Enhancing medication safety with System Approach to Verifying Electronic Prescriptions (SAV E-Rx): pharmacists’ review of product selection outcomes between prescribed and dispensed medications

Abstract

Objectives

Electronic prescriptions (e-prescriptions) introduce drug product selection mismatches during pharmacy data entry. System Approach to Verifying Electronic Prescriptions (SAV E-Rx) detects and alerts pharmacy staff to clinically significant occurrences. This study evaluates outcomes of the identified mismatches.

Methods

A retrospective analysis was conducted using 1 year of e-prescriptions and dispensing data from 14 community pharmacies across 9 US states. SAV E-Rx screened the data, and flagged mismatches were reviewed by pharmacists using the Common Formats for Event Reporting. Data were analysed using descriptive statistics, the Mann-Whitney U test and χ² tests.

Results

Of 1 250 804 records processed, 699 662 included sufficient data for comparison. Pharmacists classified 587 (88.7%) flagged records as intended mismatches and 75 (11.3%) as unintended. Intended mismatches involved ingredients (26.2%), strengths (53.7%) and dosage forms (47.4%), mainly due to prescriber-approved substitutions (62.4%). Unintended mismatches stemmed from ingredients (42.7%), strengths (36.0%) and dosage forms (54.7%) discrepancies, primarily reported as human error (82.7%) and labelling issues (76.0%). Future alerts were favoured for unintended mismatches (96.0%) compared with intended mismatches (56.7%) (p<0.001).

Discussion

While routine substitutions are a normal part of quality and timely care, unintended mismatches may pose clinical risks. These errors can arise from human factors and workflow challenges, including high prescription volumes and manual overrides. SAV E-Rx serves as an independent, automated safety net that flags mismatches, catching postdispensing errors that would otherwise go unnoticed.

Conclusions

E-prescription errors remain a safety concern. Routine implementation of SAV E-Rx could enhance error detection and enable timely interventions.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• This study demonstrates that the System Approach to Verifying Electronic Prescriptions algorithm can effectively detect clinically significant mismatches in prescribed and dispensed medications. It reveals that most mismatches are intended substitutions, but a notable proportion are unintended errors linked to human and system-level issues, which can be intercepted through automated alerts.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Findings support the integration of automated mismatch detection tools in pharmacy workflows and highlight opportunities to enhance labelling standards and pharmacy staff training to reduce the incidence of dispensing errors.

Introduction

Electronic prescriptions (e-prescriptions) enable prescribers to electronically transmit patients’ prescription information directly to pharmacy systems. They have revolutionised medication prescribing by streamlining the process, improving medication tracking and efficiency and eliminating errors associated with handwriting.^{1 2} However, e-prescriptions also introduce new types of medication errors, including incorrect drug selection, dosage discrepancies and system-related errors such as transmission failures or incorrect autopopulated fields.^{3 4}

Medication errors in e-prescription processing can occur when the prescribed medication differs in ingredient, strength or dosage form from what is dispensed.^{5 6} The incidence of these mismatch errors ranges from 4% to 40%.⁷ The mismatches between prescribed and dispensed medications often stem from inconsistencies between National Drug Codes (NDCs) of prescribed and dispensed drug products, unique product identifiers for drugs in the USA, or incorrect product selection within prescribing and dispensing software.^{8 9} Medication product data entry is typically a human-dependent process that involves matching an e-prescription drug product description to the corresponding medication product in a pharmacy’s inventory.¹⁰ A study revealed that 20.3% of e-prescriptions still required pharmacy staff to intervene, and mismatched NDCs between the electronic health records (EHR) and the pharmacy platform were identified as the most common cause of prescription failures, accounting for around 60% of failed cases.¹¹ Meanwhile, poor-quality data exchange between EHR and pharmacy dispensing software can lead to incorrect medication prescribing or dispensing, posing severe risks to patients and making recovery from such errors challenging for healthcare organisations.¹²

