Abstract
Objective
Describe real-world obesity medication (OM) prescribing and dispensing among adolescents and young adults (AYAs) and examine factors associated with dispensing.
Methods
Retrospective cohort linking Nemours Children’s Health electronic health record to Surescripts dispensing (2020–2025). AYAs aged 12–20 with prescriptions for liraglutide, semaglutide, phentermine, phentermine–topiramate, or tirzepatide were included; youth with diabetes were excluded. Primary outcome was ever dispensed. Multilevel logistic regression assessed the odds of dispensing by age, sex, and ethnicity, language, Child Opportunity Index (COI), insurance, prescription (Rx) coverage, drug, prescription year, and obesity class.
Results
Among 1,194 AYAs with ≥1 OM prescription, 56.7% received ≥1 fill. Versus semaglutide, odds were higher for liraglutide (OR 1.73), phentermine (OR 3.60), and phentermine–topiramate (OR 2.66), and lower for tirzepatide (OR 0.52; all p≤0.003). Hispanic AYAs had lower odds than non-Hispanic White peers (OR 0.61; p≤0.001). Public (OR 1.31) and mixed insurance (OR 1.63), and Rx coverage (OR 2.00; all p≤0.05), were associated with higher odds. Despite increased rates of prescribing each year, rates of dispensing declined.
Conclusions
Nearly half of AYA OM prescriptions were never dispensed. Barriers to initiation persist and inequities affect Hispanic youth. Addressing insurance/Rx coverage constraints may improve equitable access.
Keywords: adolescent obesity, obesity medications, GLP-1 receptor agonists, electronic health records, medication dispensing
Introduction
Obesity affects one in five United States adolescents and young adults (AYAs) and contributes to early cardiometabolic disease and psychosocial burden.1,2 Intensive Health Behavior and Lifestyle Treatment (IHBLT) remain first-line care, but effects are modest and access is limited.2,3 For AYAs who do not achieve adequate weight loss with IHBLT alone, pharmacotherapy is a critical component of care, with approved options including incretins (GLP-1 receptor agonists (GLP-1 RA) semaglutide, liraglutide, and tirzepatide), phentermine, and phentermine–topiramate. Randomized trials demonstrate clinically meaningful weight loss and cardiometabolic improvements in youth with all OMs, with incretin–based therapies producing the largest effects.4,5
Although GLP-1 RA prescribing to adolescents increased >300% from 2020–2023, fewer than 1% of AYAs with obesity received a prescription, and many faced coverage denials, high out-of-pocket costs, and supply shortages.6,7 While pharmacotherapy can produce meaningful BMI reductions, expected health gains are not realized when prescriptions are not translated into dispensed medications.
Existing evidence comes from plan-policy reviews, claims analyses (mostly adults), and Food and Drug Administration shortage communications,8–10 not from linked pediatric datasets that follow prescriptions to verified dispensing. EHR prescribing data reflect provider intent, whereas dispensing records capture post-prescription barriers such as nonadherence, prior-authorization denial, copayment burden, and supply interruptions.11 Few studies have linked patient-level prescribing and dispensing data,12,13 and none have focused on obesity medications (OM) in AYAs.14
To address this gap, we linked EHR data from a large pediatric health system to pharmacy dispensing records (2020–2025) to (1) describe real-world OM prescribing-dispensing patterns among AYAs and (2) examine sociodemographic and clinical characteristics associated with dispensing. This linked approach provides a more accurate, policy-relevant assessment of AYA access to pharmacologic obesity treatment.
Methods
Study Design and Data Collection
We conducted a retrospective cohort study of AYAs (12–20 years) prescribed OMs between January 1, 2020, and July 28, 2025, in the Nemours Children’s Health System. Reporting follows STROBE for cohort studies and considers the RECORD extension for studies using routinely collected health data (Supplemental File 1).15
Setting
Nemours Children’s Health operates two freestanding children’s hospitals (Wilmington, Delaware and Orlando, Florida) and an extensive primary and specialty outpatient network across Delaware, Florida, Pennsylvania, and New Jersey. A single EHR is used for all sites. The cohort comprises AYAs who received an OM prescription in any Nemours ambulatory and specialty clinics. Observation began at each patient’s first OM prescription during 2020–2025 and continued through July 28, 2025. Data collection reflects routine clinical care and pharmacy dispensing during this period.
Study Population
Eligible participants were Nemours Children’s Health patients aged 12–20 years who had at least one OM prescription in the EHR during 2020–2025. The OMs included were liraglutide, semaglutide, phentermine, phentermine–topiramate, or tirzepatide. Patient prescriptions were identified from the EHR medication orders table using RxNorm and National Drug Codes (NDCs).We excluded patients with type 1 or type 2 diabetes because GLP-1 RAs are often prescribed for glycemic control rather than weight management in this population and because weight management is often complicated by concurrent insulin use.2,16 Diabetes exclusions used ICD-10-CM diagnosis codes recorded in problem lists or encounters prior to or on the index date. Where multiple codes could represent the same construct, we used prespecified hierarchies and clinician review.
Data Sources and Linkage
Data from the Nemours EHR included demographics, encounters, measurements, and medication orders, standardized to the PEDSnet common data model.17 The Nemours Biomedical Research Informatics Center (BRIC) linked patient-level data to dispensing data from (1) Nemours Outpatient Pharmacy and (2) community pharmacies reported to the EHR via Surescripts.18 Surescripts connects most U.S. EHRs, pharmacies, and payers (>99% population coverage). Deterministic linkage used patient and prescription attributes available in both sources (for example, medical record number, date of birth, sex, order and dispense dates, RxNorm/NDC, and dispensing pharmacy). Linkage quality was evaluated using agreement on patient identifiers and allowable date windows. We constructed a longitudinal table with one row per prescription order and per dispensing event.
Primary Outcome
Our primary, patient-level outcome was ever dispensed during 2020–2025. We coded this as yes if the patient with ≥1 OM prescription in the EHR had any linked pharmacy dispensing record for an OM; otherwise no (never dispensed).
