This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Medical Informatics, is properly cited. The complete bibliographic information, a link to the original publication on http://medinform.jmir.org/, as well as this copyright and license information must be included.
Computable phenotypes have the ability to utilize data within the electronic health record (EHR) to identify patients with certain characteristics. Many computable phenotypes rely on multiple types of data within the EHR including prescription drug information. Hypertension (HTN)-related computable phenotypes are particularly dependent on the correct classification of antihypertensive prescription drug information, as well as corresponding diagnoses and blood pressure information.
This study aimed to create an antihypertensive drug classification system to be utilized with EHR-based data as part of HTN-related computable phenotypes.
We compared 4 different antihypertensive drug classification systems based off of 4 different methodologies and terminologies, including 3 RxNorm Concept Unique Identifier (RxCUI)–based classifications and 1 medication name–based classification. The RxCUI-based classifications utilized data from (1) the Drug Ontology, (2) the new Medication Reference Terminology, and (3) the Anatomical Therapeutic Chemical Classification System and DrugBank, whereas the medication name–based classification relied on antihypertensive drug names. Each classification system was applied to EHR-based prescription drug data from hypertensive patients in the OneFlorida Data Trust.
There were 13,627 unique RxCUIs and 8025 unique medication names from the 13,879,046 prescriptions. We observed a broad overlap between the 4 methods, with 84.1% (691/822) to 95.3% (695/729) of terms overlapping pairwise between the different classification methods. Key differences arose from drug products with multiple dosage forms, drug products with an indication of benign prostatic hyperplasia, drug products that contain more than 1 ingredient (combination products), and terms within the classification systems corresponding to retired or obsolete RxCUIs.
In total, 2 antihypertensive drug classifications were constructed, one based on RxCUIs and one based on medication name, that can be used in future computable phenotypes that require antihypertensive drug classifications.
Electronic health records (EHRs) contain a wealth of clinical information, and the development of tools to perform structured queries of common data models (CDMs) to standardized data formats has allowed researchers to utilize EHR data for discriminating complex clinical phenotypes with high measurement validity [
Accordingly, we aimed to design a set of standardized antihypertensive drug codes to be usable and maintainable by both ourselves and others in the research community [
RxNorm is a standardized terminology to represent drugs. It was developed by the US National Library of Medicine (NLM) in 2002 and represents medications through normalized names for clinical drugs, which include ingredient or ingredients strength or strengths, and dose form [
The Drug Ontology (DrOn) is a formal representation of drug products, drug ingredients, mechanisms of action, therapeutic indications, strengths, and dosage forms based on the OWL2 Web Ontology Language [
The OneFlorida Clinical Research Consortium is a statewide network of health systems, providers, and payers covering more than 74% of Florida’s population [
All prescription drug data for the HTN population were extracted from the Prescribing Table, including information on raw medication name and RxCUI. A total of 2 drug lists were created from this dataset. The first contained all unique raw medication names (see Drug Classification by Ingredient Name) derived directly from source EHRs, ie, not cleaned or curated during mapping to the PCORnet CDM. The second contained all unique RxCUIs (see Drug Classification by RxCUI utilizing DrOn and Drug Classification by RxCUI utilizing RxClass), which may be derived from source EHRs or created during mapping to the PCORnet CDM.
A Medication Name Classification was constructed for antihypertensive medications, utilizing drug ingredient names. A summary of the features included is available in
To apply the Medication Name Classification to the OneFlorida raw medication list, the first word in the field was extracted as the ingredient name, with additional coding to capture combination medications (eg, adding underscores between the first and second words) and strings where the ingredient was not the first word. The antihypertensive Medication Name Classification was merged with the unique raw medication names from the OneFlorida dataset to map the drugs by antihypertensive drug class. All of the raw medication names that did not merge with the Medication Name Classification were discarded (eg, statins, insulin, etc).
The DrOn RxCUI Classification was constructed through multiple steps (
The unique RxCUIs from the OneFlorida dataset were merged with the DrOn RxCUI Classification to map the drugs by antihypertensive drug class. All of the RxCUIs that did not merge with the DrOn RxCUI Classification were discarded (eg, statins, insulin, etc).