To systematically identify unsafe e-prescription transactions, we developed the System Approach to Verifying Electronic Prescriptions (SAV E-Rx) tool, which verifies medication selection and identifies mismatches between prescribed and dispensed medications based on ingredient, strength and dosage form using the RxNorm terminology, a standardised clinical drug terminology by the National Library of Medicine (NLM) for consistent medication representation. SAV E-Rx flags records for pharmacist review when the ingredient, strength and dosage form information do not fully align. By monitoring these mismatches, the system helps pharmacy staff identify and address potential issues, enhancing patient safety.¹⁰ SAV E-Rx also incorporates pharmacist feedback through iterative cycles of design, implementation and evaluation.

This study aims to assess the effectiveness of automated mismatch detection with SAV E-Rx by analysing the outcomes of product selection mismatches reviewed by pharmacists. It seeks to identify the sources of these mismatches, investigate the underlying causes and evaluate their impact on patient safety. By analysing these outcomes, the study aims to provide valuable insights into the potential areas for improvement in automated systems for detecting prescription mismatches and increasing the success rate of selecting the prescribed medication product.

Methods

Study design

We conducted a retrospective analysis of electronically transmitted e-prescriptions and dispensed medication transactions over 1 year from 2023 to 2024 to identify any mismatches. As part of this study, the SAV E-Rx tool flagged records with discrepancies between prescribed and dispensed medications, defined here as individual medication items listed within e-prescriptions. Pharmacists then reviewed the flagged records and completed a survey form based on the Agency for Healthcare Research and Quality (AHRQ) Common Formats for Event Reporting to document their assessments (see online supplemental appendix form). The report of this study follows Standards for Quality Improvement Reporting Excellence V.2.0, which is a revised publication guideline based on a detailed consensus process.¹³

Data sources

E-prescription and medication dispensing data associated with e-prescriptions from 14 pharmacies across 9 states were included. Each participating pharmacy provided a complete year of data. The required variables for analysis include the ‘e-prescription NDC’, capturing the NDC in e-prescription data, the ‘pharmacy NDC’, which denotes the NDC of the dispensed medications and the ‘deidentified prescription ID’, an alphanumeric code that the pharmacist used to locate the prescription in their dispensing software.

The e-prescriptions and dispensed medication transactions included electronically transmitted e-prescriptions and non-e-prescription records. Non-e-prescription records encompassed prescriptions received via paper, fax, telephone orders, pharmacy-to-pharmacy transfers and certain refill or renewal requests processed without structured e-prescription data. Since SAV E-Rx relies on standardised e-prescription data formats, it was applied exclusively to e-prescription records and non-e-prescription data were excluded from the mismatch analysis.

SAV E-Rx data processing

A standardised report was developed by the study team and the software vendor to extract deidentified data from each e-prescription and dispensed medication pair in a .csv format. Multiple NDCs for medications can have identical active ingredients and dosages and be linked to the same RxNorm concept unique identifiers (RxCUI).¹⁴ For each complete record pair, the two NDCs were queried via the NLM’s publicly available application programming interface to retrieve the corresponding RxCUI.¹⁵ The RxCUI provided information on ingredients, dosage forms, strength values and units. Data cleaning involved removing rows that: (1) lacked e-prescription data, (2) were missing the NDC for the prescribed item, (3) had an NDC not linked to an RxCUI or (4) were missing ingredient, dosage form or strength information from RxNorm.

The SAV E-Rx algorithm evaluated medication pairs sequentially to determine match status (match or mismatch). First, the NDCs were compared, and matching NDCs were labelled as a match. If the NDCs differed, the RxCUIs were compared, and matching RxCUIs were labelled as matches. If both NDCs and RxCUIs differed, the ingredients, dosage forms and strengths were analysed for equivalence. Details are displayed in online supplemental appendix figure 1. Note that cases where the only difference between the prescribed and dispensed medications is the brand versus generic name were classified as ‘different substitutions’ and were automatically excluded by SAV E-Rx from the final analysis (eg, Ativan 1 mg oral tablet vs lorazepam 1 mg oral tablet).

