:10.1097/PEC.0000000000001847