Covariates
Demographic and contextual covariates were measured at first prescription: age (continuous and grouped 12–15 and 16–20 years), sex (female or male), race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, other/unknown), preferred language (English, Spanish, other), insurance type (private, public, mixed, other/unknown including none, self-pay, and charity care). The “mixed insurance” category includes patients who had overlapping public and private coverage within a given year (e.g., dual eligibility or transitions between coverage types). Additional covariates included prescription (Rx) insurance (binary indicator of an active pharmacy benefit recorded in the EHR), neighborhood conditions using the Child Opportunity Index (COI) 2.0 national percentile grouped into quintiles (very low, low, moderate, high, very high),19 year of first prescription, and drug category. COI captures census-tract-level conditions across education, health/environment, and social/economic domains and may influence the likelihood of obtaining a fill (e.g., pharmacy proximity, public transportation access, household resources). Clinical adiposity was calculated from measured height and weight recorded at the same visit within −90 to +30 days of the index prescription, and derived BMI-for-age as a percentage of the sex-specific CDC 95th percentile (BMIp95) using the CDC growth-chart SAS code.20 Obesity severity was classified as Class I (100≤120% BMIp95), Class II (120≤140% BMIp95), or Class III (≥140% BMIp95).21,22 Complete code lists and algorithms for the analytic dataset are provided in Supplementary File 2.
Statistical Analysis
We built two analysis files. (1) Patient-level file (one record per AYA) was used to summarize baseline cohort characteristics (Table 1) with counts/percentages for categorical variables and means/standard deviations for continuous variables.23 (2) Patient–drug file (one record per AYA–drug combination) was used for all other summaries and models (Tables 2–6). In the patient–drug file, each AYA contributed a separate observation for every unique obesity medication (liraglutide, semaglutide, phentermine, phentermine–topiramate, tirzepatide) with an index prescription during 2020–2025; the outcome “ever dispensed” was defined within that drug. AYA prescribed ≥2 different OMs appear in multiple rows (n=550).
Table 1:
Participant characteristics at index prescription, overall and by ever dispensed (patient-level)
| Overall | Dispensed | Never Dispensed | |
|---|---|---|---|
| n(%)/mean(SD) | n(%)/mean(SD) | n(%)/mean(SD) | |
| Total | 1194 | 677 (56.7%) | 517 (43.3%) |
| Age, years | 16.81 (2.23) | 17.10 (2.18) | 16.43 (2.25) |
| Sex | |||
| Female | 730 (61.1%) | 428 (58.6%) | 302 (41.4%) |
| Male | 464 (38.9%) | 249 (53.67%) | 215 (46.3%) |
| Race/Ethnicity | |||
| Non-Hispanic White | 451 (37.8%) | 269 (59.6%) | 182 (40.3%) |
| Non-Hispanic Black | 399 (33.4%) | 231 (57.9%) | 168 (42.1%) |
| Hispanic | 286 (23.9%) | 144 (50.3%) | 142 (49.6%) |
| Other | 58 (4.9%) | 33 (56.9%) | 25 (43.1%) |
| Preferred Language | |||
| English | 1072 (89.8%) | 613 (57.2%) | 459 (42.8%) |
| Spanish | 109 (9.1%) | 56 (51.4%) | 53 (48.6%) |
| Other | 13 (1.1%) | 8 (61.5%) | 5 (38.5%) |
| Health Insurance | |||
| Mixed (Public/Private) | 53 (4.4%) | 30 (56.6%) | 23 (43.4%) |
| Private | 456 (38.2%) | 268 (58.8%) | 188 (41.2%) |
| Public (Medicaid/ CHIP/TRICARE) | 449 (37.6%) | 250 (55.7%) | 199 (44.3%) |
| Other (self-pay/charity /unknown) | 236 (19.8%) | 129 (54.7%) | 107 (45.3%) |
| Documented pharmacy benefit | |||
| Yes | 410 (34.3%) | 264 (64.4%) | 146 (35.6%) |
| No | 784 (65.7%) | 413 (52.68%) | 371 (47.32%) |
| Childhood Opportunity Index (national quintile) | |||
| Very low (0–20) | 207 (17.3%) | 111 (53.6%) | 96 (46.4%) |
| Low (21–40) | 199 (16.7%) | 111 (55.8%) | 88 (44.2%) |
| Moderate (41–60) | 337 (28.2%) | 182 (54.0%) | 155 (46.0%) |
| High (61–80) | 224 (18.8%) | 125 (55.8%) | 99 (44.2%) |
| Very High (81–100) | 227 (19.0%) | 148 (65.2%) | 79 (34.8%) |
| BMIp95 | 141.98 (30.30) | 140.16 (28.90) | 144.89 (32.24) |
| Weight Status | |||
| Class I Obesity | 112 (9.4%) | 70 (62.5%) | 42 (37.5%) |
| Class II Obesity | 179 (15.0%) | 111 (62.0%) | 68 (38.0%) |
| Class III Obesity | 304 (25.5%) | 179 (58.9%) | 125 (41.1%) |
| Other/Missing | 599 (50.2%) | 317 (52.9%) | 282 (47.1%) |
| Number of Obesity Medication Classes Prescribed | |||
| 1.46 (0.68) | 1.66 (0.75) | 1.10 (0.45) | |
Note: Values are n (%) or mean (SD). Percentages in the “Dispensed” and “Never dispensed” columns are row percentages. “Ever dispensed” = ≥1 linked pharmacy fill during follow-up. Public insurance includes Medicaid/CHIP/TRICARE. “Other” insurance includes self-pay, charity care, or unknown coverage. Documented pharmacy benefit indicates an active BIN/PCN/group recorded at the index prescription. Child Opportunity Index 2.0 quintiles: very low (0–20), low (21–40), moderate (41–60), high (61–80), very high (81–100). BMIp95 = BMI-for-age expressed as percent of the CDC 95th percentile; weight status: Class I (≥100% to <120%), Class II (≥120% to <140%), Class III (≥140%).
