PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of jamiaAlertsAuthor InstructionsSubmitAboutJAMIA - The Journal of the American Medical Informatics Association
 
J Am Med Inform Assoc. 2016 April; 23(e1): e99–e107.
Published online 2015 October 28. doi:  10.1093/jamia/ocv131
PMCID: PMC4954631

Evaluating the implementation of RxNorm in ambulatory electronic prescriptions

Abstract

Objective RxNorm is a standardized drug nomenclature maintained by the National Library of Medicine that has been recommended as an alternative to the National Drug Code (NDC) terminology for use in electronic prescribing. The objective of this study was to evaluate the implementation of RxNorm in ambulatory care electronic prescriptions (e-prescriptions).

Methods We analyzed a random sample of 49 997 e-prescriptions that were received by 7391 locations of a national retail pharmacy chain during a single day in April 2014. The e-prescriptions in the sample were generated by 37 801 ambulatory care prescribers using 519 different e-prescribing software applications.

Results We found that 97.9% of e-prescriptions in the study sample could be accurately represented by an RxNorm identifier. However, RxNorm identifiers were actually used as drug identifiers in only 16 433 (33.0%) e-prescriptions. Another 431 (2.5%) e-prescriptions that used RxNorm identifiers had a discrepancy in the corresponding Drug Database Code qualifier field or did not have a qualifier (Term Type) at all. In 10 e-prescriptions (0.06%), the free-text drug description and the RxNorm concept unique identifier pointed to completely different drug concepts, and in 7 e-prescriptions (0.04%), the NDC and RxNorm drug identifiers pointed to completely different drug concepts.

Discussion The National Library of Medicine continues to enhance the RxNorm terminology and expand its scope. This study illustrates the need for technology vendors to improve their implementation of RxNorm; doing so will accelerate the adoption of RxNorm as the preferred alternative to using the NDC terminology in e-prescribing.

Keywords: RxNorm, E-Prescribing, electronic prescriptions, NDC, quality

INTRODUCTION AND BACKGROUND

Transmitting clear and complete electronic prescription (e-prescription) messages between prescribers and dispensing pharmacies is essential for the benefits of electronic prescribing (e-prescribing) to be fully realized and to minimize the potential risks of this practice.1 Therefore, communicating an unambiguous numeric drug identifier in e-prescription messages is also of critical importance. An accurate, interpretable drug identifier in an e-prescription enables the receiving pharmacy system to definitively identify the prescribed medication, validate the associated free-text drug description, pre-populate internal user screens, and trigger automated clinical decision support checks.1–5

Accomplishing this goal requires a standard, interoperable terminology system that provides a single, unique identifier for each clinically distinct medication currently available for prescribing by clinicians.3 Desirable characteristics of such a system include, but are not necessarily limited to:6,7

  • Comprehensive content: Includes all available drugs at the level of granularity required for prescribing;
  • Concept permanence: The meaning of an identifier never changes;
  • Non-ambiguity: Each identifier has only one meaning, and vice versa;
  • Multiple granularities: Represents classes, ingredients, brand names, dosage forms, etc., in a computable manner;
  • Graceful evolution: Historical data using retired identifiers can still be unambiguously interpreted; and
  • Context-specific information: Links to key knowledge sources, such as the Drug Enforcement Administration’s schedule of controlled substances, drug classes, and commercial knowledge bases, are available.

Currently, the predominant drug identifier terminology used in e-prescribing is the Food and Drug Administration’s National Drug Code (NDC).3,8 The NDC scheme of identifiers does not meet all of the desirable characteristics of terminology system listed above, and, although widely used in e-prescribing transactions, NDC identifiers have been criticized for a number of shortcomings relative to their application in e-prescribing.3,9 An NDC identifier, by definition, is specific to a particular manufacturer/labeler and product, and, hence, a single e-prescribing drug description concept (ie, drug name, strength, and dosage form) can have multiple assigned NDC identifiers, which makes the NDC scheme restrictive, cumbersome, and potentially confusing to implement, because the receiving e-prescribing system may not contain a particular NDC identifier in its drug database or the receiving pharmacy may not stock a particular manufacturer’s product.9,12 Furthermore, concerns regarding the absence of a single authoritative and up-to-date repository source to identify and cross-reference NDC identifiers have been reported.1,3,5,9

