Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors
Filter by Categories
Case Report
Drug Evaluation and Clinical Trials, Original Research Article
Epidemiology and Public Health, Original Research Article
Ethnomedicine and Phytomedicines, Original Research Article
Ethnomedicine and Phytomedicines, Review Article
Medical Science and Practice, Case Report
Medical Science and Practice, Original Research Article
Pharmaceutical Technology and Manufacturing, Original Research Article
Pharmacology and Toxicology, Original Research Article
Pharmacotherapy/Pharmaceutical Care, Case Report
Pharmacotherapy/Pharmaceutical Care, Original Research Article
Pharmacy Practice, Original Research Article
Pharmacy Profession and Advocacy, Original Research Article
Review Article
View/Download PDF

Translate this page into:

Epidemiology and Public Health
Original Research Article

Should insulin billing procedure codes be considered for exposure ascertainment in pharmacoepidemiology studies?

Division of Epidemiology, US Food and Drug Administration, Silver Spring, Maryland, United States
Regeneron Pharmaceuticals Inc., Tarrytown, New York, United States
Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, Massachusetts, United States
Corresponding author: Efe Eworuke, Ph.D., Division of Epidemiology, US Food and Drug Administration, Silver Spring, Maryland, United States.
This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-Share Alike 4.0 License, which allows others to remix, transform, and build upon the work non-commercially, as long as the author is credited and the new creations are licensed under the identical terms.

How to cite this article: Eworuke E, Hampp C, Menzin TJ, Burk J, Kluberg SA. Should insulin billing procedure codes be considered for exposure ascertainment in pharmacoepidemiology studies? Am J Pharmacother Pharm Sci 2022;3.



The purpose of the study was to examine whether Healthcare Common Procedure Coding System (HCPCS) billing codes should be used in conjunction with National Drug Codes (NDCs) to establish insulin exposure episodes.

Materials and Methods:

We identified insulin claims billed by NDCs or HCPCS codes in FDA’s Sentinel System from 2013 to 2018. We created insulin exposure episodes separately based on NDCs only, HCPCS only, and a combination of both NDC and HCPCS. We considered gaps of <30 days between valid billing codes as continuous exposure. Patients were followed until the earliest of (1) episode end date, (2) death, (3) disenrollment, (4) query end date, and (5) evidence of the opposite exposure defining code type (for cohorts defined by only NDCs or only HCPCS). We examined the median duration of incident episodes, requiring no NDC or HCPCS codes in the 183 days (washout period) before the first billing code and prevalent episodes (no washout period required). For patients with more than 1 treatment episode, we calculated median gap length between episodes.


We identified 107,528,855 insulin claims using NDCs or HCPCS. Of these, 98.5% were billed using NDCs. HCPCS were largely billed during emergency and ambulatory visits (52.5% and 38%, respectively). We identified 6,350,872 incident and 12,922,593 prevalent NDC episodes; and 6,821,075 incident and 13,465,108 prevalent NDC-HCPCS episodes. The median length of the first incident NDC. NDC episodes was 110 days (IQR: 60; 212); 31 (IQR: 19; 31) days for HCPCS only and 90 (IQR: 19-31) for NDC-HCPCS episodes. The median gap between the first and second episodes was shorter for incident NDC episodes than HCPCS episodes (NDC: 49 [IQR: 17; 132]; HCPCS: 249 [IQR: 93; 550]). Prevalent episodes showed similar trends.


HCPCS insulin codes appeared to indicate either 1 time or sporadic occurrences with long gaps between two codes. HCPCS codes in conjunction with NDC codes increased the number but reduced the median length of insulin episodes. Unless studies investigate the effects of insulin administered in specific settings to identify transient adverse reactions treated with insulin, we do not encourage the use of HCPCS to establish insulin exposure episodes.