Table 2:
Dispensing Characteristics by Obesity Medication, 2020–2025.
| Overall | Dispensed ≥1 | Never Dispensed | |
|---|---|---|---|
| n (%) | n (%) | n (%) | |
| All Medications | 1744 | 854 (48.97%) | 890 (51.03%) |
| Liraglutide | 437 (25.06%) | 245 (56.06%) | 192 (43.94%) |
| Phentermine | 244 (13.99%) | 159 (65.16%) | 85 (34.84%) |
| Phentermine-Topiramate | 85 (4.87%) | 46 (54.12%) | 39 (45.88%) |
| Semaglutide | 897 (51.43%) | 382 (42.59%) | 515 (57.41%) |
| Tirzepatide | 81 (4.64%) | 22 (27.16%) | 59 (72.84%) |
Note: Data represent per-patient means within each drug category. The dispensing rate is calculated as the ratio of total dispensed to total prescribed fills, averaged within each category. “Ever dispensed” indicates ≥1 fill during 2020–2025; “Never dispensed” indicates none.
Table 6:
Association between Patient-Level Characteristics and Ever Dispensed Obesity Medication (Yes/No) By Drug Category, 2020–2025
| Obesity Medication Dispensing | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Liraglutide | Phentermine | Phentermine-Topiramate | Semaglutide | Tirzepatide | ||||||
| OR (CI) | p | OR (CI) | p | OR (CI) | p | OR (CI) | p | OR (CI) | p | |
| Race/Ethnicity (ref=White) | ||||||||||
| Black | 0.87 (0.52, 1.47) | 0.6 | 0.53 (0.24, 1.19) | 0.12 | 0.40 (0.11, 1.52) | 0.18 | 0.89 (0.62, 1.26) | 0.5 | 5.12 (1.07, 24.40) | 0.04* |
| Hispanic | 1.16 (0.67, 2.01) | 0.59 | 0.23 (0.09, 0.58) | <.002* | 0.25 (0.05, 1.20) | 0.08 | 0.46 (0.30, 0.69) | <.001* | 0.42 (0.05, 3.32) | 0.41 |
| Other | 2.45 (0.55, 10.88) | 0.24 | 0.28 (0.07, 1.14) | 0.07 | 0.33 (0.04, 2.92) | 0.32 | 1.37 (0.71, 2.63) | 0.35 | 2.60 (0.09, 74. 92) | 0.58 |
| Insurance (ref=Private) | ||||||||||
| Mixed | 1.73 (0.78, 3.84) | 0.18 | 2.28 (0.48, 10.77) | 0.3 | 0.35 (0.03, 4.78) | 0.43 | 1.46 (0.77, 2.75) | 0.25 | 2.61 (0.19, 35.82) | 0.47 |
| Public | 1.41 (0.87, 2.29) | 0.16 | 2.47 (1.15, 5.33) | 0.02* | 0.92 (0.21, 4.09) | 0.92 | 0.95 (0.64, 1.40) | 0.79 | 0.69 (0.10, 4.69) | 0.71 |
| Other | 1.51 (0.59, 3.83) | 0.39 | 8.12 (3.00, 22.03) | <.001* | 31.52 (2.46, 404.78) | 0.008* | 3.21 (2.13, 4.82) | <.001* | 8.97 (1.25, 64.21) | 0.03* |
| Rx Insurance (ref=No) | ||||||||||
| Yes | 2.67 (1.61, 4.43) | .0002* | 2.78 (1.32, 5.86) | .007* | 7.04 (1.43, 34.63) | 0.02* | 2.12 (1.50, 3.00) | <.001* | 1.64 (0.29, 9.25) | 0.58 |
| Childhood Opportunity Index Quintiles (ref=Very High) | ||||||||||
| Very Low | 1.19 (0.56, 2.55) | 0.65 | 1.27 (0.44, 3.65) | 0.65 | 0.51 (0.05, 5.49) | 0.58 | 0.61 (0.36, 1.04) | 0.07 | 0.23 (0.02, 3.08) | 0.27 |
| Low | 0.93 (0.45, 1.92) | 0.84 | 0.86 (0.30, 2.42) | 0.77 | 0.32 (0.04, 2.73) | 0.3 | 0.86 (0.53, 1.40) | 0.55 | 0.22 (0.01, 3.59) | 0.3 |
| Moderate | 0.72 (0.38, 1.34) | 0.3 | 1.03 (0.41, 2.61) | 0.94 | 0.27 (0.05, 1.35) | 0.11 | 0.72 (0.47, 1.10) | 0.12 | 1.80 (0.36, 9.01) | 0.48 |
| High | 0.59 (0.29, 1.21) | 0.15 | 0.55 (0.20, 1.51) | 0.25 | 0.13 (0.02, 0.80) | 0.03* | 0.83 (0.53, 1.30) | 0.41 | 0.48 (0.08, 3.11) | 0.44 |
| Prescription Initiation Year | ||||||||||
| 2021 | 0.54 (0.25, 1.68) | 0.37 | ||||||||
| 2022 | 0.78 (0.30, 2.03) | 0.62 | 0.58 (0.13, 2.56) | 0.47 | 0.98 | |||||
| 2023 | 0.60 (0.23, 1.53) | 0.28 | 0.39 (0.10, 1.59) | 0.19 | 0.34 (0.07, 1.60) | 0.17 | 1.29 (0.62, 2.69) | 0.49 | ||
| 2024 | 0.12 (0.04, 0.41) | <.001* | 0.16 (0.04, 0.66) | 0.01* | 0.19 (0.04, 1.01) | 0.05* | 0.400 (0.12, 0.84) | 0.01* | 0.73 (0.06, 9.01 | 0.81 |
| 2025 | 0.09 (0.015, 0.55) | 0.01* | 0.08 (0.02, 0.38) | 0.002* | 1.48 (0.08, 28.00) | 0.8 | 0.354 (0.16, 0.8 | 0.01* | 2.88 (0.21, 38.78) | 0.42 |
Note: ORs (95% CIs) from multivariable logistic regression adjusted for all covariates listed. Significant if CI excludes 1.00; p < 0.05. Each model was adjusted for all covariates shown in the table (race/ethnicity, insurance type, prescription insurance, Childhood Opportunity Index quintile, prescription initiation year).