RxNorm is a standardized, non-proprietary clinical drug nomenclature developed and maintained by the United States National Library of Medicine (NLM) to facilitate the electronic exchange of clinical health information between various systems that use different drug nomenclatures.10 Full RxNorm distribution files are updated every month and are available at no cost from the NLM RxNorm website.11 RxNorm has been recommended by the Centers for Medicare and Medicaid Services and other organizations as a preferred alternative to the NDC identifier scheme for use as a standardized terminology system in e-prescribing, because it more closely approximates an “ideal” system of drug identifiers.1,3,5,12,13

Most e-prescriptions exchanged between prescribers and pharmacies in the United States use the SCRIPT standard, a message format maintained by the National Council for Prescription Drug Programs (NCPDP). At present, version 10.6 of the NCPDP SCRIPT standard, which is widely used in ambulatory practice, does not require the use of a coded drug identifier in order to successfully transmit an e-prescription message.14 Rather, all that is required is a free-text drug description in the drug segment of the message (Figure 1).

Figure 1:
E-prescription data fields.

If an NDC identifier is included in the e-prescription, then it is transmitted in the Drug Product Code field. Alternatively, the Drug Database Code field and the corresponding qualifier field are the designated fields in the SCRIPT standard for transmitting other (non-NDC) identifiers, including RxNorm Concept Unique Identifiers (RxCUIs), RxNorm Term Types (called “Term Types” hereafter), as well as proprietary commercial drug database codes and their corresponding qualifiers. The SCRIPT standard mandates the inclusion of corresponding qualifier values if the Drug Product Code field or the Drug Database Code field in the e-prescription message is populated. The free-text drug descriptions associated with the drug identifiers and the free-text drug description string in the e-prescription message must point to the same drug concept.1,5,15

RxNorm uses an interrelated set of Term Types to denote multiple levels of descriptions and relationships that can be used to express a clinical drug concept that most closely maps to what clinicians prescribe and also uses broader concepts to represent multiple related clinical drugs. The SCRIPT standard allows for the transmission of four specific Term Types: BPCK (“brand name pack,” a branded drug delivery device), SBD (“semantic branded drug,” a semantic brand drug description), GPCK (“generic pack,” a generic drug delivery device), and SCD (“semantic clinical drug,” a semantic clinical generic drug description), along with corresponding RxCUIs that most closely resemble the prescription medication concepts that are most familiar to clinicians. These Term Types are described in Table 1.

Table 1:
RxNorm Term Types Allowed in NCPDP SCRIPT Standard

OBJECTIVES

The objectives of this study were to: (1) determine the frequency of the use of RxNorm drug identifiers in ambulatory e-prescriptions; (2) evaluate the quality of RxCUIs and their corresponding Term Types that are transmitted in e-prescriptions; (3) assess the agreement of RxNorm RxCUI drug identifier concepts submitted in e-prescriptions with the corresponding mandatory free-text drug description in the message; (4) compare the drug description associated with an NDC drug identifier to the drug description associated with the RxCUI drug identifier sent in an e-prescription message and identify discrepancies between the two; and (5) assess the comprehensiveness of RxNorm by calculating the proportion of e-prescriptions that could be mapped to RxCUIs.

METHODS

Design and Data Source

The study design was a retrospective analysis of a random sample of e-prescriptions received in a single day in April 2014 by 7391 pharmacies in a national retail drugstore chain. The data elements extracted for analysis included: (1) individual pharmacy store identifier; (2) unique prescriber identifier; (3) free-text drug description; (4) NDC value and corresponding qualifier; and (5) drug database code value and corresponding qualifier. Prior to analysis, extracted data were de-identified by the pharmacy chain – as required by the Health Insurance Portability and Accountability Act Privacy Rule, at 45 C.F.R. §164.514. Prescriber identifiers were provided to the researchers only as unique 13-digit code numbers, and no protected health information data were made available to the investigation team.

The study investigators had access to two reference drug database source files: (1) the April 7, 2014 RxNorm release from NLM, containing all available RxCUI values and their corresponding RxNorm free-text drug descriptions, Term Types, and associated NDC identifiers (ie, the RxNorm file) and (2) the First Databank’s April 22, 2014 MedKnowledge™ (formerly NDDF Plus™) file containing a list of NDC identifiers with the corresponding Label Name drug description (ie, the First Databank file), for analysis purposes.

Quality of RxCUIs and Term Types

To assess the quality of the RxCUIs, all e-prescriptions that had a numeric value populated in the Drug Database Code field were matched to RxCUIs in the RxNorm file. If a match occurred, the corresponding value from the Drug Database Code Qualifier field was also selected, to further evaluate the quality of the Term Types.