Pharmacists’ reviews

The prescribing and dispensing data were initially analysed by SAV E-Rx, which flagged mismatches and forwarded them to pharmacists for further evaluation. Before processing the results, pharmacists completed an intake form to provide pharmacy demographic information and customise their results.¹⁶ This included specifying therapeutic equivalents they wished to exclude, even if predefined rules flagged them as mismatches. Examples of optional exclusions included substituting an oral capsule for an oral tablet or treating differences in salt forms (eg, hydroxyzine pamoate vs hydroxyzine hydrochloride) as equivalent. The selected rules were incorporated into SAV E-Rx for additional processing before it generated a final mismatch list for that pharmacy. Note that all Schedule II medications were required to match the prescription exactly without exception. The US Drug Enforcement Administration defines Schedule II medications as substances with a high potential for misuse and a risk of severe psychological or physical dependence. Examples include oxycodone, methadone and fentanyl.

Once customisation was complete, pharmacists reviewed the flagged records and completed a medication event form (see online supplemental appendix form)¹⁷ within the Research Electronic Data Capture (REDCap) platform. REDCap served as a secure, web-based application to facilitate data entry, management and sharing, supporting compliance with privacy and security regulations. The event form adhered to relevant components of the AHRQ Common Formats for Event Reporting, which provided a standardised framework for capturing and categorising medication-related events, ensuring consistency and reliability in data collection across all participating sites. The event form included identifying incorrect actions, explaining reasons for intended mismatches and detailing the causes of unintended mismatches. As a second phase of customization, pharmacists also indicated whether they would like similar mismatches to be flagged in the future.

Outcome measures

The following outcome metrics were used: (1) Incorrect actions: the number and percentage of errors, including dispensing different ingredients, incorrect strengths or concentrations and incorrect dosage forms. (2) Intended mismatches: the number and percentage of intended mismatches and the documented reasons behind these substitutions. (3) Unintended mismatches: the number and percentage of unintended mismatches, the contributing factors leading to these errors, and any associated harms caused by such mismatches. (4) Future alerts: the number and percentage of mismatch pairs for which pharmacists preferred to receive future alerts.

Statistical analysis

Descriptive statistics, including counts and percentages, were used to summarise all outcomes. Mann-Whitney U and χ² tests were conducted to compare differences between intended and unintended mismatches, with p values of less than 0.05 considered statistically significant. We performed the statistical analyses using ‘readxl’¹⁸ and ‘tidyverse’¹⁹ packages in R V.4.3.1.

Results

Of the 1 250 804 records sent for processing, 699 662 contained all the required information to compare the prescribed item with the dispensed, and 29 119 mismatches were reported. After cleaning, 662 records with complete pharmacist reviews remained, comprising 587 (88.7%) intended and 75 (11.3%) unintended mismatches (figure 1). The overall dispensing error rate was 0.01% (75/699,662).

Figure 1. Flow chart of e-prescription and dispensing medication record processing and analysis. NDC, National Drug Code; RxCUI, RxNorm concept unique identifiers.

One pharmacy was excluded due to incomplete pharmacist reviews. Among the remaining 13 pharmacies, 69% were independent pharmacies, with 67% operating for over a decade. Weekly prescription volumes ranged from 500 to 3000, with a median of 1543. All pharmacies had at least one full-time pharmacist, 77% also employed part-time pharmacists and 92% had full-time technicians. Approximately half had overlapping pharmacist shifts (16–35+ hours per week). Standard services included free home delivery (92%), medication therapy management (92%), care plan development (85%), home visits (48%), 24-hour emergency services (48%) and compounding services (62%). Additional pharmacy demographic details are provided in online supplemental appendix table 1.