Bivariate associations with the primary outcome (ever dispensed vs never) were assessed using chi-square tests for categorical variables. Variables were carried forward to multivariable modeling based on statistical significance in bivariate analyses (p<0.05) and a priori relevance.
We then fit multilevel logistic regression models to estimate the association between patient-level factors and the odds of ever being dispensed at least one OM during 2020–2025. Because some individuals contributed more than one drug record, models included a patient-level random intercept to account for within-person clustering.24 The pooled model included drug as a categorical covariate (reference semaglutide) and adjusted for race and ethnicity (reference non-Hispanic White),25 insurance type (private), prescription (Rx) insurance (no), Child Opportunity Index (COI) quintile (very high), and year of index prescription (2020). ORs with 95% CIs and two-sided p-values are reported.26
To examine heterogeneity by medication, we conducted drug-stratified multilevel logistic regressions (one model per drug; same covariates except drug), reported in Table 6 and corresponding figures. Missingness categories (e.g., “other/unknown”) were retained where applicable to avoid listwise deletion. Analyses used SAS 9.4 with α=0.05; reproducible code is provided in Supplementary File 3.27 The Nemours IRB approved procedures and waived consent for use of deidentified EHR data.
Results
From 2020–2025, 1,194 AYAs received ≥1 OM prescription and 56.7% (n=677) were ever dispensed at least one fill and 43.3% (n=517) were never dispensed (Table 1). The mean age at first prescription was 16.8 (SD 2.2) years. The cohort was 61.1% female. By race and ethnicity, 37.8% were non-Hispanic White, 33.4% non-Hispanic Black, 23.9% Hispanic, and 4.9% other/unknown. Most patients reported English as their preferred language (89.8%). Insurance at first prescription was 38.2% private, 37.6% public, 4.4% mixed, and 19.8% other/unknown; 34.3% had prescription (Rx) insurance documented. Only 50.8% had a BMIp95 recorded within the prespecified window around the first prescription; among those with a measure, 9.4% had Class I obesity, 15.0% had Class II obesity, and 25.5% had Class III obesity.
Among 1,744 prescriptions (Table 2), semaglutide comprised 51.4%, followed by liraglutide (25.1%), phentermine (14.0%), phentermine–topiramate (4.9%), and tirzepatide (4.6%). Ever dispensed proportions varied by drug, highest for phentermine (65.2%) and lower for liraglutide (56.1%), phentermine–topiramate (54.1%), semaglutide (42.6%), and tirzepatide (27.2%); mean dispensing rates showed the same ordering (Table 2). Prescription initiations shifted over time (Table 3): liraglutide dominated early (2020–2022), whereas semaglutide became the majority from 2023 onward (57.9% in 2023, 68.8% in 2024, 59.8% in 2025); tirzepatide appeared in 2023 and increased to 14.1% by 2025, while phentermine maintained a stable 12–22% share after introduction in 2021 and phentermine–topiramate remained ~5–6% after its 2022 introduction.
Table 3:
Patient Drug Prescription Initiation by Year (2020–2025)
| Year | Liraglutide (n/%) | Phentermine (n/%) | Phentermine-Topiramate (n/%) | Semaglutide (n/%) | Tirzepatide (n/%) | Total (n/%) |
|---|---|---|---|---|---|---|
| 2020 | 25 (100%) | 0 | 0 | 0 | 0 | 25 (1.43%) |
| 2021 | 120 (84.51%) | 20 (14.08%) | 0 | 2 (1.41%) | 0 | 142 (8.14%) |
| 2022 | 111 (52.36%) | 47 (22.17%) | 18 (8.49%) | 36 (16.98%) | 0 | 212 (12.16%) |
| 2023 | 139 (23.8%) | 74 (12.67%) | 26 (4.45%) | 338 (57.88%) | 7 (1.2%) | 584 (33.49%) |
| 2024 | 31 (5.19%) | 72 (12.06%) | 35 (5.86%) | 411 (68.84%) | 48 (8.04%) | 597 (34.23%) |
| 2025 | 11 (5.98%) | 31 (16.85%) | 6 (3.26%) | 110 (59.78%) | 26 (14.13%) | 184 (10.55%) |
| Total | 437 (25.06%) | 244 (13.99%) | 85 (4.87%) | 897 (51.43%) | 81 (4.64%) | 1744 |
In bivariate analyses (Table 4), race and ethnicity, insurance type, Rx insurance, drug category, and year of first prescription were significantly associated with ever dispensed (all p≤0.05) and thus were retained in the multivariable regression models. A borderline trend was observed for COI quintile (p=0.06). However, it was evaluated further in adjusted models because of its conceptual relevance. Age category, sex, BMI category, and preferred language were not significantly associated with dispensing, and were thus not included in the regression models.