The unique list of RxCUIs extracted from OneFlorida were also mapped utilizing the RxClass API on RxMix [
The different drug classification methods were compared pairwise by calculating percent coverage and by reviewing the overlapping and nonoverlapping sets of RxCUIs among them. For all classification methods, the percent of antihypertensive drugs covered was calculated as the number of antihypertensive medications mapped by the classification method divided by the total number of unique terms (Raw Name or RxCUI) contained in the OneFlorida Prescribing Table. Within the ATC relationship source from RxClass, the antihypertensive classes were selected from the “name” field. A complete list of the ATC relationships included as antihypertensive drugs is available in
On the basis of our review of the overlapping and nonoverlapping sets of names and RxCUIs among the classifications, we created a version 1.0 of 2 finalized classifications—one for ingredient names and one for RxCUIs—for use by us and other researchers. We published them on GitHub [
The Medication Name Classification and the DrOn RxCUI Classification were applied to all of the prescription drug data from the OneFlorida HTN population. All available prescriptions from January 2011 onward were considered. Antihypertensive coverage by each method was calculated as number of prescription records mapped to an antihypertensive drug class by each map divided by the total number of prescriptions. Differences between the coverage results were identified by pairwise comparison. Additionally, summary-level frequency counts and percentages by the antihypertensive drug class were also calculated. All coding for mapping the drug data and the summary statistics was conducted using SAS version 9.4 (SAS).
At the time of the data extraction, there were 1,188,977 patients in the OneFlorida Data Trust with an ICD-9 or ICD-10 diagnosis code for HTN (
In total, there were 13,879,046 prescriptions from these patients over the study period (January 2011 to July 2017; approximately 6.6 years). These prescriptions consisted of 13,627 unique RxCUIs and 8025 unique first words from the raw medication name string (
Counts of data attributes in the OneFlorida hypertensive patient prescribing table dataset.
Data attributes | Values, N |
Hypertensive patients | 1,188,977 |
Prescription records | 13,879,046 |
Unique RxCUIsa | 13,627 |
Unique Raw Med Nameb | 8025 |
aRxCUI: RxNorm Concept Unique Identifier.
bUnique first word of the Raw_Med_Name field, after additional data cleaning and quality control steps.
The initial Medication Name Classification contained 286 antihypertensive medications that are mapped to 35 antihypertensive medication classes or combination medication classes (eg, ACE inhibitors, ARBs, CCBs, CCB-ARB combinations, etc). We chose to exclude timolol to be conservative. On the basis of this classification system, it is impossible to distinguish between oral and ophthalmic products. An excerpt from the Medication Name Classification is shown in
Excerpt from Medication Name Classification.
Codea | Drug_Name | Genericb | Drug_Class |
1 | Benazepril | —c | ACEd |
2 | Lotensin | benazepril | ACE |
3 | Captopril | — | ACE |
4 | Capoten | captopril | ACE |
5 | Enalapril | — | ACE |
6 | Enalaprilat | enalapril | ACE |
7 | Fosinopril | — | ACE |
8 | Monopril | fosinopril | ACE |
9 | Lisinopril | — | ACE |
10 | Prinivil | lisinopril | ACE |
11 | Zestril | lisinopril | ACE |
aFull classification list available in
bGeneric drug name included for brand name drugs.
cNot applicable for generic drugs.
dACE: angiotensin-converting enzyme inhibitor.
The initial DrOn RxCUI Classification contained 2543 antihypertensive RxCUIs that were mapped to 46 antihypertensive medication classes or combination medication classes. Each RxCUI entry contains the RxCUI, the drug product, and the drug class. An excerpt of the DrOn RxCUI Classification is shown in
Excerpt from the Drug Ontology RxNorm Concept Unique Identifier classification.
RxCUIa,b | Drug_Product | Rx_Norm_Drug_Class |
858926 | Enalapril Maleate 20 MG Chewable Tablet | ACEc |
378269 | Enalapril Chewable Tablet | ACE |
858810 | Enalapril Maleate 20 MG Oral Tablet | ACE |
858804 | Enalapril Maleate 2.5 MG Oral Tablet | ACE |
858821 | Enalapril Maleate 1.25 MG/ML Injectable Solution | ACE |
858817 | Enalapril Maleate 10 MG Oral Tablet | ACE |
858813 | Enalapril Maleate 5 MG Oral Tablet | ACE |
372007 | Enalapril Oral Tablet | ACE |
378288 | Enalapril Injectable Solution | ACE |
246264 | Enalaprilat 1 MG/ML Injectable Solution | ACE |
374378 | Enalaprilat Injectable Solution | ACE |
252820 | Enalaprilat 0.625 MG/ML Injectable Solution | ACE |
204404 | Enalaprilat 1.25 MG/ML Injectable Solution | ACE |
aFull map available in
bRxCUI: RxNorm Concept Unique Identifier.
cACE: angiotensin-converting enzyme inhibitor.
The percent of drugs covered by each antihypertensive map is shown in
When the DrOn RxCUI Classification was compared with the other classifications, they broadly overlapped; however, there were some key differences (
Of the 135 unique RxCUIs from the RxClass ATC and MED-RT maps, 29 had a term type of
Drug coverage by each classification method.