Three members of the investigation team independently reviewed all the e-prescriptions in the sample that contained an RxCUI and compared the drug description associated with each RxCUI (from the RxNorm file) to the free-text drug description from the e-prescription, to identify possible discrepancies between the two. The reviewers were certified pharmacy technicians, each with over 3 years of experience processing e-prescriptions in the community practice setting. All reviewers completed training under the supervision of a pharmacist to confirm their subject matter proficiency.

The reviewers were given six possible “Match Type” codes to use for the comparison: (1) match on all parameters, (2) brand/generic mismatch, (3) brand/brand mismatch, (4) drug name, strength, or form mismatch, (5) commercial unit dose packaging mismatch, and (6) “unable to determine.” Additionally, the three reviewers were asked to independently review e-prescriptions in which the numeric code value from the Drug Database Code field exactly matched the RxCUI from the RxNorm file, but the corresponding qualifier did not match the RxNorm Term Type. Reviewers were asked to determine the accuracy of each RxCUI and Term Type based on the similarity between the free-text drug description in the e-prescription message and the drug description associated with the matched RxCUI.

Agreement among the three reviewers was measured using the kappa coefficient with Light’s modification:

[κ=P(0)P(e)/1P(e)] 

where P(0) equals the proportion of observed agreements and P(e) equals the proportion of agreements expected by chance. The kappa coefficient is a measure of agreement between two raters for nominal data that is corrected for the agreement that would be expected by chance alone. Light’s modification allows for application to multiple raters by computing scores for all rating pairs, then calculating the arithmetic mean.16

The kappa coefficient was calculated based on the reviewers’ agreement on assigning one of the six possible Match Type codes described above to each e-prescription in the sample after the independent review and preceding reconciliation. In cases in which there was disagreement among the three reviewers, the e-prescriptions were discussed as a group to reach a consensus. Two licensed pharmacists with ambulatory care experience and residency training served as final adjudicators in cases in which the three primary reviewers could not reach a consensus after discussion or the agreed-upon Match Type code was “unable to determine.”

Comparison of Textual Drug Descriptions Associated with RxNorm RxCUIs and NDC Drug Identifiers

To begin the analysis, e-prescriptions containing an NDC and a Drug Database Code field value and the corresponding qualifier that exactly matched the RxCUI and Term Types from the RxNorm file were identified. The First Databank MedKnowledge™ file was then accessed to identify e-prescription NDC identifiers and their corresponding First Databank Label Name drug descriptions.

The First Databank Label Name drug description is defined as a combination of the drug name appearing on the package label, the strength description, and the dosage form description, for a specified product.17 The NDC-associated drug descriptions were compared to the RxCUI-associated drug descriptions from the RxNorm file. The three primary pharmacy technician reviewers independently reviewed the drug descriptions for discrepancies, as they had with the RxCUI and Term Type analysis. Once again, the two secondary pharmacist reviewers served as final adjudicators in cases in which the three primary reviewers could not reach a consensus after discussion.

Scope of RxNorm Coverage

To assess the comprehensiveness of RxNorm, e-prescriptions that contained a value in the Drug Database Code field that mapped to an RxCUI from the RxNorm file were first identified and assumed to contain an RxCUI. Then, for e-prescriptions that did not include an RxCUI but contained an NDC identifier, an RxCUI value was assigned using the NDC-to-RxCUI mapping available from the RxNorm file. Finally, all remaining e-prescriptions (including those without an NDC identifier) were manually reviewed and assigned an RxCUI from the RxNorm file based on the free-text drug description in the e-prescription. When there was uncertainty about the equivalency of the free-text drug description in the e-prescription message to the free-text drug description in the RxNorm drug file, specific drug product information was sought from NLM’s RxNorm expert resource.

RESULTS

Data for the study included 49 977 e-prescriptions issued by 37 801 ambulatory care prescribers practicing in all 50 United States and the District of Columbia, that had been received during a single day by 7391 unique chain pharmacy store locations. E-prescriptions in the sample had been transmitted by users of 519 different e-prescribing software applications, including both standalone applications and those integrated within electronic health record systems.