Table 4:
Bivariate Associations with Patient-Level Dispensing (Ever vs Never)
| Never Dispensed | Dispensed | Total | ||
|---|---|---|---|---|
| n (%) | n (%) | n (%) | ||
| Age Category | ||||
| 12–15 years | 232 (52.73) | 208 (47.27) | 440 (25.23) | 0.41 |
| 16–20 years | 658 (50.46) | 646 (49.54) | 1304 (74.77) | |
| Sex | ||||
| Male | 355 (52.75) | 318 (47.25) | 673 (38.59) | 0.25 |
| Female | 535 (49.95) | 536 (50.05) | 1071 (61.41) | |
| Race/Ethnicity | ||||
| Black | 300 (51.11) | 287 (48.89) | 587 (33.66) | 0.02* |
| Hispanic | 242 (56.94) | 183 (43.06) | 425 (24.37) | |
| Other/Unknown | 37 (46.25) | 43 (53.75) | 80 (4.59) | |
| White | 311 (47.70) | 341 (52.30) | 652 (37.39) | |
| Bmip95 Category | ||||
| Class I Obesity | 81 (48.21) | 87 (51.79) | 168 (9.63) | 0.28 |
| Class II Obesity | 130 (46.76) | 148 (53.24) | 278 (15.94) | |
| Class III Obesity | 249 (51.13) | 238 (48.87) | 487 (27.92) | |
| Missing | 430 (53.02) | 381 (46.98) | 811 (46.50) | |
| Language | ||||
| English | 796 (50.67) | 775 (49.33) | 1571 (90.08) | 0.54 |
| Other | 7 (46.67) | 8 (53.33) | 15 (0.86) | |
| Spanish | 87 (55.06) | 71 (44.94) | 158 (9.06) | |
| Rx Insurance | ||||
| No | 646 (52.99) | 573 (47.01) | 1219 (69.90) | 0.01* |
| Yes | 244 (46.48) | 281 (53.52) | 525 (30.10) | |
| Insurance | ||||
| Mixed (Public + Private) | 50 (45.45) | 60 (54.55) | 110 (6.31) | <0.001* |
| Other/Unknown | 166 (42.89) | 221 (57.11) | 387 (22.19) | |
| Private only | 362 (54.68) | 300 (45.32) | 662 (37.96) | |
| Public only | 312 (53.33) | 273 (46.67) | 585 (33.54) | |
| Childhood Opportunity Index (national quintile) | ||||
| Very low (0–20) | 146 (53.28) | 128 (46.72) | 274 (15.71) | 0.06 |
| Low (21–40) | 147 (51.58) | 138 (48.42) | 285 (16.34) | |
| Moderate (41–60) | 263 (52.08) | 242 (47.92) | 505 (28.96) | |
| High (61–80) | 181 (54.35) | 152 (45.65) | 333 (19.09) | |
| Very High (81–100) | 153 (44.09) | 194 (55.91) | 347 (19.90) | |
| Drug Category | ||||
| Liraglutide | 192 (43.94) | 245 (56.06) | 437 (25.06) | <0.001* |
| Phentermine | 85 (34.84) | 159 (65.16) | 244 (13.99) | |
| Phentermine-Topiramate | 39 (45.88) | 46 (54.12) | 85 (4.87) | |
| Semaglutide | 515 (57.41) | 382 (42.59) | 897 (51.43) | |
| Tirzepatide | 59 (72.84) | 22 (27.16) | 81 (4.64) | |
| Prescription Initiation Year b | ||||
| 2020 | 8 (32) | 17 (68) | 25 (1.43) | <0.001* |
| 2021 | 53 (37.32) | 89 (62.68) | 142 (8.14) | |
| 2022 | 79 (37.26) | 133 (62.74) | 212 (12.16) | |
| 2023 | 255 (43.66) | 329 (56.34) | 584 (33.49) | |
| 2024 | 371 (62.14) | 226 (37.86) | 597 (34.23) | |
| 2025 | 124 (67.39) | 60 (32.61) | 184 (10.55) | |
Note: This analysis was conducted with patient-drug level data (n=1744). Individual patients may be repeated if they had multple types of obesity medication drug prescriptions. BMIp95 = BMI-for-age percentile. “Drug initiation year” refers to the calendar year of the first OM prescription during 2020–2025.
In adjusted models (Table 5), dispensing differed by OM, insurance, Rx coverage, race and ethnicity, and COI. Relative to semaglutide, odds of ever being dispensed were higher for liraglutide, phentermine, and phentermine–topiramate, and lower for tirzepatide (all p≤0.003). Hispanic AYAs had lower odds of ever being dispensed an OM than non-Hispanic White AYAs (OR 0.61 [95% CI 0.46–0.82]; p<0.001). Compared with private insurance, public (OR 1.31 [95% CI 1.00–1.70]; p<0.05) and mixed coverage (OR 1.63 [95% CI 1.04–2.56]; p<0.05) showed modestly higher odds, and other/unknown coverage (OR 3.06 [95% CI 2.23–4.20]; p<0.01) showed the highest odds of being dispensed an OM. Having Rx insurance was also associated with double the odds of dispensing (OR 2.00 [95% CI 1.55–2.56]; p<0.01). Residence in high COI areas (61–80th percentile) was associated with lower odds of dispensing compared with very high COI (81–100th percentile) (OR 0.67 [95% CI 0.48–0.94]; p<0.05).
Table 5:
Association between Patient-Level Characteristics and Ever Dispensed Obesity Medication (Yes/No), 2020–2025
| Ever Dispensed | |||
|---|---|---|---|
| OR (95% CI) | P-Value | ||
| Drug Category (ref = Semaglutide) | |||
| Liraglutide | 2.40 | 1.85, 3.12 | <.001 * |
| Phentermine | 3.32 | 2.38, 4.62 | <.001 * |
| Phentermine-Topiramate | 2.16 | 1.33, 3.50 | 0.002 * |
| Tirzepatide | 0.45 | 0.25, 0.75 | 0.003 * |
| Race/Ethnicity (ref = White) | |||
| Black | 0.86 | 0.67, 1.12 | 0.27 |
| Hispanic | 0.61 | 0.46,0.82 | <0.001 * |
| Other/Unknown | 1.06 | 0.64, 1.76 | 0.82 |
| Insurance (ref = Private) | |||
| Mixed (Public/Private) | 1.63 | 1.04, 2.56 | 0.03 * |
| Public | 1.31 | 1.00, 1.70 | 0.05 * |
| Other/Unknown | 3.06 | 2.23, 4.20 | <.001 * |
| Rx Insurance (ref = No) | |||
| Yes | 2.00 | 1.55, 2.56 | <.001 * |
| Childhood Opportunity Index Quintiles (ref = Very High [5]) | |||
| Very low (0–20) | 0.70 | 0.48, 1.01 | 0.06 |
| Low (21–40) | 0.78 | 0.54, 1.11 | 0.17 |
| Moderate (41–60) | 0.76 | 0.56, 1.03 | 0.08 |
| High (61–80) | 0.67 | 0.48, 0.94 | 0.02 * |
ORs (95% CIs) from multivariable logistic regression adjusted for all covariates listed. Significant if CI excludes 1.00; p < 0.05. Each model was adjusted for all covariates shown in the table (drug category, race/ethnicity, insurance type, prescription insurance, and Childhood Opportunity Index quintile).