Classification method | Input term | Input, N | Antihypertensive | |
Mapped, n | Coverage, % | |||
Medication Name | Raw Name | 8025 | 207 | 2.58 |
DrOna RxCUIb | RxCUI | 13,627 | 729 | 5.35 |
RxClass-ATCc | RxCUI | 13,602 | 822 | 6.04 |
RxClass–MED-RTd,e | RXCUI | 13,602 | 792 | 5.82 |
aDrOn: Drug Ontology.
bRxCUI: RxNorm Concept Unique Identifier.
cATC: Anatomical Therapeutic Chemical.
dMED-RT Antihypertensive coverage was mapped using 2 steps: (1) "may_treat" hypertension and (2) mechanism of action of an antihypertensive drug class.
eMED-RT: Medication Reference Terminology.
Comparisons of the Drug Ontology RxNorm Concept Unique Identifier Classification to the other classification methods. Results are shown for the number overlapping between the methods (Both) and the numbers unique to each method. ATC: Anatomical Therapeutic Chemical; MED-RT: Medication Reference Terminology.
Examples of RxNorm Concept Unique Identifiers inappropriately mapped as antihypertensives utilizing the RxMix tools.
RxCUIa | DrugName | hasMOA | ConceptName | DoseForm |
207371 | Minoxidil | Potassium Channel Interactions | Minoxidil 20 MG/ML Topical Solution (Rogaine) | Topical solution |
208560 | Timolol | Adrenergic beta1-Antagonists | Timolol 2.5 MG/ML Ophthalmic Solution (Betimol) | Ophthalmic solution |
213729 | Betaxolol | Adrenergic beta1-Antagonists | Betaxolol 2.5 MG/ML Ophthalmic Suspension (Betoptic S) | Ophthalmic suspension |
aRxCUI: RxNorm Concept Unique Identifier.
Of the remaining 82 RxCUIs, 11 were RxCUIs corresponding to alpha-blockers primarily indicated for benign prostatic hyperplasia, 6 were RxCUIs for sacubitril/valsartan (Entresto) indicated for chronic heart failure with reduced ejection fraction, 1 was an RxCUI corresponding to an injectable antihypertensive product that is rarely used to treat HTN outside of a hypertensive crisis, and 1 was the
When the DrOn RxCUI Classification was compared with the Medication Name Classification using the DrOn RxCUI Classification as the reference, 695 terms overlapped, and 34 terms were unique to the DrOn RxCUI Classification (
From the comparisons, the DrOn RxCUI Classification had 71 unique RxCUIs/terms that were not present in the RxClass MED-RT, RxClass ATC, and/or Medication Name classifications. Many were combination products (n=29), and the remaining 42 RxCUIs were spread across the other antihypertensive drug classes: ACE inhibitor (n=1), alpha-blocker (n=1), ARB (n=1), beta-blocker (n=17), CCB (n=11), loop diuretic (n=1), thiazide diuretic (n=3), and vasodilator (n=7).
Following the comparisons between the different drug classifications, 96 RxCUIs were added to the DrOn RxCUI Classification, and 15 antihypertensive medication names were added to the Medication Name Classification. The DrOn RxCUI Classification version 1.0 contains 2639 RxCUIs and is available in
The results from applying the DrOn RxCUI Classification v1.0 and the Medication Name Classification v1.0 to the prescribing information from the OneFlorida HTN population are shown in
Overall, the methods performed very similarly, with approximately 15% (2,080,685 versus 2,089,557/13,879,046) of the total prescriptions mapping to antihypertensive drugs. The DrOn RxCUI Classification v1.0 was able to map 8872 more prescription records to an antihypertensive drug class compared with the Medication Name Classification v1.0 (
Summary of the methods to the OneFlorida Antihypertensive Prescribing Dataset.
Classification method | Antihypertensive (N=13,879,046) | |
Mapped, n | Coverage, % | |
Medication Name | 2,080,685 | 14.99 |
DrOna RxCUIb | 2,089,557 | 15.06 |
aDrOn: Drug Ontology.
bRxCUI: RxNorm Concept Unique Identifier.
Application of the Drug Ontology RxNorm Concept Unique Identifier Classification to the OneFlorida antihypertensive prescribing dataset (N=2,089,557).
Antihypertensive Drug Class or Classesa | Frequency, n (%) |
ACEb | 443,835 (21.24) |
Beta-blocker | 411,721 (19.70) |
Calcium Channel Blocker | 360,653 (17.26) |
Diuretic (Thiazide/Thiazide Like) | 217,474 (10.41) |
ARBc | 178,252 (8.53) |
Loop Diuretic | 115,931 (5.55) |
Diuretic/ACE Combo | 92,275 (4.42) |
Diuretic/ARB Combo | 63,010 (3.02) |
Centrally Acting | 56,501 (2.70) |
Vasodilator | 30,665 (1.47) |
Aldosterone Antagonist | 29,214 (1.40) |
Alpha Blocker | 26,995 (1.29) |
aResults are shown for antihypertensive drug classes that represent ≥1% of all antihypertensive prescriptions among the OneFlorida HTN population.
bACE: angiotensin-converting enzyme inhibitor
cARB: angiotensin II receptor inhibitor.