Frequency of Drug Identifiers in the Sample

As indicated in Figure 2, 45 906 (91.8%) e-prescriptions in the sample included one or more numeric drug identifiers in addition to the required free-text drug description. Of these, 42 602 (85.2%) included a numeric value in the Drug Product Code field that is intended to be an NDC identifier, while 29 990 (60%) contained a numeric value in the Drug Database Code field that is intended to be for RxNorm RxCUI or an identifier from another proprietary code set. A total of 4091 (8.2%) e-prescriptions only included a free-text drug description and no numeric drug identifiers. Of the 29 990 e-prescriptions that contained a numeric value in the Drug Database field, we were able to match 16 919 to an RxCUI from the RxNorm file, while 13 071 could not be matched. Finally, 16 488 of the 16 919 e-prescriptions’ RxCUIs and the corresponding Term Types could be exactly matched to the corresponding entries from the RxNorm file. We therefore classified these 16 488 (33.0%) as the true frequency of the use of the RxNorm terminology in our data sample.

Figure 3:
RxNorm coverage.

Figure 2:
Drug identifiers in the data sample.

Quality of RxCUIs and RxNorm Term Types

As indicated in Table 2, an exact match on both RxCUI and RxNorm Term Type was found in 16 488 of the e-prescriptions in our dataset, while the remaining 431 e-prescriptions contained some type of discrepancy, as described in Table 2.

Table 2:
E-prescriptions with a Matching RxNorm RxCUI

A total of 336 out of the 431 discrepant e-prescriptions were found to have an allowed Term Type. Upon reviewing the corresponding drug descriptions, NLM’s classification of the RxNorm Term Type was determined to be justifiable in these 336 examples.

Upon further analysis of these 336 e-prescriptions, we found that 221 had a free-text drug description that exactly mapped to the drug description associated with the RxCUI, and another 108 had brand-generic, brand-brand, or unit dose vs bulk package mismatches. The remaining seven e-prescriptions had completely different drug descriptions that represented either a different ingredient, strength, or dosage form.

Another 70 e-prescriptions, although positively matched to a corresponding RxCUI, included one of the commercial drug database code qualifiers instead of the anticipated RxNorm Term Type value. Upon reviewing the corresponding drug descriptions associated with these 70 e-prescriptions, 51 were found to have a free-text drug description that exactly matched the drug description associated with the RxCUI, and 17 had a completely different drug description and contained either a different ingredient, strength, or dosage form. The two remaining e-prescriptions only differed on dosage packaging.

The remaining 25 of the 431 discrepant e-prescription did not have a corresponding drug database qualifier. The free-text drug descriptions from these messages were matched to the drug descriptions associated with the corresponding RxCUIs and perfect matches were found in all instances.

The average pairwise kappa coefficient value of inter-rater agreement between the three reviewers was determined to be 0.92 for the above review, before reconciliation.

13 071 (26.1%) of the e-prescriptions in our dataset included a numeric value in the Drug Database Code field but could not be matched to an RxCUI from the RxNorm file. Of these, 12 899 (25.5%) e-prescriptions had a valid corresponding qualifier, as listed in the NCPDP External Code list, that represented a proprietary commercial drug database code, and these codes were therefore assumed to be proprietary drug database codes.18 There were 20 e-prescriptions that did not have any corresponding drug database code qualifier and, thus, could not be attributed to any drug database source.

Another 152 e-prescriptions in the dataset included a valid RxNorm Term Type in the Drug Database Qualifier field, but the corresponding drug database code value could not be matched to the RxCUIs from the April 7, 2014 RxNorm file. Upon further investigation, 125 of these e-prescriptions contained an RxCUI that had been discontinued by NLM, and a new RxCUI was available for the same drug description concept in the April 7, 2014 RxNorm release. Eighteen of these e-prescriptions did include a valid RxCUI but were not associated with one of the four Term Type concepts recommended for e-prescribing (ie, SBD, SCD, BPCK, GPCK); hence, the corresponding Term Types sent with these messages were incorrect. An additional seven of these e-prescriptions contained an RxCUI that had been discontinued by NLM, and no replacement codes had been subsequently published. Finally, two of these e-prescriptions were determined to have an invalid numeric value in the Drug Database Code field that did not exist in the RxNorm database.

Analysis of Free-Text Drug Descriptions Based on RxNorm RxCUI Concept

A total of 16 488 e-prescriptions were found to contain both an RxCUI and RxNorm Term Type that exactly matched with an RxCUI and RxNorm Term Type from the April 7, 2014 RxNorm file. Further details of this analysis appear in Table 3.