Associations varied by drug (Table 6). Hispanic AYAs had significantly lower odds of dispensing for phentermine (OR 0.23 [95% CI 0.09–0.58]; p<0.005) and semaglutide (OR 0.46 [95% CI 0.30–0.69]; p<0.001), while Black AYAs had higher odds for tirzepatide (OR 5.12 [95% CI 1.07–24.40]; p<0.05) compared to non-Hispanic White peers. Public insurance was associated with higher odds of dispensing for phentermine (OR 2.47 [95% CI 1.15–5.33]; p<0.05) vs. private. Other/unknown insurance showed higher odds of dispensing across multiple drugs, including phentermine, phentermine–topiramate, semaglutide, and tirzepatide (see Table 6 for estimates). Having Rx insurance was consistently associated with higher odds for liraglutide, phentermine, phentermine–topiramate, and semaglutide (all p≤0.05), but not for tirzepatide. COI was generally not associated with dispensing by drug except for phentermine–topiramate, where high COI had lower odds vs. very high (OR 0.13 [95% CI 0.02–0.80]; p<0.05). Dispensing declined for prescriptions initiated in 2024–2025 across liraglutide, phentermine, and semaglutide (see Table 6 for estimates).
Discussion
This study is among the first to link pediatric prescribing data with external pharmacy dispensing, offering a more complete view of real-world access to OMs. Only about half prescribed an OM obtained at least one fill during 2020–2025, demonstrating substantial attrition between prescribing and treatment initiation. Our outcome of ever dispensed captures this first step in the treatment cascade. These results reveal gaps, not in prescribing intent, but in the ability to translate a prescription into actual medication dispensing.13 Our results demonstrate significant sociodemographic disparities in dispensing, signaling barriers to access for OMs among certain subgroups that could exacerbate disparities in obesity treatment and outcomes.
Our observed dispensing rate aligns with published data on new prescription fill rates in pediatric populations generally. Nationally, 38–56% of prescriptions for pediatric chronic conditions are ever filled, a range that is consistent with our findings.28 This context suggests that while obesity medications face unique coverage and cost barriers, their initial fill rates are not lower than expected for comparable treatments for chronic conditions. Similar patterns have also been reported in adults, with studies showing that approximately half of adults with obesity fill an OM prescription and about 60% of GLP-1 receptor agonist orders are filled within 90 days.14,29
Not surprisingly, prescribing shifted during the period studied toward incretin therapies (semaglutide, later tirzepatide) yet dispenses were less likely for incretin agents than for lower-cost oral options (phentermine, phentermine-topiramate).9,10 Likely contributors include intermittent supply shortages of semaglutide,6 stricter insurance coverage and higher cost sharing for incretins compared with oral drugs,11,30,31 and age-based labeling and administration route (injections vs oral).10 Initial prescriptions written in 2024–2025 were less likely to result in dispensing than those written earlier, which may be due to tightening payer controls amid rising demand and episodic supply constraints.
The observed racial and ethnic differences in medication dispensing reflect structural barriers that occur after a prescription is written. Hispanic AYAs were less likely to be dispensed an OM, a pattern consistent with broader inequities in access to obesity treatment and other chronic disease medications. Prior studies have shown that Hispanic families face higher rates of coverage denials, more complex prior-authorization processes, and greater financial and logistical barriers to filling prescriptions. Language and cultural factors, transportation limitations, and limited availability of bilingual pharmacy or clinic staff may further complicate access.14 In addition, cultural perceptions of body weight may contribute to lower uptake of obesity medications. For example, Hispanic parents are more likely to perceive a higher BMI to be normal or healthy compared with non-Hispanic White parents, which may reduce perceived need for pharmacologic treatment.32 Our findings, supported by the literature, suggest that greater attention to reducing these systematic barriers and providing culturally competent care is needed to ensure equitable access to OMs.
In the primary model, AYAs with mixed or public insurance had higher odds of being dispensed any OM than those with private insurance, and the other/unknown group had the highest odds of being dispensed any OM. Several features may explain dispensing heterogeneity by drug and insurance plan. First, utilization management and cost sharing for incretin therapies are common and vary across plans, which can be a major barrier to dispensing of semaglutide and tirzepatide. Clinic support (for example, a specialty pharmacy initiating prior authorizations) can mitigate barriers but may not be available in all settings. Second, the availability of lower-cost generic options, for example Phentermine, may facilitate dispensing. It is likely that the earlier approval of liraglutide explains its comparatively higher odds of being dispensed compared to the other incretins. Also, most of the patients in our cohort were from our main site in Delaware where Medicaid has historically been favorable to covering OMs. Finally, it is likely that patients listed as having no insurance were able to pay out of pocket for newer, higher cost medications, exacerbating inequities in access to OMs.
Interestingly, documentation of an active pharmacy benefit in the EHR at the index prescription was associated with higher odds of dispensing across all OMs (except tirzepatide). In theory, everyone with insurance should also have Rx insurance. However, the EHR does not reflect this in all cases, which may be related to patient understanding and knowledge of their insurance plans. While the Rx insurance may just be noted in the EHR for billing purposes, it is possible that this notation could expedite insurance or billing authorization processes for medication coverage or be a proxy for health literacy and a patient’s ability to navigate the healthcare system.