We created 2 different medication classification systems for antihypertensive drugs: one utilizing medication names and the other utilizing RxCUIs. After comparing these classification systems to each other, and to existing drug class terminologies available through RxNorm, we identified key areas that can lead to misclassification of antihypertensive medications and drug classes. Most misclassifications stemmed from failure to discriminate between dosage forms or issues related to primary indications of drugs (eg, selection of drugs that are primarily indicated for benign prostatic hyperplasia).
To illustrate, timolol is a beta-blocker that has both oral and ophthalmic dosage forms. The oral form is used to treat HTN, whereas the ophthalmic form is used to treat glaucoma [
Retired RxCUIs, or those that have been remapped to other classes, were classified by DrOn but not by MED-RT or ATC. This illustrates the need for maps and terminologies that include retired and obsolete RxCUIs as many of our longitudinal data sources include these. The data source used in this study contains data from January 2011 to July 2017, and contained 1170 RxCUIs that have been retired and 421 that have been remapped.
We conducted this work in the context of optimizing antihypertensive medication classification for use in HTN-related computable phenotypes. In other disease states where there are not as many options for pharmaceutical treatment, the classification of drug products may not be as complicated. However, in the case of HTN, and particularly the complex clinical phenotype of resistant hypertension (RHTN) [
We also observed differences in the classification of combination drug products. There were some combination antihypertensive drug products that were not identified in the MED-RT terminology through the methods that we utilized in this study. Additionally, when utilizing a classification system based on medication name, all possible permutations of the combination name must be considered and included in the map (eg, HCTZ-metoprolol, hydrochlorothiazide-metoprolol, metoprolol-HCTZ, and metoprolol-hydrochlorothiazide). Without each of these permutations, it is possible to miss certain combination products. Finally, as a phenotype such as RHTN is determined partially, or fully, based on the antihypertensive drug count, a correct drug count must be assigned to each combination product. We have added this field into our DrOn RxCUI Classification v1.0.
Our study is not without limitation; currently, we do not have a gold standard antihypertensive medication classification list. We selected different data sources and compared classification based on these sources. However, we only used terminologies available through the US NLM. We have not included terminologies maintained by other groups (eg, American Hospital Formulary Service Pharmacologic-Therapeutic Classification) [
In conclusion, we created and compared 4 different drug classification methods, focusing on the classification of antihypertensive drug products. We observed key differences in how each classification system handled drug products with multiple dosage forms, drug products with indications for benign prostatic hyperplasia, drug products that contain multiple antihypertensive ingredients (combination drug products), and RxCUIs that have been retired or remapped. We constructed 2 antihypertensive drug classification systems, 1 based off of RxCUIs and 1 based off of medication names. These are available for public use [
Features of each Classification Method.
List of Anatomical Therapeutic Chemical relationships included as antihypertensive drugs.
List of Medication Reference Terminology has_MoA relationships included as antihypertensive drugs.
Initial Medication Name Classification.
Initial Drug Ontology RxNorm Concept Unique Identifier Classification.
Drug Ontology RxNorm Concept Unique Identifier Classification version 1.0.
Medication Name Classification version 1.0.
angiotensin-converting enzyme
application programming interface
angiotensin II receptor inhibitor
anatomical therapeutic chemical
blood pressure
calcium channel blocker
common data model
drug ontology
electronic health record
hypertension
International Classification of Diseases
Medication Reference Terminology
mechanism of action
National Institutes of Health
National Library of Medicine
Patient-Centered Outcomes Research Institute
The National Patient Centered Clinical Research Network
resistant hypertension
RxNorm Concept Unique Identifier
Support for this project came from National Institutes of Health (NIH) grants KL2 TR001429 (CM), K01 HL141690 (CM), and K01 HL138172 (SS). Additionally, the research reported in this publication was supported in part by the OneFlorida Clinical Data Network, funded by the Patient-Centered Outcomes Research Institute (PCORI) #CDRN-1501-26692, in part by the OneFlorida Cancer Control Alliance, funded by the Florida Department of Health’s James and Esther King Biomedical Research Program #4KB16, and in part by the University of Florida Clinical and Translational Science Institute, which is supported in part by the NIH National Center for Advancing Translational Sciences under award number UL1TR001427. The content is solely the responsibility of the authors and does not necessarily represent the official views of the PCORI, its Board of Governors or Methodology, the OneFlorida Clinical Research Consortium, the University of Florida’s Clinical and Translational Science Institute, the Florida Department of Health, or the NIH.
WRH is one of the creators of DrOn.