Table 3:
Comparison of Free-Text Drug Descriptions and RxNorm Drug Descriptions for E-Prescriptions That Exactly Matched with RxNorm RxCUIs and Term Types

In 15 734 of the e-prescriptions in our dataset, the drug description associated with the RxCUI and the free-text drug description pointed to the same semantic drug concept. In 446 instances, a mismatch was noted between the free-text drug description and the description from the RxNorm file, but the difference was limited to brand vs generic distinctions for the same drug ingredient. In another 283 e-prescriptions in the dataset, the difference between these descriptions was limited to drug manufacturing package discrepancies. In 15 instances, the free-text drug description was not adequately specified (the strength or form of the drug was missing) and an accurate comparison could not be made between the free-text drug description and the RxCUI description. Finally, in 10 of the e-prescriptions in our dataset, the free-text drug description and the RxCUI drug description pointed to completely different drug concepts.

Comparison of Drug Descriptions Between RxNorm RxCUIs and NDC Drug Identifiers

A total of 16 488 e-prescriptions in our dataset had an RxCUI and a Term Type that matched with the corresponding RxCUI and Term Type from the RxNorm file and also included an NDC value. A total of 16 288 (98.8%) out of the 16 488 e-prescriptions with NDC identifiers were found in the First Databank MedKnowledge™ file. The Label Name drug descriptions associated with these NDC identifiers from the First Databank MedKnowledge™ file were compared with the drug descriptions associated with the e-prescription RxCUIs obtained from the April 7, 2014 RxNorm file. The results of this comparison are shown in Table 4.

Table 4:
Comparison of Drug Descriptions Associated with RxNorm RxCUIs and NDC Drug Identifiers

In 15 297 e-prescription messages in our dataset, the drug description associated with the RxCUI and the NDC value sent in the message pointed to the same drug concept. In 743 e-prescriptions in the dataset, the drug descriptions associated with the two drug identifiers sent in the message were different, but the difference was limited to brand-generic or brand-brand mismatches only. Similarly, 273 e-prescriptions in the dataset were different, but the difference was limited to manufacturer unit dose and bulk package mismatches. Eighteen e-prescriptions in the dataset had differences due to differences in the editorial policies of NLM and First Databank. Finally, seven e-prescriptions in the dataset had drug descriptions associated with completely different drug concepts (eg, different active ingredients, strengths, or dosage forms).

The average pairwise kappa coefficient value of inter-rater agreement between the three reviewers was determined to be 0.88 for the above review, before reconciliation.

Scope of RxNorm

As illustrated in Figure 2, we found 1134 (2.3%) e-prescriptions from the sample that could not be assigned an RxCUI. A total of 99 out of these 1134 e-prescriptions contained inadequately specified drug description names in the free-text drug description string of the message. An inadequately specified drug description name is defined as a drug description for which the specific drug concept being prescribed could not be determined from the textual drug description, the representative NDC identifier, or the RxCUI included in the message. The remaining 1035 e-prescriptions were for over-the-counter medications, vitamins, medical supplies, prescription medical devices, vaccines, and medical food products that are currently outside the scope of RxNorm.

DISCUSSION

We found the use of RxNorm in ambulatory e-prescribing to be surprisingly low. Only one-third of our sample (33.0%) contained an RxCUI and Term Type pair that exactly matched to the RxNorm file. In contrast, and as expected, the use of NDC identifiers (85.2%) was considerably higher. Disappointingly, a small but significant fraction of e-prescriptions in the sample (8.2%) did not include any drug identifiers at all.

Our results suggest that the implementation of drug identifiers, especially RxNorm, by software vendors needs improvement to ensure their accuracy and meaningful use by recipient systems. In our study, 2.5% of the e-prescriptions in the dataset that included an RxCUI that matched to a corresponding RxCUI from the RxNorm file had either a discrepancy in the corresponding Drug Database Code qualifier field (ie, the wrong corresponding Term Type or proprietary drug database code qualifiers sent vs appropriate Term Types) or did not have a qualifier at all. Upon review of the free-text drug description in these e-prescription vs the drug description associated with the matched RxCUI, it was noted that, in a majority of these e-prescriptions, the drug descriptions exactly matched, suggesting that erroneous Database Code qualifiers had been sent.

For the purposes of refining the editorial policy of RxNorm, correcting errors, and for other terminology maintenance reasons, NLM periodically discontinues RxCUIs and reassigns the drug concepts from the discontinued RxCUIs to new ones. In this study, we found e-prescriptions that contained discontinued RxCUIs that both did and did not have replacement RxCUIs assigned by NLM.