Healthcare Common Procedure Coding System codes
National Drug Codes
Exposure ascertainment


Pharmacoepidemiology studies often rely on National Drug Codes (NDCs) for the identification of insulin products and the creation of insulin exposure episodes in administrative databases. Like non-biologic products, insulin is mostly self-administered and pharmacy dispensing has historically been billed as drug products using NDCs. However, there are some instances where Healthcare Common Procedure Coding System (HCPCS) codes may be considered for billing for instance, in the reimbursement of insulin administered using a pump or reusable pen or in situations where insulin is administered at the emergency department or physician offices. The use of HCPCS codes for insulin reimbursement has not been quantified and it remains unknown whether incorporating HCPCS codes into the claims-based algorithms for defining insulin exposure, improves the capture of insulin claims and, hence, the accuracy of insulin exposure. Thus, we examined the frequency of different billing codes (NDCs or HCPCS codes) that may be used for insulin product reimbursement. This study was a descriptive analysis to quantify the use of billing codes in a large national administrative database and determine whether HCPCS codes should be used in conjunction with NDC codes to ascertain insulin exposure.


Data source

We included data from 16 Sentinel Data Partners, between January 1, 2013, and December 31, 2018. The Sentinel Distributed Database is a curated data source composed of data partners, primarily large national and regional insurers and integrated delivery care networks in the United States. Each data partner contributed medical data, including inpatient and outpatient diagnoses, claims for procedures, and pharmacy claims data, including retail and mail order dispensing records.[1] These data are routinely transformed into the Sentinel Common Data Model[2,3] to facilitate queries. Of note for the purposes of this study, the common data model includes care settings for procedures (HCPCS), but not for dispensing (NDC). In general, NDCs are routinely billed out of outpatient pharmacies. This study was conducted as part of the public health surveillance activities under the auspices of the Food and Drug Administration and are, therefore, not under the purview of Institutional Review Boards.[4,5] Adults ages 18 and older were included in our study population; no restrictions on diabetes type or status were applied.

Distribution of billing codes for insulin reimbursement

During the study period, we identified every insulin claim generated by brand NDC (a total of 24 insulin products) or HCPCS insulin code [Online Supplement Table 1]. For each insulin product billed through NDC, we obtained the number of HCPCS codes within 3 days of the dispensed NDC products for the same patient. We also examined the distribution of all insulin NDCs within 3 days of a HCPCS code for the same patient. Finally, we obtained patient characteristics, encounter settings for non-pharmacy insulin claims, and the distribution of type of diabetes diagnosis codes (Type 1 or 2) 6 months before the date of the NDC or HCPCS code.

Table 1:: Insulin claims distribution and selected patient characteristics among insulin users.
All insulin Insulin by NDCs Insulin by HCPCS
Number of claims 107,528,855 105,941,433 1,682,070
Insulin claims billed by NDC codes
Lantus and Lantus Solostar 39,326,260 36.6 39,326,260 37.1 N/A
Novolog 17,167,717 16.0 17,167,717 16.2
Levemir 16,966,889 15.8 16,966,889 16.0
Humalog 14,842,175 13.8 14,842,175 14.0
Insulin claims billed by HCPCS codes
J1815 1,635,916 1.5 N/A 1,635,916 97.3
J1817 55,272 0.1 55,272 3.3
J1820 11 0.0 11 0.0
K0548 0 0.0 0 0.0
Encounter settingsa
Emergency department 887,937 0.8 10,108 0.0 883,308 52.5
Ambulatory visit 666,931 0.6 63,477 0.1 639,676 38.0
Other ambulatory visit† 22,125 0.0 10,160 0.0 18,787 1.1
Non-acute institutional‡ 11,695 0.0 83 0.0 11,624 2.1
Inpatient 137,483 0.1 12,737 0.0 130,048 5.6
Pharmacy 105,815,851 98.4 105,851,527 99.9 0 0.0
Number of unique patients 5,944,918 5,617,952 984,969
Mean age (years) 66.4 12.7 66.4 12.7 64.9 14
Age: 18–44 years 8,269,097 7.7 8,117,463 7.7 166,254 9.9
Age: 45–64 years 33,749,596 31.4 33,216,124 31.4 75,566 34.2
Age: 65+years 65,510,162 60.9 64,607,846 61.0 940,250 55.9
Sex (female) 3,045,963 51.2 2,879,780 51.3 517,254 52.5
Sex (male) 2,898,955 48.8 2,738,172 48.7 467,715 47.5
Presence of diabetes codes –3–180 days from Index:
Type 1 diabetes code only 4,197,383 3.9 4,187,536 4.0 11,765 0.7
Type 2 diabetes code only 77,375,206 72.0 76,195,803 71.9 1,247,411 74.2
Type 1 or Type 2 diabetes code 102,259,039 95.1 100,722,071 95.1 1,628,421 96.8
Neither Type 1 nor Type 2 diabetes code 5,269,816 4.9 5,219,362 4.9 53,649 3.2