Finally, AYAs with very high COI quintiles had the highest dispensing rates for all OMs, exacerbating disparities in OM access. Overall, COI gradients were modest, suggesting that neighborhood advantage may help with transportation and pharmacy access, but the effect was limited in this cohort, suggesting that other factors like ethnicity and insurance were primary drivers of sociodemographic disparities. Future work should pair COI with direct measures of pharmacy access, transportation, and plan policies to clarify mechanisms.2,13
Several limitations should be acknowledged. Findings reflect a single health-system and may not generalize to settings with different payer/prescribing contexts. Dispensing may have been underestimated if dispenses occurred outside SureScripts-linked pharmacies, through cash payments, or via mail-order. The outcome of ever dispensed (yes vs. no) captures initial access but not the timing of fills. Future studies should link prescriptions and dispenses temporally to assess primary adherence within a certain timeframe and identify delays in treatment initiation. Some covariates, including BMI-for-age percentile (calculated from height and weight at the same visit), were incomplete likely reflecting variability in visit timing and record-keeping across clinics, and insurance type was defined at first prescription only. Smaller subgroups, particularly for newer drugs, produced wide confidence intervals. Finally, 2025 data were truncated in July, limiting capture of recent dispensing activity. Future studies should assess longitudinal dispensing patterns and other patterns of medication adherence across different OMs, as well as state- or insurer-level differences in coverage and authorization practices. Linking outcomes such as weight change, cardiometabolic risk, and health-care utilization will also be critical to evaluate the effectiveness and equity of pharmacologic obesity treatment in youth.
Conclusion
In this multi-state pediatric cohort linking electronic health record data with pharmacy dispensing data, only half of AYAs prescribed an obesity medication were ever dispensed the medication. Incretin therapies were prescribed increasingly over time but were dispensed less often than lower-cost oral options. Hispanic AYAs were less likely to be dispensed an OM than non-Hispanic White peers and insurance features, including documentation of an active pharmacy benefit and not having insurance, were strongly associated with dispensing. Our findings suggest that systemic barriers to accessing OMs exist and exacerbate disparities in obesity treatment among certain populations. More work is needed to understand barriers to OM access in the ever-changing landscape of pharmacologic treatment for obesity and to advocate for increased access for all children.
Supplementary Material
STUDY IMPORTANCE:
Few studies have successfully linked prescribing and dispensing data in adolescents and young adults, and none have examined this relationship with obesity medications. Our study links EHR prescribing data from a large pediatric health system with external pharmacy dispensing records to provide a more accurate and policy-relevant view of adolescent and young adult access to pharmacologic obesity treatment during a period of rapid expansion.
We found a substantial proportion of prescriptions are not being dispensed and the odds of being dispensed are significantly associated with medication class, ethnicity, insurance, and year of medication prescription.
FUNDING:
BRIC is supported by PEDSnet (RI-CHOP-01-PS1). Dr. Thao-Ly Phan is supported by the REACH Center of Biomedical Research Excellence (P20GM144270), and IDeA States Pediatric Clinical Trials Network (UG1OD090915 and UG1OD024958).
DISCLOSURE:
The authors declared no conflicts of interest. K. Arasteh reports research funding from the National Institutes of Health, the American Cancer Society, and JDRF, outside the submitted work. H. Timothy Bunnell reports support for the present manuscript from PCORI through PEDSnet and Nemours Children’s Health. Thao-Ly T. Phan is supported by the REACH Center of Biomedical Research Excellence (P20GM144270) and the IDeA States Pediatric Clinical Trials Network (UG1OD090915 and UG1OD024958). All other authors declare no potential conflicts of interest relevant to this article.
References
- 1.Skinner AC, Ravanbakht SN, Skelton JA, Perrin EM, Armstrong SC. Prevalence of obesity and severe obesity in US children, 1999–2016. Pediatrics. 2018;141(3) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hampl SE, Hassink SG, Skinner AC, et al. Clinical Practice Guideline for the Evaluation and Treatment of Children and Adolescents With Obesity. Pediatrics. 2023;151(2)doi: 10.1542/peds.2022-060640 [DOI] [PubMed] [Google Scholar]
- 3.Grossman DC, Bibbins-Domingo K, Curry SJ, et al. Screening for Obesity in Children and Adolescents: US Preventive Services Task Force Recommendation Statement. Jama. Jun 20 2017;317(23):2417–2426. doi: 10.1001/jama.2017.6803 [DOI] [PubMed] [Google Scholar]
- 4.Weghuber D, Barrett T, Barrientos-Pérez M, et al. Once-Weekly Semaglutide in Adolescents with Obesity. N Engl J Med. Dec 15 2022;387(24):2245–2257. doi: 10.1056/NEJMoa2208601 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kelly AS, Auerbach P, Barrientos-Perez M, et al. A Randomized, Controlled Trial of Liraglutide for Adolescents with Obesity. New England Journal of Medicine. 2020;382(22):2117–2128. doi:doi: 10.1056/NEJMoa1916038 [DOI] [PubMed] [Google Scholar]
- 6.(FDA) USFDA. Declaratory Order: Resolution of Shortages of Semaglutide Injection Products (Ozempic and Wegovy). Updated February 21, 2025. Accessed October 3, 2025. https://www.fda.gov/drugs/drug-safety-and-availability/fda-clarifies-policies-compounders-national-glp-1-supply-begins-stabilize
- 7.Resolution of Tirzepatide Injection Product Shortage and Supply Status. (2024).