Our results appear to support concerns that have been raised in the literature, as well as anecdotal feedback from pharmacists, concerning discrepancies between drug identifiers and free-text drug descriptions in e-prescriptions.19–23 At best, such discrepancies can result in workflow disruptions for pharmacists and prescribers alike. At worst, these discrepancies can threaten patient safety. Our results revealed that 2.7% of e-prescriptions in the dataset had brand vs. generic name differences. In another 1.7%, noted disparities were related to differences in manufacturer-packaged unit dose medications vs traditional bulk medications (eg, “methylprednisolone 4 mg tablets” vs “methylprednisolone dose pack”). Most disturbingly, in 0.06% of cases, the drug description associated with the RxCUI and the free-text drug description string pointed to a completely different drug concept.

If sent in the e-prescription, drug identifiers from different terminology systems must point to the exact same drug concept and agree with the free-text drug description provided as part of the e-prescription message, so that data can be auto-populated onto user interface screens and confidently used to conduct clinical decision support checks.1,5,15 Our findings revealed that, in 0.04% of e-prescriptions in the sample, the drug descriptions associated with the NDC identifiers and RxCUIs pointed to completely different drug concepts. While this incidence rate may appear small, the potential problems it can present to dispensing pharmacies and patients are not insignificant.

RxNorm provides standardized concepts for nearly all prescriptions that are prescribed in the ambulatory care setting. In our study sample, 97.9% of the e-prescriptions could have been accurately represented by the RxNorm terminology. It is important to note, however, that some non-prescription medications, vitamins, medical devices/supplies, and medical food are currently outside the scope of RxNorm. For that reason, it is recommended that NDC identifiers, manufacturer numbers, or Universal Product Codes that are supported by the NCPDP SCRIPT standard be used when RxNorm RxCUIs and corresponding Term Types are not available for prescription products.

In this study, we found e-prescriptions that had an assigned NDC identifier but did not comply with NLM’s current RxNorm editorial criteria and, hence, did not have a valid RxNorm RxCUI and Term Type. In order for NLM to generate an RxNorm RxCUI and Term Type, two specific criteria must be met: (1) RxNorm must be informed of that drug by one of RxNorm’s data sources, such as the Food and Drug Administration, through the Structured Product Labeling data feed or one of the commercial drug compendia that provide NLM with these data, such as First Databank, and, (2) the drug product’s active ingredient, strength, and dosage form must be unambiguously indicated. In our data sample, we found drug products that had assigned NDC codes and appeared in the Structured Product Labeling but whose label did not unambiguously indicate an active ingredient. For example: at the time of the study, the label for HylatopicPlus® Emollient Foam named 17 ingredients but did not specify any active ingredient, and, hence, RxNorm has not created a valid RxNorm RxCUI and Term Type for this product.24 NLM continues to work with industry stakeholder groups to appropriately address gaps in RxNorm’s content.

Our results suggest that more comprehensive end-user implementation guidance around RxNorm should be developed, with input and assistance from existing users and industry stakeholders. In addition to the RxNorm release notes and documentation, NLM uses a listserv to alert subscribers to proposed changes to and new features in RxNorm. We recommend that additional end-user communication options be explored, to share user success stories and workflow cases that will help accelerate RxNorm’s adoption.

Although there are many commonalities in the international uses of medications, there are also many differences. Among these are the different medication classes available for use (ie, prescription, non-prescription, pharmacy-only, etc.) as well as drug nomenclatures and naming conventions, which can vary significantly from country to country. Although we know of no countries outside the United States that currently use RxNorm, other countries have developed their own standard clinical terminologies and prescription drug naming conventions to facilitate interoperability between disparate data sources and enable safe medication dispensing. Examples include the United Kingdom’s National Health Service dictionary of medicines and devices (dm + d), and the Australian Medicines Terminology.25,26 Canada uses a computer-generated eight-digit number for drug identification (DIN) similar to the NDC drug identifier.27 Research is needed to better understand the relative strengths, weaknesses, and limitations of the clinical terminologies used elsewhere in the world, to aid in the development of best practice recommendations for a standardized, interoperable terminology system in the United States and globally.