NDC: National Drug Code, HCPCS: Healthcare Common Procedure Coding System, † includes other non-overnight AV encounters such as hospice visits, home health visits, skilled nursing facility (SNF) visits, other non-hospital visits, as well as telemedicine, telephone, and email consultations, ‡ includes hospice, SNF, rehab center, nursing home, residential, overnight non-hospital dialysis, and other non-hospital stays. aNDC claims generally do not have an associated encounter setting, however, two data partners do record their NDCs in the procedure table, thus capturing a clinical encounter setting

Insulin exposure episodes

For this analysis, we created insulin episodes using the days’ supply on the NDC claim or 1 day for a HCPCS claim. We allowed for 30-day gaps between the calculated end of supply and subsequent claim and added a 30-day extension at the end of each episode’s last claim to account for short delays in medication refills. We created insulin episodes separately, based on NDCs, HCPCS, and combination of NDCs and/or HCPCS. With the 30-day extension, all HCPCS exposures had a uniform length of 31 days unless bridged with a subsequent administration during the episode. An incident insulin episode was defined as no NDC (for NDC episodes) or no HCPCS (for HCPCS episodes) or no NDC and no HCPCS (for NDCHCPCS episode) 6 months before the first claim. We removed this restriction for prevalent insulin episodes. We censored treatment episodes after the 30-day gap allowance expired without a new claim for insulin. We also censored episodes in the event of a new claim for HCPCS during NDC-only, or NDC during HCPCS-only cohorts; death; disenrollment; end of available data; or the study end date, whichever came first. We assessed the median duration (in days) of cumulative and first NDC, HCPCS, and NDC-HCPCS episodes separately. For patients with more than 1 episode of use, we assessed the length of time in days between each episode for both a patient’s first observed gap and for all gaps. We also present descriptive statistics to characterize the gaps.


During the evaluation period, we identified 107,528,855 insulin claims using NDCs or HCPCS. Of these, 98.5% were billed using NDCs and 1.6% using HCPCS codes. The most frequent NDCs were for Lantus (37.1%), Novolog (16.2%), and Levemir (16.0%), while the most frequent HCPCS was J1815 (injection, insulin, per 5 units or insulin injection) (97.3%) and J1817 (insulin for administration through DME [i.e., insulin pump] per 50 units) (3.3%) [Table 1]. While almost all NDCs were billed by default as pharmacy claims, the majority of HCPCS codes were billed in the emergency department (52.5%) or during an ambulatory visit (38.0%). In 94.3% of the 1,682,070 HCPCS codes identified, patients had no NDC code on the same date as the HCPCS codes. Thus, the insulin administration could not be associated with any product. The remaining 94,673 (5.6%) HCPCS codes had at a least one branded insulin NDC on the same date as the HCPCS claims date. The most frequent NDCs billed within 3 days of the J1815 code were Lantus (5.6%), Humalog and Novolog (3.3% each), and Levemir (2.5%). The top three NDCs associated with J1817 were Humalog (6.3%), Novolog (5.3%), and Lantus (4.7%). For about 95% of all insulin billed, through NDC or HCPCS, patients had at least one claim with a Type 1 or Type 2 diabetes code in the 3 days before through 180 days after (3.9% Type 1 only; 72.0% Type 2 only; and 95.1% Type 1 or 2).