Table 1 presents a basic comparison of intended and unintended mismatches across the included pharmacies. Median mismatches per pharmacy were higher for intended mismatches (32, IQR 14–52) than unintended mismatches (5, IQR 1–14). When comparing intended and unintended mismatches, significant differences were observed for different ingredients (26.2% vs 42.7%, p=0.004) and strengths (53.7% vs 36.0%, p=0.005). In comparison, dosage form differences were not statistically significant (47.4% vs 54.7%, p=0.27). Preferences for future alerts to the mismatches revealed a significant contrast, with 96.0% of respondents favouring alerts for unintended mismatches compared with 56.7% for intended mismatches (p<0.001).

Table 1. Comparison of intended and unintended mismatches across pharmacies.

	Intended mismatches (n=587)	Unintended mismatches (n=75)	P value
Incorrect actions
Different drug	154 (26.2%)	32 (42.7%)	0.004
Different strength or concentration	315 (53.7%)	27 (36.0%)	0.005
Different dosage form	278 (47.4%)	41 (54.7%)	0.27
Notification preference for recurring mismatches
No	254 (43.3%)	3 (4.0%)	<0.001
Yes	333 (56.7%)	72 (96.0%)

A single record may include multiple incorrect actions.

As figure 2 shows, among intended mismatches, key reasons behind intended substitutions included prescriber-approved substitutions (366, 62.4%), dosage form changes (155, 26.4%), equivalent strength switches (124, 21.1%), stock issues (98, 16.7%), insurance or patient preferences (73, 12.4%) and other reasons (72, 12.3%). Multiple responses were permitted in a single record, with a median selection of 1 (IQR: 1–2, range: 0–4). The ‘other’ category overlapped with the previously mentioned factors. At the same time, additional causes included miscommunication or prescription errors, where prescribers had to clarify inconsistencies in strength or dosage forms and operational constraints, such as hospitals not completing prior authorisations, necessitating medication adjustments (eg, switching suspensions to capsules for g-tube use).

Figure 2. Reasons for intended mismatches between prescribed and dispensed medications.

Regarding the 75 unintended mismatches, the majority of reported unintended mismatches (72, 96.0%) originated during the data entry, order or transcription stage. Most incidents (66, 88.0%) were categorised as events that reached the patient, while 9.3% were near misses and 2.7% were unsafe conditions. Among reported harm levels, pharmacists reported whether the patient sustained harm from the incident: 90.7% involved no harm, while 2.7% resulted in mild harm and 1.3% in moderate harm.

Figure 3 indicates that among unintended mismatches, contributors included human factors (62, 82.7%), drug name, label or packaging issues (57, 76.0%), unknown causes (23, 30.7%), workflow issues (12, 16.0%) and rare cases of staffing/scheduling or technology problems (1, 1.3% each). Multiple selections were also allowed in an individual record, with a median of 2 choices (IQR: 2–3, range: 0–3).

Figure 3. Contributors to unintended mismatches between prescribed and dispensed medications.

Table 2 demonstrates the top 10 most frequent mismatches between prescribed and dispensed medications. Common discrepancies include formulation differences (eg, ophthalmic vs otic solutions), extended-release versus immediate-release formulations and variations in ingredient compositions and dosage strengths. Notable discrepancies include metformin hydrochloride 500 mg oral tablet being substituted with an extended-release tablet (n=21) and ofloxacin 3 mg/1 mL ophthalmic solution being dispensed as an otic solution (n=13). Other mismatches involve famotidine 10 mg oral tablet being replaced with a 20 mg strength (n=9) and albuterol sulfate 5 mg/1 mL inhalation solution substituted with a 0.83 mg/1 mL formulation (n=5).

Table 2. Top 10 most frequent medication mismatches by ingredient, dosage form and strength.