Limitations

The sample of e-prescriptions used in this analysis were received by the stores of only one national retail pharmacy chain during a single day and, thus, may not be representative of all ambulatory e-prescriptions. In addition, the investigators used the standard April 2014 RxNorm Full Release available for download from NLM. It is unclear whether the content of this file is the same as that which each prescriber used on the day they issued their e-prescriptions. Finally, the investigators used one specific drug compendium to identify drug concepts, based on NDC identifiers to compare drug descriptions. We do not know which drug compendium was used by each prescriber to generate their e-prescriptions, and it is possible there may be drug description variances between different drug compendia that we did not consider.

CONCLUSION

NLM continues to refine and enhance RxNorm’s product offering. RxNorm’s nomenclature has been argued to be superior, and, therefore, its use has been recommended as the preferred alternative to using NDC identifiers in e-prescribing.1,3,5,12,13 If properly implemented, RxNorm holds great promise and should become the primary drug identifier used in e-prescriptions. For its promise to be realized, prescriber and pharmacy e-prescribing technology vendors must: (1) send valid and accurate RxNorm RxCUI and Term Types; (2) ensure that RxNorm Term Types exactly correspond to the RxCUIs sent in e-prescriptions; and (3) confirm that the drug description associated with the RxNorm RxCUI and Term Type exactly matches the free-text drug description string in the e-prescription.

To facilitate a transition from using the NDC terminology to using RxNorm, e-prescribing technology vendors should maintain current RxNorm drug files from their data sources. Furthermore, vendors should conduct validation checks to ensure that the drug description associated with the RxNorm identifier and the free-text drug description sent in the e-prescription point to the same exact drug concept before the e-prescription is transmitted. For their part, all commercial drug compendia must ensure that the drug identifier mappings they provide to e-prescribing technology vendors that subscribe to their products and services are accurate and include detailed implementation guidance. Finally, given the recognized status of RxNorm in Meaningful Use regulations and its relative advantages to alternative drug identifier systems in e-prescribing, consideration should be given to adding designated fields for RxNorm drug identifiers in the NCPDP SCRIPT standard.

CONTRIBUTORS

The corresponding author, A.A.D., takes responsibility for the integrity of this manuscript.

Study concept and design: A.A.D., S.W-C., M.T.R., and V.P.A.

Acquisition of data: A.A.D., V.P.A, and J.R.

Analysis and interpretation of data: A.A.D., S.W-C., J.K., M.T.R., J.R., and V.P.A.

Drafting of manuscript: A.A.D., S.W-C., and M.T.R.

Critical revision of manuscript for important intellectual content: A.A.D., J.K., M.T.R., J.R., and V.P.A.

Statistical analysis: S.W-C., A.A.D., and J.R.

Administrative, technical, or material support: S.W-C.

Supervision: A.A.D. The views expressed in this article are those of the authors and do not necessarily represent those of Surescripts LLC, Midwestern University, the National Library of Medicine, and CVS Health.

FUNDING

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

COMPETING INTERESTS

Dr Rupp reports receipt of consulting fees from Surescripts, LLC during the conduct of the study. All other authors declare that they have no conflicts of interest.

ACKNOWLEDGEMENTS

We would like to thank Keith Fisher, Reem Mohamed, Patrick McLaughlin, Steve Franko, Yuze Yang, Sara Juster, Richard Lee, Valentina Lengkong, and Jade Pressley for their help and assistance with this research study.