- 8.Czepiel KS, Perez NP, Campoverde Reyes KJ, Sabharwal S, Stanford FC. Pharmacotherapy for the treatment of overweight and obesity in children, adolescents, and young adults in a large health system in the US. Frontiers in Endocrinology. 2020;11:290. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kompaniyets L, Pierce SL, Porter R, Autrey K, Chua KP, Belay B, Blanck HM, Goodman AB. Prescriptions for Obesity Medications Among Adolescents Aged 12–17 Years with Obesity — United States, 2018–2023. Vol. 74. 2025;74:337–344. . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lee JM, Sharifi M, Oshman L, Griauzde DH, Chua K-P. Dispensing of Glucagon-Like Peptide-1 Receptor Agonists to Adolescents and Young Adults, 2020–2023. JAMA. 2024;331(23):2041–2043. doi: 10.1001/jama.2024.7112 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Fischer MA, Stedman MR, Lii J, et al. Primary medication non-adherence: analysis of 195,930 electronic prescriptions. J Gen Intern Med. Apr 2010;25(4):284–90. doi: 10.1007/s11606-010-1253-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Zweigoron RT, Binns HJ, Tanz RR. Unfilled prescriptions in pediatric primary care. Pediatrics. Oct 2012;130(4):620–6. doi: 10.1542/peds.2011-3480 [DOI] [PubMed] [Google Scholar]
- 13.Wu P, Hurst JH, French A, Chrestensen M, Goldstein BA. Linking Electronic Health Record Prescribing Data and Pharmacy Dispensing Records to Identify Patient-Level Factors Associated With Psychotropic Medication Receipt: Retrospective Study. JMIR Med Inform. Mar 4 2025;13:e63740. doi: 10.2196/63740 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gasoyan H, Pfoh ER, Schulte R, Sullivan E, Le P, Rothberg MB. Association of patient characteristics and insurance type with anti-obesity medications prescribing and fills. Diabetes, Obesity and Metabolism. 2024;26(5):1687–1696. doi: 10.1111/dom.15473 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Benchimol EI, Smeeth L, Guttmann A, et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med. Oct 2015;12(10):e1001885. doi: 10.1371/journal.pmed.1001885 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Alfaris N, Waldrop S, Johnson V, Boaventura B, Kendrick K, Stanford FC. GLP-1 single, dual, and triple receptor agonists for treating type 2 diabetes and obesity: a narrative review. EClinicalMedicine. Sep 2024;75:102782. doi: 10.1016/j.eclinm.2024.102782 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Forrest CB, Margolis PA, Bailey LC, et al. PEDSnet: a National Pediatric Learning Health System. J Am Med Inform Assoc. Jul-Aug 2014;21(4):602–6. doi: 10.1136/amiajnl-2014-002743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Surescripts. Why it matters: prescribers can now access accurate medication history data for virtually every American. Accessed October 2, 2025. https://surescripts.com/news-center/intelligence-in-action/interoperability/why-it-matters-prescribers-can-now-access-accurate-medication-history-data-for-virtually-every-american/.
- 19.Acevedo-Garcia D, Noelke C, McArdle N, et al. Racial and ethnic inequities in children’s neighborhoods: evidence from the new Child Opportunity Index 2.0: study uses the Child Opportunity Index 2.0 to examine geographic and racial/ethnic inequities children are exposed to in the one hundred largest metropolitan areas of the United States. Health Affairs. 2020;39(10):1693–1701. [DOI] [PubMed] [Google Scholar]
- 20.SAS Program for WHO Growth Charts. Updated September 24, 2024. Accessed October 2, 2025, https://www.cdc.gov/growth-chart-training/hcp/computer-programs/sas-who.html
- 21.Li X, Keown-Stoneman CDG, Lebovic G, et al. The association between body mass index trajectories and cardiometabolic risk in young children. Pediatr Obes. Aug 2020;15(8):e12633. doi: 10.1111/ijpo.12633 [DOI] [PubMed] [Google Scholar]
- 22.Pratt JS, Browne A, Browne NT, et al. ASMBS pediatric metabolic and bariatric surgery guidelines, 2018. Surgery for Obesity and Related Diseases. 2018;14(7):882–901. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bursac Z, Gauss CH, Williams DK, Hosmer DW. Purposeful selection of variables in logistic regression. Source Code for Biology and Medicine. 2008/12/16 2008;3(1):17. doi: 10.1186/1751-0473-3-17 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Stawski RS. Multilevel analysis: an introduction to basic and advanced multilevel modeling. Taylor & Francis; 2013. [Google Scholar]
- 25.Flanagin A, Frey T, Christiansen SL, Committee AMoS. Updated Guidance on the Reporting of Race and Ethnicity in Medical and Science Journals. JAMA. 2021;326(7):621–627. doi: 10.1001/jama.2021.13304 [DOI] [PubMed] [Google Scholar]
- 26.Vittinghoff E, Shiboski S, Glidden D, McCulloch C. Regression Methods in Biostatistics: Linear, Logistic, Survival and Repeated Measures Models. New York:Springer. November/30 2004;doi: 10.1007/b138825 [DOI] [Google Scholar]
- 27.Inc SI. SAS/STAT® 15.3 User’s Guide. SAS Institute Inc.; 2023. [Google Scholar]
- 28.El-Rachidi S, LaRochelle JM, Morgan JA. Pharmacists and Pediatric Medication Adherence: Bridging the Gap. Hosp Pharm. Feb 2017;52(2):124–131. doi: 10.1310/hpj5202-124 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sarpatwari A, Soto MJ, Ganguli I, Sloan CE, Goss F, Sinaiko AD. Glucagon-Like Peptide-1 Receptor Agonist Order Fills and Out-of-Pocket Costs by Race, Ethnicity, and Indication. JAMA Health Forum. Oct 3 2025;6(10):e254258. doi: 10.1001/jamahealthforum.2025.4258 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Williams ERR, Bell C. Medicaid Coverage of and Spending on GLP-1s. 2024. Nov 4, 2024. https://www.kff.org/medicaid/medicaid-coverage-of-and-spending-on-glp-1s/
- 31.Khandelwal N, Duncan I, Rubinstein E, et al. Medication adherence for 90-day quantities of medication dispensed through retail and mail order pharmacies. Am J Manag Care. Nov 2011;17(11):e427–34. [PubMed] [Google Scholar]
- 32.Martin MA, May AL, Frisco ML. Equal weights but different weight perceptions among US adolescents. J Health Psychol. May 2010;15(4):493–504. doi: 10.1177/1359105309355334 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.