Insulin exposure episodes

We identified 6,350,872 incident NDC episodes, 1,217,664 incident HCPCS episodes, and 6,821,075 incident NDCHCPCS episodes. Similarly, we identified 12,922,593 prevalent NDC episodes, 1,240,888 prevalent HCPCS episodes, and 13,465,108 prevalent NDC-HCPS episodes. The median duration (days) of the first NDC episodes was longer than for HCPCS episodes, for both incident (NDC: 110 [IQR: 60; 212]; NDC-HCPCS: 90 [IQR: 58; 191]; and HCPCS: 31 [IQR: 19; 31]) and prevalent (NDC: 120 [IQR: 60; 251]; NDC-HCPCS: 114 [IQR: 60; 237]; and HCPCS: 31 [IQR: 19; 31]) episodes. Similar trends were observed for the cumulative [Figure 1] and subsequent episodes (data not shown). Of all incident NDC episodes, 45% represented single treatment episodes with no gap (no subsequent NDC); of all incident HCPCS episodes, 78% represented single episodes with no gap. The median gap length (in days) between the first and second episodes was much shorter for NDC and NDC-HCPCS episodes than HCPCS episodes, for both incident (NDC: 49 [IQR: 17; 132]; NDC-HCPCS: 51 [IQR: 17; 143]; and HCPCS: 249 [IQR: 93; 550] days) and prevalent episodes (NDC: 35 [IQR: 12; 92]; NDC-HCPCS: 35 [IQR: 12; 93]; and HCPCS: 245 [IQR: 91; 546]) [Figure 1]. We observed similar distributions for all gaps between all observed episodes.

Figure 1:: Gap between insulin episodes. (a) Duration of the first gap (days [IQR]) for incident Insulin episodes and (b) for prevalent insulin episodes.


In this large observational study of insulin users, we found that only 1.5% of all insulin claims were billed using HCPCS codes. The J1815 code accounted for more than 90% of the HCPCS codes for insulin administration. The majority (~90%) of HCPCS codes were used in emergency departments or in the outpatient office setting. Because of the high proportion of HCPCS as solitary occurrences for the patient, and long gaps between HCPCS when subsequent episodes were observed, HCPCS codes appeared to be either 1 time administrations or sporadic occurrences. In contrast, episodes based on NDC codes are generally consistent with chronic use.

Online Supplement Table 1:: List of Healthcare Common Procedure Coding System (HCPCS) procedure codes used to define the exposure.
Code Description
J1815 Injection, insulin, per 5 units
J1817 Insulin for administration through DME (i.e., insulin pump) per 50 units
S5552 Insulin, intermediate acting (NPH or LENTE); 5 units
S5553 Insulin, long acting; 5 units
S5561 Insulin delivery device, reusable pen; 3 ml size
S5551 Insulin, most rapid onset (Lispro or Aspart); 5 units
S5550 Insulin, rapid onset, 5 units
S5571 Insulin delivery device, disposable pen (including insulin); 3 ml size
S5560 Insulin delivery device, reusable pen; 1.5 ml size
J1820 Injection, insulin, up to 100 units
S5565 Insulin cartridge for use in insulin delivery device other than pump; 150 units
S5570 Insulin delivery device, disposable pen (including insulin); 1.5 ml size
S5566 Insulin cartridge for use in insulin delivery device other than pump; 300 units
K0548 Injection, insulin lispro, up to 50 units

While we were able to capture a greater number of patients and treatment episodes when including HCPCS, relative to assessment through NDCs alone, this might not be a desirable patient identification strategy for insulin use studies. Combining these code types resulted in shorter median episode duration relative to the NDC-only cohorts, illustrating the fact that NDC-HCPCS cohorts represented a mix of two different types of exposure. In studies that examine incident insulin use, the inclusion of HCPCS codes may capture brief emergency department use rather than subsequent chronic use episodes, which are more likely to be of interest in safety assessment.

Our data reveal that emergency administration of insulin products is best captured using HCPCS codes. For studies that examine insulin utilization in these settings or the impact of emergency administration of insulin, one might consider the use of HCPCS codes. However, such studies might be unable to evaluate specific insulin products.