Prescribed medication	Dispensed medication	N
Metformin hydrochloride—500 mg/1 each—oral tablet, oral product, pill	Metformin hydrochloride—500 mg/1 each—extended release oral tablet, oral product, pill	21
Ofloxacin—3 mg/1 mL—ophthalmic solution, ophthalmic product	Ofloxacin—3 mg/1 mL—otic solution, otic product	13
Famotidine—10 mg/1 each—oral tablet	Famotidine—20 mg/1 each—oral tablet	9
Metformin hydrochloride—1000 mg/1 each—extended release oral tablet, oral product, pill	Metformin hydrochloride—1000 mg/1 each—oral tablet, oral product, pill	9
Dextroamphetamine sulfate—5 mg/1 each—extended release oral capsule, oral product, pill	Dextroamphetamine saccharate, dextroamphetamine sulfate, amphetamine aspartate, amphetamine sulfate—1.25 mg/1 each 1.25 mg/1 each 1.25 mg/1 each 1.25 mg/1 each—extended release oral capsule, oral product, pill	7
Trimethoprim sulfate, polymyxin B sulfate—1 mg/1 mL, 10 000 UNT/1 mL—ophthalmic solution, ophthalmic product	Ofloxacin—3 mg/1 mL—ophthalmic solution, ophthalmic product	7
Ethinyl estradiol, ethinyl estradiol, levonorgestrel—0.01 mg/1 each, 0.03 mg/1 each, 0.15 mg/1 each—pack, oral product, pill	Ethinyl estradiol, levonorgestrel, inert ingredients—0.03 mg/1 each 0.15 mg/1 each 1 mg/1 each—pack, oral product, pill	6
Albuterol sulfate—5 mg/1 mL—inhalation solution, inhalant product	Albuterol sulfate—0.83 mg/1 mL—inhalation solution, inhalant product	5
Lamotrigine—25 mg/1 each—extended release oral tablet	Lamotrigine—25 mg/1 each—oral tablet	5
Acetaminophen, caffeine, butalbital—300 mg/1 each, 40 mg/1 each, 50 mg/1 each—oral capsule, oral product, pill	Acetaminophen, caffeine, butalbital—325 mg/1 each 40 mg/1 each 50 mg/1 each—oral tablet, oral product, pill	4
Aspirin, caffeine, butalbital—325 mg/1 each, 40 mg/1 each, 50 mg/1 each—oral capsule	Acetaminophen, caffeine, butalbital—325 mg/1 each 40 mg/1 each 50 mg/1 each—oral tablet	4

Discussion

SAV E-Rx is a novel, automated and independent tool that verifies medication product selection and identifies medication selection errors after they have occurred, providing additional validation to support pharmacists. This study evaluated its performance in detecting discrepancies between prescribed and dispensed medications using retrospective data. To our knowledge, this study is the largest analysis of its kind to automatically identify mismatches between e-prescribed and dispensed medications, revealing key patterns in both intended and unintended discrepancies reviewed by pharmacists. Of the 662 pharmacist-reviewed cases, 88.7% were intended mismatches due to reasons such as prescriber-approved substitutions and supply constraints, while 11.3% were unintended errors from human factors, labelling issues and transcription mistakes. Most respondents preferred future alerts, with a significantly higher preference for unintended mismatches (96.0%) compared with intended mismatches (56.7%). Notably, dispensing errors remain rare, with an estimated rate of 0.01%. Implementing automated verification tools like SAV E-Rx could reassure pharmacists that most prescriptions are dispensed correctly while identifying true dispensing errors and improving the overall reliability of e-prescription dispensing, thereby enhancing patient safety.

Pharmacist-initiated interventions play a crucial role in facilitating medication substitutions. Our study found that prescriber-approved substitutions (62.4%) were the leading cause of intended mismatches, followed by dosage form changes (26.4%) and stock issues (16.7%). These intended substitutions are essential for addressing medication availability, optimising therapy and ensuring timely patient access. Many intended substitutions done manually by pharmacy staff are routine, expected changes, such as equivalent formulations (eg, oral tablets vs oral capsules).¹⁰ However, in some cases, manual product selection leads to unintended mismatches, which are more difficult to justify and may pose greater clinical risks. SAV E-Rx can detect medication selection errors that may otherwise go unnoticed by serving as an independent, automated safety net. By identifying postdispensing mismatches, SAV E-Rx enhances medication verification and patient safety while acknowledging the necessity of pharmacist-driven substitutions. This distinction is reflected in pharmacists’ preferences for alerts, with 96% favouring notifications for unintended mismatches, compared with 56.7% for intended ones, reinforcing the need for automated verification to enhance patient safety.