REFERENCES

1. Dhavle AA, Rupp MT. Towards creating the perfect electronic prescription. JAMIA. Published online July 18, 2014, doi: 10.1136/amiajnl-2014-002738. [PubMed]
2. Teich JM, Osheroff JA, Pifer EA, et al. CDS Expert Review Panel. Clinical decision support in electronic prescribing: recommendations and an action plan: report of the joint clinical decision support workgroup. JAMIA. 2005;12:365–376. [PMC free article] [PubMed]
3. Bell DS, O’Neill S, Reynolds K, et al. Evaluation of RxNorm in Ambulatory Electronic Prescribing. TR-941-CMS. Santa Monica, CA: Rand Corporation; 2011. http://www.rand.org/pubs/technical_reports/TR941. Accessed September 2, 2014.
4. Board of Directors of the American Medical Informatics Association. Standards for medical identifiers, codes, and messages needed to create an efficient computer-stored medical record. JAMIA. 1994;1:1–7. [PMC free article] [PubMed]
5. Grossman JM, Cross DA, Boukus ER, et al. Transmitting and processing electronic prescriptions: experiences of physician practices and pharmacies. JAMIA. 2012;19:353–359. [PMC free article] [PubMed]
6. Cimino JJ. Desiderata for controlled medical vocabularies in the twenty-first century. Methods Inf Med. 1998;37:394–403. [PMC free article] [PubMed]
7. Lau LM, Lam SH. Applying the desiderata for controlled medical vocabularies to drug information databases. Proc AMIA Annu Symp. 1999:97–101. [PMC free article] [PubMed]
8. Sterling RB. How e-Prescribing Can Land You in Court. Medscape. July 10, 2014. http://www.medscape.com/viewarticle/827532. Accessed October 5, 2014.
9. Levinson DR. The Food and Drug Administration’s National Drug Code Directory. Department of Health and Human Services, Office of Inspector General. http://oig.hhs.gov/oei/reports/oei-06-05-00060.pdf. Accessed October 2014.
10. Liu S, Ma W, Moore R, et al. “RxNorm: Prescription for electronic drug information exchange,” IEEE IT Professional, Vol. 7, no. 5, September 2005.
11. Nelson SJ, Zeng K, Kilbourne J, et al. Normalized names for clinical drugs: RxNorm at 6 years. JAMIA. 2011;18:441–448. [PMC free article] [PubMed]
12. Halamka JD. The Benefits of RxNorm [online] November 2, 2011. Life as a Healthcare CIO. http://geekdoctor.blogspot.com/2011/11/benefits-of-rxnorm.html. Accessed November 17, 2014.
13. Bell DS, Schueth AJ, Guinan JP, et al. Evaluating the technical adequacy of electronic prescribing standards: Results of an expert panel process. AMIA Annu Symp Proc. 2008;6:46–50. [PMC free article] [PubMed]
14. National Council for Prescription Drug Programs. SCRIPT Standard Implementation Guide. Version 10.6. Scottsdale, AZ. October 2014.
15. National Council for Prescription Drug Programs (NCPDP). SCRIPT Implementation Recommendations. April 2014 Version 1.28. http://www.ncpdp.org/NCPDP/media/pdf/SCRIPTImplementationRecommendationsV128.pdf. Accessed October 5, 2014.
16. Light RJ. Measures of response agreement for qualitative data: some generalizations and alternatives. Psychol Bull. 1971;76(5):365–377.
17. First Databank MedKnowledge Documentation, First Databank, Inc. San Francisco, CA, August 2014.
18. National Council for Prescription Drug Programs. The External Code List October 2008 Release. Scottsdale, AZ, October 5, 2014.
19. Odukoya OK, Chui MA. Retail pharmacy staff perceptions of design strengths and weaknesses of electronic prescribing. JAMIA. 2012;19:1059–1065. [PMC free article] [PubMed]
20. Warholak TL, Rupp MT. Analysis of community chain pharmacists’ interventions on electronic prescriptions. J Am Pharm Assoc. 2009;49 (1):59–64. [PubMed]
21. Nanji KC, Rothschild JM, Boehne JJ, et al. Unrealized potential and residual consequences of electronic prescribing on pharmacy workflow in the outpatient pharmacy. JAMIA. 2014;21(3):481–486. [PMC free article] [PubMed]
22. Hincapie AL, Warholak T, Altyar A, et al. Electronic prescribing problems reported to the Pharmacy and Provider ePrescribing Experience Reporting (PEER) portal. Res Social Adm Pharm 2013;pii:S1551–S7411. [PubMed]
23. Odukoya OK, Stone JA, Chui MA. E-prescribing errors in community pharmacies: exploring consequences and contributing factors. Int J Med Inform. 2014;83(6):427–437. [PMC free article] [PubMed]
24. Hylatopicplus Emollient. [package insert]. Onset Dermatologics LLC. US National Library of Medicine. [updated November 2014] http://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=aefede53-c089-403c-ac7d-881d39f245e2. Accessed November 17, 2014.
25. Department of Health. (2012) Common medicines dictionary approved for NHS [press release]. https://www.gov.uk/government/news/common-medicines-dictionary-approved-for-nhs. Accessed June 15, 2015.
26. National E-Health Transition Authority. Australian Medicines Terminology (AMT). http://www.nehta.gov.au. Accessed June 15, 2015.
27. Health Canada. Drug Identification Number (DIN). http://www.hc-sc.gc.ca/dhp-mps/alt_formats/hpfb-dgpsa/pdf/prodpharma/dinfs_fd-eng.pdf. Accessed June 15, 2015.

Articles from Journal of the American Medical Informatics Association : JAMIA are provided here courtesy of American Medical Informatics Association