Insulin products were billed primarily using NDCs in administrative claims databases. We observed that the frequency of NDC codes alone reflected a pattern of dispensing consistent with chronic insulin use. A small fraction of insulin claims was billed using HCPCS codes, mostly as solitary occurrences for patients in emergency situations and during office visits. NDCs alone best depict chronic insulin exposure for safety assessment studies; however, one might consider the use of HCPCS codes in the assessment of emergency department administration of insulin products, or in situations where there is interest in identifying transient adverse reactions that are treated with insulin, for example, steroid-induced hyperglycemia.


Many thanks are due to those who participated in this project: Data partners who provided data used in the analysis: Aetna, a CVS Health company, Blue Bell, PA; Blue Cross Blue Shield of Massachusetts, Boston, MA; Duke University School of Medicine, Department of Population Health Sciences, Durham, NC, through the Centers for Medicare and Medicaid Services which provided data; Harvard Pilgrim Health Care Institute, Boston, MA; HealthCore, Inc., Translational Research for Affordability and Quality, Alexandria, VA; HealthPartners Institute, Minneapolis, Minnesota; Humana Healthcare Research Inc., Louisville, KY; Kaiser Permanente Colorado Institute for Health Research, Aurora, CO; Kaiser Permanente Center for Integrated Health Care Research Hawai’i, Honolulu, HI; Kaiser Permanente Mid-Atlantic States, Mid-Atlantic Permanente Research Institute, Rockville, MD; Kaiser Permanente Northern California, Division of Research, Oakland, CA; Kaiser Permanente Northwest Center for Health Research, Portland, OR; Kaiser Permanente Washington Health Research Institute, Seattle, WA; Marshfield Clinic Research Institute, Marshfield, WI; Optuminsight Life Sciences Inc., Boston, MA; Vanderbilt University Medical Center, Department of Health Policy, Nashville, TN, through the TennCare Division of the Tennessee Department of Finance and Administration which provided data.

Take-home message

Unless studies investigate the effects of insulin administered in specific settings or aim to evaluate transient adverse reactions treated with insulin, we do not encourage the use of Healthcare Common Procedure Coding System (HCPCS) to establish insulin exposure episodes.


The views expressed in this publication are those of the authors and do not necessarily reflect the official policy of the U.S. Food and Drug Administration. Christian Hampp was an employee of the Food and Drug Administration during the time of study conduct. He is now employed by Regeneron Pharmaceuticals, Inc., and owns company stock. Regeneron did not provide study funding, does not hold a marketing license for any of the study drugs, and had no role in manuscript development or decision to publish.

Declaration of patient consent

Patient’s consent not required as patient’s identity is not disclosed or compromised.

Financial support and sponsorship

This project was supported by Task Order HHSF22301012T under Master Agreement HHSF223201400030I from the US Food and Drug Administration (FDA). The FDA approved the study protocol including statistical analysis plan and reviewed and approved this manuscript. Coauthors from the FDA participated in the results interpretation and in the preparation and decision to submit the manuscript for publication. The FDA had no role in data collection, management, or analysis.

Conflicts of interest

There are no conflicts of interest.


  1. , , , et al. Development and application of two semi-automated tools for targeted medical product surveillance in a distributed data network. Curr Epidemiol Rep. 2017;4:298-306.
    [CrossRef] [PubMed] [Google Scholar]
  2. , , , et al. Design considerations, architecture, and use of the mini-sentinel distributed data system. Pharmacoepidemiol Drug Saf. 2012;21:23-31.
    [CrossRef] [PubMed] [Google Scholar]
  3. , , , et al. Sentinel Initiative Principles and Policies: HIPAA and Common Rule Compliance in the Sentinel Initiative. Available from: Accessed 01 Mar 2020
    [Google Scholar]
  4. , et al. Federal policy for the protection of human subjects. Final rule Fed Regist. 2017;82:7194-7274.
    [Google Scholar]
Show Sections