Manual product selection remains a necessary step for a significant portion of e-prescriptions received by community pharmacies, potentially leading to unintended mismatches.¹¹ Previous interviews with pharmacists revealed that they often must manually input the dispensed product, which can introduce these unexpected medication errors, as e-prescriptions contain varying levels of detail and are processed differently across pharmacy dispensing software platforms.¹⁰ Meanwhile, studies have shown that drug product descriptions often exhibit significant variability and poor readability.^{20 21} This inconsistency makes it more difficult for pharmacists to interpret the information accurately, particularly when simultaneously processing large volumes of e-prescriptions each day. Our study identified 662 records with complete pharmacist reviews, of which 75 (11.3%) were unintended mismatches. These findings align with previous concerns about product selection challenges and indicate how discrepancies translate into actual mismatches. Additionally, although the National Council for Prescription Drug Programs’ (NCPDP) SCRIPT standards for transmitting e-prescription data components recommend using RxNorm identifiers on prescriber platforms, community pharmacies predominantly rely on NDCs.²² However, the accuracy of the NDC identifier is sometimes unreliable due to outdated information or discrepancies where the listed NDC corresponds to a different drug than the accompanying description.^{23 24} These challenges complicate the product selection process and increase the risk of medication errors.

Extensive efforts have been made to enhance accuracy in product fulfilment (ie, retrieving the correct product from a pharmacy’s shelves) through technologies such as optical and barcode scanning, which can solve issues including look-alike/sound-alike medications and inconsistent labelling. However, comparatively less attention and technologies have been directed toward improving accuracy at the data entry stage, where 96% of the unintended mismatches occurred in our analysis. Human errors accounted for more than 80% of unintended mismatches in total. Human factors may include inefficiencies from high prescription volumes and inconsistencies in e-prescription and dispensing systems that require pharmacists to correct information manually.^{25 26} Additionally, look-alike/sound-alike drug names increase the risk of incorrect product selection during data entry, further contributing to dispensing errors. To reduce unintended mismatches, targeted interventions should focus on training, workflow optimisation and decision support tools in dispensing software.^{23 27 28} A multifaceted approach is needed, as not all interventions are equally effective in the long term. Additional pharmacy personnel training can improve error detection and accuracy in prescription processing.²⁹ Optimising workflows can minimise inefficiencies, reduce handoff-related errors and enhance coordination.³⁰ Implementing decision support tools, such as automated verification systems like a real-time SAV E-Rx in the future, can proactively monitor product selection errors. It could shift incidents to near misses and reduce patient harm while allowing pharmacists to focus on high-priority fills. Together, these strategies can reduce the risk of delayed dispensing and dispensing a medication different from what the prescriber intended, potentially improving patient safety.

This study has several limitations. First, the retrospective design, which relies on historical data from a single year, may not capture temporal trends or evolving pharmacy practices that could influence mismatch rates over a longer period. Second, the findings may not fully represent all e-prescriptions issued in community pharmacies nationwide, as the data were limited to pharmacies in nine states. Third, pharmacists self-reported their review results in this study, which may introduce social desirability bias, leading to potential underreporting of errors or an underestimation of potential harm.1,3 Lastly, these analysed mismatches include only medications used in outpatient settings and do not cover injectable or infusion products administered in inpatient or long-term care facilities.

Conclusions

Mismatch errors between e-prescribed and dispensed medications continue to pose a risk to medication safety, emphasising the need for enhanced verification processes. The routine implementation of SAV E-Rx, an automated double-check system, could improve error detection and facilitate timely interventions, reducing the likelihood of unintended mismatches. Future efforts should focus on standardising labelling practices, optimising workflow processes, integrating advanced decision-support tools and refining e-prescription and dispensing systems to ensure greater accuracy and reliability in medication management.

Data availability statement

Data are available on reasonable request.